d3.6 dynamic radio resource management for hybrid ... · pdf filed3.6: dynamic radio resource...

SANSA-645047 D3.6

D3.6

Dynamic Radio Resource Management for Hybrid Terrestrial-Satellite Backhauling

Grant Agreement nº: 645047 Project Acronym: SANSA Project Title: Shared Access Terrestrial-Satellite Backhaul

Network enabled by Smart Antennas Contractual delivery date:

30/04/2017

Actual delivery date: 28/04/2017 Contributing WP WP3 Dissemination level: Public Editors: ULUX Contributors: ULUX, AIT, CTTC

Abstract: This deliverable includes the investigations and findings on Task 3.2 Dynamic Radio Resource Management. Based on the defined work plan and scenario definitions in WP2, this deliverable focuses on performance assessment and adaptation of scheduling techniques, carrier allocation, power control, flow control and delay-aware scheduling in the SANSA scenarios. In evaluating the performance results, specific attention is given to the Spectral Efficiency (SE), which relates system throughput and total bandwidth. Based on the main conclusions of this deliverable, the proposed radio resource management techniques is shown to reach SE gains of the order of 2-3x in terms of SE when compared to the benchmark.

D3.6: Dynamic Radio Resource Management for Hybrid Terrestrial-Satellite Backhauling

Date: 21/04/2017

of 87

SANSA-645047 D3.6

Document History Version Date Editor Modification 0.0 01/08/2016 ULUX Initial ToC 0.1 12/01/2017 ULUX Initial input from ULUX regarding introduction,

SANSA scenarios and requirements. 0.2 10/02/2017 AIT, CTTC,

ULUX First round of contributions (before F2F). Version 1 circulated to partners.

0.3 09/03/2017 ULUX Version 2 circulated to partners. 0.4 20/03/2017 ULUX Last day to send inputs 0.5 28/03/2017 ULUX Document Ready for QA 0.6 28/03/2017 AIT Input on Section 7 0.7 31/03/2017 AIT Input on Section 8 0.8 07/04/2017 ULUX QA by AIT, FRA, CTTC 1.0 21/04/2017 ULUX Final document sent to coordinator 1.1 26/04/2017 CTTC Final document ready for submission


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Contributors Name Company Contributions included Eva Lagunas ULUX All document [main editor] Symeon Chatzinotas ULUX All document Miguel Ángel Vázquez CTTC Section 3, Section 4 Musbah Shaat CTTC Section 3, Section 7 Ana Isabel Pérez-Neira CTTC Section 7 and QA Kostas Voulgaris AIT Section 3, Section 8 George Papageorgiou AIT Section 5 and QA Constantinos B. Papadias AIT Section 3, Section 5, Section 8 Venkatesh Ramireddy FRA QA


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Table of Contents List of Figures ................................................................................................................................ 6

List of Tables .................................................................................................................................. 8

List of Acronyms ............................................................................................................................ 9

Executive Summary ..................................................................................................................... 10

1 Introduction ........................................................................................................................ 12

2 Summary of SANSA scenarios and radio resource management requirements ................. 14

3 Motivation of radio resource management ........................................................................ 16

4 Scheduling, precoding and power control for single BS point-to-multipoint backhaul connections ................................................................................................................................. 18

4.1 Motivation ....................................................................................................................... 18

4.2 System Model .................................................................................................................. 19

4.3 Low-complexity precoding .............................................................................................. 21

4.4 Scheduling ....................................................................................................................... 22

4.5 Power Control ................................................................................................................. 23

4.6 Numerical Results ............................................................................................................ 23

5 Power control and precoding design for single BS point-to-point backhaul connections .. 26

5.1 Motivation ....................................................................................................................... 26

5.2 Model and Problem Formulation .................................................................................... 27

5.3 Solution and Algorithm ................................................................................................... 30

5.4 Experimental Evaluation ................................................................................................. 32

6 Carrier allocation for hybrid satellite-terrestrial backhaul networks ................................. 34

6.1 Problem formulation ....................................................................................................... 36

6.1.1 Terrestrial segment model .................................................................................. 36

6.1.2 Satellite segment model ...................................................................................... 37

6.2 Proposed carrier allocation algorithm ............................................................................ 38

6.2.1 Step 1: Carrier Allocation for the Satellite Backhaul Network ............................ 39

6.2.2 Step 2: Carrier Allocation for the Terrestrial Backhaul Network ........................ 40

6.3 Simulation results ............................................................................................................ 43

6.3.1 Benchmark carrier allocation .............................................................................. 43

6.3.2 Simulation set-up: Extended Helsinki topology .................................................. 45


Date: 21/04/2017

of 87

SANSA-645047 D3.6

6.3.3 Extended Benchmark (without satellite links) .................................................... 47

6.3.4 Spectral efficiency analysis .................................................................................. 50

7 Carrier assignment and flow control for hybrid satellite-terrestrial backhaul networks ... 56

7.1 Introduction .................................................................................................................... 56

7.2 Related work ................................................................................................................... 56

7.3 System model .................................................................................................................. 57

7.4 Problem formulation and solution approach .................................................................. 60

7.5 Performance evaluation and comparison with Helsinki topology benchmark ............... 63

7.5.1 Extended Benchmark (without satellite links) .................................................... 63

7.5.2 Performance evaluation considering 56 MHz channels ...................................... 65

7.5.3 Performance evaluation considering variable channel width............................. 67

8 QoS-aware satellite scheduling ........................................................................................... 69

8.1 Motivation ....................................................................................................................... 69

8.2 Description of the satellite communications system and problem definition ................ 70

8.2.1 Satellite system description ................................................................................ 70

8.2.2 Adaptive Coding and Modulation ....................................................................... 71

8.2.3 Generic Stream Encapsulation in DVB-S2 ........................................................... 71

8.3 Scheduling of SANSA backhaul packets .......................................................................... 72

8.3.1 Scheduling algorithm for the Satellite forward link ............................................ 72

8.4 Simulation setup ............................................................................................................. 73

8.4.1 ACM implementation .......................................................................................... 74

8.4.2 Channel variation implementation ..................................................................... 75

8.4.3 Generic Stream Encapsulation implementation ................................................. 76

8.4.4 QoS parameters of user packets ......................................................................... 76

8.5 Simulation results ............................................................................................................ 77

8.5.1 Proportional fairness ........................................................................................... 77

8.5.2 Probability to transmit packets from each class ................................................. 78

8.5.3 Effect of the scheduling algorithms on the total throughput ............................. 79

9 Conclusions ......................................................................................................................... 81

References ................................................................................................................................... 83


Date: 21/04/2017

of 87

SANSA-645047 D3.6

List of Figures Figure 1-1. Database-assisted resource management envisioned in SANSA project ................. 13 Figure 2-1. SANSA inter-system interference scenarios ............................................................. 15 Figure 2-2. CEPT Band segmentation for FSS and FS in the band 27.5-29.5 GHz. The band labelled Sat* indicates that there are legacy FS in a few countries. ........................................... 15 Figure 4-1. Scheduling designed for M=3 .................................................................................... 20 Figure 4-2 Sum-rate versus transmit power for different δs and δt. It is considered L = 1, Q = 64 and NRF = 4. ................................................................................................................................ 24 Figure 4-3. Helsinki point-to-multipoint scenario under evaluation........................................... 25 Figure 5-1: Spectrum reuse MIMO setup considering a desired and an interfering link. ........... 28 Figure 5-2: Representation of the optimizations task’s feasible region for 𝑟𝑟 = 2. .................... 30 Figure 5-3: Capacity versus various values of interference constraints, 𝑃𝑃𝑃𝑃 , for a 4 × 4 MIMO link. (a) The dashed line corresponds to the maximum capacity of a single unconstrained link (ignoring the interference) and the solid one corresponds to the achieved capacity of the proposed method, which guarantees that no interference is caused to the protected link. (b) The penalty, which corresponds to the percentage of capacity loss, for different values of interference level 𝑃𝑃𝑃𝑃. .................................................................................................................. 33 Figure 5-4: CDF's of a 4 × 4 MIMO link for different capacity values, which correspond to interference levels 𝑃𝑃𝑃𝑃. The dashed line corresponds to the case where no interference constraint exists. ......................................................................................................................... 34 Figure 6-1. Hybrid satellite-terrestrial backhaul network topology ............................................ 35 Figure 6-2. Helsinki topology ....................................................................................................... 43 Figure 6-3. Benchmark carrier assignment ................................................................................. 44 Figure 6-4. Extended Helsinki topology ...................................................................................... 46 Figure 6-5. Simulated HSTB topology with the simulated satellite beam pattern. ..................... 46 Figure 6-6. Benchmark carrier allocation for the extended Helsinki topology (22 bi-directional links) ............................................................................................................................................ 49 Figure 6-7. SE of the hybrid terrestrial-satellite network as a function of the carrier bandwidth ..................................................................................................................................................... 53 Figure 6-8. CDF of terrestrial SINR distribution ........................................................................... 53 Figure 6-9. EE as a function of the carrier bandwidth ................................................................. 54 Figure 6-10. SE of the hybrid terrestrial-satellite network as a function of the number of carriers ........................................................................................................................................ 55 Figure 6-11. CDF of terrestrial SINR distribution ......................................................................... 55 Figure 7-1 Example of wireless backhauling network ................................................................. 58 Figure 7-2 The ratio of the delivered traffic per node 𝛼𝛼 against the number of used 56 MHz channels with enabled satellite links. ......................................................................................... 66 Figure 7-3 The ratio of the delivered traffic per node 𝛼𝛼 against the number of used 56 MHz channels with disabled satellite links. ......................................................................................... 67 Figure 7-4 The SE of the system against the number of channels. ............................................. 68 Figure 7-5 EE of the terrestrial links in system against the number of channels. ....................... 69


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Figure 8-1: Satellite access network ............................................................................................ 70 Figure 8-2: MODCOD combinations supported by the DVB-S2 standards along with the C/N conditions they can be applied in, and the spectral efficiency they achieve [19]. ..................... 71 Figure 8-3: Illustration of the functionality of the Generic Stream Encapsulation mechanism [21]. ............................................................................................................................................. 72 Figure 8-4: MODCOD combinations considered in our simulations ........................................... 74 Figure 8-5: Probability density function for the probability of a given MODCOD to be used. ... 76 Figure 8-6: Performance of the compared scheduling algorithms in terms of sharing the available resources ...................................................................................................................... 78 Figure 8-7: Probability to serve packets from the different classes ........................................... 79 Figure 8-8: Overall throughput when using different scheduling algorithms ............................. 80 Figure 8-9: Zooming in Figure 8-8 shows the advantage and trade-off of the considered schedulers ................................................................................................................................... 80


Date: 21/04/2017

of 87

SANSA-645047 D3.6

List of Tables Table 4-1. Helsinki point-to-multipoint scenario description ..................................................... 25 Table 4-2. Helsinki simulation setting for the point-to-multipoint case ..................................... 25 Table 4-3. Numerical results for point-to-multipoint Helsinki sub-set scenario ......................... 26 Table 6-1. SINR and link-rate for the benchmark topology ........................................................ 44 Table 6-2. Terrestrial Segment Parameters ................................................................................ 47 Table 6-3. Satellite Segment Parameters .................................................................................... 47 Table 6-4. Benchmark carrier allocation for the extended Helsinki topology ............................ 48 Table 6-5. SINR and link-rate for the extended benchmark topology ........................................ 49 Table 6-6. Bandwidth values ....................................................................................................... 51 Table 6-7. Results obtained with the proposed carrier allocation algorithm ............................. 52 Table 6-8. Results obtained with the proposed carrier allocation algorithm ............................. 54 Table 7-1 Terrestrial system Parameters .................................................................................... 63 Table 7-2 Benchmark links SINR and rates values. ...................................................................... 63 Table 7-3 Delivered rate and SE vs. No of 56 MHz channels ...................................................... 65 Table 7-4 Delivered rate and SE vs. No of 56 MHz channels ...................................................... 66 Table 7-5 Delivered rate and SE vs. No of channels .................................................................... 68 Table 8-1: List of reduced MODCOD subset used in our simulations. ........................................ 74 Table 8-2: Probability of arrival and QoS requirements for the packet classes considered ....... 76


Date: 21/04/2017

of 87

SANSA-645047 D3.6

List of Acronyms ACM Adaptive Coding and Modulation BBF Baseband Frame BS Base station CDF Cumulative Density Function CL Close Loop CN Core Network C/N Carrier to Noise Ratio DVB Digital Video Broadcast EC European Commission EE Energy Efficiency EIRP Effective Isotropic Radiated Power eNB eNodeB FEC Forward Error Correction FS Fixed-Service FSS Fixed-Satellite Service GS Gateway Station GSO Geostationary Orbit HNM Hybrid Network Management iBN Intelligent Backhaul Node ITU International Telecommunication Union KPI key performance indicators LOS Line-of-Sight MIMO Multiple-Input-Multiple-Output mmWave Millimeter Wave MU-MIMO Multiuser-MIMO OL Open Loop QoS Quality-of-Service REM Radio Environment Map RF Radio Frequency RRM Radio Resource Management RX Receiver SE Spectral Efficiency SINR Signal to Interference plus Noise Ratio SNR Signal to Noise Ratio ST Satellite Terminal SVD Singular Value Decomposition TCP Transmission Control Protocol TX Transmitter WF Water-Filling WRR Weighted Round Robin


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Executive Summary This document summarizes the work carried out in the Task 3.2 of the SANSA project. The aim of the Task 3.2 is to study joint radio resource management techniques for the hybrid terrestrial-satellite backhauling system.

Based on the defined work plan, the key performance indicators (KPIs) and scenario definitions in WP2, this deliverable focuses on performance assessment and adaptation of different Radio Resource Management (RRM) techniques. More specifically, this deliverable addresses developing new technologies for terrestrial and satellite scheduling, carrier allocation, power and flow control.

First, we focus on the beamforming design defined in D3.5 and we combine it with scheduling and power control that tries to maximize the overall sum-rate of a point-to-multipoint SANSA MISO links while imposing interference constraints onto the non-intended receivers. In particular, unlike conventional greedy scheduling approaches that sequentially select the users to be served, here we consider the scheduling design based on the co-linearity between their channel vectors and the non-intended users’ channel vectors. Once the users are scheduled, the transmitted power is assigned according to an introduced optimal power control scheme.

Similarly, we study the coexistence of a terrestrial MIMO links sharing spectral resources with other MIMO links in a point-to-point fashion. Specific interference constraints are considered, that should be satisfied in order to ensure good performance of links sharing the same spectral resources. We formulate the problem as a maximization of the desired link throughput subject to interference constraints. Next, we provide a power allocation algorithm, as well as the precoding technique that achieves the optimal capacity. We provide simulation results for the scenario where the interference constraint is imposed by a satellite link, which we aim to protect.

The previous contributions assumed fixed carrier assignment by considering full frequency reuse. However, proper carrier allocation may relax the interference constraints and thus, reduce the requirements of the previous techniques. As a consequence, we propose an efficient carrier assignment approach to solve the radio resource sharing problem in hybrid satellite-terrestrial backhaul networks. Given the intractability of the problem, we solved the joint optimization task by first optimizing the satellite carrier allocation and, in a next step, optimizing the terrestrial carrier allocation subject to the previous satellite carrier assignment.

As an extension, we investigate the problem of cross-layer design of the frequency assignment together with the link scheduling and flow control. Considering network limitations and system requirements, the target is to maximize the traffic that can be delivered by the network in a given period of time by deciding the active backhauling links that can transmit simultaneously as well as the amount of traffic that should be forwarded in these active links and frequencies.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

The previous contribution aims at avoiding the terrestrial network congestion as much as possible given the available resources. However, there can be specific cases where the amount of traffic exceeds the terrestrial resources provision or when there has been a problem in the terrestrial infrastructure that prevents the traffic to be redirected via the multi-hop terrestrial links. In these cases, the provision of mobile backhaul should be redirected to the satellite. The last contribution of this deliverable focuses on a satellite scheduling design aiming to optimise the allocation of resources to meet the defined Quality-of-Service (QoS) requirements of different cellular services, and in particular maximum delay and jitter. The main goal of the proposed algorithm is to take into account QoS requirements in order to enable services of diverse constraints; to serve in a fair manner all the base stations; and to improve the achievable throughput by reducing the percentage of dropped packets.

In evaluating the performance results, this deliverable pays specific attention to the Spectral Efficiency (SE) measured in bps/Hz, which is related to SANSA objective no. 4 of improving the SE 10-times within the considered Ka-band segments. In this regard, two techniques have focused on different parameters: 1) For the point-to-point power and precoding design, we provide results in terms of rate vs interference constraint trade-off; and 2) For the delay-aware scheduling on the satellite links, we focused on the packet delivery rate.

In summary, the main conclusions of this deliverable is that scheduling and power allocation combined with the multi-antenna beamforming designs of D3.5 can provide 2.83x improvement in terms of SE with respect to the benchmark. In order to reduce the interference constraints and the beamforming requirements, optimal carrier allocation can be designed. We show that carrier allocation by itself can provide 2.09x improvement in terms of SE with respect to the benchmark. Moreover, when the carrier allocation is combined with power and flow control, this gain can be pushed put to 2.47x. Last but not least, we proposed a novel, adaptable, scheduling algorithm for the satellite forward link which allows the SANSA system to meet the QoS requirements of different services under high traffic conditions, and to increase by approximately 30% the offered traffic for which packet delivery is guaranteed.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

1 Introduction The overall scope of this document is to take the SANSA scenario’s defined in WP2, D2.3 [1], and to evaluate radio resource management (RRM) techniques, which have been identified as a key technological enabler to facilitate the interference control within the SANSA network.

SANSA considers that both terrestrial and satellite segments share the same spectrum in order to enhance the overall system spectral efficiency. Focusing on interference characteristics, SANSA scenarios can be grouped into two types:

1) Satellite downlink in the 17.7-19.7 band, where satellite terminals receive the interference caused by the terrestrial transmitters. 2) Satellite uplink in the 27.5-29.5 band, where the satellite terminals cause interference to the terrestrial receivers.

Moreover, in this deliverable we will investigate the possibility of aggressive frequency reuse schemes within the terrestrial segment, which will generate interference among the terrestrial links. Therefore, both interference among terrestrial links and between terrestrial and satellite links need to be taken into account in order to guarantee operation of the overall SANSA network. All these aforementioned types of interference are considered as the SANSA intra-system interference, since they are caused by transmitters within the SANSA network. Investigation on the external sources of interference are provided in SANSA deliverable D3.7 [2]. To deal with the intra-system interference, SANSA foresees two technological enablers: The deployment of smart antennas at the terrestrial backhaul nodes which will enable

spatial interference mitigation capabilities. The use of RRM techniques.

The former refers to placing radiation nulls on the directions of unintended terrestrial and satellite receivers. This work is considered in SANSA deliverable D3.5 [3] as a result of Task 3.1.

In this deliverable, we focus on the second technological enabler: the use of RRM techniques. For managing the intra-system interference, in SANSA we follow a database-assisted shared spectrum access approach as well as dynamic resource management. The Hybrid Network Manager (HNM), as defined in deliverable D2.3 [1], is a new entity introduced by SANSA which includes functionalities to manage jointly satellite and terrestrial backhaul radio resources. Based on its central node definition, the HNM gathers complete SANSA network information which is used not only to reconfigure the SANSA topology (if needed), but also to manage the shared resources among terrestrial and satellite segments, as depicted in Figure 1-1.

Therefore, the HNM is assumed to store a complete spectrum database of the SANSA network, including description details for each of the links, which can be further exploited to optimally allocate the available resources. Most of the environment parameters stored in the HNM


Date: 21/04/2017

of 87

SANSA-645047 D3.6

database are received directly from the different nodes, while some others will be extracted from the HNM memory, like the network topology or the satellite orbit.

The proposed resource allocation techniques operate on a network management level and it aim at optimizing the allocation of the SANSA resources so that the Spectral Efficiency (SE) is maximized, assuming a given topology and assuming that the interference management techniques proposed in D3.5 [3] are in place.

Figure 1-1. Database-assisted resource management envisioned in SANSA project

The remaining of this deliverable is structured as follows: Section 4 and Section 5 consider the optimization of terrestrial multi-antenna links subject to interference constraints at neighboring receivers. In particular, Section 4 extends the point-to-multipoint analog-digital beamforming of D3.5 with scheduling and power control. Precisely, considering that an iBN shares the available spectrum with its neighboring nodes, we design the scheduling and power control technique that maximizes the spectral efficiency, while keeping the interference to non-intended nodes under a certain threshold. On the other hand, Section 5 focuses on power and precoding design considering point-to-point backhaul connections, where the rate of the desired link is to be maximized but guaranteeing the caused interference below to certain threshold.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Carrier allocation techniques will be explored in Section 6 and 7 in order to allocate the shared spectrum among satellite and terrestrial links. This is particularly important since a proper carrier assignment can reduce the level of interference and, thus, relax the antenna requirements of the previous techniques. In Section 6 we show that the joint carrier allocation problem of terrestrial and satellite segment is intractable due to the nonlinear coupling between each other. This is because the terrestrial allocation disturbs the sum-rate of the satellite backhaul network directly. As alternative technique to solve the carrier we proposed to decompose the problem into two sub-problems: we first deal with the carrier assignment for the satellite backhaul network and, on a second step and assuming the resulting satellite segment allocation, we design a sub-optimal carrier assignment for the terrestrial part of the network. This contribution has been published in the ICC Workshop on Satellite Communications co-organized by SANSA [4]. Section 7, we consider the channel assignment together with joint flow control in a Spatial Time Division Multiple Access (STDMA) based wireless backhaul network, where both terrestrial and satellite links are available in the system. The problem is solved by taking into account the use of the satellite links in cases of traffic offloading, link failure and remote access. However, the system can be configured to have all unrestricted usage of the satellite links. This contribution has also been published in the ICC Workshop on Satellite Communications co-organized by SANSA [5].

Finally, Section 8 focus in the satellite segment. In particular, Section 8 presents QoS-aware scheduling techniques which tries to fairly route all the traffic to the iBN taking into account both channel conditions and QoS requirements based on delay-sensitive applications and channel conditions.

2 Summary of SANSA scenarios and radio resource management requirements

The SANSA scenarios have been defined in detail in WP2, deliverable D2.3 [1]. However, in this section we review them paying special attention to the interference issues. SANSA WP3 Task T3.2 analyses whether radio resource allocation techniques could be helpful to resolve the specific inter-system coexistence conditions associated to each scenario.

In WP2, 6 scenarios of the SANSA architecture were selected for further investigation. From the resource allocation perspective, the most important parameter is how the frequencies are reused. In this sense, SANSA scenarios focus on two different sharing scenarios, namely satellite downlink scenario and satellite uplink scenario. These two scenarios are depicted in Figure 2-1. In both, we depicted the terrestrial-to-terrestrial interference since this kind of interference is related to the high frequency reuse aimed at SANSA and it is independent of the satellite transmission.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Figure 2-1. SANSA inter-system interference scenarios

The satellite downlink scenario considers the GSO satellite terminals reusing frequency bands of terrestrial links in the 17.7 – 19.7 GHz band (Space to Earth communication). As illustrated in Figure 1-1, there is one types of interference associated to the satellite downlink: the interference from terrestrial transmitter to satellite receiver

Note that the interference caused by the satellite towards the Earth is neglected. The interference from space to Earth already takes into account the defined power flux density limits established by ITU, as outlined in previous research projects [6]. Considering that current satellite systems are usually in line with these power density limitations, we decided to neglect this interference component.

Regarding the satellite uplink, it considers the GSO satellite terminals reusing frequency bands of terrestrial links in the 27.5 – 29.5 GHz band (Earth to Space communication). As illustrated in Figure 2-1, the main interference associated to this scenario is the interference from the satellite terminal emitter to the terrestrial receiver.

However, this scenario is not possible in the context of current European regulation since the proposed CEPT fragmentation of the 28GHz bands (see Figure 2-2) forces the satellite Earth transmitters to operate in exclusive spectrum.

Figure 2-2. CEPT Band segmentation for FSS and FS in the band 27.5-29.5 GHz. The band labelled Sat*

indicates that there are legacy FS in a few countries.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Operating in exclusive regime only is insufficient to meet future capacity demands [7][8]. Therefore, SANSA has selected this scenario as a potential scenario where terrestrial and satellite systems are able to share spectrum if proper resource allocation is design to take care of the resulting interference.

3 Motivation of radio resource management Conventional wireless backhaul solutions operate mostly on the 18GHz band (mainly for point-to-point), followed by the 28GHz band (mainly used for point-to-multipoint). In general, there is an increasing trend of adopting new wireless solutions such as large-scale antenna array-based systems, also known as massive Multiple Input Multiple Output (MIMO) [9] systems and millimeter wave (mmWave) technology [10]. On one hand, mmWave (including the Ka-band), which although very susceptible to blockage and weather impairments, the short transmission paths and high propagation losses allow for spectrum reuse by limiting the interference caused at adjacent stations. On the other hand, the short wavelength allows modest size antennas to have a small beam width, allowing the BSs to be equipped with antenna arrays that can simultaneously serve multiple users using the same frequency resource. Therefore, interference mitigation and resource management techniques are mandatory for enabling the shared use of the spectrum in mm-wave bands [11][12]. Note that, the use of hybrid analog-digital precoding schemes (as suggested in T3.1) will require more computationally demanding operations compared to the fully-digital solutions of sub-6 GHz deployments.

In order to enhance the user data rates over the conceived interference mitigation techniques, the system designer could opt to perform resource allocation (power, frequency, etc.) and scheduling techniques. Regarding resource allocation, there is a vast literature available for general wireless communication systems [13]. In general, the need to simultaneously and reliably provide multiple users with high-rate communication links leads to challenging optimization problems. The reader is referred to [13] for recent developments in this field. The main challenge of SANSA is the combination of the satellite segment together with the terrestrial segment, which renders even more challenging optimization problems.

Regarding the scheduling techniques, the most popular approach is based on the intended and non-intended channel vectors [14]. Indeed, the seminal works in [15][16] show that in a spectrum sharing multi-user multi-antenna scenario, scheduling can substantially increase the achievable data rates. So far, the mm-wave scheduling proposals have not considered the spectrum sharing scenario described in [15][16]. The work in [17] proposes a zero-forcing digital precoding where the users are greedily selected whenever they increase the sum-rate. On the other hand, the work in [18] proposes an indirect scheduling algorithm where the user selection is done considering the analog beamforming design phase shifters resolution.

While terrestrial systems can take advantage of several spatial and temporal degrees of freedom, these degrees of diversity are not available to the satellite segment. This is due to the strong line-of-sight channel component dominating the communications link, and the typically


Date: 21/04/2017

of 87

SANSA-645047 D3.6

stationary location of satellite terminals. This leaves the satellite systems with a more constrained set of parameters that can be optimised to the communication channel’s conditions. The latest satellite communications standard DVB-S2 [19], along with its subsequent amendments, has introduced the Adaptive Coding and Modulation (ACM) technique for the forward link. This method allows the network operator to adjust the transmission rate over a satellite link depending on the channel’s conditions by adaptively choosing between a predefined list of Forward Error Correction (FEC) Coding and Modulation combinations according to the Carrier to Noise (C/N) ratio of the received signal at the Satellite terminal. This allows the satellite system to transmit at higher rates to Satellite Terminals that have good channel conditions while being able to also serve those Satellite Terminals that have experience lower C/N.

The DVB-S2 standard has defined the supported ACM combinations as a function of the C/N, as well as the Baseband Frame (BBF) characteristics in detail. The length of each BBF is variable, but is significantly larger than the frames typically defined for terrestrial networks. This is due to the long FEC codes used in DVB-S2. However, all the user data packets included in the same BBF have to be transmitted using the same ACM combination. This means that some packets may be transmitted at a rate lower than the maximum rate they could be transmitted at. It is therefore important to schedule packet transmissions taking into account the ACM combination, i.e. in a cross-layer fashion.

Cross layer optimization has been considered extensively in the literature, primarily for fixed networks. The authors in [20] introduce the concept of cross-layer design, revealing the performance benefits it can bring (when compared to the standard, layer-based, design) but also the caveats it has that could lead to performance degradation. In satellite communications, the DVB-S2 standard proposes a Weighted Round Robin scheduling algorithm without, however, defining in detail the operation of the scheduling algorithm. An overview of the standardized QoS architectures for DVB-S2/RCS systems is provided in [21]. In [22], the authors propose a two-stage scheduler to balance between satisfying QoS requirements while maintaining a degree of fairness in a DVB-S2 satellite environment. In [24] and [25] a cross-layer protocol was proposed for the forward link packet scheduler that introduces fairness as a tuneable parameter. In this way, the authors try to control fairness according to the channel conditions of the different Satellite Terminals. In addition, the authors show that their proposed approach could be used to offer QoS adaptation for different services. The authors in [26] further extended this and propose an adaptive algorithm for managing its weights in order to optimize bandwidth utilization while satisfying QoS requirements.

In general, resource allocation in shared terrestrial and satellite networks is considered to be a rather new research area where previous contributions are limited. Most of the published works have focused on the Mobile Satellite Services (MSS), which refers to networks of communications satellites intended for use with mobile and portable wireless telephones. Essentially, the literature has centered its attention on the co-existence of terrestrial cellular system with MSS. One example of this is [27], where the resource allocation for the satellite and terrestrial components is coordinated to optimize spectral efficiency and increase the overall


Date: 21/04/2017

of 87

SANSA-645047 D3.6

system capacity. This is accomplished by controlling the inter-system interference through transmission power control. However, most commercial voice and some data mobile satellite services are provided by systems operating in the L-band by Iridium, Inmarsat, Globalstar and Thuraya. The L-band spectrum allocated for MSS is between 1.5 and 2.5 GHz, with the upper portion often referred to as the S-band.

In SANSA, we focus on the application of satellite communications for backhaul deployment, where Fixed Satellite Services (FSS) are considered. Co-existence of terrestrial and satellite system for backhauling has been mostly examined in the context of Cognitive Radio (CR). Essentially, the satellite spectrum utilization is enhanced by enabling dynamic spectrum access between two satellite systems [28][29] or between satellite and terrestrial systems [7][8][30][31][32][33]. For the downlink case, [31] and [7] proposed a joint beamforming and carrier allocation scheme to enable cognitive Space-to-Earth communications. For the uplink, [7][8][32][33] proposed a joint power and carrier allocation strategy which guarantees protection of the terrestrial system while maximizing the satellite total throughput. However, the scenario proposed in all the aforementioned works considers that the terrestrial microwave links have priority of operation, and thus, should not be adversely affected by the presence of the satellite system. In other words, the satellite system has to adapt its resource management design in order to protect the terrestrial system.

However, the space segment is expected to operate in the future in collaboration with the terrestrial component in a seamless interworking environment. This is the view of SANSA, where a HNM jointly manages satellite and terrestrial backhaul radio resources. An HNM was considered in [34], with the main peculiarity that both segments operate in their respective exclusive spectrum bands.

To the best of the authors' knowledge, radio resource allocation for hybrid satellite-terrestrial backhaul networks operating in the Ka-band has never been considered before.

4 Scheduling, precoding and power control for single BS point-to-multipoint backhaul connections

4.1 Motivation

This Section aims at describing scheduling and power control techniques of SANSA point-to-multipoint systems. The proposed techniques assume that the involved iBN share the same spectral resources, and rely on the low-complexity precoding design presented in Section 4.3 in D3.5 [3]. In other words, bearing in mind that the transceiver employs a low-complexity precoding design, we propose a scheduling and power control technique able to increase the achievable rates.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

The proposed approach is non-iterative. This is, the scheduling and power control design is not coupled with the precoding computation leading to a substantial reduction of the overall computational complexity.

As reported in the numerical results section, we observe a significant spectral efficiency increase with respect to the case where there is no scheduling and power control. This superior performance is shown not only in the random case scenario, but also in the Helsinki topology considered as a benchmark in SANSA.

4.2 System Model

Let us consider a base station equipped with Q antennas transmitting K independent symbols to K receivers. The received signal by the k-th user can be modelled as

𝑦𝑦𝑘𝑘 = 𝒉𝒉𝑘𝑘𝐻𝐻𝒗𝒗𝒌𝒌𝑠𝑠𝑘𝑘 + � 𝒉𝒉𝑘𝑘

𝐻𝐻𝒗𝒗𝒋𝒋𝑠𝑠𝑗𝑗

|𝓚𝓚|

𝒋𝒋\𝒌𝒌 ,𝒋𝒋 ∊ 𝓚𝓚

+ 𝑛𝑛𝑘𝑘, (1)

where 𝒉𝒉𝑘𝑘 is the channel vector between the base station and the k-th receiver, vector 𝒗𝒗𝒌𝒌 denotes the beamforming that supports the transmission of the symbol sent to the k-th which is denoted by 𝑠𝑠𝑘𝑘 and assumed to be zero mean and unit norm. The set of served users is denoted by 𝓚𝓚 and its cardinality is denoted by |𝓚𝓚|. Note that in the notation we have implicitly assumed that 𝑘𝑘 ∊ 𝓚𝓚 . Finally, 𝑛𝑛𝑘𝑘is the additive white Gaussian noise with zero mean and variance equal to 𝜎𝜎2 for all intended receivers.

In contrast to all-digital designs where 𝒗𝒗𝒌𝒌 for k=1, …, K are designed to fulfil a sum-power constraint, in here, we consider that each beamformer consists of an analog processing part 𝑷𝑷, and a digital processing part 𝒘𝒘𝒌𝒌 so that

𝒗𝒗𝒌𝒌 = 𝑷𝑷𝒘𝒘𝒌𝒌 , (2)

We assume that the analog processing part is formed by phase shifters which cannot modify their amplitude, but their phase. We consider three different beamforming network architectures; namely, fully-connected, localized and interleaved. These architectures are mathematically expressed as

𝔓𝔓𝑃𝑃𝑃𝑃−𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓: �[𝑷𝑷]𝒎𝒎,𝒏𝒏�2 = 1 , (3)

𝔓𝔓𝑃𝑃𝑃𝑃−𝑓𝑓𝑙𝑙𝑙𝑙𝑙𝑙𝑓𝑓𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙: �[𝑷𝑷]𝒎𝒎,𝒏𝒏�2 = �𝟏𝟏𝜷𝜷 ⊗ 𝑰𝑰𝑵𝑵𝑵𝑵𝑵𝑵�𝒎𝒎,𝒏𝒏

, (4)

𝔓𝔓𝑃𝑃𝑃𝑃−𝑙𝑙𝑖𝑖𝑖𝑖𝑙𝑙𝑖𝑖𝑓𝑓𝑙𝑙𝑙𝑙𝑖𝑖𝑙𝑙𝑙𝑙: �[𝑷𝑷]𝒎𝒎,𝒏𝒏�2 = �𝑰𝑰𝑵𝑵𝑵𝑵𝑵𝑵 ⊗ 𝟏𝟏𝜷𝜷�𝒎𝒎,𝒏𝒏

. (5)


Date: 21/04/2017

of 87

SANSA-645047 D3.6

for m=1,…,Q and n=1,…, 𝑁𝑁𝑟𝑟𝑁𝑁. The parameter 𝑁𝑁𝑟𝑟𝑁𝑁 denotes the number of RF chains and 𝛽𝛽 =

𝑄𝑄 𝑁𝑁𝑟𝑟𝑁𝑁� is assumed to be an integer value.

We aim at developing a scheduling technique that selects the set of served users 𝓚𝓚 out of all possible intended users set 𝓜𝓜 (as illustrated in Figure 4-1) in order to maximize the sum-rate while keeping the interference to the non-intended receivers low. Furthermore, we also optimize the transmit power so that the sum-rate is maximized while maintaining the interference to the non-intended user low.

Figure 4-1. Scheduling designed for M=3

Mathematically, we focus on the following optimization problem

(6)

where

𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑘𝑘 =𝑞𝑞𝑘𝑘�𝒉𝒉𝑘𝑘

𝐻𝐻𝑷𝑷𝒘𝒘𝒌𝒌�2

∑ 𝑞𝑞𝑗𝑗�𝒉𝒉𝑘𝑘𝐻𝐻𝑷𝑷𝒘𝒘𝒋𝒋�2 + 𝜎𝜎2|𝓚𝓚|

𝒋𝒋\𝒌𝒌 ,𝒋𝒋 ∊ 𝓚𝓚

, (7)

which is the Signal-to-Interference and Noise Ratio (SINR) of the k-th user and 𝑞𝑞𝑘𝑘 its corresponding transmit power, 𝑃𝑃𝑚𝑚𝑙𝑙𝑚𝑚 is the maximum available power. The parameter 𝒈𝒈𝒍𝒍 represents the channel vectors to the non-intended users from l = 1, …, L.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

As described in the optimization problem, we impose a maximum interference power level to the non-intended receivers 𝑃𝑃𝑓𝑓. This threshold can be arbitrarily chosen by the system designer in order to maximize the overall inter-system capacity.

The optimization problem in (6) presents a combinatorial computational complexity due to the optimization of 𝓚𝓚 over 𝓜𝓜. In contrast to other multiuser MIMO scheduling techniques, the optimization problem (6) additionally presents the challenge of designing the analog beamforming network which is known to be a difficult non-convex problem even for the single user case.

Scheduling, power control and precoding operations are coupled optimizations as it can be observed in the optimization problem. In this work, we opt to decouple the different optimizations tasks, i.e. precoding, power control and scheduling.

In the next sections we describe how these different operations are performed. We first describe the precoding algorithm assuming that we know the set of users to be served. Posteriorly, we present the scheduling and power control schemes that deal with the optimization problem.

4.3 Low-complexity precoding

We propose the following closed-form analog beamforming design

(8)

For m=1,…,Q and n=1,…, 𝑁𝑁𝑟𝑟𝑁𝑁. Matrix 𝑯𝑯 collapses on its rows all user channel vectors and ∠a operator extracts the angle of the complex number a. Matrix 𝑪𝑪 describes the beamforming network connectivity matrix.

The proposed analog beamforming design aims at pointing the intended users by considering the beamforming network restrictions included in matrix 𝑪𝑪, which takes different values depending on the assumed underlying connectivity matrix:

𝑪𝑪𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓−𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖𝑙𝑙𝑙𝑙𝑖𝑖𝑙𝑙𝑙𝑙 = 𝟏𝟏𝑵𝑵𝑵𝑵𝑵𝑵 ⊗ 𝟏𝟏𝜷𝜷 , (9)

𝑪𝑪𝑓𝑓𝑙𝑙𝑙𝑙𝑙𝑙𝑓𝑓𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 = 𝟏𝟏𝜷𝜷 ⊗ 𝑰𝑰𝑵𝑵𝑵𝑵𝑵𝑵, (10)

𝑪𝑪𝑙𝑙𝑖𝑖𝑖𝑖𝑙𝑙𝑖𝑖𝑓𝑓𝑙𝑙𝑙𝑙𝑖𝑖𝑙𝑙𝑙𝑙 = 𝑰𝑰𝑵𝑵𝑵𝑵𝑵𝑵 ⊗ 𝟏𝟏𝜷𝜷. (11)

Once the analog design is settled, the digital beamforming observes an equivalent channel matrix


Date: 21/04/2017

of 87

SANSA-645047 D3.6

(12)

Bearing this in mind, we can proceed with the digital design using the well-known zero-forcing technique.

(13)

Where ρ is a scaling factor to guarantee that matrix 𝑾𝑾𝑍𝑍𝑍𝑍 has unit norm. The digital precoding vectors are the columns of matrix 𝑾𝑾𝑍𝑍𝑍𝑍 .

4.4 Scheduling

While greedy scheduling approaches sequentially select the users to be served and compute the resulting beamforming and power allocation iteratively [14], we consider the indirect approach where the scheduler performs its task of the beamforming and power control design separately. In particular, we select the users based on the co-linearity between their channel vectors and the non-intended receivers’ channel vectors.

The procedure consists of two parts. First, the scheduler randomly selects one user i ∈ 𝓜𝓜 Next, the scheduler selects additional users so that

(14)

for j ∈ 𝓜𝓜 −{i}. The parameter 𝛿𝛿𝑖𝑖 denotes maximum co-linearly acceptable value. This approach was first presented in [35] and coined as semi-orthogonal user clustering.

Let us denote 𝓢𝓢, the selected receivers at the first scheduling phase. If | 𝓢𝓢| ≤ 𝑁𝑁𝑟𝑟𝑁𝑁 the stations to be served are the ones in 𝓢𝓢. There might be case where the set S becomes empty. In such case, the value of 𝛿𝛿𝑖𝑖 should be reduced. On the other hand, in case | 𝓢𝓢| > 𝑁𝑁 , we propose that the selected 𝑁𝑁𝑟𝑟𝑁𝑁 users consider the spatial signature of the non-intended receiver terminals.

Mathematically, we will select the receivers n ∈ 𝓢𝓢 so that

(15)

For all l=1,… , L. The scheduling algorithm is summarized in the following.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

4.5 Power Control

Finally, once the users are selected and the analog and digital precoding parts are optimized, the transmitter can optimize the power control so that

(16)

This is a convex problem solvable with conventional water-filling based methods.

4.6 Numerical Results

In this section, we numerically evaluate the proposed scheduling mechanism. We consider a transmitter equipped with Q = 64 antennas and 𝑁𝑁𝑟𝑟𝑁𝑁 = 4 in presence of | 𝓜𝓜 | = 100 possible receivers and L = 1 non-intended receiver. The simulations results are obtained with an average of 500 Monte Carlo runs and a variable transmit power of 𝑃𝑃𝑚𝑚𝑙𝑙𝑚𝑚 = 2,…,16 dBW. Both the intended and non-intended user channel vectors are obtained with the channel model that is described in D3.5 Section 2.2 [3]. Finally, we consider 𝑃𝑃𝑓𝑓 = 0.01.

In all cases, we evaluate the sum-rate which can be described as follows


Date: 21/04/2017

of 87

SANSA-645047 D3.6

𝑆𝑆𝑖𝑖𝑙𝑙𝑖𝑖𝑖𝑖𝑙𝑙𝑒𝑒𝑖𝑖𝑖𝑖𝑙𝑙𝑙𝑙𝑓𝑓 = 1𝐾𝐾

� log2�1 + 𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑖𝑖𝑙𝑙𝑖𝑖𝑖𝑖𝑙𝑙𝑒𝑒𝑖𝑖𝑖𝑖𝑙𝑙𝑙𝑙𝑓𝑓𝑘𝑘 �

𝐾𝐾

𝑘𝑘=1

(17)

where 𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑖𝑖𝑙𝑙𝑖𝑖𝑖𝑖𝑙𝑙𝑒𝑒𝑖𝑖𝑖𝑖𝑙𝑙𝑙𝑙𝑓𝑓𝑘𝑘 is the SINR of the k-th user.

Figure 4-2 Sum-rate versus transmit power for different δs and δt. It is considered L = 1, Q = 64 and NRF = 4.

Bearing in mind that the fully-connected alternative with the proposed analog optimization is the one that offers the largest sum-rate (see D3.5 [3] for further details), we next evaluate the performance of the scheduling technique. In Figure 4-2, we can oberve the sum-rate when no scheduling is performed as well as when the proposed scheduling method is used for different values of δt = δs.

Remarkably, for the lowest value δt = δs = 0.05, the scheduling technique yields to a lower sum-rate compared to the case where the users are randomly selected. On the other hand, for higher values (i.e. 0.06 and 0.07) a sum-rate gain can be appreciated over all transmit power. Precisely a 6% gain in sum-rate is observable over the transmit power.

This behavior of the sum-rate versus δt, δs values was reported in [15], where it can be observed that for very low values of δt, δs the sum-rate is low and it increases as δt, δs increase. However, at some point, increasing δt, δs decreases the sum-rate. This is due to the fact that large number of receivers have been scheduled simultaneously and this negatively impacts the achievable data rates.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Now, we evaluate the conceived technique in a sub-set of the Helsinki topology reference. In particular, we consider the operation of the transmitter node 9 to the nodes 3, 10, 11 and 12. The considered point-to-multipoint scenarios is highlighted in red in Figure 4-3.

Figure 4-3. Helsinki point-to-multipoint scenario under evaluation

The following table describes the node relative position and the link distances.

Table 4-1. Helsinki point-to-multipoint scenario description Link Elevation angle

[degrees] Azimuth angle

[degrees] Distance [meters]

9-3 0.99 -29 720.45 9-11 -0.91 -10 595.89 9-12 -0.45 -1 1075 9-10 -1.16 38 524

In addition, the following parameters have been used for the different realizations:

Table 4-2. Helsinki simulation setting for the point-to-multipoint case Parameter Value for urban scenario Maximum EIRP 24 dBWatts Receive antenna gain 38 dB Noise power level -121 dBWatts Channel Model 3 cluster 5 scatters

The results have been obtained via averaging 1000 Monte Carlo runs. At each realization, we assume a random scattered environment obtained via the channel modelling presented in D3.5 [3].

24.64 24.65 24.66 24.67 24.68 24.69

60.195

60.2

60.205

60.21

60.215

60.22

1

2

3

4

5

6

7

8

9

10

11 12

13 14

15

Longitude (deg)

Latit

ude

(deg

)

LinksNodes


Date: 21/04/2017

of 87

SANSA-645047 D3.6

As a benchmark scenario, we consider that each intended base station is served in a FDMA fashion (i.e. each transmission takes place in disjoint frequency beam). On the contrary, the proposed scheme considers that each transmission takes places on the same frequency bin and the interference mitigation is performed via spatial processing.

Considering that the benchmark setting leads to an average SE of 35.14 bps/Hz, Table 4-3 summarizes the gains of using the proposed precoding, scheduling and power control scheme for two different array configurations. In both cases, a uniform planar array is assumed, where Nx is the number of elements in the vertical direction and Ny in the horizontal. In the considered setting, both array configurations lead to nearly the same SE result, independently of the total number of antenna elements, which provides a 2.83x gain with respect to the benchmark. The resulting high gains motivate the use of the proposed technique in spectrum sharing point-to-multipoint scenarios.

Table 4-3. Numerical results for point-to-multipoint Helsinki sub-set scenario Array Configuration

(Number of antennas on x and y axis) SE [bps/Hz] SE Gain with respect to the

benchmark (Fully-connected)

Nx = 2, Ny = 200. 99.72 2.83 Nx = 10, Ny = 36. 99.79 2.83

5 Power control and precoding design for single BS point-to-point backhaul connections

5.1 Motivation

MIMO communication methods enhance the capacity of a system by enabling the transmission of multiple data streams destined to a single user or a group of active users on the same time-frequency resource [36]. The capacity of open-loop (OL) and closed-loop (CL) single-user MIMO (SU-MIMO) systems has been derived in the middle 1990's [37][38]. In the latter case, it has been proven that the use of singular value decomposition (SVD) based pre- and post-coding is required, in order to reach the capacity. Then, the mutual information maximization problem reduces to one of finding the optimal power allocation under a sum-power constraint. The optimal power levels are obtained via the water-filling (WF) algorithm [39]. A WF power allocation solution also maximizes the mutual information in CL SU-MIMO systems operating under interference, assuming that the TX knows perfectly the interference plus additive white Gaussian noise (AWGN) covariance matrix, [40], as well as the sum-rate (SR) throughput in multi-user MIMO (MU-MIMO) setups, [36], and cooperative multi-cell MIMO (Co-MC-MIMO) paradigms for given linear precoding and user selection schemes [41].

The lesson learned from spectrum sharing / reuse and Co-MC-MIMO is that the cooperation between the interested parties allows for efficient coexistence as in a typical spectrum sharing


Date: 21/04/2017

of 87

SANSA-645047 D3.6

setup. This can be achieved via the use of beamforming / precoding, power allocation or/and power control techniques. The objective in this section is to maximize the mutual information / sum-rate throughput of the Desired Link (DL) under some given power constraint, while at the same time ensuring that the interference link (IL) does not exceed a predefined power threshold.

One of the first works dealing with power control taking into account the total received power constraints was presented in [42]. However, [42] assumes a decentralized power control. Differently to [42], this section assumes the possibility to exchange information among iBNs.

More precisely, in this section, our goal is to maximize the sum-rate for the DL under an interfered receiver power constraint, which is imposed by the DL, at the cost of an exchange of information between the two links. Thus, coexistence of both links is ensured, i.e., without harmful interference between the links. Finally, we provide a power allocation algorithm, as well as the precoding technique that achieves the optimal capacity. The performance of the system is evaluated for various interference thresholds and an insight on the effect of this parameter is provided.

5.2 Model and Problem Formulation

We consider a MIMO setup in which two entities share the same resources. The system architecture is depicted in Figure 5-1. Thus, provided a DL, TXDL − RXDL and an IL, TXIL − RXIL, we model the signals from each link, by taking into account the possible interference that is caused from the cross channels. Our goal is to maximise the mutual information of the DL (in what follows, we refer to mutual information as the rate of the DL), under an additional interference constraint, which is imposed by the IL and provides QoS guarantees to its receiver. As in standard MIMO setups, we assume that each transmitter / receiver is equipped with an antenna array of multiple antennas. Let the IL consist of 𝑘𝑘 elements for the transmitter and ℓ for the receiver, while the DL consists of 𝑚𝑚 elements for the transmitter and 𝑛𝑛 for its receiver. The received signals at both DL and IL are modelled as:

𝒚𝒚𝐷𝐷𝐷𝐷 = 𝑯𝑯𝐷𝐷𝐷𝐷𝒔𝒔 + 𝑯𝑯𝐼𝐼𝐷𝐷𝒙𝒙 + 𝜼𝜼, (18)

𝐲𝐲𝐼𝐼𝐷𝐷 = 𝑯𝑯𝐼𝐼𝐷𝐷𝒙𝒙 + 𝑯𝑯𝐷𝐷𝐼𝐼𝒔𝒔 + 𝒗𝒗, (19)


Date: 21/04/2017

of 87

SANSA-645047 D3.6

respectively1. The transmitted signals for the DL and the IL are denoted as 𝒔𝒔 ∈ ℂ𝑚𝑚 and 𝒙𝒙 ∈ ℂ𝑘𝑘, respectively, and are zero mean complex Gaussian (uncorrelated); the channel gain from the i-th transmitter to the j-th receiver element is denoted as ℎ𝑗𝑗𝑙𝑙, thus, for the channels of each link of Figure 5-1 we have: 𝑯𝑯𝐷𝐷𝐷𝐷 ∈ ℂ𝑖𝑖×𝑚𝑚 , 𝑯𝑯𝐼𝐼𝐷𝐷 ∈ ℂℓ×𝑘𝑘 , 𝑯𝑯𝐷𝐷𝐼𝐼 ∈ ℂℓ×𝑚𝑚 and 𝑯𝑯𝐼𝐼𝐷𝐷 ∈ ℂ𝑖𝑖×𝑘𝑘 and are assumed fixed and frequency flat. We have also considered that 𝒗𝒗~𝑁𝑁(𝟎𝟎, 𝑰𝑰ℓ) and 𝜼𝜼~𝑁𝑁(𝟎𝟎, 𝑰𝑰ℓ) are additive white circularly complex Gaussian noise processes. Both the signals and the noise are assumed uncorrelated with each other.

Let 𝒛𝒛 = 𝑯𝑯𝐼𝐼𝐷𝐷𝒙𝒙 + 𝜼𝜼. According to (45), it can be readily seen that the covariance matrix of the signal received by the SL is:

𝑹𝑹𝒚𝒚𝐷𝐷𝐷𝐷 : = E�𝒚𝒚𝐷𝐷𝐷𝐷𝒚𝒚𝐷𝐷𝐷𝐷† � = 𝑯𝑯𝐷𝐷𝐷𝐷𝑹𝑹𝑒𝑒𝑯𝑯𝐷𝐷𝐷𝐷

† + 𝑹𝑹𝑙𝑙, (20)

where 𝑹𝑹𝑒𝑒 is the covariance matrix of the DL's transmitted signal; 𝑹𝑹𝑙𝑙 is the covariance matrix of the vector 𝒛𝒛, i.e., 𝑹𝑹𝑙𝑙 = 𝑯𝑯𝐼𝐼𝐷𝐷𝑹𝑹𝑚𝑚𝑯𝑯𝐼𝐼𝐷𝐷

† + 𝑰𝑰𝑖𝑖, and 𝑹𝑹𝑚𝑚 is the covariance matrix of the interfering link, which is assumed to be known to the system.

The mutual information of the DL (disregarding any constraint on the interference caused to RXIL), is given by:

𝑃𝑃(𝒚𝒚𝐷𝐷𝐷𝐷; 𝒛𝒛) = 𝑙𝑙𝑙𝑙𝑔𝑔2 det� 𝜋𝜋𝜋𝜋𝑹𝑹𝒚𝒚𝐷𝐷𝐷𝐷� − 𝑙𝑙𝑙𝑙𝑔𝑔2 𝑑𝑑𝜋𝜋𝑑𝑑(𝜋𝜋𝜋𝜋𝑹𝑹𝑙𝑙)

= 𝑙𝑙𝑙𝑙𝑔𝑔2 𝑑𝑑𝜋𝜋𝑑𝑑 (𝑰𝑰𝑚𝑚 + 𝑯𝑯𝐷𝐷𝐷𝐷† 𝑹𝑹𝑙𝑙

−1𝑯𝑯𝐷𝐷𝐷𝐷𝑹𝑹𝑒𝑒) (21)

for 𝑛𝑛 ≥ 𝑚𝑚. At this point we consider the eigen-decomposition of matrix 𝑯𝑯𝐷𝐷𝐷𝐷† 𝑹𝑹𝑙𝑙

−1𝑯𝑯𝐷𝐷𝐷𝐷, which is given by:

1 The dependency of the signals and random variables over time are omitted for simplicity.

Figure 5-1: Spectrum reuse MIMO setup considering a desired and an interfering link.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

𝑯𝑯𝐷𝐷𝐷𝐷† 𝑹𝑹𝑙𝑙

−1𝑯𝑯𝐷𝐷𝐷𝐷 = 𝑼𝑼𝚲𝚲𝑼𝑼†, (22)

where 𝑼𝑼 is a unitary matrix and 𝚲𝚲 is the diagonal matrix with its positive eigenvalues, i.e., diag(𝜆𝜆1, … , 𝜆𝜆𝑖𝑖), where 𝑟𝑟 is the rank of the decomposed matrix. Therefore, by imposing the DL's transmitted signal to be of the form 𝒔𝒔 = 𝑼𝑼𝒔𝒔𝑤𝑤, where 𝒔𝒔𝑤𝑤 is spatially white, leads to 𝑹𝑹𝑒𝑒 = E�𝒔𝒔𝒔𝒔†� = 𝑼𝑼𝑼𝑼𝑼𝑼†, 𝑼𝑼 = E�𝒔𝒔𝑤𝑤𝒔𝒔𝑤𝑤

† � is diagonal. Thus, (21) is simplified to:

𝑃𝑃(𝒚𝒚𝐷𝐷𝐷𝐷; 𝒛𝒛) = 𝑙𝑙𝑙𝑙𝑔𝑔2 𝑑𝑑𝜋𝜋𝑑𝑑 (𝑰𝑰𝑖𝑖 + 𝚲𝚲𝑼𝑼). (23)

Hence, the standard mutual information maximization task for the DL's transmitted signal is given by:

max

𝑼𝑼log2 det (𝑰𝑰𝑖𝑖 + 𝚲𝚲𝑼𝑼)

𝑠𝑠. 𝑑𝑑. 𝑼𝑼 ≽ 0 tr(𝑼𝑼) ≤ 1

(24)

where without loss of generality (avoiding an equivalent normalization) we have considered that the maximum transmission power of the DL's MIMO antenna array is 1.

The optimization task in (24) obtains the standard water-filling solution2, which is given by:

𝑑𝑑𝑙𝑙 = �ρ − λi−1�+, i = 1, … , 𝑟𝑟, (25)

where 𝜌𝜌 is the water-level chosen to satisfy the power constraint with equality, i.e., ∑ 𝑑𝑑𝑙𝑙 = 1𝑖𝑖𝑙𝑙=1 .

However, in the presence of the IL, i.e., TXIL − RXIL, computing the optimum value for the transmission power in (24) does not take into account the interference that will be caused to the IL. In order to avoid causing excessive interference to the IL's receiver, RXIL , an additional constraint should be satisfied, which can be expressed in view of (19) as:

tr(𝑯𝑯𝐷𝐷𝐼𝐼𝑹𝑹𝑒𝑒𝑯𝑯𝐷𝐷𝐼𝐼† ) = tr(𝑯𝑯� 𝐷𝐷𝐼𝐼𝑹𝑹𝑒𝑒𝑯𝑯� 𝐷𝐷𝐼𝐼

† ) ≤ 𝑃𝑃𝐼𝐼 , (26)

where 𝑯𝑯� 𝐷𝐷𝐼𝐼 = 𝑯𝑯𝐷𝐷𝐼𝐼𝑼𝑼 and 𝑃𝑃𝐼𝐼 > 0 is the maximum value of interference that is tolerable to the IL receiver due to TXDL. Thus, our goal is to find a solution for (24) under the additional constraint given by (26).

2 The task can be equivalently transformed to a convex optimization one (since the cost function is concave), thus a unique solution exists.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

5.3 Solution and Algorithm

The cost function in (24) can be further simplified and thus the new optimization power allocation with the interference constraint task is now formulated as:

min 𝑙𝑙𝑖𝑖

− � log2(1 + λ𝑙𝑙𝑑𝑑𝑙𝑙)𝑖𝑖

𝑙𝑙=1𝑠𝑠. 𝑑𝑑. 𝑑𝑑𝑙𝑙 ≥ 0, 𝑖𝑖 = 1, … , 𝑟𝑟,

� 𝑑𝑑𝑙𝑙 ≤ 1𝑖𝑖

𝑙𝑙=1

,

� 𝑎𝑎𝑙𝑙𝑑𝑑𝑙𝑙 ≤ 𝑃𝑃𝐼𝐼

𝑖𝑖

𝑙𝑙=1

(27)

where 𝑎𝑎𝑙𝑙 = �𝒉𝒉�𝑙𝑙�22

, 𝑖𝑖 = 1, … , 𝑟𝑟, is the squared norm of the column vectors of matrix 𝑯𝑯� 𝐷𝐷𝐼𝐼. The objective function in (54) is convex and the constraints define a polyhedron, as demonstrated in Figure 5-2 for 𝑟𝑟 = 2. Thus, the optimization task we are attempting to solve is convex and hence it attains a unique minimum and its solution can be derived via the Karush-Kuhn-Tucker (KKT) (optimality) conditions, see [43][44].

Theorem 1: The solution to the optimization task (27) is:

Figure 5-2: Representation of the optimizations task’s feasible region for 𝑟𝑟 = 2.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

𝑑𝑑𝑙𝑙 =

⎩⎪⎨

⎪⎧ �

log2 𝜋𝜋𝜈𝜈

−1𝜆𝜆𝑙𝑙

�+

𝑖𝑖𝑁𝑁 � 𝑎𝑎𝑙𝑙 �log2 𝜋𝜋

𝜈𝜈−

1𝜆𝜆𝑙𝑙

�+

≤ 𝑃𝑃𝐼𝐼 ,𝑖𝑖

𝑙𝑙=1

�log2 𝜋𝜋

𝜈𝜈 + 𝜇𝜇𝑎𝑎𝑙𝑙−

1𝜆𝜆𝑙𝑙

�+

𝑙𝑙𝑑𝑑ℎ𝜋𝜋𝑟𝑟𝑒𝑒𝑖𝑖𝑠𝑠𝜋𝜋

, (28)

where the Lagrange multipliers are obtained from a two-stage procedure. First, 𝜈𝜈 is obtained by solving the standard water-filling algorithm and, if required, 𝜇𝜇 is obtained by solving the interfered power constraint with equality.

For the first case the value 𝜌𝜌 = 𝑓𝑓𝑙𝑙𝑔𝑔2𝑙𝑙𝜈𝜈

can be interpreted as the standard water level of the WF method. However, for the second case, the initial water level violates the additional condition, and the initial water level is penalized by the term 𝜇𝜇𝑎𝑎𝑙𝑙, which is different for each channel, since it depends on 𝑎𝑎𝑙𝑙 's. Moreover, it can be readily seen that, for the new power level and the 𝜈𝜈 obtained at the first stage, ∑ 𝑑𝑑𝑙𝑙

𝑖𝑖𝑙𝑙=1 < 1, for any 𝜇𝜇 > 0.

The established iterative scheme for the power allocation task under interfered receiver constraint is presented in the following algorithm. It should be noted that this is a generic method, whose standard WF algorithmic part is only a special case. Thus, the case of greater interest is when the interference constraint is not satisfied.

At the first stage, the algorithm computes a 𝜈𝜈, which is related to a specific water level, according to the standard WF solution. At the second stage, a decision is taken; the derived solution can either satisfy the interference power constraint or not. In the latter case, given the 𝜈𝜈 that is already computed, the algorithm computes a 𝜇𝜇 > 0 from the solution of the interfered power constraint, satisfied with equality. At this point one should notice that the function

𝑔𝑔𝑝𝑝(𝜇𝜇): = �𝑎𝑎𝑙𝑙

𝜈𝜈 + 𝜇𝜇𝑎𝑎𝑙𝑙

𝑖𝑖−𝑝𝑝+1

𝑙𝑙=1

− 𝛾𝛾𝑝𝑝, (29)


Date: 21/04/2017

of 87

SANSA-645047 D3.6

is strictly decreasing for 𝜇𝜇 ≥ 0. Moreover, 𝑔𝑔𝑝𝑝(0) > 0. Thus, 𝑔𝑔𝑝𝑝(𝜇𝜇) = 0 has a unique solution for every 𝜈𝜈 obtained from the first stage of the algorithm, which can be directly derived via an iterative method, such as the bisection or the Newton's method.

5.4 Experimental Evaluation

For the evaluation of the derived power allocation technique, we perform the following experiment within the SANSA scenario of spectrum reuse in hybrid satellite-terrestrial links.

To this end, we consider the desired link to be a 4 × 4 MIMO one (𝑛𝑛, 𝑚𝑚 = 4), equipped with omnidirectional antennas, operating at 18GHz, and transmitting at a narrow band. The distance between the TXDL and RXDL is 500m. For the interfering link, which we aim at protecting, we have considered a satellite SISO link (𝑘𝑘, ℓ = 1) with antennas of 40dB gain, operating at the same frequency band. The link corresponds to a GEO satellite, which is in orbit at the height of 35,786 ∙ 106m. The satellite ground terminal, RXIL, is collocated (in the 𝑥𝑥, 𝑦𝑦 plane) with the RXDL at a height of 10m above the latter receiver.

The channels between the terrestrial links are provided by SANSA’s Channel Model Simulator [3], whereas for the satellite apart from the antenna gains we have also considered the free-space path loss. For each value of interference constraint, 𝑃𝑃𝐼𝐼, we perform 2000 Monte Carlo runs and average the maximum achieved capacity. Due to the chosen normalization3, we have considered 𝑃𝑃 = 1 and vary the power constraint 𝑃𝑃𝐼𝐼; however, if the sum-power constraint was chosen equal to 𝑃𝑃 ≠ 1 one should measure the capacity for different values of the ratio 𝑃𝑃𝐼𝐼

𝑃𝑃.

In Figure 5-3 (a) we have evaluated the achieved capacity for various values of interference constraint 𝑃𝑃𝐼𝐼 . The solid line corresponds to the power allocation under the interference constraint by the IL, which is achieved via the proposed algorithm, and the dashed one to the power allocation without the interference constraint ( 𝑃𝑃𝐼𝐼 = ∞ ), which is achieved by the standard WF algorithm. Moreover, in Figure 5-3 (b), we have computed the percentage of capacity loss that is caused by the interference constraint imposed by the IL. It is evident that a tighter constraint translates to a greater penalty; however, it ensures that no interference is caused to the satellite receiver.

Finally, in Figure 5-4 we present the empirical cumulative distribution function's (CDF's) for capacities achieved with various interference levels 𝑃𝑃𝐼𝐼 's. The dashed line corresponds to the unconstrained power allocation (WF). It is observed that the CDF's for small 𝑃𝑃𝐼𝐼 (the extra condition is not directly satisfied) are far from the ideal case of the unconstrained optimization task.

3 The channels are also normalized to unit power.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

(a) Capacity

(b) Penalty due to satisfaction of the interference constraint

Figure 5-3: Capacity versus various values of interference constraints, 𝑃𝑃𝐼𝐼 , for a 4 × 4 MIMO link. (a) The dashed line corresponds to the maximum capacity of a single unconstrained link

(ignoring the interference) and the solid one corresponds to the achieved capacity of the proposed method, which guarantees that no interference is caused to the protected link. (b)

The penalty, which corresponds to the percentage of capacity loss, for different values of interference level 𝑃𝑃𝐼𝐼 .

1 2 3 4 5 6 71.7

1.75

1.8

1.85

1.9

1.95

2

2.05

2.1

2.15

Ave

rage

Cap

acity

(bps

/Hz)

Interfering MIMO linkNon-interfereing MIMO link

1 2 3 4 5 6 70

2

4

6

8

10

12

14

16

18

20

Pen

alty

%


Date: 21/04/2017

of 87

SANSA-645047 D3.6

6 Carrier allocation for hybrid satellite-terrestrial backhaul networks

In the previous section, carrier assignment was assumed to be known.

In this section, we consider the problem of carrier allocation in the SANSA Hybrid Satellite-Terrestrial Backhaul (HSTB) network. The peculiarity of SANSA is that terrestrial and satellite segments share the same spectrum in order to enhance the overall spectrum efficiency. As indicated in previous sections, due to the spectrum sharing condition, both systems are subject to interference constraints which should be properly taken into account in the carrier allocation algorithm design.

In a first attempt, we consider the satellite downlink scenario which corresponds to the satellite and terrestrial system sharing the 17.7-19.7 GHz band.

Figure 5-4: CDF's of a 4 × 4 MIMO link for different capacity values, which correspond to interference levels 𝑃𝑃𝐼𝐼 . The dashed line corresponds to the case where no interference constraint

exists

0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4

Capacity bps/Hz

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1E

mpi

rical

CD

F


Date: 21/04/2017

of 87

SANSA-645047 D3.6

In the considered scenario, which is depicted in Figure 6-1, the interference from terrestrial backhauling transmitters to the satellite backhauling terminals needs to be taken into account in order to guarantee operation of the satellite segment of the system. Moreover, in this section we consider aggressive frequency reuse schemes within the terrestrial segment which generate terrestrial intra-system interference as well.

Figure 6-1. Hybrid satellite-terrestrial backhaul network topology

Hence, it is crucial to properly design the frequency allocation of the HSTB that aims at sharing the available and limited spectrum resources between terrestrial and satellite segments as efficiently as possible.

In this deliverable, we formulate the carrier allocation optimization problem focusing on maximizing the worst link in terms of interference. We show that the latter is a NP-hard problem, due to the coupling between the satellite and terrestrial carrier allocation. To overcome this hurdle, we propose to optimize each part of the HTSB network in a sequential manner, so that the first optimization step provides the satellite carrier allocation that is taken into account in the second optimization step, which tackles the terrestrial part of the HTSB network. This work was presented in [4].


Date: 21/04/2017

of 87

SANSA-645047 D3.6

6.1 Problem formulation

We consider a multi-hop wireless backhaul network composed of several terrestrial stations. Some of them are equipped with a satellite dish antenna and, therefore, can receive backhaul traffic through the satellite network. The considered scenario is the one illustrated in Figure 6-1. The illustrated terrestrial topology is an extension of a true backhaul network in Finland, which was obtained from the Finnish regulator. We assume that all the nodes in the HSTB network have access to the core through multi-hop wireless links, either through the terrestrial segment or the satellite segment of the network.

6.1.1 Terrestrial segment model

Let us consider a hybrid backhaul network with 𝑁𝑁 terrestrial nodes indexed by 𝑛𝑛 = 1, … , 𝑁𝑁, which can send, receive and relay backhaul traffic. We consider the terrestrial nodes to be interconnected through 𝐿𝐿 unidirectional communication links, indexed by 𝑙𝑙 = 1, … , 𝐿𝐿 as illustrated in Figure 6-1. Following the notation in the literature [45][46], we represent the set of terrestrial links that are outgoing from node 𝑛𝑛 with 𝑂𝑂(𝑛𝑛) and the set of terrestrial links that are incoming to node 𝑛𝑛 with 𝑃𝑃(𝑛𝑛).

Let us assume that the 𝐿𝐿 terrestrial links operate in the 17.7-19.7 GHz band. We assume that the 17.7-19.7 GHz band is divided into 𝐾𝐾 equally sized frequency carriers of bandwidth size 𝐵𝐵𝑖𝑖 which represent a standard bandwidth supported by the terrestrial system. For the sake of clarity, we assign a carrier identification number to each of the 𝐾𝐾 frequency carriers, namely 𝑘𝑘 =1, … , 𝐾𝐾. We assume 𝐿𝐿 > 𝐾𝐾, which is the general and most challenging case.

Let 𝒂𝒂𝑖𝑖 ∈ 𝑆𝑆𝐷𝐷 be the terrestrial carrier allocation vector, whose elements 𝑎𝑎𝑖𝑖(𝑖𝑖) ∈ [1, 𝐾𝐾] contain the terrestrial carrier identification number of the carrier that has been assigned to the i-th terrestrial link. Note that for implementation issues, only one carrier should be assigned to each link. Additionally, full-duplex scenarios in which the same carrier is used for transmission and reception at the same terrestrial station should be avoided. The latter constraint can be mathematically expressed as follows:

𝒂𝒂𝑖𝑖(𝑗𝑗) ≠ 𝒂𝒂𝑘𝑘(𝑖𝑖) 𝑗𝑗 ∈ I(n), i ∈ O(n), n = 1, … , 𝑁𝑁 (30)

On the other hand, one carrier can be simultaneously assigned to multiple links. This is in-line with the current trend of targeting aggressive frequency reuse schemes, which are expected to increase the spectrum efficiency and network capacity at the expense of increased interference levels. From the total 𝐾𝐾 carriers, the carrier allocation algorithm chooses 𝐾𝐾′ ≤ 𝐾𝐾 to design the terrestrial backhaul network.

Regarding the interference modeling, the terrestrial intra-system interference signal level seen by the l-th terrestrial link operating at the k-th carrier can be expressed as follows:


Date: 21/04/2017

of 87

SANSA-645047 D3.6

𝑖𝑖𝑖𝑖

𝑓𝑓 (𝑘𝑘) = � 𝑃𝑃𝑖𝑖𝑚𝑚𝑇𝑇𝑇𝑇𝑇𝑇(𝑖𝑖) ∙ 𝐺𝐺𝑖𝑖𝑚𝑚

𝑇𝑇𝑇𝑇𝑇𝑇(𝜃𝜃𝑙𝑙,𝑓𝑓) ∙ ℎ(𝑖𝑖, 𝑘𝑘, 𝑙𝑙) ∙𝑙𝑙∈𝒯𝒯(𝑓𝑓)

𝑙𝑙≠𝑓𝑓

𝐺𝐺𝑖𝑖𝑚𝑚𝑇𝑇𝑇𝑇𝑇𝑇(𝜃𝜃𝑓𝑓,𝑙𝑙) (31)

where,

𝒯𝒯(𝑙𝑙): Set of terrestrial links sharing the same frequency carrier with the l-th terrestial link.

𝑃𝑃𝑖𝑖𝑚𝑚𝑇𝑇𝑇𝑇𝑇𝑇(𝑖𝑖): Transmit power of the i-th link transmit station.

𝐺𝐺𝑖𝑖𝑚𝑚𝑇𝑇𝑇𝑇𝑇𝑇(𝜃𝜃) and 𝐺𝐺𝑖𝑖𝑚𝑚

𝑇𝑇𝑇𝑇𝑇𝑇(𝜃𝜃) : Gain of the terrestrial transmitting/receiving antenna at an offset angle 𝜃𝜃. The radiation pattern can be obtained from ITU-R F.1245-2.

𝜃𝜃𝑙𝑙,𝑓𝑓 : Offset angle (from the boresight direction) of the i-th transmit antenna in the direction of the l-th link receiver antenna.

ℎ(𝑖𝑖, 𝑘𝑘, 𝑙𝑙): Propagation loss considering free space path loss.

According to the previous expression, the Signal-to-Interference plus Noise Ratio (SINR) of the l-th terrestrial backhaul link operating at the k-th carrier can be computed as follows,

𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑖𝑖(𝑙𝑙, 𝑘𝑘) =𝑃𝑃𝑖𝑖𝑚𝑚

𝑇𝑇𝑇𝑇𝑇𝑇(𝑙𝑙) ∙ 𝐺𝐺𝑖𝑖𝑚𝑚𝑇𝑇𝑇𝑇𝑇𝑇(0) ∙ ℎ(𝑙𝑙, 𝑘𝑘, 𝑙𝑙) ∙ 𝐺𝐺𝑖𝑖𝑚𝑚

𝑇𝑇𝑇𝑇𝑇𝑇(0)𝑖𝑖𝑖𝑖

𝑓𝑓 (𝑘𝑘) + 𝑁𝑁𝑖𝑖 (32)

where 𝑁𝑁𝑖𝑖 is the noise thermal power.

6.1.2 Satellite segment model

In this section, we focus on the Satellite-to-Earth (downlink) transmission direction. Let us consider 𝑀𝑀 ≤ 𝑁𝑁 terrestrial nodes equipped with satellite dish antennas. Note that in the satellite segment, the number of links is equal to the number of hybrid satellite-terrestrial nodes.

The 𝑀𝑀 satellite links operate in the 17.7-19.7 GHz band, which is the same also for the terrestrial segment. We assume that the 17.7-19.7 GHz band is divided into 𝐶𝐶 equally sized frequency carriers of bandwidth size 𝐵𝐵𝑒𝑒, which represent the satellite forward carrier bandwidth. For the sake of clarity, we assigned a carrier identification number to each of the 𝐶𝐶 frequency carriers, namely 𝑐𝑐 = 1, … , 𝐶𝐶. Similarly, from the total 𝐶𝐶 carriers, the carrier allocation algorithm chooses 𝐶𝐶′ ≤ 𝐶𝐶 to design the satellite backhaul network.

Let 𝒂𝒂𝑒𝑒 ∈ 𝑆𝑆𝑀𝑀 be the satellite carrier allocation vector, whose elements 𝑎𝑎𝑒𝑒(𝑖𝑖) ∈ [1, 𝐶𝐶] contain the carrier identification number of the satellite carrier that has been assigned to the i-th satellite link. The forward satellite links work on a single carrier communication mode and, thus, they can only be assigned one carrier frequency and this cannot be shared with others satellite links. Based on this discussion, 𝐶𝐶′ = 𝑀𝑀 and the following constraint is considered:

𝒂𝒂𝑒𝑒(𝑖𝑖) ≠ 𝒂𝒂𝑒𝑒(𝑗𝑗) 𝑖𝑖, 𝑗𝑗 = 1, … , 𝑀𝑀; 𝑖𝑖 ≠ 𝑗𝑗 (33)


Date: 21/04/2017

of 87

SANSA-645047 D3.6

which ensures that a single carrier is not assigned to multiple links.

On the other hand, and due to the spectrum sharing situation, satellite links will be affected by the interference caused by the terrestrial backhaul links. Let us denote 𝑆𝑆(𝑚𝑚) as the set of terrestrial links that share the same carrier frequency than the m-th satellite link. The interference level seen at the m-th satellite link operating at the c-th carrier can be written as,

𝑖𝑖𝑒𝑒𝑚𝑚(𝑐𝑐) = � 𝑃𝑃𝑖𝑖𝑚𝑚

𝑇𝑇𝑇𝑇𝑇𝑇(𝑖𝑖) ∙ 𝐺𝐺𝑖𝑖𝑚𝑚𝑇𝑇𝑇𝑇𝑇𝑇(𝜃𝜃𝑙𝑙,𝑚𝑚) ∙ ℎ(𝑖𝑖, 𝑐𝑐, 𝑚𝑚) ∙ 𝐺𝐺𝑖𝑖𝑚𝑚

𝑃𝑃𝑆𝑆𝑇𝑇(𝜃𝜃𝑚𝑚,𝑙𝑙)𝑙𝑙∈𝑃𝑃(𝑚𝑚)

(34)

where 𝐺𝐺𝑖𝑖𝑚𝑚𝑃𝑃𝑆𝑆𝑇𝑇(𝜃𝜃) is the gain of the satellite dish receiver antenna at an offset angle 𝜃𝜃 . The

radiation pattern of the dish antenna can be obtained from ITU-R S.465-6. The variable ℎ(𝑖𝑖, 𝑐𝑐, 𝑚𝑚) denotes the propagation loss between the transmit antenna of the i-th terrestrial link and the satellite terminal receiver of the m-th satellite link when operating at the c-th carrier. For the computation of the later, path loss considered. Note that thus far, we assumed that the interfering signal falls within the victim bandwidth. If the spectra do not overlap completely,

then a compensation factor of 𝐵𝐵𝑙𝑙𝑖𝑖𝑙𝑙𝑖𝑖𝑓𝑓𝑙𝑙𝑝𝑝

𝐵𝐵𝑒𝑒� is applied, where 𝐵𝐵𝑙𝑙𝑖𝑖𝑙𝑙𝑖𝑖𝑓𝑓𝑙𝑙𝑝𝑝 stands for the portion of

the interfering signal spectral density within the receive modem filter bandwidth given by 𝐵𝐵𝑒𝑒.

Finally, the SINR of the satellite backhaul links can be computed as follows,

𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑒𝑒(𝑚𝑚, 𝑐𝑐) =𝑃𝑃𝑒𝑒 ∙ 𝐺𝐺𝑒𝑒 ∙ ℎ𝑒𝑒(𝑚𝑚, 𝑐𝑐) ∙ 𝐺𝐺𝑖𝑖𝑚𝑚

𝑃𝑃𝑆𝑆𝑇𝑇(0)𝑖𝑖𝑒𝑒

𝑚𝑚(𝑐𝑐) + 𝑃𝑃𝑙𝑙𝑙𝑙 + 𝑁𝑁𝑒𝑒 (35)

where 𝑃𝑃𝑒𝑒 refers to the satellite transmit power, 𝐺𝐺𝑒𝑒 denotes the satellite antenna gain, ℎ𝑒𝑒(𝑚𝑚, 𝑐𝑐) denotes the satellite channel gain of the m-th satellite link operating at the c-th carrier (only free space path loss is considered for the satellite links), 𝑃𝑃𝑙𝑙𝑙𝑙 is the co-channel interference due to the use of multibeam satellite, and 𝑁𝑁𝑒𝑒 is the thermal noise power seen at the satellite dish antenna.

6.2 Proposed carrier allocation algorithm

The goal is to design jointly the terrestrial and satellite carrier assignment for which the interference impact on both the terrestrial and satellite link performance is minimal. The allocation processing is assumed to take place in a centralized controller that has access to the overall network planning information. To be more specific, in this section, we will focus on the maximization of the worst link in terms of interference. We discarded the sum-rate utility function because, although it was proposed as one of the key performance indicators identified in WP2, it might end up sacrificing some links in detriment of others with good rate performance. It should be noted that, we assume that the topology configuration has been decided in a previous step by the HNM and resource allocation cannot change that by sacrificing links.

Therefore, the corresponding max-min problem is formulated as follows:


Date: 21/04/2017

of 87

SANSA-645047 D3.6

max𝒂𝒂𝑡𝑡,𝒂𝒂𝑠𝑠

min𝑗𝑗,𝑙𝑙

{𝑆𝑆𝑖𝑖(𝑗𝑗), 𝑆𝑆𝑒𝑒(𝑖𝑖)}

𝑠𝑠. 𝑑𝑑. 𝑎𝑎𝑖𝑖(𝑗𝑗) ∈ [1, 𝐾𝐾], 𝑗𝑗 = 1, … , 𝐿𝐿𝑎𝑎𝑒𝑒(𝑖𝑖) ∈ [1, 𝐶𝐶], 𝑖𝑖 = 1, … , 𝑀𝑀

𝑎𝑎𝑖𝑖(𝑖𝑖) ≠ 𝑎𝑎𝑖𝑖(𝑗𝑗), 𝑗𝑗 ∈ I(n), i ∈ O(n), n = 1, … , 𝑁𝑁𝑎𝑎𝑒𝑒(𝑖𝑖) ≠ 𝑎𝑎𝑒𝑒(𝑗𝑗), 𝑖𝑖, 𝑗𝑗 = 1, … , 𝑀𝑀; 𝑖𝑖 ≠ 𝑗𝑗

(36)

where

𝑆𝑆𝑖𝑖(𝑗𝑗) = 𝐵𝐵𝑖𝑖log2(1 + 𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑖𝑖(𝑗𝑗, 𝒂𝒂𝑖𝑖)) (37)

𝑆𝑆𝑒𝑒(𝑖𝑖) = 𝐵𝐵𝑒𝑒log2(1 + 𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑒𝑒(𝑖𝑖, 𝒂𝒂𝑖𝑖 , 𝒂𝒂𝑒𝑒)) (38)

Clearly, the considered joint carrier allocation problem of terrestrial and satellite segment is intractable due to the non-linear coupling between each other. This is because the terrestrial allocation directly disturbs the sum-rate of the satellite backhaul network. Even the max-min link rate maximization of only terrestrial links remains intractable since the link allocation per carrier amounts to determine which terrestrial link should be grouped into an interfering channel such that the sum-rate is maximized, and this problem is in general NP-hard [47]. More specifically, solving the carrier allocation problem would require that all possible carrier combinations are considered, which is computationally prohibitive in general.

To handle the problem, we make use of decomposition methods in which the coupled problem is split into several small sub-problems. More precisely, in this document we first deal with the carrier assignment for the satellite backhaul network and, on a second step, assuming the resulting satellite segment allocation, we design a sub-optimal carrier assignment for the terrestrial part of the network. As discussed before, the latter is not a tractable problem and, thus, finding the optimal solution is very challenging. Therefore, we propose an algorithm for solving the terrestrial assignment iteratively.

The order in which each part of the hybrid network is optimized first is driven by the degrees of freedom of the resource allocation problem, which in our case is higher for the terrestrial network since it has more flexibility to adapt to the existing spectral environment.

6.2.1 Step 1: Carrier Allocation for the Satellite Backhaul Network

In this first step, the carrier allocation for the satellite backhaul network is designed by ignoring the terrestrial networks. Given the expected number of terrestrial links, orders of magnitude higher than the satellite link, we expect that the terrestrial link will be able to design its carrier allocation so that the interference seen from the satellite links is minimized.

Given the single carrier transmission of the satellite segment, the satellite spectrum usage is dictated by the number of satellite links. Therefore, the carrier allocation of the satellite segment


Date: 21/04/2017

of 87

SANSA-645047 D3.6

reduced to find the combination of 𝑀𝑀 satellite links with 𝐶𝐶′ satellite carriers, where 𝑀𝑀 = 𝐶𝐶′. In this first step, we proceed as if there were no terrestrial network and, thus, the satellite carrier allocation is based on each satellite link’s SINR conditions, as in [7][31]. Unlike [7][31], we first select the satellite link with the lowest SINR and then we assign it with the carrier that provides the higher SINR. The SINR is computed as in (35) with 𝑖𝑖𝑒𝑒

𝑚𝑚(𝑐𝑐) = 0, 𝑚𝑚 = 1, … , 𝑀𝑀, 𝑐𝑐 = 1, … , 𝐶𝐶. The problem can be formulated as follows:

max 𝒂𝒂𝑠𝑠

min 𝑗𝑗

𝑆𝑆𝑒𝑒(𝑗𝑗)

𝑠𝑠. 𝑑𝑑. 𝑎𝑎𝑒𝑒(𝑖𝑖) ≠ 𝑎𝑎𝑒𝑒(𝑗𝑗), 𝑖𝑖, 𝑗𝑗 = 1, … , 𝑀𝑀; 𝑖𝑖 ≠ 𝑗𝑗𝑎𝑎𝑒𝑒(𝑗𝑗) ∈ [1, 𝐶𝐶], 𝑗𝑗 = 1, … , 𝑀𝑀 (39)

6.2.2 Step 2: Carrier Allocation for the Terrestrial Backhaul Network

In this section, we aim at designing the carrier allocation of the terrestrial network so that the rate of the terrestrial link in the worst interference conditions is maximized and the interference caused to the satellite receivers is minimized.

The terrestrial carrier allocation optimization problem is a full search space problem, i.e. for its solution one needs to search for the best among all the possible combinations of carrier assignments, which is a tedious and time-consuming process. In this section, we propose an iterative algorithm that solves the carrier assignment in a sequential manner so that at each step the search space is reduced.

In general, terrestrial intra-system interference should be minimized so that the SINR of the terrestrial links, which is given in (32), is maximized. In other words, the optimal carrier allocation will tend to be carrier hungry. That is, if the channel condition on an empty carrier is acceptable for a particular link, it should be allocated to the empty carrier instead of being allocated to one carrier that is already being used by other links. On the other hand, the terrestrial interference caused to the satellite receivers should be minimized. In other words, the terrestrial carrier allocation should not only maximize the SINR of the terrestrial links but also maximize the SINR of the satellite links, which is given in (65). Therefore, the terrestrial assignment should take into account both types of interference.

Regarding the terrestrial intra-system interference, let us define the SINR matrix of the terrestrial links as follows,

𝑺𝑺𝑰𝑰𝑵𝑵𝑹𝑹𝑖𝑖 = �𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑖𝑖(1,1) ⋯ 𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑖𝑖(1, 𝐾𝐾)

⋮ ⋱ ⋮𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑖𝑖(𝐿𝐿, 1) ⋯ 𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑖𝑖(𝐿𝐿, 𝐾𝐾)

� (40)

where the columns indicate the carrier frequencies and the rows indicate the terrestrial links.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Regarding the interference caused at the satellite terminals, let us define 𝑮𝑮(𝑙𝑙) ∈ 𝑆𝑆𝑀𝑀×𝐾𝐾 as a matrix containing the interference level 𝑔𝑔𝑓𝑓(𝑚𝑚, 𝑘𝑘), which is defined as the interference caused by the l-th terrestrial link operating at k-th carrier and received at the m-th satellite link, which can be expressed as,

𝑔𝑔𝑓𝑓(𝑚𝑚, 𝑘𝑘) = 𝑃𝑃𝑖𝑖𝑚𝑚𝑇𝑇𝑇𝑇𝑇𝑇(𝑙𝑙) ∙ 𝐺𝐺𝑖𝑖𝑚𝑚

𝑇𝑇𝑇𝑇𝑇𝑇(𝜃𝜃𝑓𝑓,𝑚𝑚) ∙ ℎ(𝑙𝑙, 𝑘𝑘, 𝑚𝑚) ∙ 𝐺𝐺𝑖𝑖𝑚𝑚𝑃𝑃𝑆𝑆𝑇𝑇(𝜃𝜃𝑚𝑚,𝑓𝑓) (41)

The information in 𝑮𝑮(𝑙𝑙) is used for identifying the satellite link that receives the highest level of interference when the l-th terrestrial link operates in carrier k. That is, for each k-th carrier, 𝑀𝑀𝑤𝑤(𝑙𝑙, 𝑘𝑘) = max [𝑮𝑮(𝑙𝑙)]𝑘𝑘, where [𝐺𝐺(𝑙𝑙)]𝑘𝑘 denotes the k-th column of matrix 𝑮𝑮(𝑙𝑙) and 𝑀𝑀𝑤𝑤(𝑙𝑙, 𝑘𝑘)

indicates the worst satellite link in terms of interference when terrestrial link l operates in carrier k. For convenience, let us define the following matrix containing these worst satellite links as follows:

𝑴𝑴𝑤𝑤 = �𝑀𝑀𝑤𝑤(1,1) ⋯ 𝑀𝑀𝑤𝑤(1, 𝐾𝐾)

⋮ ⋱ ⋮𝑀𝑀𝑤𝑤(𝐿𝐿, 1) ⋯ 𝑀𝑀𝑤𝑤(𝐿𝐿, 𝐾𝐾)

� (42)

Next, the SINR level of these worst satellite links is computed considering the satellite carrier allocation of the first step, where only the l-th terrestrial link is active. These SINR values are captured in the following matrix,

𝑺𝑺𝑰𝑰𝑵𝑵𝑹𝑹𝑒𝑒 = �𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑒𝑒(1,1) ⋯ 𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑒𝑒(1, 𝐾𝐾)

⋮ ⋱ ⋮𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑒𝑒(𝐿𝐿, 1) ⋯ 𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑒𝑒(𝐿𝐿, 𝐾𝐾)

� (43)

which captures the individual interference effect of each terrestrial link operating at the different carriers.

The two SINR matrices, namely 𝑺𝑺𝑰𝑰𝑵𝑵𝑹𝑹𝑖𝑖 and 𝑺𝑺𝑰𝑰𝑵𝑵𝑹𝑹𝑒𝑒 can be transformed into rate matrices, 𝑹𝑹𝑖𝑖 and 𝑹𝑹�𝑒𝑒, by computing namely 𝑙𝑙𝑙𝑙𝑔𝑔2(1 + 𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆) of each matrix element. The proposed terrestrial assignment algorithm reduces to maximize the worst rate represented by these two matrices, one corresponding to the terrestrial segment performance and one corresponding to the satellite segment performance. The bi-objective optimization problem can be reduced to a single-objective optimization problem with the weighted sum technique in which the two rate matrices are combined in a single one by performing simple weighted addition, i.e. 𝑒𝑒𝑖𝑖𝑹𝑹𝑖𝑖 +𝑒𝑒𝑒𝑒𝑹𝑹�𝑒𝑒. In SANSA, the weights are all set to one so that no priority is given to either terrestrial or satellite links.

Therefore, the problem can be formulated as follows:


Date: 21/04/2017

of 87

SANSA-645047 D3.6

max

𝒂𝒂𝑡𝑡 min

𝑗𝑗�𝑆𝑆𝑖𝑖(𝑗𝑗) + 𝑆𝑆�𝑒𝑒(𝑗𝑗)�

𝑠𝑠. 𝑑𝑑. 𝑎𝑎𝑖𝑖(𝑗𝑗) ∈ [1, 𝐾𝐾], 𝑗𝑗 = 1, … , 𝐿𝐿𝑎𝑎𝑖𝑖(𝑖𝑖) ≠ 𝑎𝑎𝑖𝑖(𝑗𝑗), 𝑗𝑗 ∈ I(n), i ∈ O(n), n = 1, … , 𝑁𝑁 (44)

The proposed algorithm for the terrestrial carrier assignment is summarized in Algorithm 2, which makes use of the previous ideas. Essentially, the proposed algorithm is based on a process that assigns carriers in an iterative manner to the unassigned terrestrial links in the worst interference condition, according to 𝑹𝑹𝑖𝑖 and 𝑹𝑹�𝑒𝑒. This process is repeated by taking into account previous assignments until all L links are assigned. Once all L links have been assigned, the algorithm performes the previous procedure iteratively by considering the previous carrier assignment. Note that for the computation of 𝑹𝑹𝑖𝑖 and 𝑹𝑹�𝑒𝑒 , only the links that have been already assigned are considered. Due to the sequential assignment considered in Algorithm 2, the algorithm is expected to take few iterations until the carrier assignment converges to a steady state.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

6.3 Simulation results

6.3.1 Benchmark carrier allocation

In order to compare the proposed carrier allocation strategy, we consider the HSTB network topology depicted in Figure 6-2 where we have 15 terrestrial nodes that are interconnected via L=28 unidirectional communication links (forming 14 bi-directional links). This is a true backhaul topology that is used in Finland. The complete database related to the terrestrial links has been obtained from the Finnish communications regulatory authority (FICORA). Apart from details on the geographical location, pointing and antenna gains, it also includes a benchmark carrier allocation.

The database indicates that 8 carriers of 56 MHz are used to allocate the L=28 terrestrial links. These carriers are divided into two blocks of 4 carriers each: one block from 17700 to 17924 MHz and another block from 18708 and 18934 MHz. The assignment is as illustrated in Figure 6-3.

Figure 6-2. Helsinki topology


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Figure 6-3. Benchmark carrier assignment

According to the benchmark carrier assignment included in the database, we computed the SINR values and rates achieved by each link, which are detailed in Table 6-1.

Table 6-1. SINR and link-rate for the benchmark topology

Link Number

Initial Node

End Node

Number of interfering

links SINR (dB) Link Rate

(Mbps)

1 1 2 4 60.44 1124.32 2 2 1 4 60.96 1133.97 3 1 3 3 52.98 985.55 4 3 1 3 49.14 914.20 5 1 4 3 49.18 914.93 6 4 1 3 54.42 1012.44 7 1 5 3 60.83 1131.61 8 5 1 3 61.11 1136.86 9 1 6 3 59.25 1102.17

10 6 1 3 59.78 1112.08 11 7 3 3 54.13 1006.89 12 3 7 3 51.42 956.59 13 8 1 0 61.16 1137.73 14 1 8 0 60.68 1128.80 15 9 10 4 56.42 1049.55 16 10 9 4 58.68 1091.66 17 9 11 3 61.43 1142.84 18 11 9 3 61.01 1135.02 19 9 12 3 61.43 1142.75


Date: 21/04/2017

of 87

SANSA-645047 D3.6

20 12 9 3 60.95 1133.91 21 13 10 3 61.00 1134.77 22 10 13 3 60.52 1125.90 23 8 10 4 53.03 986.54 24 10 8 4 46.46 864.27 25 8 14 4 48.44 901.15 26 14 8 4 49.00 911.53 27 8 15 4 45.91 854.11 28 15 8 4 49.33 917.67

Sum-Rate = 29189.81

In this document we will focus on the evaluation based on the SE computed as,

𝑆𝑆𝑆𝑆 [𝑏𝑏𝑏𝑏𝑠𝑠/𝐻𝐻𝐻𝐻] =SumRate

Total bandwidth (45)

For the benchmark scenario considered here, 8 carriers of 56MHz are used, which correspond to 448 MHz of total bandwidth. Therefore, the benchmark SE is 65.16 bps/Hz.

We will also consider the Energy Efficiency (EE) defined as,

𝑆𝑆𝑆𝑆 [𝑏𝑏𝑏𝑏𝑠𝑠/𝑊𝑊] =SumRate

Total Power (46)

Regarding the EE, the benchmark considers 28 links each operating at different transmit powers as indicated in the database. The resulting EE is 83.44 Mbps/mW.

6.3.2 Simulation set-up: Extended Helsinki topology

In order to evaluate the proposed carrier allocation method, we consider an extension of the Helsinki topology, which is depicted in Figure 6-4. The changes on the Helsinki topology were decided within the WP3-WP4 collaboration in order to include the hybrid nodes and complicate a bit the network by introducing higher number of links. In particular, as depicted in Figure 6-4, the changes are: - 9 new bi-direction links have been included (in green) - 1 bi-direction link between node 1 and node 5 have been removed (in red) - Node 4 and node 12 have satellite transmission capabilities. These are nodes that are far away from each other and represent alternative offloading points from the terrestrial backhaul. The final extended HSTB network topology is illustrated in Figure 6-5, where we have N=15 terrestrial nodes that are interconnected via L=44 unidirectional communication links. As


Date: 21/04/2017

of 87

SANSA-645047 D3.6

mentioned above, we assume that M=2 out of N=15 terrestrial nodes are equipped with a satellite dish antenna. The hybrid nodes are indicated with a red dot in Figure 6-5.

We assume a multi-beam satellite located at orbital position 13E. The beam pattern has been simulated as in [48] and it is depicted as well in Figure 6-5. A summary of the system parameters that have been considered for the simulation set-up for the terrestrial and satellite segment is given in Table 6-2 and Table 6-3, respectively.

Figure 6-4. Extended Helsinki topology

Figure 6-5. Simulated HSTB topology with the simulated satellite beam pattern.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Table 6-2. Terrestrial Segment Parameters Terrestrial Segment

Parameter Value Antenna Pattern ITU-R F.1245-2

Max. antenna gain 38 dBi Channel Model Path Loss + Diffraction Loss Noise Density -139 dBW/MHz

Transmit Power Density -38.13 dBW/MHz Total bandwidth 17.7-19.7 GHz

Table 6-3. Satellite Segment Parameters Satellite Segment

Parameter Value Antenna Pattern ITU-R S. 465

Max. antenna gain 42.1 dBi Channel Model Path loss

Antenna Temperature 262 K Height of the Antenna Half of the terrestrial antenna

Transmit Power 9.23 dBW Link Distance 35.786 km

C/I 10.53 dB Total bandwidth 17.7-19.7 GHz

The partial database related to the true terrestrial links have been obtained from the Finnish communications regulatory authority (FICORA), and this includes information listed on a station by station basis with the geographical location, maximum antenna gain, transmit power, channel bandwidth, height, etc.

For the new links, we extended the database by mapping the true altitude and height of the true nodes to the corresponding new links; we fixed the antenna gain to 38 dBi (as it is for all the antennas in the Helsinki topology) and we computed the elevation and azimuth angle together with the distance according to the transmit and receive antenna locations. The information included in the database will be used here to properly model the interference levels and to properly compute the link budget.

6.3.3 Extended Benchmark (without satellite links)

For the computation of the benchmark SE, we consider the benchmark carrier allocation indicated at the database as shown in Figure 6-3.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

For the new links, we propose the carrier assignment described in Table 6-4 and Figure 6-6, which ensures low interference values while using the same set of 8 carriers.

Table 6-4. Benchmark carrier allocation for the extended Helsinki topology

Initial Node

End Node

Min. Frequency

(GHz)

Max. Frequency

(GHz)

Bandwidth (MHz)

5 2 17.756 17.812 56 2 5 18.764 18.820 56 5 6 18.708 18.764 56 6 5 17.700 17.756 56 5 8 17.756 17.812 56 8 5 18.764 18.820 56 9 3 18.708 18.764 56 3 9 17.700 17.756 56 9 8 18.708 18.764 56 8 9 17.700 17.756 56

10 11 17.812 17.868 56 11 10 18.820 18.876 56 7 12 18.876 18.934 56

12 7 17.868 17.924 56 7 13 18.876 18.934 56

13 7 17.868 17.924 56 13 15 17.756 17.812 56 15 13 18.764 18.820 56


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Figure 6-6. Benchmark carrier allocation for the extended Helsinki topology (22 bi-directional links)

According to the proposed carrier assignment, we computed the SINR values and rates achieved by each link, which are detailed in Table 6-5.

Table 6-5. SINR and link-rate for the extended benchmark topology

Link Number

Initial Node

End Node



(Mbps)

1 1 2 6 60.44 1124.31 2 2 1 6 60.96 1133.97 3 1 3 4 52.98 985.53 4 3 1 4 49.14 914.20 5 1 4 4 49.18 914.93 6 4 1 4 54.42 1012.44 7 1 6 5 58.79 1093.56 8 6 1 5 59.83 1112.96 9 1 8 3 50.01 930.29

10 8 1 3 60.41 1123.84 11 2 5 5 48.50 902.21 12 5 2 5 48.12 895.23 13 7 3 4 54.12 1006.78 14 3 7 4 51.42 956.58 15 9 3 3 48.16 895.82 16 3 9 3 51.11 950.74 17 5 6 3 52.87 983.61


Date: 21/04/2017

of 87

SANSA-645047 D3.6

18 6 5 3 58.44 1087.07 19 5 8 5 47.18 877.75 20 8 5 5 47.80 889.26 21 7 12 6 48.48 901.94 22 12 7 6 58.19 1082.44 23 7 13 6 47.63 886.00 24 13 7 6 38.40 714.28 25 9 8 3 48.13 895.27 26 8 9 3 44.05 819.40 27 10 8 6 46.45 864.15 28 8 10 6 53.03 986.52 29 14 8 6 49.00 911.50 30 8 14 6 48.05 893.85 31 15 8 6 49.33 917.63 32 8 15 6 45.91 854.09 33 9 10 6 56.42 1049.53 34 10 9 6 58.58 1089.74 35 9 11 4 46.49 864.89 36 11 9 4 55.92 1040.33 37 9 12 5 61.43 1142.68 38 12 9 5 60.94 1133.64 39 11 10 4 48.45 901.39 40 10 11 4 57.99 1078.83 41 13 10 5 60.37 1123.02 42 10 13 5 42.61 792.69 43 13 15 5 48.33 899.05 44 15 13 5 65.84 1224.73

Sum-Rate = 42858.67

For the benchmark extended topology considered we have a benchmark SE of 95.67 bps/Hz.

The increase of the SE compared with the non-extended Helsinki topology is given by the fact that we are adding 8 links and within the same 448 MHz band.

Regarding the EE, the extended benchmark considers 44 links each operating at different transmit powers. The resulting EE is 83.75 Mbps/mW.

6.3.4 Spectral efficiency analysis

In this section, we evaluate the proposed carrier assignment algorithm. In particular, we would like to shed some light into how the frequency re-use can improve the backhaul efficiency when the resulting intra-carrier interference is properly managed by the carrier allocation design.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

To this aim, we divide the 448 MHz band (corresponding to the benchmark) into sub-bands of 𝐵𝐵𝑊𝑊 [MHz] each. Here, we assume that 𝐵𝐵𝑊𝑊 = 𝐵𝐵𝑖𝑖 = 𝐵𝐵𝑒𝑒. The values of 𝐵𝐵𝑊𝑊 that are going to be evaluated are illustrated in Table 6-6, together with the corresponding number of resulting carriers.

Note that, increasing the channel bandwidth translates into an increase on the received noise, which automatically implies lower SINR values for a fixed transmit power. Therefore, in order to prevent SINR drops, we fixed the transmit power density to -38.13 dBW/MHz. This value has been set so that the achieved SINR is of similar values to the ones obtained in the benchmark Helsinki topology. The transmit power corresponding to the evaluated channel bandwidths are shown in Table 6-6.

Table 6-6. Bandwidth values BW (MHz) No. Carriers Tx Power (dBW)

14 32 -26.67

28 16 -23.66

56 8 -20.65 112 4 -17.64

224 2 -14.63

For the evaluation, we consider the extended topology depicted in Figure 6-5 and the simulation set-up described in Section 6.3.2. Therefore, 2 satellite links and 44 terrestrial links should be allocated. We have applied the algorithm described in 6.2 and the obtained results are summarized in Table 6-7. The results are based on the spectral efficiency in [bps/Hz]. We detail the SE obtained with the satellite and the terrestrial segment separately and the overall SE obtained with the SANSA hybrid terrestrial-satellite network. Moreover, for the satellite case, we provide two results: a) Assuming no-terrestrial interference (noted in Table 6-7 as “w/o Terrestrial”), and b) Assuming terrestrial interference (noted in Table 6-7 as “w/ Terrestrial”). It can be observed that for 𝐵𝐵𝑊𝑊 = {14,28} MHz, which corresponds to a total number of available carriers of {32,16}, respectively, the algorithm concludes that there is no need to use all available carriers (14 and 13 carriers are enough to solve the carrier allocation for this specific example). In other words, there is no need to make use of the remaining empty carriers because the interference is efficiently managed by the carrier allocation and will not be furtherly reduced by increasing the number of carriers. Therefore, for these three cases, the satellite system is not affected by the spectral co-existence with the terrestrial backhauling network as indicated by the SE values achieved by the “w/ terrestrial” and “w/o terrestrial” columns of Table 6-7 . Moreover, the overall SE of the SANSA network is maintained to around 70 bps/Hz. Note that this is already lower than that obtained by the benchmark only-terrestrial carrier assignment described in Section 6.3.3, which considered 56 MHz carriers.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Table 6-7. Results obtained with the proposed carrier allocation algorithm BW

(MHz) Max No. Carriers

No. Carriers

Used

SE Satellite w/o

Terrestrial [bps/Hz]

SE Satellite w/


SE Terrestrial [bps/Hz]

SE Hybrid [bps/Hz]

14 32 14 3.61 3.61 64.46 64.97

28 16 13 3.59 3.59 69.37 69.93

56 8 8 3.56 3.56 107.44 108.33 112 4 4 3.50 3.10 179.88 181.43

224 2 3 3.38 2.80 106.45 109.26

For the values of bandwidth 𝐵𝐵𝑊𝑊 > 56MHz, all the available carriers have to be used and, thus, the effect of the interference on the satellite link starts becoming visible on the satellite SE. This can be observed on the “SE Satellite w/ Terrestrial” column, whose entries decrease as the number of available carriers reduces. The satellite SE drop is justified by the fact that, as the number of carriers reduces, higher frequency re-use should be done on the terrestrial segment, which translate in higher interference to the satellite. On the contrary, the SE of the terrestrial network increases as the number of carriers reduces and this is because the carrier allocation algorithm is able to efficiently manage the resulting terrestrial-to-terrestrial interference. This has a positive effect on the overall hybrid network, whose SE increases as the spectrum re-use increases, as illustrated in Figure 6-7. The most important conclusion of Table 6-7 and Figure 6-7 is the fact that, with high frequency reuse, the SE goes to 181.43 bps/Hz, which compared to the 95.67 bps/Hz of the benchmark (indicated with a red dot in Figure 6-7) translates into 1.90x increase. Note that there is a saturation point in 𝐵𝐵𝑊𝑊 = 112, which correspond to 4 available carriers. The next point on the SE curve of Figure 6-7 drastically drops breaking the growth trend. This point corresponds to the case where the carrier allocation algorithm is not able to manage the interference of the overall network. This is basically due to the Full-Duplex scenarios that come up when using 2 carriers for the whole network.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Figure 6-7. SE of the hybrid terrestrial-satellite network as a function of the carrier bandwidth

The effect of the terrestrial-to-terrestrial interference on the SINR values of the terrestrial system is depicted in Figure 6-8 in terms of CDF. It can be observed that the SINR distribution degrades as the spectral re-use increases, especially when the number of available carrier is way below the 12 limit (cases 𝐵𝐵𝑊𝑊 = {112,224}MHz).

Figure 6-8. CDF of terrestrial SINR distribution

Figure 6-9 illustrates the EE values achieved with respect to the carrier bandwidth. With the 𝐵𝐵𝑊𝑊 = 112 MHz, which was the case considering 4 carriers, we get a 1.27x gain in terms of EE when comparing with the benchmark.

0 50 100 150 200 25060

80

100

120

140

160

180

200

Carrier Bandwidth [MHz]

Hyb

rid S

E [b

ps/H

z]

Benchmark

25 30 35 40 45 50 55 60 65 700

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Terrestrial SINR (dB)

CD

F

B=224MHzB=112MHzB=56MHzB=28MHzB=14MHz


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Figure 6-9. EE as a function of the carrier bandwidth

Now, we present some results where the BW value is kept to 56 MHz, while the variable that changes is the number of carriers or, equivalently, the total available bandwidth. Table 6-8 summarizes the different values of total available bandwidth examined in this simulation experiment, together with the SE values obtained. The SE values are also depicted in Figure 6-10. It is interesting to highlight that the fixed carrier bandwidth makes the satellite SE “w/o Terrestrial” to be constant to 3.56 bps/Hz. However, these values decrease as the terrestrial system increases the frequency reuse, thereby creating more interference to the satellite terminals. The maximum SE is achieved with 3 carriers, reaching 200.28 bps/Hz. Compared to the benchmark, this translates into 2.09x gain in terms of SE.

Table 6-8. Results obtained with the proposed carrier allocation algorithm Total BW

(MHz)

Max No. Carriers

No. Carriers

Used

SE Satellite w/o


SE Satellite w/


SE Terrestrial [bps/Hz]

SE Hybrid [bps/Hz]

448 8 8 3.56 3.56 107.44 108.33

336 6 6 3.56 3.55 133.72 134.91 280 5 5 3.56 3.56 156.17 157.59

224 4 4 3.56 3.56 178.97 180.75

168 3 3 3.56 3.32 198.07 200.28

112 2 2 3.56 3.32 194.32 197.64

0 50 100 150 200 25080

90

100

110

120

130

140

Carrier Bandwidth [MHz]

EE

[Mbp

s/m

W]

Terrestrial SANSA NetworkBenchmark


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Figure 6-10. SE of the hybrid terrestrial-satellite network as a function of the number of carriers

Next, Figure 6-11 illustrates the effect of the terrestrial-to-terrestrial interference on the SINR values of the terrestrial system. As expected, the SINR values degrade as the frequency re-use increases.

Figure 6-11. CDF of terrestrial SINR distribution

234567880

100

120

140

160

180

200

220

No. of carriers

Hyb

rid S

E [b

ps/H

z]

Benchmark

25 30 35 40 45 50 55 60 65 700

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Terrestrial SINR (dB)

CD

F

8 carriers5 carriers4 carriers3 carriers2 carriers


Date: 21/04/2017

of 87

SANSA-645047 D3.6

7 Carrier assignment and flow control for hybrid satellite-terrestrial backhaul networks

7.1 Introduction

Achieving maximum performance of the SANSA network is not a question of optimizing only the carrier assignment. There is an essential requirement for performing cross-layer optimization between the different layers to handle the link scheduling in an intelligent way as well as the traffic that should be scheduled over these links together with the carrier assignment. The different backhauling nodes need to transmit/receive their own traffic in addition to routing the other nodes traffic. Accordingly, an adequate link scheduling and frequency assignment policy allows the network nodes to have simultaneous transmissions, which increases the network spectrum efficiency and guarantee minimum throughput in the different links.

This chapter tackles the problem of cross-layer design of flow control, frequency assignment, and link scheduling of the SANSA hybrid terrestrial-satellite wireless backhaul network. The frequency and the link scheduling are performed in a way that guarantees the satisfaction of the radio and the interference constraints. The flow control uses the information related to the network link capacities to decide the amount of data that should be transmitted over the different links. The overall target is to deliver the highest amount of traffic considering the characteristics of the hybrid satellite-terrestrial backhauling network.

7.2 Related work

A survey paper that summarizes the frequency assignment and flow control in wireless mesh network can be found in [49].

The problem of channel scheduling in multi-radio multi-channel is proved to be an NP-Hard problem [50]. In [51][52], a work on assigning a subset of subcarriers to a given link in wireless mesh networks has been developed. The assignment is based on the channel conditions. The frequency reuse is not allowed and thus the frequency is assigned to a single link at a given time. In [53], the achievable rates in multi-hop wireless mesh networks with orthogonal channels are determined together with tight necessary and sufficient conditions for the achievability of the rate vector. In [54], the frequency reuse is enabled in an OFDMA-based network, where heuristic algorithms for joint power control, frequency allocation and link scheduling are developed. This work is developed by the same authors in [55] by relaxing part of the constrains and estimating the links’ capacities in low and high SNR regimes.

In [56], fair end-to-end transmission is achieved by developing a distributed algorithm for joint power and subcarrier allocation in OFDMA-based wireless mesh network. Based on a known


Date: 21/04/2017

of 87

SANSA-645047 D3.6

transmitting paths, the joint rate and power control is treated as a network utility maximization problem with interference consideration.

In [57], interference aware resource allocation in OFDMA-based wireless mesh network is proposed. The frequencies are allocated to the different links in a way that guarantees the interference mitigation. The frequency is assumed not to be used in all the nodes that are in the interference distance (the interference model protocol). Based on this work, the same authors adopted a decoupled routing and scheduling scheme [58] to maximize the network overall throughput. The interference aware resource allocation in [57] is applied and afterwards the flow is calculated for maximizing the delivered throughput. Based on the resulted flow, the time-frequency assignment is performed for the evaluation of the final scheduling. A general framework based on [57] and [58] is published in [59].

In [60], the resource allocation problem is modeled as the bankruptcy game by considering the interference between the different nodes. Different solutions are identified based on the cooperative game theory. The authors in [61] consider optimally partitioning the spectrum into a set of non-overlapping bands channels with non-uniform widths to allow more parallel transmission; moreover, a low complexity heuristic algorithm has been developed.

The work in this chapter is different from the already mentioned papers in the sense that it considers the joint flow control and link scheduling in multi-frequency wireless backhaul network, where both terrestrial and satellite links are available in the system. The problem is solved by taking the radio limitations of the network into account. The frequency reuse is enabled in the network under pre-determined link scheduling constraints. The optimization problem is formulated in a way that enables adding more restriction to a given nodes traffic path and allow the network to provide some priority between nodes, if needed.

7.3 System model

The system considered in this work is a multi-hop wireless backhaul network, where there are several backhauling nodes that serve multiple users at a given geographical area. Not all backhauling nodes have direct connection to the core network. Accordingly, these distant nodes reach the core network through multi-hop links. Some of the backhauling nodes have hybrid terrestrial-satellite communication capabilities, which enable them to reach the core network through satellite. An example of the considered multi-hop network is depicted in Figure 7-1, which corresponds to the extended Helsinki topology.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Figure 7-1 Example of wireless backhauling network

The network is modelled as a directed graph 𝒢𝒢 = {𝒩𝒩, ℒ} where 𝒩𝒩 = {0,1, ⋯ , 𝑁𝑁} is the set of all backhauling nodes (vertices) 𝑛𝑛. The zero node represents the core network and ℒ is the set of all available links (edges). A link between two nodes exists if those that are in the transmission range d, i.e. ℒ = {𝑙𝑙, 𝑑𝑑𝑖𝑖(𝑓𝑓),𝑖𝑖(𝑓𝑓) ≤ 𝑇𝑇𝑖𝑖(𝑓𝑓), 𝑑𝑑(𝑙𝑙) ∈ 𝑁𝑁, 𝑟𝑟(𝑙𝑙) ∈ 𝑁𝑁, 𝑑𝑑(𝑙𝑙) ≠ 𝑟𝑟(𝑙𝑙)} where 𝑑𝑑𝑖𝑖(𝑓𝑓),𝑖𝑖(𝑓𝑓) is the distance between the transmitter on the 𝑙𝑙𝑖𝑖ℎ link, 𝑑𝑑(𝑙𝑙), and the receiver on the 𝑙𝑙𝑖𝑖ℎ link, 𝑟𝑟(𝑙𝑙) and 𝑇𝑇𝑖𝑖(𝑓𝑓) is the transmission range of the transmitting node 𝑑𝑑(𝑙𝑙). The incidence matrix of the graph 𝒢𝒢 is defined as follows:

𝑃𝑃(𝑛𝑛, 𝑙𝑙) = �1, if the node 𝑛𝑛 is transmitting on 𝑙𝑙,

−1, if the node 𝑛𝑛 is receiving on 𝑙𝑙, 0, otherwise

(47)

The backhauling nodes are assumed to have access to 𝐾𝐾 channels of bandwidth 𝑊𝑊 Hz each. Let ℱ = {𝑁𝑁1, 𝑁𝑁2, ⋯ , 𝑁𝑁𝐾𝐾} denote the set of available channels. We are interested in the task of upstreaming the backhauling nodes traffic to the core network. A similar approach can be followed for the downstreaming as well. The satellite system is assumed to share the available frequency band with the terrestrial system, under conditions that will be described in the next section.

The Multi-Commodity Flow (MCF) model is assumed here, considering that the queuing effect is negligible at each backhauling node and the aggregate traffic per node is constant over the


Date: 21/04/2017

of 87

SANSA-645047 D3.6

execution time. Let 𝑥𝑥𝑓𝑓,𝑙𝑙𝑓𝑓𝑘𝑘 represent the amount of flow assigned to the 𝑙𝑙𝑖𝑖ℎ link over the

frequency 𝑁𝑁𝑘𝑘 and corresponding to the commodity 𝑑𝑑. The generated traffic at each backhauling node is considered as a single commodity 𝑠𝑠𝑖𝑖. The flow conservation law should be satisfied so that it ensures that the sum of ingoing and outgoing flows belonging to commodity 𝑑𝑑 at each relaying backhauling node is equal; which can be expressed mathematically as

� � 𝑥𝑥𝑓𝑓,𝑙𝑙

𝑓𝑓𝑘𝑘 =𝑓𝑓𝑘𝑘∈ℱ𝑓𝑓∈𝑂𝑂𝑛𝑛

+

� � 𝑥𝑥𝑓𝑓,𝑙𝑙𝑓𝑓𝑘𝑘 ,

𝑓𝑓𝑘𝑘∈ℱ𝑓𝑓∈𝑂𝑂𝑛𝑛−

∀𝑛𝑛 ∈ 𝒩𝒩\{0, 𝑑𝑑}, 𝑑𝑑 ∈ 𝒩𝒩\{0}

(48)

where 𝑂𝑂𝑖𝑖+ is the set of ongoing links of node 𝑛𝑛, i.e. 𝑂𝑂𝑖𝑖

+ = {𝑙𝑙: 𝑃𝑃(𝑛𝑛, 𝑙𝑙) = 1} while 𝑂𝑂𝑖𝑖− is the set of

ingoing links to node 𝑛𝑛, i.e. 𝑂𝑂𝑖𝑖− = {𝑙𝑙: 𝑃𝑃(𝑛𝑛, 𝑙𝑙) = −1}. The nodes 𝑑𝑑 and 0 are excluded in (48) as

they are representing the source and destination nodes respectively. For each source backhauling node, the sum of the amount of the outgoing and ingoing traffic should be equal to the amount of the traffic generated at the node given in the following equation

� � 𝑥𝑥𝑓𝑓,𝑙𝑙𝑓𝑓𝑘𝑘 −

𝑓𝑓𝑘𝑘∈ℱ𝑓𝑓∈𝑂𝑂𝑛𝑛+

� � 𝑥𝑥𝑓𝑓,𝑙𝑙𝑓𝑓𝑘𝑘 =

𝑓𝑓𝑘𝑘∈ℱ𝑓𝑓∈𝑂𝑂𝑛𝑛−

𝑠𝑠𝑙𝑙 , ∀𝑑𝑑 ∈ 𝒩𝒩\{0} (49)

The rate that should be transmitted over any link should be below or equal the capacity of that link 𝐶𝐶𝑓𝑓

𝑓𝑓𝑘𝑘 , which can be expressed mathematically as

� 𝑥𝑥𝑓𝑓,𝑙𝑙

𝑓𝑓𝑘𝑘

𝑙𝑙

≤ 𝐶𝐶𝑓𝑓𝑓𝑓𝑘𝑘 , ∀𝑁𝑁𝑘𝑘 ∈ ℱ, 𝑙𝑙 ∈ ℒ

(50)

Each backhauling node is assumed to be half-duplex, which means that the node cannot receive and transmit simultaneously over the same channel. The backhauling nodes can communicate with multiple nodes simultaneously. However, it’s assumed that the backhauling node cannot transmit to different nodes over the same frequency, i.e. there is no broadcasting or multicasting. Additionally, the backhauling node cannot receive from multiple nodes over the same frequency. These constraints, in addition to the capacity constraint given in (50), can be expressed mathematically as

� �

𝑥𝑥𝑓𝑓,𝑙𝑙𝑓𝑓𝑘𝑘

𝐶𝐶𝑓𝑓𝑓𝑓𝑘𝑘

+ � �𝑥𝑥𝑓𝑓,𝑙𝑙

𝑓𝑓𝑘𝑘


≤ 𝛾𝛾𝑙𝑙𝑓𝑓∈𝑂𝑂𝑛𝑛

−𝑙𝑙𝑓𝑓∈𝑂𝑂𝑛𝑛+

, ∀𝑁𝑁𝑘𝑘 ∈ ℱ, ∀𝑛𝑛 ∈ 𝒩𝒩\{0}

(51)

where 𝛾𝛾 = 23

defines a sufficient condition for feasible schedule as discussed in [53][55][62]

while 𝛾𝛾 = 1 defines a necessary condition and 𝑥𝑥𝑓𝑓,𝑙𝑙𝑓𝑓𝑘𝑘 𝐶𝐶𝑓𝑓

𝑓𝑓𝑘𝑘� represents the fraction of time the channel 𝑁𝑁𝑘𝑘 is active on the 𝑙𝑙𝑖𝑖ℎ link.

The capacity of the channel is given by 𝐶𝐶𝑓𝑓𝑓𝑓𝑘𝑘 = 𝑊𝑊 log2(1 + 𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑓𝑓

𝑓𝑓𝑘𝑘), where 𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑓𝑓𝑓𝑓𝑘𝑘 is the

received signal-to-interference-plus-noise-ratio (SINR) on the 𝑙𝑙𝑖𝑖ℎ link over the frequency 𝑁𝑁𝑘𝑘 and can be expressed as follows


Date: 21/04/2017

of 87

SANSA-645047 D3.6

𝑆𝑆𝑃𝑃𝑁𝑁𝑆𝑆𝑓𝑓

𝑓𝑓𝑘𝑘 =𝑃𝑃𝑖𝑖(𝑓𝑓). 𝐺𝐺𝑖𝑖𝑚𝑚(0). ℎ𝑖𝑖(𝑓𝑓),𝑖𝑖(𝑓𝑓)

𝑓𝑓𝑘𝑘 . 𝐺𝐺𝑖𝑖𝑚𝑚(0)

𝑃𝑃𝑃𝑃𝑆𝑆𝑇𝑇𝑓𝑓𝑘𝑘 + 𝑁𝑁𝑖𝑖 + ∑ 𝑃𝑃𝑖𝑖(𝑓𝑓∗). 𝐺𝐺𝑖𝑖𝑚𝑚�𝜃𝜃𝑖𝑖(𝑓𝑓∗),𝑖𝑖(𝑓𝑓) �. ℎ𝑖𝑖(𝑓𝑓),𝑖𝑖(𝑓𝑓∗)

𝑓𝑓𝑘𝑘 . 𝐺𝐺𝑖𝑖𝑚𝑚�𝜃𝜃𝑖𝑖(𝑓𝑓),𝑖𝑖(𝑓𝑓∗) �𝑓𝑓∗∈ℒ𝑓𝑓𝑘𝑘𝑓𝑓∗≠𝑓𝑓

(52)

where,

• 𝑃𝑃𝑇𝑇𝑥𝑥 : Transmit power of the 𝑇𝑇𝑚𝑚 backhauling node. • 𝐺𝐺𝑖𝑖𝑚𝑚�𝜃𝜃𝑇𝑇𝑥𝑥,𝑇𝑇𝑥𝑥� and 𝐺𝐺𝑖𝑖𝑚𝑚(𝜃𝜃𝑇𝑇𝑥𝑥,𝑇𝑇𝑥𝑥): Gain of the transmitting and receiving antenna at offset

angle 𝜃𝜃𝑇𝑇𝑥𝑥,𝑇𝑇𝑥𝑥. • 𝜃𝜃𝑇𝑇𝑥𝑥,𝑇𝑇𝑥𝑥: Boresight direction offset angle of the 𝑇𝑇𝑚𝑚 backhauling node transmit antenna in

the direction of 𝑆𝑆𝑚𝑚 receiver antenna. • ℎ𝑇𝑇𝑥𝑥,𝑇𝑇𝑥𝑥

𝑓𝑓𝑘𝑘 : Channel attenuation considering free space path-loss between the backhauling node 𝑇𝑇𝑚𝑚 and backhauling node 𝑆𝑆𝑚𝑚.

• ℒ𝑓𝑓𝑘𝑘: Set of links transmitting over the frequency 𝑁𝑁𝑘𝑘.

• 𝑃𝑃𝑃𝑃𝑆𝑆𝑇𝑇𝑓𝑓𝑘𝑘 : The interference introduced by the satellite links transmitting over the frequency

𝑁𝑁𝑘𝑘. • 𝑁𝑁𝑖𝑖: Noise thermal noise.

In this work, the satellite system is assumed to be designed properly so that the interference introduced to the neighboring terrestrial nodes is small. The satellite is also assumed to be connected to the core network.

7.4 Problem formulation and solution approach

The main objective of this work is to maximize the net incoming traffic to the core network, while satisfying the net flow constraints, the link scheduling constraints and the link capacity constraints. The considered scenario is the upstreaming one, but the down-streaming can be solved in a similar way, by exchanging the core and the backhauling nodes role to be the source and destinations, respectively. The problem can be mathematically formulated as follows:


Date: 21/04/2017

of 87

SANSA-645047 D3.6

where 𝛼𝛼 is the ratio of the generated traffic at each backhauling node that can be delivered to the core. Δ is a positive constant that is multiplied by the sum of flows of the commodities in the set 𝐷𝐷 that are transmitted over the links in the set 𝒮𝒮. The set 𝒮𝒮 can be the satellite system links while 𝐷𝐷 can contain the set of nodes with delay sensitive traffic that should not be routed through the satellite links to avoid having excessive delay, due to the satellite links’ long propagation delay. Δ can be used to activate and deactivate the second part of the objective function as well as calibrating the ratio of the flow that can be transmitted over the links in set 𝒮𝒮. 𝐶𝐶1 and 𝐶𝐶2 satisfies the flow conservation law while 𝐶𝐶3 represents the capacity and the link scheduling constraints. 𝐶𝐶4 ensures that only one frequency is used on the link at a given time. This constraint can be removed in cases where multi-frequency transmission is allowed. Constraints 𝐶𝐶5 and 𝐶𝐶6 ensures positive values of the amount of flow and the delivered traffic ratio.

The objective function as well as the constraints (𝐶𝐶1, 𝐶𝐶2, 𝐶𝐶5, 𝐶𝐶6) are linear. Constraints (𝐶𝐶3, 𝐶𝐶4) are nonlinear constraints, which makes the problem non-linear with non-convex constraints. In order to solve the problem, one can try to find an estimate value of the link capacities in order to linearize the constraints (𝐶𝐶3, 𝐶𝐶4).

(53)


Date: 21/04/2017

of 87

SANSA-645047 D3.6

The capacity of the link depends on the transmit power of the link and the interference received from other nodes transmitting over the same frequency. If the transmit power for all the backhauling nodes is fixed to a pre-defined value, the interference introduced to the link can be evaluated, if the links that share the same frequency are identified. To go forward with the interference calculation, we start by evaluating the interference introduced by a given link to rest of the links if they are assumed to share the same frequency. Accordingly, the matrix 𝑃𝑃𝑙𝑙𝑖𝑖𝑖𝑖 is constructed where 𝑰𝑰𝑙𝑙𝑖𝑖𝑖𝑖(𝑙𝑙∗, 𝑙𝑙) element represents the received interference on the 𝑙𝑙∗𝑖𝑖ℎ link due to the transmission occurring on the 𝑙𝑙𝑖𝑖ℎ link, i.e. matrix 𝑰𝑰𝑙𝑙𝑖𝑖𝑖𝑖 elements are the amount of interference introduced to the row-link when the column link is transmitting on the same frequency. The matrix 𝑰𝑰𝑙𝑙𝑖𝑖𝑖𝑖 can be written as

𝑰𝑰𝑙𝑙𝑖𝑖𝑖𝑖 = �0 ⋯ 𝑃𝑃𝑙𝑙𝑖𝑖𝑖𝑖(1, 𝐿𝐿)⋮ ⋱ ⋮

𝑃𝑃𝑙𝑙𝑖𝑖𝑖𝑖(𝐿𝐿, 1) ⋯ 0�

(54)

Afterwards, the links that generate high interference with each other are considered as conflict links. To this end, an additional constraint is added to the original formulation to count for this conflict, which enforces the system not to assign the same frequency to these links. To do so, the conflict matrix 𝑳𝑳𝑙𝑙𝑙𝑙𝑖𝑖𝑓𝑓 should be constructed where the element 𝑳𝑳𝑙𝑙𝑙𝑙𝑖𝑖𝑓𝑓(𝑙𝑙∗, 𝑙𝑙) is equal to one when the links 𝑙𝑙∗ and 𝑙𝑙 are in conflict and should not share the same frequency. If the ingoing and outgoing links are added to the 𝑳𝑳𝑙𝑙𝑙𝑙𝑖𝑖𝑓𝑓(𝑙𝑙∗, 𝑙𝑙) matrix as a conflict links as well, the constraint 𝐶𝐶3 can be replaced by the following constraint, which counts for both the link scheduling constraints as well as the links conflict constraints as indicated in equation (51)

� �𝑥𝑥𝑓𝑓,𝑙𝑙

𝑓𝑓𝑘𝑘


𝑳𝑳𝑙𝑙𝑙𝑙𝑖𝑖𝑓𝑓(𝑙𝑙∗, 𝑙𝑙) ≤ 𝛾𝛾𝑙𝑙𝑓𝑓∈ ℒ

, ∀𝑙𝑙∗ ∈ ℒ, ∀𝑁𝑁𝑘𝑘 ∈ ℱ

(55)

where

𝑳𝑳𝑙𝑙𝑙𝑙𝑖𝑖𝑓𝑓 = �1 ⋯ 𝐿𝐿𝑙𝑙𝑙𝑙𝑖𝑖𝑓𝑓(1, 𝐿𝐿)⋮ ⋱ ⋮

𝐿𝐿𝑙𝑙𝑙𝑙𝑖𝑖𝑓𝑓(𝐿𝐿, 1) ⋯ 1�

(56)

By assuming the worst-case scenario where all the channels with no conflict to a given link are using the same frequency, the total interference introduced to a given link can be evaluated. Accordingly, the capacity for each link can be calculated and nonlinear constraints can be converted into linear ones, which reformulates the problem into a convex one, which can be solved more easily. Once the different fraction of time variables is determined, the algorithm presented in [53] can be used to extract the link scheduling.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

7.5 Performance evaluation and comparison with Helsinki topology benchmark

To show the performance of the proposed scheme, the extended Helsinki topology depicted in Figure 6-4 is used. The detailed description of this topology is already presented in section 6.3.2. We assume that the node 8 and satellite are connected to the core. The satellite node is forwarding the incoming traffic and has no generated traffic. Moreover, the interference from the satellite system to the terrestrial one is assumed to be negligible as soon as the outgoing and ingoing links to the hybrid backhauling node do not transmit over the same frequency used by the satellite segment. The capacities of the satellite links are assumed to be the average value of the terrestrial links capacities. For the proposed scheme, the worst case SINR is considered by assuming that all the links which have no conflict with a given link are sharing the same frequency and inducing interference to it. The link between the nodes and the core is assumed to have sufficient capacity to deliver the data arrived at node 8 and the satellite. This is assumed to avoid the case that the performance of network under study is limited by the limitation of the connection to the core but not by the characteristics of network links. The traffic per node is assumed to be 420 Mbps for the backhauling nodes. The proposed technique can have a solution for any traffic value. However, this value is chosen so that compact figures with 𝛼𝛼 ∈ [0,2] are obtained. A summary of the terrestrial system parameters can be found in Table 7-1.

Table 7-1 Terrestrial system Parameters Parameter Value

Antenna Pattern As given in [38] Max. antenna gain 38 dBi

Channel Model Free Space Path Loss Noise Density -139 dBW/MHz

Transmit Power Density -38.13 dBW/MHz Total bandwidth 17.7-19.7 GHz

Rate per backhauling node 420 Mbps

7.5.1 Extended Benchmark (without satellite links) The carrier assignment described in Table 6-4 and Figure 6-6 of section 6.3.3 is also used here for the benchmark scenario. Accordingly, based on the applied system parameters, the SINR values and the rates achieved by each link are detailed in Table 7-2. The values are a bit different from those presented in the previous section as free space path loss model is considered with no diffraction. In the following simulation results, the satellite links is enabled and disabled in the benchmark by the same way that is done for the proposed scheme.

Table 7-2 Benchmark links SINR and rates values.

Link Number

Initial Node

End Node



(Mbps)

1 1 2 6 59.34 1103.99


Date: 21/04/2017

of 87

SANSA-645047 D3.6

2 2 1 6 59.40 1105.09 3 1 3 4 52.50 976.75 4 3 1 4 52.51 976.93 5 1 4 4 49.05 912.44 6 4 1 4 49.07 912.77 7 1 6 5 54.23 1008.79 8 6 1 5 55.84 1038.71 9 1 8 3 52.72 980.82

10 8 1 3 52.76 981.56 11 2 5 5 59.27 1102.72 12 5 2 5 59.28 1102.77 13 7 3 4 55.38 1030.26 14 3 7 4 55.34 1029.57 15 9 3 3 49.42 919.39 16 3 9 3 49.42 919.43 17 5 6 3 52.30 972.98 18 6 5 3 52.32 973.46 19 5 8 5 57.29 1065.86 20 8 5 5 57.27 1065.45 21 7 12 6 52.18 970.74 22 12 7 6 52.23 971.76 23 7 13 6 56.30 1047.36 24 13 7 6 56.24 1046.28 25 9 8 3 62.20 1157.27 26 8 9 3 62.20 1157.22 27 10 8 6 55.71 1036.46 28 8 10 6 55.68 1035.87 29 14 8 6 62.04 1154.21 30 8 14 6 62.04 1154.14 31 15 8 6 59.95 1115.31 32 8 15 6 59.96 1115.41 33 9 10 6 56.41 1049.37 34 10 9 6 50.25 934.87 35 9 11 4 51.20 952.46 36 11 9 4 51.21 952.62 37 9 12 5 46.63 867.39 38 12 9 5 48.10 894.75 39 11 10 4 59.61 1108.82 40 10 11 4 59.61 1108.95 41 13 10 5 57.66 1072.70 42 10 13 5 57.68 1073.00 43 13 15 5 65.41 1216.89 44 15 13 5 65.46 1217.70


Date: 21/04/2017

of 87

SANSA-645047 D3.6

7.5.2 Performance evaluation considering 56 MHz channels In this part of the simulation, we fix the channel BW to 56MHz as applied in the benchmark scheme. Based on the data rate values given in Table 7-2, the maximum net flow that can be delivered to the core considering the frequency assignment used in the benchmark scheme is evaluated. This result is compared with the proposed scheme considering different number of channels (similar to the results presented in section 6). In the simulation results, we consider the cases where the satellite link are/(are not) activated as given in the sequel subsections.

7.5.2.1 Active Satellite links

In this part of the simulation, the satellite links are active and can deliver traffic to the core network. This is mathematically achieved by considering Δ = 0.

By solving the net flow maximization problem for the benchmark scheme, the delivered data rate is equal to 4888.10 Mbps. Considering that the benchmark scheme uses 8 channels, each with 56 MHz, the total used bandwidth is 448 MHz. Accordingly, the benchmark spectral efficiency is SE = 4888.10 /448 = 10.91 bps/Hz. For the proposed scheme, the delivered rate and the SE as function of the number of the used channel is summarized in Table 7-3.

Table 7-3 Delivered rate and SE vs. No of 56 MHz channels No. of 56

MHz Channels

Delivered data rate (Mbps)

SE (bps/Hz)

1 1508.57 26.94 2 3017.14 26.94 3 4524.54 26.93 4 5309.61 23.70 5 6074.37 21.69 6 6776.30 20.16 7 7477.50 19.07 8 8179.39 18.25

It can be noticed that the proposed scheme can achieve the delivered rate by the benchmark network by using only four frequency bands. In particular, the SE gain that can be achieved corresponds to 2.47x compared to the benchmark. Additionally, considering the same bandwidth for both system, i.e. 8 channels, the SE is approximately 1.6x improved.

Figure 7-2 depicts the ratio of the delivered traffic per node α against the number of used 56 MHz channels. The delivered rate increases as the number of channel increases due to the additional transmission bands and reduced interference. After 9 channels, no increment is achieved by increasing the number of channels as the system is able to manage the link scheduling problem in the network. Recall that every link can only use one frequency and the value of the capacities for a given link over the different frequencies is equal assuming that a similar worst case SINR scheme is applied in all the frequencies.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Figure 7-2 The ratio of the delivered traffic per node 𝛼𝛼 against the number of used 56 MHz channels with enabled satellite links.

7.5.2.2 Disabled Satellite links

The satellite links now are not active and the traffic can be delivered to the core network only by node 8. This is mathematically achieved by considering a high value of Δ.

By solving the net flow maximization problem for the benchmark scheme, the delivered data rate is equal to 2772 Mbps. Considering that the benchmark scheme uses 8 channels, each with 56 MHz, the total used bandwidth is 448 MHz. Accordingly, the benchmark SE 2772/448 = 6.1875 bps/Hz. For the proposed scheme, the delivered rate and the SE as function of the number of the used channel is summarized in Table 7-4.

Table 7-4 Delivered rate and SE vs. No of 56 MHz channels No. of 56

MHz Channels

Delivered data rate (Mbps)

SE (bps/Hz)

1 809.66 14.46 2 1619.31 14.46 3 2428.95 14.30 4 3202.65 14.15 5 3961.06 13.90 6 4672.08 13.73

1 2 3 4 5 6 7 8 9 10 11 12

No. of 56MHz Channels

0.2

0.4

0.6

0.8

1

1.2

1.4

Rat

io o

f the

del

iver

ed tr

affic

per

nod

e (

)

Proposed

Benchmark (8 channels)


Date: 21/04/2017

of 87

SANSA-645047 D3.6

7 5383.10 13.51 8 6054.08 12.04

The proposed scheme can achieve the delivered rate by the benchmark network by using only four frequency bands, and the SE gain goes up to 2.34x when considering 2 channels only. Additionally, considering the same B.W for both system, i.e. 8 channels, the SE is approximately 2x improved.

Figure 7-3 depicts the ratio of the delivered traffic per node 𝛼𝛼 against the number of used 56 MHz channels. It is observed that the link scheduling in the network with 8 channels and no rate increment can be achieved by using higher values. It is one channel less in this case as the constraints of the problem is reduced by removing the satellite links. Additionally, it should be noted that –as expected-disabling the satellite links will reduce the total amount of the delivered traffic as the number of links connecting the network to the core is reduced.

Figure 7-3 The ratio of the delivered traffic per node 𝛼𝛼 against the number of used 56 MHz channels with disabled satellite links.

7.5.3 Performance evaluation considering variable channel width

For this case, we assume that the satellite links are active. Additionally, we assume that the channel bandwidth is varying with the number of channels considered in the system. However, all the links are using channels of a fixed bandwidth. The 8*56MHz = 448 MHz bandwidth is

1 2 3 4 5 6 7 8 9 10 11 12

No. of 56MHz Channels

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Rat

io o

f the

del

iver

ed tr

affic

per

nod

e (

)

Proposed



Date: 21/04/2017

of 87

SANSA-645047 D3.6

considered as the total bandwidth and the per channel bandwidth is 448 / (No. of channels). The delivered data rates as well as the SE summarized in Table 7-5. The SE versus the number of channels is depicted in Figure 7-4.

Table 7-5 Delivered rate and SE vs. No of channels No. of

Channels Delivered data rate (Mbps)

SE (bps/Hz)

2 11933.82 26.64 4 10618.46 23.70 8 8179.39 18.25

16 4180.05 16.59 32 2081.92 16.52

Figure 7-4 The SE of the system against the number of channels.

It can be remarked that the SE decreases with the increase in the number of channels as the frequency is reused less. As the number of channels becomes more than 9, the system is using only 9 channels as described in Figure 7-2 and hence, the SE of the system keeps constant. The date rate of the lower number of channels is higher than that of higher number of channels due to the use of more transmission power, which increases the capacity of the links if the links are scheduled adequately. This is observed from Figure 7-5 where the EE of the terrestrial links of the system is depicted against the number of the channels. We should emphasize that the EE increases with the increment of the number of the channels. It should also be noted that the

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 32

No. of Subchannels

10

12

14

16

18

20

22

24

26

28

Spe

ctru

m e

ffici

ency

(bps

/Hz)


Proposed


Date: 21/04/2017

of 87

SANSA-645047 D3.6

benchmark scheme has EE =12.3365 Mbps /mW while the proposed system with the same channels and B.W has EE= 20.6416 Mbps/mW.

Figure 7-5 EE of the terrestrial links in system against the number of channels.

8 QoS-aware satellite scheduling

8.1 Motivation

There can be specific cases where the amount of traffic exceeds the terrestrial provision or when there has been a problem in the terrestrial infrastructure that prevents the traffic to be redirected via the multi-hop terrestrial links. In this section, we investigate the provision of mobile backhaul in these situations, where the mobile backhaul traffic is redirected to the satellite. However, the satellite connection impose severe delay constraints which should be properly considered in the satellite packet scheduling decisions. This is particularly important for the evolution of the traffic nature, which encompasses multiple services and service of different types. This has created the need for QoS compliance more important than ever.

In the rest of this section we present the system architecture in more detail and define the scheduling problem we are solving. Then, we present our scheduling priorities and derive the scheduling algorithm. In subsection 8.4 we describe our simulation parameters, comparing them to the DVB-S2 standards, and present our results in Section 8.5. Finally, we draw our conclusions.

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 32

No . of Channels

0

5

10

15

20

25

Terre

stria

l Ene

rgy

Effi

cien

cy (M

bps/

mW

)

Proposed



Date: 21/04/2017

of 87

SANSA-645047 D3.6

8.2 Description of the satellite communications system and problem definition

8.2.1 Satellite system description

We assume a Satellite communication system using a Geostationary Earth Orbit (GEO) satellite. The Satellite system follows the DVB-S2 standards and is illustrated in Figure 8-1. A satellite acts as a transponder, connecting the satellite terminals of the SANSA network to a Core Network (CN) via one or more satellite gateways. The broadcast channel, or the forward link is defined as the communications link between the gateway and the satellite terminals. The Gateway sends through the Satellite station a carrier per beam, which can be received by multiple Satellite terminals. Therefore, a certain division in time, power, or bandwidth allocation is necessary. The resource allocation function, including the definition the ACM combination and the packet scheduling, is carried out at the Gateway station. Finally, the logical link on the opposite direction is called the Return Link. As our work in this section accounts only for the forward link, we have omitted the return link from Figure 8-1.

Figure 8-1: Satellite access network

For the sake of simplicity, we assume the probability of PHY layer errors is negligible and assume transparent data transmission at the transport layer. This means that we assume that the chosen MODCOD combination is sufficiently good to eliminate the probability of unrecoverable errors. In addition, there is no automatic retransmission request (ARQ) mechanism to request packet retransmissions. Another important assumption made is that the downlink carrier bandwidth is considered constant, and therefore the symbol rate is constant independently of the transmitted coding rate and modulation format. We further assume that the Gateway is able to keep track of channel variations for each Satellite Terminal, in real time. This assumption can be justified by the fact that satellite channels vary slowly due to the relatively slow variation of

Satellite

CN

Scheduling and ACM functions

Satellite Gateway

Satellite Terminals (iBN)


Date: 21/04/2017

of 87

SANSA-645047 D3.6

atmospheric conditions. This assumption further means that even the long propagation delay of GEO satellite communication links is short compared to the coherence time of the channel.

The total received signal power to noise plus interference ratio is a stochastic process and depends on both the channel between the gateway to the satellite (the feeder link) and the channel between the satellite and the satellite terminals (the user link). However, the feeder link is typically designed for a low outage probability and high SINR thanks to the use of large parabolic antennas and the use of space diversity (several gateways can be used) [63]. We can therefore assume that the overall carrier to noise ratio (C/N) depends only on the user link.

8.2.2 Adaptive Coding and Modulation

The possible ACM combinations defined for DVB-S2 are considered, according to [19]. Figure 8-2 shows the various Modulation and Coding (MODCOD) combinations defined in [19]. The table in Figure 8-2(a) lists the MODCOD combinations along with the theoretical spectral efficiency and the ideal Ec/No when each combination should be used. The same information is plotted in Figure 8-2(b) along with the spectral efficiency achieved by previous satellite communication standards and the modulation-constrained Shannon limit.

(a) (b) Figure 8-2: MODCOD combinations supported by the DVB-S2 standards along with the C/N conditions

they can be applied in, and the spectral efficiency they achieve [19].

8.2.3 Generic Stream Encapsulation in DVB-S2

One of the features introduced in DVB-S2 to support dynamic resource allocation and to offer multiple services over the same carrier and BBFs is the Generic Stream Encapsulation [21]. The GSE is a layer introduced between the Network and Transport layers, responsible to encapsulate several packets on the same BBF. The GSE operation is illustrated in Figure 8-3. The GSE layer


Date: 21/04/2017

of 87

SANSA-645047 D3.6

supports both concatenation of many packets on a single BBF but also the splitting of a single higher layer packet (protocol data unit – PDU) in more than one BBFs. GSE therefore allows the more efficient packing of BBFs with user data, reducing satellite link overheads.

Figure 8-3: Illustration of the functionality of the Generic Stream Encapsulation mechanism [21].

8.3 Scheduling of SANSA backhaul packets

The vision of SANSA is to facilitate the provision of mobile backhaul in situations where the network is congested, or where there is terrestrial backhaul link failure. In such conditions it is important to be able to prioritise user traffic according to defined QoS metrics. Traditionally mobile networks have relied on Best Effort data delivery, which means that all packets have the same priority over the access channels. However, the evolution of multiple services and service types has made the need for QoS compliance more important than ever. This becomes evident in the evolution of mobile standards. Whereas UMTS defined four different QoS classes (Conversational, Streaming, Interactive, and Background), LTE has defined nine different QoS classes each with different priority levels and packet delay constraints. The SANSA system has, therefore, to be able to adapt to the QoS settings of different user packets to guarantee the highest end user satisfaction.

8.3.1 Scheduling algorithm for the Satellite forward link

When designing our algorithm, we are taking into account the following requirements:

i. All base stations should be served, even those that witness bad weather conditions affecting the received C/N power;

ii. The QoS requirements of individual user data packets have to be considered in the


Date: 21/04/2017

of 87

SANSA-645047 D3.6

scheduling algorithm;

iii. Packets scheduled in the same BBF have to experience the minimum level of spectral inefficiency, i.e. packets for Base Stations with similar channel rates should be grouped together to avoid underusing the satellite’s channel capacity; and,

iv. Exhaustive search should be avoided as it is computationally heavy.

To satisfy the above requirements we propose the following scheduling algorithm:

Scheduling algorithm • For the first packet scheduled in a BBF, select user data packet 𝑢𝑢 at BBF 𝑛𝑛 s.t.

𝑢𝑢𝑓𝑓 = argmax𝑘𝑘=1…𝐾𝐾𝑖𝑖𝑘𝑘,𝑛𝑛⋅𝑝𝑝𝑘𝑘

𝑇𝑇𝑘𝑘,𝑛𝑛−1(𝑇𝑇𝑇𝑇𝐷𝐷𝑘𝑘𝑚𝑚𝑖𝑖𝑛𝑛)𝑚𝑚

,

• For any subsequent packet scheduled in the same BBF, select user data packet 𝑢𝑢 at BBF 𝑛𝑛 s.t.

𝑢𝑢𝑒𝑒 = argmax𝑘𝑘=1…𝐾𝐾�𝑖𝑖𝑘𝑘,𝑛𝑛−�̂�𝑖𝑘𝑘,𝑛𝑛�⋅𝑝𝑝𝑘𝑘

𝑇𝑇𝑘𝑘,𝑛𝑛−1(𝑇𝑇𝑇𝑇𝐷𝐷𝑘𝑘𝑚𝑚𝑖𝑖𝑛𝑛)𝑚𝑚

, where

𝑟𝑟𝑘𝑘,𝑖𝑖 is the achievable rate for user 𝑘𝑘 in slot 𝑛𝑛,

�̂�𝑟𝑘𝑘,𝑖𝑖 is the achievable rate of the first user scheduled in the BBF,

𝑏𝑏𝑘𝑘 is a priority weight so that packets with higher 𝑏𝑏𝑘𝑘 have higher priority over those with lower 𝑏𝑏𝑘𝑘,

𝑇𝑇𝑇𝑇𝐿𝐿 is the remaining time that a packet can remain in the GS’s queue,

𝑚𝑚 is a parameter that changes how important the delay constraint is for the scheduling,

𝑆𝑆𝑘𝑘,𝑖𝑖 = ��1 − 1

𝑤𝑤� 𝑆𝑆𝑘𝑘,𝑖𝑖−1 + 𝑖𝑖𝑘𝑘,𝑛𝑛

𝑤𝑤, if k is scheduled at n

�1 − 1𝑤𝑤

� 𝑆𝑆𝑘𝑘,𝑖𝑖−1, otherwise, and

𝑒𝑒 is the size of a rolling window during which R is considered.

The proposed algorithm reduces to a type of Proportional Fair algorithm when all the packets have the same priority (𝑏𝑏𝑘𝑘 = 1, ∀𝑘𝑘 ∈ [1, 𝐾𝐾]) and the delay constraint parameter 𝑚𝑚 = 0.

8.4 Simulation setup

In this subsection we describe the assumptions made in our simulator.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

8.4.1 ACM implementation

While the standards define in total 28 MODCOD combinations we observe that there is a degree of overlapping between them. For example, it is spectrally more efficient to use 16APSK with 2/3 coding rate than 8PSK with 5/6 coding rate. Therefore, in our link level simulator we use a reduced set of MODCOD combinations so that the most efficient MODCOD between overlapping combinations is always chosen. The reduced set is shown in Figure 8-4: MODCOD combinations considered in our simulations

Figure 8-4: MODCOD combinations considered in our simulations

Table 8-1: List of reduced MODCOD subset used in our simulations.

Mode Spectral efficiency

Ideal Es/No(dB) for FECFRAME length = 64800

Normalisation multiplier

QPSK 1/4 0.490243 -2.35 0.17 QPSK 1/3 0.656448 -1.24 0.22 QPSK 2/5 0.789412 -0.3 0.26 QPSK 1/2 0.988858 1 0.33 QPSK 3/5 1.188304 2.23 0.4 QPSK 2/3 1.322253 3.1 0.44 QPSK 3/4 1.487473 4.03 0.5 QPSK 4/5 1.587196 4.68 0.53 QPSK 5/6 1.654663 5.18 0.55 8PSK 3/5 1.779991 5.5 0.6 8PSK 2/3 1.980636 6.62 0.66 8PSK 3/4 2.228124 7.91 0.75

16APSK 2/3 2.637201 8.97 0.88 16APSK 3/4 2.966728 10.21 1

-2 0 2 4 6 8 10 12 14 16

C/N [dB] in R s

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

Sim

ulat

ed R

u[b

its/s

] per

uni

t sym

bol r

ate

Rs

Applicable Modulation and Coding combination

Modulation and Coding combination according to DVB-S2


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Mode Spectral efficiency

Ideal Es/No(dB) for FECFRAME length = 64800

Normalisation multiplier

16APSK 4/5 3.165623 11.03 1.07 16APSK 5/6 3.300184 11.61 1.11 32APSK 3/4 3.703295 12.73 1.25 32APSK 4/5 3.951571 13.64 1.32 32APSK 5/6 4.11954 14.28 1.38 32APSK 8/9 4.397854 15.69 1.48

32APSK 9/10 4.453027 16.05 1.5

8.4.2 Channel variation implementation As discussed in section 8.2, the received C/N power is affected by weather conditions. These are further affected by the location of the receiver base stations and the local climate. Clear skies offer the lower signal attenuation (higher C/N), while heavy rain leads to higher attenuation (lower C/N). Both the clear skies and the heavy rain conditions are extreme and not very common. Instead, there is usually some attenuation due to clouds or the presence of dust in the atmosphere. Furthermore, while under clear skies the power attenuation due to weather reduces to 0, the same does not hold for heavy rain where there is a long tail at the distribution of the attenuation.

The exact power attenuation distribution in an area can be obtained through statistical measurements over a long period of time covering all the seasons in a year. Such measurement falls beyond the scope of this work. For the purpose of our simulations, though, we have chosen to simulate the channel conditions by modelling the random variable that defines the selected MODCOD combination. The chosen distribution (based on a Poisson random variable) is shown in Figure 8-5. We have set the parameters so that the expected MODCOD combination is the 16APSK with 3/4 coding rate. While we chose this distribution for its realistic characteristics, the choice of any other distribution will not affect the results in a qualitative way.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Figure 8-5: Probability density function for the probability of a given MODCOD to be used.

8.4.3 Generic Stream Encapsulation implementation We have assumed that all BBFs have the same length in symbols (or, equivalently, time), 𝑇𝑇𝐵𝐵𝐵𝐵𝑍𝑍. However, the number of user data packets that fit in a BBF depends on the MODCOD combination which affects the spectral efficiency of the transmitted symbols. Since we also assume that all user data packets have fixed length in bits, ℓ𝑓𝑓, we calculate the size of a user data packet in symbols, ℓ�𝑓𝑓, using the “Normalisation Multiplier” field of Table 8-1. This field effectively normalizes the gain (or loss) in terms of spectral efficiency against the spectral efficiency of the expected MODCOD combination. Furthermore, we assume that, on average, three user packets can fit in a BBF, i.e.:

𝑇𝑇𝐵𝐵𝐵𝐵𝑍𝑍 = 3 × ℓ�𝑓𝑓, when MODCOD = 16APSK 3/4

The scheduler keeps filling up a BBF as long as there is more than 10% of 𝑇𝑇𝐵𝐵𝐵𝐵𝑍𝑍 free. Every time a new user packet is added, the total length (in symbols) of the user packets is re-evaluated according to the lower supported MODCOD combination. Finally, if the addition of a user packet is exceeding the length of a BBF, the last packet is split with the second part being scheduled for the next BBF.

8.4.4 QoS parameters of user packets

For our simulations we assume four packet classes with different delay constraints (𝑇𝑇𝑇𝑇𝐿𝐿𝑙𝑙) and different priority level 𝑏𝑏𝑙𝑙. Packets from the four classes are generated randomly for each base station according to the distribution shown in Table 8-2. The 𝑇𝑇𝑇𝑇𝐿𝐿𝑙𝑙 is normalised in 𝑇𝑇𝐵𝐵𝐵𝐵𝑍𝑍.

Table 8-2: Probability of arrival and QoS requirements for the packet classes considered

C/N [dB]

-5 0 5 10 15 20

Pro

babi

lity

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Arrival probability 𝑻𝑻𝑻𝑻𝑳𝑳𝒊𝒊 𝒑𝒑𝒊𝒊 Class 1 10% 3 3 Class 2 20% 5 2 Class 3 30% 7 1 Class 4 40% 10 1

8.5 Simulation results

We performed Monte Carlo simulations to investigate the performance of our proposed scheduling algorithm. In particular, we compare the performance of four different scheduling approaches; (i) a maximum rate scheduler that always schedules the packets that can be transmitted at the highest rate; (ii) the proportional fair (PF) scheduling algorithm; (iii) the PF algorithm with the delay constraint; and, (iv) the PF algorithm with the delay constraint and the priority index. The four scheduling algorithms are compared in terms of the way they share the resources between different users, the probability of dropping packets from different classes, and the overall throughput achieved with each scheduling algorithm.

8.5.1 Proportional fairness

The first objective of our scheduling algorithm is to avoid penalizing base stations that experience bad channel conditions. Figure 8-6 shows the distribution of resources between the 20 base stations considered in the simulations. These simulations assume completely saturated conditions – all users have packets to transmit at all instances. We further assume static channels for the duration of the simulation run and across the different scheduling algorithms. The maximum rate scheduler (Figure 8-6(a)) allocates all the resources to the base station that has the highest rate, in this case base station 15. Instead, the other three algorithms tested share the resources more evenly across the base stations, reducing the resources allocated to base station 15 to just over 10% of the total available resources. Furthermore, we can see that the distribution of resources does not significantly change when we apply the Delay constraint and the priority index in the scheduling algorithm.

User

0 5 10 15 20

Sha

re o

f res

ourc

es

0

0.2

0.4

0.6

0.8

1

User

0 5 10 15 20

Sha

re o

f res

ourc

es

0

0.02

0.04

0.06

0.08

0.1


Date: 21/04/2017

of 87

SANSA-645047 D3.6

(a) Maximum rate scheduling (b) Proportional fairness

(c) PF + Delay Constraint (d) PF + Delay Constraint + Priority

Figure 8-6: Performance of the compared scheduling algorithms in terms of sharing the available resources

8.5.2 Probability to transmit packets from each class

This is the most important evaluation since it relates to the driver behind our analysis and proposed algorithm. Figure 8-7 compares the four scheduling algorithms for various levels of offered traffic. The maximum rate and the proportional fairness scheduling algorithms have very similar performance. In both cases Class 1 packets are less likely to be served. This is because Class 1 packets have shorter delay constraints and are therefore more likely to be dropped by the GS. However, when the Delay constraint is added in our scheduling function we notice that all the packets have the same probability of being served. This means that the delay constraint is satisfied to the extent that the offered traffic does not overhaul the channel’s capacity. Finally, adding the priority index in our scheduling function allows us to significantly increase the probability that Class 1 and Class 2 packets are served. This shows that our proposed scheduling algorithm has the ability to satisfy QoS requirements for different services, even when at high offered traffic conditions.

User0 5 10 15 20

Sha

re o

f res

ourc

es

0

0.02

0.04

0.06

0.08

0.1

User

0 5 10 15 20

Sha

re o

f res

ourc

es

0

0.02

0.04

0.06

0.08

0.1


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Figure 8-7: Probability to serve packets from the different classes

8.5.3 Effect of the scheduling algorithms on the total throughput

Finally, Figure 8-8 plots the total throughput, expressed as user packets / BBF, achieved with the considered scheduling algorithms. Overall, the four scheduling algorithms perform very similarly. Figure 8-9 zooms in two regions of interest from Figure 8-8, the region where traffic saturation starts occurring (Figure 8-9(a)), and the region where the offered traffic is saturated (Figure 8-9(b)).

Figure 8-9(a) shows that in the case of the two QoS-aware scheduling algorithms the throughput scales linearly for higher offered traffic than in the maximum rate and the proportional fair scheduling algorithms. This is because of the QoS-aware scheduling algorithm present in the former two algorithms. The QoS-aware algorithm prioritises packets that are close to expiring and delays those that can be scheduled at a later slot. As the offered traffic has not yet reached saturation levels, there is high probability that the delayed packets will still be able to be scheduled before their TTL expires. This effectively means that the QoS-aware scheduling algorithm increases the range of offered traffic for which packet delivery can be guaranteed.

Figure 8-9(b), on the other hand, shows that at saturated conditions the overall throughput achieved by the QoS-aware scheduling algorithms is lower than the throughput achieved by the maximum rate algorithm. This is because the PF and the QoS aware algorithms sacrifice higher rate packets to achieve the levels of fairness, and service delivery that they are designed to achieve. In a way this loss of capacity is the price to pay for the superior management of the QoS requirements.

Offered traffic

0 1 2 3 4

Tran

smitt

ed p

acke

ts (r

atio

)

0.6

0.7

0.8

0.9

1Maximum rate

Class 1

Class 2

Class 4

Class 3

0 1 2 3 40.5

0.6

0.7

0.8

0.9

1Proportional Fairness

Class 1

Class 2

Class 4

Class 3

Offered traffic

0 1 2 3 4

Tran

smitt

ed p

acke

ts (r

atio

)

0.6

0.7

0.8

0.9

1PF + Delay Constraint

Class 1

Class 2

Class 4

Class 3

Offered traffic

0 1 2 3 4

Tran

smitt

ed p

acke

ts (r

atio

)

0.5

0.6

0.7

0.8

0.9

1PF + Delay Constraint + Priority Scheduling

Class 1 - p=3

Class 2 - p=2

Class 4 - p=1

Class 3 - p=1


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Figure 8-8: Overall throughput when using different scheduling algorithms

(a) (b)

Figure 8-9: Zooming in Figure 8-8 shows the advantage and trade-off of the considered schedulers

Offered traffic

0 0.5 1 1.5 2 2.5 3 3.5 4

Pac

kets

/ B

aseb

and

Fram

e

0

0.5

1

1.5

2

2.5

3Overall Throughput

Maximum Rate

Proportional Fairness (PF)

PF + Delay Constraint

PF + Delay Constraint + Priority

Offered traffic

1.5 2 2.5 3 3.5

Pac

kets

/ B

aseb

and

Fram

e

1.4

1.6

1.8

2

2.2

2.4

2.6

2.8Overall Throughput

Maximum Rate




Offered traffic

3 3.2 3.4 3.6 3.8 4

Pac

kets

/ B

aseb

and

Fram

e

2.6

2.65

2.7

2.75

2.8

2.85

2.9Overall Throughput

Maximum Rate





Date: 21/04/2017

of 87

SANSA-645047 D3.6

9 Conclusions In this deliverable, we have proposed different resource allocation techniques specifically designed to be applied in the considered SANSA scenarios. Each technique is verified in several different setups considering realistic and modeled parameters, most of them keeping the true Helsinki topology as a benchmark for comparison purposes.

As a first step, we have focused on the power control and beamforming design assuming fixed carrier assignment. In Section 4 we presented a novel power and beamforming strategy combined with scheduling for point-to-multipoint communications; while Section 5 presented and discussed the power and beamforming design for point-to-point communications.

For the multi-antenna point-to-multipoint transmission for backhaul links, we relied on the hybrid analog-digital precoding design detailed in D3.5 and we aimed to improve their performance by combining it with link scheduling and proper power allocation. The numerical results reveal that there is a significant performance gain in terms of spectral efficiency with the considered results. In particular, in a sub-set of the real Helsinki topology, it is observed a 2.38x gain with respect to the original deployment.

For the multi-antenna point-to-point transmission for backhaul links, scheduling is not possible. Therefore, we proposed a power allocation strategy combined with precoding techniques. The proposed scheme maximizes the rate of the desired link, while keeping the caused interference below some specific threshold. The evaluation of the method is performed over a set of experiments, where it is shown that the level of the interference constraint determines the loss on the desired link capacity. Moreover, the method guarantees a gain up to 68.32% in terms of capacity for the interference link.

The performance of the aforementioned techniques is highly dependent on the number of antennas and the quality of their beamforming implementation. One way to relax these constraints is to design proper carrier allocation in order to minimize the amount of intra-system interference. In the extreme case of single antenna case, Section 6 showed that carrier allocation by itself is able to provide a 2.09x gain in terms of spectral efficiency. Complementary, we have also considered in Section 7 carrier assignment together with joint flow control by allowing scheduling in the time dimension, which increases the spectral efficiency gain up to 2.47x.

The carrier and flow assignment aims at avoiding the terrestrial network congestion. In this deliverable we have investigated the provision of mobile backhaul in situations where the terrestrial network is too congested or there has been a link failure that forces the use of the satellite backhaul links. In this regard, Section 8 considers the problem of providing QoS guarantees over the satellite backhaul link. QoS is becoming increasingly important in cellular communication systems due to the large variety of new services with contrasting performance requirements that are offered to consumers. We proposed a novel scheduling algorithm that chooses which user packets are transmitted on the same Baseband Frame, taking into account the QoS requirements of the queued packets and the channel state of each Satellite-to-iBN link.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

The aim is to allocate resources in such way that base stations are served in a proportionally fair way, depending on the channel conditions, while satisfying the delay and QoS priority requirements of the different packets. Our results show that our proposed algorithm significantly increases the probability that high priority packets are served, even under saturated traffic conditions. In addition, by taking into account the delay requirements of the queued packets we increase the offered traffic for which packet delivery can be guaranteed by approximately 30%, in non-saturated traffic loads

The proposed RRM techniques are able to provide specific gains to specific topologies. In the SANSA context, the HNM is in charge of configuring the topology. To do that, it performs topology calculations to restore the hybrid network upon node congestion or failure events. In particular, the HNM periodically calculates alternative topologies which, combined with the proposed RRM techniques, can be shortlisted in terms of SE gain. With this information, the HNM is able to take a wise decision in selecting the most appropriate topology given the available radio resources. Therefore, the outcomes of WP3 on RRM technique can be taken into account in WP4, together with other parameters such as latency or energy efficiency, in order to select the best topology.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

References [1] SANSA Deliverable D2.3, “Definition of reference scenarios, overall system

architectures, research challenges, requirements and KPIs”, January 2016. [2] SANSA Deliverable D3.7, “Database-assisted spectrum access techniques for Hybrid

Terrestrial-Satellite Backhauling”, April 2017. [3] SANSA Deliverable D3.5, “Intra-system Interference Mitigation Techniques for Hybrid

Terrestrial-Satellite Backhauling”, April 2017. [4] E. Lagunas, S. Maleki, L. Lei, C. Tsinos, S. Chatzinotas, B. Ottersten, “Carrier Allocation

for Hybrid Satellite-Terrestrial Backhaul Networks”, ICC Workshop on Satellite Communications: Challenges and Integration in the 5G ecosystem, Paris, France, May 2017.

[5] M. Shaat, A.I. Pérez-Neira, “Joint Flow Control and Link Scheduling in Hybrid Terrestrial-Satellite Wireless Backhauling Network”, ICC Workshop on Satellite Communications: Challenges and Integration in the 5G ecosystem, Paris, France, May 2017.

[6] CoRaSat, Deliverable D3.2, “Scenarios definition and selection “, October 2012. [7] E. Lagunas, S.K. Sharma, S. Maleki, S. Chatzinotas, B. Ottersten, “Resource Allocation for

Cognitive Satellite Communications with Incumbent Terrestrial Networks”, IEEE Trans. On Cognitive Communications and Networking, vol. 1, no.3, Nov 2015.

[8] E. Lagunas, S.K. Sharma, S. Maleki, S. Chatzinotas, J. Grotz, J. Krause, B. Ottersten, “Resource Allocation for Cognitive Satellite Uplink and Fixed-Service Terrestrial Coexistence in Ka-band”, International Conference on Cognitive Radio Oriented Wireless Networks (CROWNCOM), Doha, Qatar, Apr 2015.

[9] Larsson EG, Tufvesson F, Marzetta TL. Massive MIMO for Next Generation Wireless Systems. IEEE Commun Mag. Feb 2014;52(2):186–195.

[10] Gao Z, Dai L, Mi D, Wang Z, Imran MA, Shakir MZ. mmWave Massive-MIMO-Based Wireless Backhaul for the 5G Ultra-Dense Network. IEEE Wireless Commun. Oct 2015;22(5):13–21.

[11] C. G. Tsinos, S. Maleki, S. Chatzinotas, and B. E. Ottersten, “Hybrid Analog-Digital Transceiver Designs for Cognitive Large-Scale AntennaArray Systems,” CoRR, vol. abs/1612.02957, 2016. [Online]. Available: http://arxiv.org/abs/1612.02957

[12] M. A. Vázquez, L. Blanco, X. Artiga, and A. Pérez-Neira, “Hybrid analog-digital transmit beamforming for spectrum sharing satellite-terrestrial systems,” in 2016 IEEE 17th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), July 2016, pp. 1–5.

[13] S. Stanczak, M. Wiczanowski, H. Boche, “Fundamentals of Resource Allocation in Wireless Networks”, Springer, 2008. ISBN 978-3-540-79386-1

[14] E. Castaneda, A. Silva, A. Gameiro, and M. Kountouris, “An Overview on Resource Allocation Techniques for Multi-User MIMO Systems,” IEEE Communications Surveys Tutorials, vol. PP, no. 99, pp. 1–1, 2016.

[15] K. Hamdi, W. Zhang, and K. B. Letaief, “Opportunistic spectrum sharing in cognitive MIMO wireless networks,” IEEE Transactions on Wireless Communications, vol. 8, no. 8, pp. 4098–4109, August 2009.

http://icc2017.ieee-icc.org/workshop/1st-international-workshop-satellite-communications-challenges-and-integration-5g-ecosystem




http://arxiv.org/abs/1612.02957


Date: 21/04/2017

of 87

SANSA-645047 D3.6

[16] E. Driouch and W. Ajib, “Downlink Scheduling and Resource Allocation for Cognitive Radio MIMO Networks,” IEEE Transactions on Vehicular Technology, vol. 62, no. 8, pp. 3875–3885, Oct 2013.

[17] T. E. Bogale, L. B. Le, A. Haghighat, and L. Vandendorpe, “On the Number of RF Chains and Phase Shifters, and Scheduling Design With Hybrid Analog-Digital Beamforming,” IEEE Transactions on Wireless Communications, vol. 15, no. 5, pp. 3311–3326, May 2016.

[18] W. Yuan, S. M. D. Armour, and A. Doufexi, “A novel user selection algorithm for multiuser hybrid precoding in mmWave systems,” in 2016 IEEE 27th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), Sept 2016, pp. 1–6.

[19] ETSI EN 302 307 V1.2.1: “Digital Video Broadcasting (DVB); Second generation framing structure, channel coding and modulation systems for Broadcasting, Interactive Services, News Gathering and other broadband satellite applications (DVB-S2)”.

[20] V. Kawadia and P.R. Kumar, “A cautionary perspective on cross-layer design,” In Wireless Communications, IEEE, 12(1):3 – 11, feb. 2005.

[21] The DVB Project Office, “Generic Stream Encapsulation: Enabling the carriage of IP directly over DVB networks”,

[22] M. Tropea, et al., “Two Step based QoS Scheduler for DVB-S2 Satellite System,” in Proc. IEEE ICC2011.

[23] L.S. Ronga, et al. "A gateway architecture for IP satellite networks with dynamic resource management and DiffServ QoS provision." International journal of satellite communications and networking 21.4-5 (2003): 351-366.

[24] F. Vieira, M.A.Vazquez-Castro, and G. Seco-Granados, "A tunable-fairness cross-layer scheduler for DVB-S2," In International Journal of Satellite Communications and Networking 24, no. 5 (2006): 437-450.

[25] M.A. Vazquez-Castro, F. Vieira. "DVB-S2 full cross-layer design for QoS provision." IEEE Communications Magazine 50.1 (2012).

[26] E. Rendon-Morales, et al. "Adaptive packet scheduling for the support of QoS over DVB-S2 satellite systems." Wired/Wireless Internet Communications. Springer Berlin Heidelberg, 2011. 15-26.

[27] U. Park, H.W. Kim, D.S. Oh, B.J. Ku, “Interference-Limited Dynamic Resource Allocation Management for an Integrated Satellite/Terrestrial System”, ETRI Journal on Information, Telecommunication and Electronics, vol. 36, no. 4, pp. 519-527, Aug. 2014.

[28] A. Vanelli-Coralli, A. Guidotti, D. Tarchi, S. Chatzinotas, S. Maleki, S. Sharma, N. Chuberre, B. Evans, M. Lopez-Benitez, W. Tang, J. Grotz, and K. Liolis, Cooperative and Cognitive Satellite Systems. Oxford, UK: Academic Press, 2015, ch. Cognitive Radio Scenarios for Satellite Communications: The CoRaSat Project.

[29] S. Sharma, S. Chatzinotas, and B. Ottersten, “Cognitive Beamhopping for Spectral Coexistence of Multibeam Satellites,” Int. J. Satellite Commun. and Networking, Mar, 2014.

[30] S. Maleki, S. Chatzinotas, B. Evans, K. Liolis, J. Grotz, A. Vanelli-Coralli, and N. Chuberre, “Cognitive Spectrum Utilization in Ka Band Multibeam Satellite Communications,” IEEE


Date: 21/04/2017

of 87

SANSA-645047 D3.6

Communication Magazine, vol. 53, no. 3, pp. 24–29, Mar, 2015. [31] S.K. Sharma, E. Lagunas, S. Maleki, S. Chatzinotas, J. Grotz, J. Krause, B. Ottersten,

“Resource Allocation for Cognitive Satellite Communications in Ka-band (17.7-19.7 GHz)”, Workshop on Cognitive Radios and Networks for Spectrum Coexistence of Satellite and Terrestrial Systems, IEEE Int. Conf. On Communications (ICC), London, UK, Jun 2015.

[32] M. Hoyhtya, J. Kyrolainen, A. Hulkkonen, J. Ylitalo, and A. Roivainen, “Application of Cognitive Radio Techniques to Satellite Communication,” IEEE Symp. on Dynamic Spectrum Access Networks (DySPAN), Bellevue, WA, USA, Oct, 2012.

[33] E. Lagunas, S. Maleki, M. Soltanalian, S. Chatzinotas, A. Pérez-Neira, B. Ottersten, “Power and Rate Allocation in Cognitive Satellite Uplink Networks”, IEEE Int. Conf. on Commun. (ICC), Kuala Lumpur, Malaysia, May 2016.

[34] L. Rosati and G. Reali, “Jointly Optimal Routing and Resource Allocation in Hybrid Satellite/Terrestrial Networks,” Int. Workshop in Satellite and Space Communications, Madrid, Spain, Sept. 2006.

[35] T. Yoo and A. Goldsmith, “On the optimality of multiantenna broadcast scheduling using zero-forcing beamforming,” IEEE Journal on Selected Areas in Communications, vol. 24, no. 3, pp. 528–541, March 2006.

[36] H. Huang, C.B. Papadias, S. Venkatesan. MIMO Communication for Cellular Networks. Springer Science & Business Media, 2011.

[37] G.J. Foschini, “Layered space-time architecture for wireless communication in a fading environment when using multi-element antennas.” Bell labs technical journal 1.2 (1996): 41-59.

[38] E. Telatar, “Capacity of Multi-antenna Gaussian Channels.” European transactions on telecommunications 10.6 (1999): 585-595.

[39] R.G. Gallager, Information theory and reliable communication. Vol. 2. New York: Wiley, 1968.

[40] D.S. Shiu, et al. “Fading correlation and its effect on the capacity of multielement antenna systems.” IEEE Transactions on communications 48.3 (2000): 502-513.

[41] E. Björnson, Emil, E. Jorswieck. “Optimal resource allocation in coordinated multi-cell systems.” Foundations and Trends in Communications and Information Theory 9.2–3 (2013): 113-381.

[42] A. Pérez-Neira, J.M. Veciana, M.A. Vázquez, E. Lagunas, “Distributed Power Control with Received Power Constraints for Time-Area-Spectrum Licenses”, ELSEVIER Signal Processing, vol. 120, pp. 141-155, March 2016.

[43] S. Theodoridis. Machine learning: a Bayesian and optimization perspective. Academic Press, 2015.

[44] S. Boyd, L. Vandenberghe. Convex optimization. Cambridge university press, 2004. [45] L. Xiao, M. Johansson, and S.P. Boyd, “Simultaneous Routing and Resource Allocation

via Dual Decomposition,” IEEE Trans. Commun., vol. 52, no. 7, pp. 1136–1144, Jul. 2014. [46] M. Cao, X. Wang, S.J. Kim, and M. Madihian, “Multi-Hop Wireless Backhaul Networks: A

Cross-Layer Design Paradigm,” IEEE J. Sel. Areas Commun., vol. 25, no. 4, pp. 738–748, May 2007.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

[47] Z.Q. Luo and S. Zhang, “Dynamic Spectrum Management: Complexity and Duality,” IEEE J. Sel. Topics Signal Process., vol. 2, no. 1, pp. 57–73, Feb. 2008.

[48] C. Caini, G.E. Corazza, G. Falciasecca, M. Ruggieri, and F. Vatalaro, “A Spectrum and Power Efficient EHF Mobile Satellite System to be Integrated with Terrestrial Cellular Systems,” IEEE J. Sel. Areas Commun., vol. 10, no. 8, pp. 1315–1325, 1992.

[49] F. Kaabi, S. Ghannay, and F. Filali. "Channel Allocation and Routing in Wireless Mesh Networks: A survey and qualitative comparison between schemes." International Journal of Wireless and Mobile Network, 2010, vol. 2, no 1, p. 132-151.

[50] W. Cheng, X. Cheng, T. Znati, X. Lu and Z. Lu, "The Complexity of Channel Scheduling in Multi-Radio Multi-Channel Wireless Networks," IEEE INFOCOM 2009, Rio de Janeiro, 2009, pp. 1512-1520.

[51] G. Li and H. Liu, “Resource allocation for OFDMA relay networks with fairness constraints,” IEEE Journal on Selected Areas in Communications, vol. 24, no. 11, pp. 2061–2069, November 2006.

[52] G. Kulkarni, S. Adlakha, and M. Srivastava, “Subcarrier allocation and bit loading algorithms for OFDMA-based wireless networks,” IEEE Transactions on Mobile Computing, vol. 4, no. 6, pp. 652–662, November 2005.

[53] M. S. Kodialam and T. Nandagopal, "Characterizing achievable rates in multi-hop wireless mesh networks with orthogonal channels", IEEE/ACM Trans. Netw. 13(4): 868-880 (2005).

[54] K. Karakayali, J. H. Kang, M. Kodialam and K. Balachandran, "Cross-Layer Optimization for OFDMA-Based Wireless Mesh Backhaul Networks," 2007 IEEE Wireless Communications and Networking Conference, Kowloon, 2007, pp. 276-281.

[55] K. Karakayali, J. H. Kang, M. Kodialam and K. Balachandran, "Joint Resource Allocation and Routing for OFDMA-Based Broadband Wireless Mesh Networks," 2007 IEEE International Conference on Communications, Glasgow, 2007, pp. 5088-5092.

[56] L. B. Le and E. Hossain, "Joint Rate Control and Resource Allocation in OFDMA Wireless Mesh Networks," IEEE Wireless Communications and Networking Conference, Kowloon, 2007, pp. 3041-3045.

[57] P. Thulasiraman and X. Shen, "Interference Aware Subcarrier Assignment for Throughput Maximization in OFDMA Wireless Relay Mesh Networks," 2009 IEEE International Conference on Communications (ICC), Dresden, 2009, pp. 1-6.

[58] P. Thulasiraman and X. Shen, "Decoupled optimization of interference aware routing and scheduling for throughput maximization in wireless relay mesh networks," 2009 6th IEEE Annual Communications Society Conference on Sensor, Mesh and Ad Hoc Communications and Networks Workshops, Rome, 2009, pp. 1-6.

[59] P. Thulasiraman, and X. Shen, "Interference aware resource allocation for hybrid hierarchical wireless networks." Computer Networks, 2010, vol. 54, no 13, p. 2271-2280.

[60] S. Hoteit, S. Secci, R. Langar and G. Pujolle, "Strategic subchannel resource allocation for cooperative OFDMA Wireless Mesh Networks," 2012 IEEE International Conference on Communications (ICC), Ottawa, ON, 2012, pp. 440-444.

[61] M.F Uddin, and C. Assi. "Joint routing and scheduling in WMNs with variable-width spectrum allocation." IEEE Transactions on Mobile Computing, 2013, vol. 12, no 11, p.


Date: 21/04/2017

of 87

SANSA-645047 D3.6

2178-2192. [62] B. Hajek and G. Sasaki, "Link scheduling in polynomial time", IEEE Trans. Inf. Theory, vol.

34, no. 5, pp. 910-917, Sep. 1988. [63] M.A. Vazquez-Castro and G.Seco-Granados, “Cross-layer packet scheduler design of a

multibeam broadband satellite system with adaptive coding and modulation,” in Wireless Communications, IEEE Transactions on, 6(1):248–258, jan. 2007.

d3.6 dynamic radio resource management for hybrid ... · pdf filed3.6: dynamic radio resource...

Documents