[ieee milcom 2013 - 2013 ieee military communications conference - san diego, ca, usa...

6
Performance Evaluation of Access Control for CRDSA and R-CRDSA under High Traffic Loads Hong-Jun Noh, Jong-Kwan Lee, and Jae-Sung Lim Tactical Network Research Center, Ajou University Suwon, South Korea {nonoboy, jklee64, jaslim}@ajou.ac.kr Abstract—In this paper, we propose access control schemes for the contention resolution diversity slotted ALOHA (CRDSA) and the reservation-CRDSA (R-CRDSA) schemes to enhance throughput under high traffic loads. We also propose an estima- tion technique for the traffic load of CRDSA. In slotted ALOHA, many schemes have been proposed for the same purpose. Some of them adjust channel access probability according to the estimated traffic load. In this paper, these techniques are extended to CRDSA and R-CRDSA by considering the number of packet replicas and the status of the reservation channels. Simulation results demonstrate that the proposed scheme enhances through- put and channel efficiency under high traffic loads. I. I NTRODUCTION In the network-centric warfare vision of the military, which focuses on transmitting the right information to the right person at the right time, satellite networks (SATNETs) play an important role because satellites can simultaneously deliver communication signals over a wide range of battlefields [1], [2]. Therefore, a considerable amount of information would be distributed in the next-generation SATNETs. This infor- mation could include many types of data, such as situation awareness, commands, and control messages for visualization and synchronization in the battlefield. Under the ever-changing battlefield conditions, bursty traffic occurs frequently. Bursty traffic is characterized by the transmission of low volumes of data by a large number of users in real-time [3]. Traditionally, the demand assignment multiple access tech- nology is responsible for the overall traffic load of SATNETs. However, it becomes less efficient with more traffic bursts, be- cause each capacity request consumes a considerable amount of time due to the lengthy propagation delay in a satellite channel [6]. In an environment with fluctuating traffic load from an unspecified number of users, random access (RA) schemes are more effective. However, slotted ALOHA (SA) [4], which is the RA scheme for SATNETs, typically suffers from low throughput and low reliability due to collisions. To overcome these shortcomings, contention resolution di- versity slotted ALOHA (CRDSA) [5] was proposed. CRDSA enhances the throughput of the RA channel by using packet replicas and successive interference cancellation (SIC). Af- ter CRDSA, CRDSA++ [6] and IRSA [7] were proposed to further improve throughput by increasing the number of packet replicas. However, these advanced RA schemes have lower critical points in traffic load and lower normalized capacity efficiency than SA. In other words, the throughput of these schemes is maximized when the traffic load is normalized to less than 1, and it decreases rapidly, despite sending more replicas. To resolve these issues, reservation- CRDSA (R-CRDSA) [8] was proposed. R-CRDSA increases the critical point and reduces the number of unnecessary replicas through slot reservation. However, in a high traffic load state where normalized traffic load is greater than 1, the packet loss ratio increases, thereby decreasing the probability of initial transmission success, which is a prerequisite for slot reservation. Therefore, access control is required to decrease the packet loss ratio at high traffic loads, which is likely to occur frequently in the next generation of SATNETs. Many access control schemes for SA have been researched since its development. Among them are the dynamic frame length ALOHA (DFLA) [9] and the adaptive traffic load (ATL) [10] which adjust the length of the frame and the channel access probability, respectively. DFLA is not suitable for a distributed environment and for use with R-CRDSA in particular, because the length of a frame must be the same for all nodes, and changing the frame length results in a change in the location of the reservation slot. In this paper, we describe an access control scheme for CRDSA based on ATL, and we propose how to extend it to R-CRDSA using the status of the reservation channel. Next, for accurate access control, the accurate traffic load should be given. Therefore, in this paper, we also propose a very simple way to extend the traffic load estimation techniques in framed SA to CRDSA. This paper is organized as follows: In Section II, we describe the system model of SATNETs and introduce CRDSA and R-CRDSA. Section III describes the proposed access control and traffic load estimation schemes. Section IV shows the performance evaluations for the proposed schemes using simulation. The conclusions are provided in Section V. II. SYSTEM MODEL DESCRIPTION A. System Model In SATNETs, a satellite terminal (ST) on the ground sends a packet on an up-link (frequency f 1, i.e., multiple access channel) to a satellite transponder. The satellite transponder returns the packet on a down-link (frequency f 2, i.e., broadcast channel) to the same ST, to adjust the co-channel interference by switching the frequencies of the up-link (f 1) and the down- link (f 2). Due to the basic characteristics of satellites that broadcast communication signals in a communication range, 2013 IEEE Military Communications Conference 978-0-7695-5124-1/13 $31.00 © 2013 IEEE DOI 10.1109/MILCOM.2013.232 1365 2013 IEEE Military Communications Conference 978-0-7695-5124-1/13 $31.00 © 2013 IEEE DOI 10.1109/MILCOM.2013.232 1365

Upload: jae-sung

Post on 28-Jan-2017

215 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: [IEEE MILCOM 2013 - 2013 IEEE Military Communications Conference - San Diego, CA, USA (2013.11.18-2013.11.20)] MILCOM 2013 - 2013 IEEE Military Communications Conference - Performance

Performance Evaluation of Access Control forCRDSA and R-CRDSA under High Traffic Loads

Hong-Jun Noh, Jong-Kwan Lee, and Jae-Sung LimTactical Network Research Center, Ajou University

Suwon, South Korea

{nonoboy, jklee64, jaslim}@ajou.ac.kr

Abstract—In this paper, we propose access control schemesfor the contention resolution diversity slotted ALOHA (CRDSA)and the reservation-CRDSA (R-CRDSA) schemes to enhancethroughput under high traffic loads. We also propose an estima-tion technique for the traffic load of CRDSA. In slotted ALOHA,many schemes have been proposed for the same purpose. Some ofthem adjust channel access probability according to the estimatedtraffic load. In this paper, these techniques are extended toCRDSA and R-CRDSA by considering the number of packetreplicas and the status of the reservation channels. Simulationresults demonstrate that the proposed scheme enhances through-put and channel efficiency under high traffic loads.

I. INTRODUCTION

In the network-centric warfare vision of the military, which

focuses on transmitting the right information to the right

person at the right time, satellite networks (SATNETs) play

an important role because satellites can simultaneously deliver

communication signals over a wide range of battlefields [1],

[2]. Therefore, a considerable amount of information would

be distributed in the next-generation SATNETs. This infor-

mation could include many types of data, such as situation

awareness, commands, and control messages for visualization

and synchronization in the battlefield. Under the ever-changing

battlefield conditions, bursty traffic occurs frequently. Bursty

traffic is characterized by the transmission of low volumes of

data by a large number of users in real-time [3].

Traditionally, the demand assignment multiple access tech-

nology is responsible for the overall traffic load of SATNETs.

However, it becomes less efficient with more traffic bursts, be-

cause each capacity request consumes a considerable amount

of time due to the lengthy propagation delay in a satellite

channel [6]. In an environment with fluctuating traffic load

from an unspecified number of users, random access (RA)

schemes are more effective. However, slotted ALOHA (SA)

[4], which is the RA scheme for SATNETs, typically suffers

from low throughput and low reliability due to collisions.

To overcome these shortcomings, contention resolution di-

versity slotted ALOHA (CRDSA) [5] was proposed. CRDSA

enhances the throughput of the RA channel by using packet

replicas and successive interference cancellation (SIC). Af-

ter CRDSA, CRDSA++ [6] and IRSA [7] were proposed

to further improve throughput by increasing the number of

packet replicas. However, these advanced RA schemes have

lower critical points in traffic load and lower normalizedcapacity efficiency than SA. In other words, the throughput

of these schemes is maximized when the traffic load is

normalized to less than 1, and it decreases rapidly, despite

sending more replicas. To resolve these issues, reservation-

CRDSA (R-CRDSA) [8] was proposed. R-CRDSA increases

the critical point and reduces the number of unnecessary

replicas through slot reservation. However, in a high traffic

load state where normalized traffic load is greater than 1, the

packet loss ratio increases, thereby decreasing the probability

of initial transmission success, which is a prerequisite for slot

reservation. Therefore, access control is required to decrease

the packet loss ratio at high traffic loads, which is likely to

occur frequently in the next generation of SATNETs.

Many access control schemes for SA have been researched

since its development. Among them are the dynamic frame

length ALOHA (DFLA) [9] and the adaptive traffic load

(ATL) [10] which adjust the length of the frame and the

channel access probability, respectively. DFLA is not suitable

for a distributed environment and for use with R-CRDSA in

particular, because the length of a frame must be the same for

all nodes, and changing the frame length results in a change in

the location of the reservation slot. In this paper, we describe

an access control scheme for CRDSA based on ATL, and we

propose how to extend it to R-CRDSA using the status of

the reservation channel. Next, for accurate access control, the

accurate traffic load should be given. Therefore, in this paper,

we also propose a very simple way to extend the traffic load

estimation techniques in framed SA to CRDSA.

This paper is organized as follows: In Section II, we

describe the system model of SATNETs and introduce CRDSA

and R-CRDSA. Section III describes the proposed access

control and traffic load estimation schemes. Section IV shows

the performance evaluations for the proposed schemes using

simulation. The conclusions are provided in Section V.

II. SYSTEM MODEL DESCRIPTION

A. System Model

In SATNETs, a satellite terminal (ST) on the ground sends

a packet on an up-link (frequency f1, i.e., multiple access

channel) to a satellite transponder. The satellite transponder

returns the packet on a down-link (frequency f2, i.e., broadcast

channel) to the same ST, to adjust the co-channel interference

by switching the frequencies of the up-link (f1) and the down-

link (f2). Due to the basic characteristics of satellites that

broadcast communication signals in a communication range,

2013 IEEE Military Communications Conference

978-0-7695-5124-1/13 $31.00 © 2013 IEEE

DOI 10.1109/MILCOM.2013.232

1365

2013 IEEE Military Communications Conference

978-0-7695-5124-1/13 $31.00 © 2013 IEEE

DOI 10.1109/MILCOM.2013.232

1365

Page 2: [IEEE MILCOM 2013 - 2013 IEEE Military Communications Conference - San Diego, CA, USA (2013.11.18-2013.11.20)] MILCOM 2013 - 2013 IEEE Military Communications Conference - Performance

Fig. 1. Example of an RA frame in a SATNET [3]. The satellite transponderswitches the frequencies between the up-link (f1, multiple access channel)and the down-link (f2, broadcast channel). The duration of a frame is greaterthan the maximum propagation delay of a packet.

all STs within a satellite beam range can track the usage status

of the shared channel without additional feedback information.

These characteristics of SATNETs are shown in Fig. 1 [8]. To

maximize this benefit, the duration of a frame is chosen to be

greater than the maximum propagation delay of a packet [11].

When a message arrives at an ST, it is fragmented into

packets with a fixed length for transmission. We assume that

the size of each packet is the same as the size of a slot, and we

consider a MAC frame that consists of N slots. Along with

the order of the packets in the message, the ST sends a packet

and its replica in each frame. We do not consider the capture

effect, and we assume that channel errors occur only when

packets collide in a slot. The performance of networks under

these channel impairments and the capture effect of CRDSA

are well documented in [5].

B. CRDSA and R-CRDSA

CRDSA is based on framed SA; therefore, so every ST

randomly selects the slots in a frame to send a packet frame

by frame. In each frame, every ST sends two replica packets,

both having the same data and the slot number of the other

packet as a pointer. Each ST has a buffer to store all packets

in a frame for the SIC iteration process, as shown in Fig. 1.

At the end of the frame, each ST performs the SIC iteration

process to resolve collisions. A packet that accesses a slot

can be a clean packet or a collided packet. A clean packet is

used to find the slot number of its own replica. If a collision

occurs in a slot, the replica is cancelled (removed) by using the

clean packet. Therefore, the additional collided packets can be

successfully decoded, and contention can be resolved by using

SIC iteratively.

R-CRDSA is based on CRDSA with the addition of slot

reservation and control of the number of replicas. If an ST has

a message with multiple fragmented packets to send succes-

sively, the number of remaining frames is added to the packets.

All the STs become aware of this because of the down-link

in the SATNET. If a packet was successfully decoded, the

slot with the smallest number among the replicas is selected

as the reserved slot. This rule forces the slot to be reserved

exclusively for the sender, and the slot cannot be accessed by

other STs in the next frame. Each ST independently decides

whether a slot is reserved using the stored packets of a frame.

(� � 2)th Frame (� � 1)th Frame (�)th Frame (� � 1)th Frame

time

Suc. Col. Idle Col. Col.

Slot

ST #1

ST #2

ST #3

ST #4

Clean packet Collided packet Cancelled packet

Res. Res. Suc. Col. Suc.

Res.

Res.

Fig. 2. Examples of CRDSA and R-CRDSA. Every ST initially sends tworeplicas. In R-CRDSA, if an ST has a reserved slot, it sends only one packetin the reserved slot. A collision between packets is resolved by both packetscancelling each other.

If an ST has a packet to send and does not have a reserved slot

in the previous frame, the ST sends two replica packets in slots

that are randomly selected from the non-reserved slots in the

previous frame. If an ST has a reserved slot, it sequentially

sends the rest of the packets without any replicas until the

end of the message. Compared to CRDSA, R-CRDSA has a

smaller number of packets in a frame, and the probability of

collisions in the non-reserved slots is greater. When sending

the final packet of the message, the ST transmits the packet

with a notification that the reserved slot will become free in

the next frame.

Fig. 2 shows the examples of CRDSA and R-CRDSA.

Suppose that four STs are contending for channel access and

that all of them have multiple packets to send over the next

few frames. In the (i−1)th frame, they do not have a reserved

slot; therfore, each ST sends two replicas of a packet. Only ST

#1 has a clean packet in the first SIC iteration of the (i− 1)th

frame In the second SIC iteration, the clean packet is used to

recover the collided packet of ST #2 by removing the replica

of ST #1. Another collided packet of ST #2 is removed in the

third SIC iteration, but the packets of ST #3 and #4 cannot be

recovered due to the loop. A loop occurs when all the replicas

from different STs are irretrievably coupled with their own

replicas. After all the SIC iterations in the (i−1)th frame, only

ST #1 and ST #2 are able to reserve slots in the (i)th frame.

The two STs (#1 and #2) each use only their own reserved

slot. The other two STs (#3 and #4) have to select a slot among

the non-reserved slots in order to send their packets.

III. PROPOSED SCHEMES

A. Access Control for CRDSA and R-CRDSA

In this section, we describe the access control scheme of

SA, which uses the access probability, and we extend it to

CRDSA and R-CRDSA. Rivero-Angeles et al. [10] proposed

an ATL multiple access control protocol to ensure maximum

throughput for SA. ATL behaves exactly the same as SA under

a traffic load that is lower than the critical point. The critical

point is the traffic load at which the network can achieve the

maximum throughput. Above the critical point, the throughput

13661366

Page 3: [IEEE MILCOM 2013 - 2013 IEEE Military Communications Conference - San Diego, CA, USA (2013.11.18-2013.11.20)] MILCOM 2013 - 2013 IEEE Military Communications Conference - Performance

of most RA schemes decreases. To solve this, ATL uses the

access probability to reduce the number of packets in an RA

channel and to maintain the maximum throughput above the

critical point. The critical point of SA has a value of 1 when

expressed as a normalized traffic load. G, which is defined as

follows:

G =n

N, (1)

where n is the number of terminals which attempt to connect

to an RA channel, and N is the number of slots in a frame.

The selection of the proper access probability makes it possible

to provide the maximum throughput under high traffic loads

greater than the critical point. The access probability is defined

as follows [10]:

pac = min

(1,

G∗

G

), (2)

where G∗ is the critical point. Since the traffic load of the RA

channel is not given, each terminal uses an estimated traffic

load (G′) instead of G to calculate pac. Each terminal uses pac

to determine whether to access the next slot. Specifically, each

terminal generates a random number uniformly between 0 and

1 and compares it to pac. If the random number is smaller than

or equal to pac, the terminal accesses the next slot and in the

opposite case, does not access the next slot.

In CRDSA, G∗ is 0.65 as a normalized traffic load, and

G′ is obtained using the status of the previous frame. Each

ST can track the usage status of the shared channel without

any additional feedback in SATNETs. How to estimate G′

is described in Section III-B. Each ST calculates pac using

(2) and determines whether to access the next frame. If the

generated random number is smaller than or equal to pac, the

ST sends two replica packets to the next frame; otherwise,

it does not send any packets. The access control scheme is

almost the same as ATL, except that it works on the frame

basis.

In R-CRDSA, we consider the reserved channel because

the behavior of a ST is different depending on whether it

has a reserved slot. All the STs that have a reserved slot do

not consider pac because they are free from contention. Only

the STs that access non-reserved slots follow access control.

Therefore, the access probability for the reservation scheme is

defined as follows:

pacR = min

(1,

G∗nn

N−Nr

), (3)

where nn is the number of STs that have no reserved slots,

and Nr is the number of reserved slots in a frame. Similarly,

we define nr as the number of STs that have a reserved slot,

and Nn is the number of non-reserved slots in a frame. The

STs that belong to nn determine whether to access the next

frame by using (3). Therefore, the number of packets in non-

reserved slots decreases and the probability of collision also

declines.

B. Traffic Load Estimation for CRDSA

Accurate traffic load estimation (TLE) is the prerequisite

for access control in Section III-A. This section describes

a representative of the TLE technique for framed SA and a

method to extend it to CRDSA. To estimate the traffic load,

the usage status of slots in a frame has to be obtained, which

can be easily done by SATNETs. Slot states are divided into

idle (I), successful (S), and collided (C) states that represent

zero, one, and more than two bursts transmitting in a slot,

respectively. Most TLE techniques utilize the probability of

slot states based on binomial distribution theory. In framed

SA, given the number of terminals n and the frame length

N , the expected numbers of slots in each state per frame are

defined as follows [12]:

mI(n,N) = N(1− p)n, (4)

mS(n,N) = n(1− p)n−1, (5)

mC(n,N) = N −mI −mS , (6)

where p = 1/N is the probability that the packet of a

single terminal is present in a given slot. Due to the mutual

dependency among the slots in a frame, regarding their states,

the probability of a set of slots being in a specific state is

defined as follows [13]:

Pn(v) =(NNs

)(N−Ns

Nc

)∑Nc

j=0

∑Nc−jl=0

(Nc

j

)(Nc−j

l

) (−1)Nc−jn!jn−l−Ns

Nn(n−l−Ns)!,

(7)

where v =< Ni, Ns, Nc > is the vector of a frame, and Ni,

Ns, and Nc are the numbers of slots in the I , S, and C states,

respectively.

TLE techniques obtain the estimated traffic load G′ by

inversely tracking equations (4)-(6) and (7). Borrowing the

notation from [12], let H(·) be a TLE function that returns G′

from the observed vector v. The classical maximum likelihood

estimator (MLE), which returns the value G′ that maximizes

the conditional probability of the vector v, given n, is defined

as follows:

HMLE(v) = argmaxn

Pn(v). (8)

In general, HMLE works accurately, but it has a very high

computational complexity in calculating Pn. To solve this,

Zanella [12] proposed a TLE function that reduces the com-

plexity of MLE by using the approximation function and dif-

ferential calculus while maintaining accuracy. The conditional

probability for the simplified MLE is as follows

PG(v) = GNse−GN (eG − 1−G)Nc . (9)

The final estimate function utilizing (9) is obtained as follows:

HZan(v) = argmaxG

PG(v). (10)

To reduce the computational complexity of (10), the differen-

tial equation of (9) is used to determine the interval of the

solution. Therefore, the computational complexity of HZan is

reduced to O(log2(N)) from O(N2) of HMLE .

13671367

Page 4: [IEEE MILCOM 2013 - 2013 IEEE Military Communications Conference - San Diego, CA, USA (2013.11.18-2013.11.20)] MILCOM 2013 - 2013 IEEE Military Communications Conference - Performance

In order to apply these TLE techniques in framed SA to

CRDSA, we simply divide the return values of H by 2,

because the number of packets in a frame of CRDSA is twice

compared the number of packets in a frame of framed SA. In

CRDSA, two replica packets are somewhat dependent on each

other, because their locations must be different. Therefore, the

actual distribution of the packet in CRDSA is different from

the situation where the number of packets in framed SA is

simply doubled. However, when the frame is moderately long,

the packets are very similarly distributed in both CRDSA and

framed SA. Thus, it is possible to estimate G through the

following equation:

HCRDSA(v) = HZan(v)/2. (11)

In R-CRDSA, we apply (11) to non-reserved slots because we

already know that one reserved slot is occupied by one ST.

IV. SIMULATION RESULTS

We implemented a simulator to observe the performance

of CRDSA and R-CRDSA system in which the SIC and

the reservation rules are implemented. In this simulator, we

assumed perfect SIC, which means that any clean packet

allows the perfect resolution of its interference with other

packets. The maximum iteration number and the number of

replicas were 10 and 2, respectively. In the simulation, each

frame had 100 slots and all the simulations were performed

with 1, 000, 000 frames. We did not consider the capture effect

and assumed that the channel error occurs only in the collided

slot. Finally, we assumed that the traffic load in the RA channel

is constant during simulation.

A. Accuracy of TLE for CRDSA

First, we investigated the accuracy of TLE for CRDSA

without any access control. Fig. 3 shows the return values from

HZan and HCRDSA in framed SA and CRDSA, respectively.

In Fig. 3, the black line is the traffic load G. Simulation

results show that the estimated traffic load G′ from HZan

and HCRDSA are very similar to G.

In order to ascertain the estimation performance in more

detail, we examine the normalized estimate error defined as

follows:

ε(v) =H(v)−G

G=

G′ −G

G. (12)

Fig. 4 shows the normalized estimate error of Fig. 3. When Gis less than 0.5, both TLE schemes have no tendency because

only a few packets are present in the network. When G is

greater than 0.5, the number of errors increase as G increases.

This is a common phenomenon in all TLE techniques because

various stochastic states of slots are observed. The normalized

estimate error of HCRDSA is more than twice as large as

HZan. Because the number of packets in CRDSA is twice

that of SA, the slots in CRDSA have more varied states under

the same traffic load. However, the normalized estimate error

of both techniques is less than 0.025, which is quite small;

therefore, the estimated traffic load is reliable until G becomes

less than 2, as shown in Fig. 3.

Fig. 3. Performance of HZan and HCRDSA (N = 100)

Fig. 4. Normalized estimate error of HZan and HCRDSA (N = 100)

B. Performance Evaluation of the Proposed Access Control

In order to verify the performance of the proposed access

control mechanism only, we assume that the traffic load of

channel is given. The critical points of SA and CRDSA are 1and 0.65, respectively. For R-CRDSA, we show cases where

the average numbers of packets in a message, γ, are 5 and 10.

Fig. 5 and Fig. 6 show the normalized throughput of basic

RA schemes and the proposed access control, respectively. Fig.

5 shows that the RA schemes based on CRDSA are worse than

the RA schemes based on SA when the traffic load is greater

than 1. Therefore, CRDSA is not suitable for a high traffic

load. Even if we use the reservation scheme, we cannot avoid

throughput degradation under high traffic loads. However, Fig.

6 shows that the RA schemes based on CRDSA maintain their

throughput by using the proposed access control with perfect

estimate. This superiority is attributable to pac, which reduces

13681368

Page 5: [IEEE MILCOM 2013 - 2013 IEEE Military Communications Conference - San Diego, CA, USA (2013.11.18-2013.11.20)] MILCOM 2013 - 2013 IEEE Military Communications Conference - Performance

Fig. 5. Normalized throughput of basic RA schemes (N = 100)

Fig. 6. Normalized throughput of the proposed access control (N = 100)

the number of packets in the channel to keep it close to the

traffic load at the critical point.

Fig. 7 and Fig. 8 show the packet loss ratio of the basic

RA schemes and the proposed access control, respectively.

We can verify that the packet loss ratio of the basic RA

schemes rapidly increases from the critical point. This causes

throughput degradation under high traffic loads. On the other

hand, the proposed access control scheme restricts the growth

of the packet loss ratio. However, even if we apply the

proposed access control, we cannot reduce the packet loss ratio

to less than that at the critical point.

To improve throughput, the packet replication of advanced

RA schemes reduces the normalized efficiency. The normal-

ized efficiency indicates the influence of RA schemes on the

Fig. 7. Packet loss ratio of basic RA schemes (N = 100)

Fig. 8. Packet loss ratio of the proposed access control (N = 100)

bandwidth and is defined as follows [4]

η =Cj

Cref, (13)

where Cref and Cj are the capacity of the Gaussian channel

and the RA scheme j, respectively. Given the average aggre-

gated signal power (PS) and noise power (PN ), the channel

capacity Cref and Cj are defined as follows:

Cref = log (1 + PS/PN ) , (14)

Cj = Tj · log(1 +

PS

PN ·Dj

), (15)

where Tj is the normalized throughput and Dj is a ratio of

the average power used for the transmission of a packet. DSA

and DR−SA are G, and DCRDSA and DR−CRDSA are 2Gand 2G−TR−CRDSA, respectively [8]. We define Dj for the

13691369

Page 6: [IEEE MILCOM 2013 - 2013 IEEE Military Communications Conference - San Diego, CA, USA (2013.11.18-2013.11.20)] MILCOM 2013 - 2013 IEEE Military Communications Conference - Performance

Fig. 9. Normalized efficiency of basic RA schemes (N = 100)

Fig. 10. Normalized efficiency of proposed access control (N = 100)

proposed access control as follows:

DacSA = Gpac, (16)

DacCRDSA = 2Gpac, (17)

DacR−SA = T ac

R−SA + (G− T acR−SA)p

acR , (18)

DacR−CRDSA = T ac

R−CRDSA + 2(G− T acR−CRDSA)p

acR . (19)

We can obtain the channel efficiency using (13)-(19) for all

schemes.

Fig. 9 and Fig. 10 show the channel efficiency of basic RA

schemes and the proposed access control respectively, when

the signal-to-noise ratio (ES/N0) is 6dB. The tendencies

of channel efficiency and throughput are similar. Above the

critical point, the channel efficiencies of basic RA schemes

decrease. However, the channel efficiencies of the proposed

access control are maintained as other performance measures.

V. CONCLUSION

In this paper, we proposed access control schemes for

CRDSA and R-CRDSA to ensure the maximum throughput

under high traffic loads. Basic RA schemes such as SA, R-

SA, CRDSA, and R-CRDSA show performance degradation

under high traffic loads greater than critical points due to the

increased probability of collision. The proposed scheme con-

trols the access probability of each ST and resolves collision

under high traffic loads. To derive the access probability, we

proposed a simple mechanism to estimate the traffic load of

CRDSA by extending the conventional traffic load estimation

technique. Simulation results demonstrated that the proposed

access control scheme enhances throughput and channel ef-

ficiency even in a high traffic load which is a substantial

characteristic of next-generation SATNETs.

ACKNOWLEDGMENT

“This work was supported by the National Research Foun-

dation of Korea (NRF) grant funded by the Korea govern-

ment (MSIP) (No. 2013R1A2A1A01016423).” “This research

was supported by the MSIP(Ministry of Science, ICT &

Future Planning), Korea, under the ITRC(Information Tech-

nology Research Center) support program supervised by the

NIPA(National IT Industry Promotion Agency (NIPA-2013-

(H0301-13-2003)).”

REFERENCES

[1] T. Maseng, and R. Landry, “Network centric military communications,”IEEE Communications Magazine, Vol. 42, No. 11, pp.: 77-78, 2004.

[2] S. L. Kota, G. Giambene, and P. Chini, “A mobile satellite systemsframe work for network centric applications,” in Proceedings on IEEEMILCOM, 2008.

[3] M.-W. Lee, J.-K. Lee, H.-J. Noh and J.-S. Lim, “Stability of reservation-contention resolution diversity slotted ALOHA for satellite networks,” inProceedings on IEEE MILCOM, 2012.

[4] N. Abramson, “The throughput of packet broadcasting channels,” IEEETransactions on Communications, Vol 25, No. 1, pp.: 117-128, 1977.

[5] E. Casini, R. De Gaudenzi, and Od. R. Herrero, “Contention resolutiondiversity slotted ALOHA (CRDSA): an enhanced random access schemefor satellite access packet network,” IEEE Transactions on WirelessCommunications, Vol. 6, No. 4, pp.: 1408-1419, 2007.

[6] R. De Gaudenzi, Od. R. Herrero, “Advances in random access protocolsfor satellite networks,” in IEEE IWSSC, 2009.

[7] G. Liva, “Graph-based analysis and optimization of contention resolutiondiversity slotted ALOHA,” IEEE Transactions on Communications, Vol.59, No. 2, pp.: 477-487, 2011.

[8] M.-W. Lee, J.-K. Lee, and J.-S. Lim, “R-CRDSA: Reservation-ContentionResolution Diversity Slotted ALOHA for Satellite Networks,” IEEECommunications Letters, Vol. 16, nN. 10, pp.: 1576-1579, 2012.

[9] F. C. Schoute, “Dynamic Frame Length ALOHA,” IEEE Transactions onCommunications, Vol. 31, No. 4, 1983.

[10] M. E. Rivero-Angeles, D. Lara-Rodriguez, and F. A. Cruz-Perez,“Random-access control mechanisms using adaptive traffic load inALOHA and CSMA strategies for EDGE,” IEEE Transactions on Ve-hicular Technology, Vol. 54, No. 3, pp.: 1160-1186, 2005.

[11] S. S. Lam, “Packet broadcast networks - A performance analysis of theR-ALOHA protocol,” IEEE Transaction on Computers, Vol. 29, No. 7,pp.: 596-603, 1980.

[12] A. Zanella, “Estimating collision set size in framed slotted aloha wirelessnetworks and RFID systems,” IEEE Communications Letters, Vol. 16, No.3, pp.: 300-303, 2012.

[13] J. Riordan, An Introduction to Combinatorial Analysis, Princeton Uni-versity Press, 1978.

13701370