1

Efficient Energy Management for Internet of Thingsin Smart Cities

Waleed Ejaz, Muhammad Naeem, Adnan Shahid, Alagan Anpalagan, and Minho Jo

Abstract—The drastic increase in urbanization over the pastfew years requires sustainable, efficient, and smart solutionsfor transportation, governance, environment, quality of life, etc.The Internet of Things (IoT) offers many sophisticated andubiquitous applications for smart cities. The energy demand ofIoT applications is increased, while the IoT devices continue togrow in both numbers and their requirements. Therefore, smartcity solutions must have the ability to efficiently utilize energy andhandle associated challenges. Energy management is consideredas a key paradigm for the realization of complex energy systemsin smart cities. In this article, we present a brief overview ofenergy management and challenges in smart cities. We thenprovide a unifying framework for energy-efficient optimizationand scheduling of IoT-based smart cities. We also discuss theenergy harvesting in smart cities, which is a promising solutionfor extending the lifetime of low power devices and its relatedchallenges. We detail two case studies, the first one targets energy-efficient scheduling in smart homes and the second covers wirelesspower transfer for IoT devices in smart cities. Simulation resultsfor case studies demonstrate the tremendous impact of energy-efficient scheduling optimization and wireless power transfer onthe performance of IoT in smart cities.

I. INTRODUCTION

The smart city solutions use communication and networkingtechnologies for dealing with the problems precipitated byurbanization and growing population. Internet of Things (IoT)is a key enabler for smart cities, in which sensing devices andactuators are major components along with communicationand network devices. The sensing devices are used for real-time detection and monitoring of city operations in variousscenarios. It is projected that in the near future, common in-dustrial, personal, office and household devices, machines, andobjects will hold the ability to sense, communicate, and pro-cess information ubiquitously [1]. However, it is challenging todesign a fully optimized framework due to the interconnectednature of the smart cities with different technologies. Further,smart city solutions have to be energy efficient from both usersand environment point of views.

W. Ejaz and A. Anpalagan are with the Department of Electrical andComputer Engineering, Ryerson University, Toronto, ON M5B 2K3, Canada(e-mail: waleed.ejaz@ieee.org; alagan@ee.ryerson.ca).

M. Naeem is with the COMSATS Institute of Information Technology-Wah, Wah Cantonment 47040, Pakistan, and also with the Department ofElectrical and Computer Engineering, Ryerson University, Toronto, ON M5B2K3, Canada (e-mail: muhammadnaeem@gmail.com).

A. Shahid is with the Department of Information Technology Internet BasedCommunication Networks and Services (IBCN) Ghent University a iMindsiGent Tower, Technologiepark Zwijnarde 15 (NEW!) 9052 Ghent, Belgium(e-mail: adnan.shahid@intec.ugent.be).

M. Jo is with Department of Computer and Information Science, KoreaUniversity, Sejong Meteropolitan City, Republic of Korea (Correspondingauthor) (e-mail: minhojo@korea.ac.kr).

This research was supported by a Korea University Grant.

Energy Management

Energy-efficient Solutions

Energy Harvesting

Lightweight

Protocols

Scheduling

Optimization

Low Power

Transceivers

Ambient Energy

Harvesting

Dedicated Energy

Harvesting

Energy

Arrival Rate

Cloud-based

Approach

Energy Harvesting

Receiver Design

Predictive Models for

Energy Consumption

Placement and Scheduling of

Dedicated Energy Sources

Multi-path

Energy Routing

Cognitive Management

Framework

Fig. 1. Classification of energy management for IoT in smart cities.

These challenges forced network designers to considera wide range of scenarios in different conditions for IoT-enabled smart cities. Thus, an efficient deployment of sensorsand optimized operational framework that can adapt to theconditions is necessary for IoT-enabled smart cities. In otherwords, smart city solutions have to be energy efficient, costefficient, reliable, secure, etc. For example, IoT devices shouldoperate in a self-sufficient way without compromising thequality of service (QoS) in order to enhance the performancewith uninterrupted network operations [2]. Therefore, energy-efficiency and life span of IoT devices are the key to the nextgeneration smart city solutions.

We classify the energy management in smart cities intotwo main types: 1) energy-efficient solutions and 2) energyharvesting operations. This classification along with few ex-amples of research topics is shown in Fig. 1. Energy-efficientsolutions for IoT-enabled smart cities include a wide rangeof topics such as lightweight protocols, scheduling optimiza-tion, predictive models for energy consumption, cloud-basedapproach, low power transceivers, and cognitive managementframework [3]–[5]. Energy harvesting allows IoT devices toharvest energy from ambient sources and/or dedicated RFsources. The aim of energy harvesting is to increase thelifetime of IoT devices. The research topics include withinboth types of energy harvesting are energy harvesting receiverdesign, energy arrival rate, placement of a minimum numberof dedicated energy sources, scheduling of dedicated energysources, and multi-path energy routing [2], [6].

2

Smart meter and display

Smart appliances

Heating and cooling control

Aggregator

Utility pricing/ consumer energy selling signal

Power distribution company

Dedicated energy

transmitter

Controller for wireless power transfer

Power line

Communication Access

Smart Parking

Smart Transportation

Fig. 2. An illustration of smart cities with focus on smart homes.

Both academia and industry are focusing on energy man-agement in smart cities. IEEE in partnership with InternationalTelecommunication Union (ITU) has smart cities communitywith an aim to provide assistance to municipalities for thetransition to smart cities. Fujitsu suggested an approach for en-ergy management for companies and has introduced an energymanagement system for smart buildings as cloud service [7].In addition, companies such as IBM, Cisco, Honeywell, Intel,and Schneider Electric are involved in various energy efficientsolutions for smart cities. There are various projects on energyefficient smart cities sponsored by the seventh frameworkprogramme (FP7) for research of the European Commissionin last few years. For example, the main objectives of “Reli-able, Resilient, and Secure IoT for Smart City Applications”project are to develop, evaluate, and test a framework ofIoT-enabled smart city applications in which smart objectscan operate energy efficiently [8]. The “ALMANAC: ReliableSmart Secure Internet of Things For Smart Cities” projectfocuses on IoT-enabled green and sustainable smart solutions[9]. Likewise, energy-saving solutions are developed for smartcities under the projects entitled “Planning for Energy EfficientCities (PLEEC)” and “NiCE-Networking Intelligent Cities forEnergy Efficiency”.

In this article, we consider energy management for IoT insmart cities. An illustration of smart cities with the focus onsmart homes is shown in Fig. 2. Our contributions can besummarized as follows:

• We provide an optimization framework for research inIoT-enabled smart cities. We present the objectives, prob-lem type, and solution approaches for energy manage-ment.

• We cover the energy-efficient solutions for IoT-enabled

smart cities. A case study is presented to show theperformance gains achieved by scheduling optimizationin smart home networks.

• Next, we devote a section to energy harvesting for IoT-enabled smart city applications. A case study is providedto investigate the performance gains achieved by thescheduling of dedicated energy sources.

• Finally, the conclusions are drawn and we provide fu-ture research directions for energy management in IoT-enabled smart cities.

II. ENERGY MANAGEMENT AND CHALLENGES FORSMART CITY APPLICATIONS

An urgent need for energy management has emerged allover the globe due to a continuous increase in the consumptiondemands. Global warming and the air pollution are seriousthreats to the future generation. This is caused by the emissionof fumes whose volume is increased with the increase inenergy demand. On the other hand, according to the statisticsprovided by Cisco, there will be more than 50 billion IoTdevices connected to the internet by 2020 [10]. This explosionin the devices will pose serious energy consumption concerns,and thus, it is imperative to manage energy for the IoT devicesso that the concept of smart cities can better be realized in asustained manner. Following are few examples where we canreduce energy consumption by effective management.

• Home Appliances: Home appliances are the majorsources of energy consumption. Demand management isa key for customizing the energy use by managing thelighting, cooling, and heating systems within the resi-dential units. On the other hand, the intelligent operation

3

Optimization framework for IoT in

smart cities

Cost

Interference

Non-linearEnergy

efficiency

Objective

Objectives AssignmentSchedulingSelection

Mixed integer programming

Type

Branch and bound, dynamic programming, Decomposition methods, outer approximation,

bender decomposition

Solution

ObjectiveType

Newton type algorithm, iterative

algorithms

Solution

Integer programming

TypeHeuristic, Meta heuristic,

branch and bound, dynamic programming

Solution

Cost and Interference

Selection, scheduling, assignment

Linear Simplex, interior point method

Objective

Type Solution

Fig. 3. A typical optimization framework for IoT in smart cities.

of activities can also facilitate the optimized managementand operation of energy.

• Education and Healthcare: Considering the importanceof educational and healthcare services, it is difficult todematerialize them. However, it is possible to demobilizeservices so as for the reduction of energy consumption.For example, exploiting remote healthcare by visualizingsensors and mobile phones, distant learning education cancreate a significant reduction in the energy consumption.

• Transportation: The energy use for transportation in-cludes public transport, daily commute to work on per-sonal vehicles, leisure travel, etc. In addition to the energyconsumed by public transport and personal vehicles, theyare also a major cause of pollution in cities. IoT-enabledsolutions can be employed for energy management, suchas traffic management, congestion control, and smartparking. This can significantly reduce the energy con-sumption as well as CO2 emission.

• Food Industry: The energy consumption in the foodindustry is related to not only the storage, purchase,and preparation of food, it also includes the visitorsmoving into restaurants in search of food. IoT-enabledsolutions can be used here for making the optimizedchoices in terms of food availability. On the other hand,the transportation of the food can also be optimized byincorporating the intelligent means of the transportation.

The IoT devices are generally battery operated and havelimited storage space. Concerning these fundamental limita-tions of sensors, it is difficult to realize the IoT solutionswith prolonged network life. In order to efficiently utilizethe limited sensor resources, an optimized energy efficientframework is of paramount importance which not only reducesthe energy consumption, but also maintains the minimum QoSfor the concerned applications.

A typical optimization framework for IoT-enabled smartcities is given in Fig. 3. This framework provides details ofthe objectives, problem type, and corresponding optimizationtechniques for the energy management. For example, an opti-mization problem for minimizing the cost of electricity usageis presented in [11]. The authors developed the optimization-based residential energy management scheme for the energymanagement of the appliances. Authors in [12] presented anoptimization framework for smart home scheduling of variousappliances and assignment of energy resources. This results ina mixed integer combinatorial problem which is transformedinto a standard convex programming problem. The goal of thisstudy is to minimize the cost and user dissatisfaction. In [13],authors presented an energy-centered and QoS-aware servicesselection algorithm for IoT environments. The objective is tominimize energy consumption while satisfying QoS require-ments. Similarly, objectives shown in Fig. 3 can be consideredand the framework can be used as a guideline to solve the

4

optimization problems.

III. ENERGY-EFFICIENT SOLUTIONS FOR SMART CITIES

With the increase in IoT applications for smart cities, energyefficient solutions are also evolving for low power devices.There are some energy-efficient solutions which can eitherreduce the energy consumption or optimize the resource uti-lization. Following are some main research trends for energy-efficient solutions of IoT-enabled smart cities.

• Lightweight Protocols: Light weight protocol means theprotocol that offers less overhead. IoT-enabled smartcities have to use various protocols for communication.There are several existing protocols in literature suchas Message Queue Telemetry Transport (MQTT), Con-strained Application Protocol (CoAP), Extensible Mes-saging and Presence Protocol (XMPP), Advanced Mes-sage Queue Protocol (AMQP), 6lowPAN, and Universalplug and play (UPnP) IoT. MQTT and CoAP are themost popular protocols. MQTT is a lightweight protocolwhich collects data from IoT devices and transmits tothe servers. Whereas CoAP is designed for constraineddevices and network for web transfer (See [14] for IoTprotocols). Each of these protocols is designed for spe-cific scenarios and applications in which it performs well.In addition, protocol conversion is an important buildingblock for IoT which may require when the IoT devices arefrom different manufacturers or using different protocols.

• Scheduling Optimization: Scheduling optimization forIoT-enabled smart cities refers to the optimization of theresources with the concern of minimizing the energyconsumption and subsequently reducing the electricityusage. In this regard, demand-side management (DSM) isof prime importance which refers to the manipulation ofresidential electricity usage by altering the system loadshape and consequently reducing the cost. Broadly speak-ing, DSM comprises of two main tasks: load shiftingand energy conservation where load shifting refers to thetransfer of customers load from high-peak to low-peaklevels and by adopting this, electricity can be conservedand provide room for other customers.

• Predictive Models for Energy Consumption: Predictivemodels for energy consumptions in IoT-enabled in smartcities are indeed of vital importance. It refers to thewide range of applications in smart cities includingpredictive models for traffic and travel, predictive modelsfor controlling temperature and humidity, etc. Variousprediction models such as neural networks and Markovdecision processes can be incorporated here. The concernof exploiting the predictive models will not only reducethe significant energy consumption but also lead to manysocietal benefits.

• Cloud-based Approach: The cloud computing has re-shaped the computing and storage services which can beused to provide energy-efficient solutions for IoT-enabledsmart cities. More precisely, the cloud-based approachhelps in managing the massive data center flexibility andin a more energy efficient manner.

• Low Power Transceivers: Since the IoT devices in smartcity application operates on the limited battery, low powerdesign architecture or operation framework is of superiorimportance for addressing the energy management inIoT-enabled smart cities. Mostly, the existing applicationprotocols for IoT devices are not in accordance with theenergy-efficiency perspective. More specifically, the radioduty cycle for IoT devices is an important factor in theenergy efficiency and, researchers are exploring methodsfor reducing the radio duty cycle of IoT devices andsubsequently to achieve the energy-efficient architecture.

• Cognitive Management Framework: The IoT devices areheterogeneous in nature and associated services are unre-liable. Therefore, it is important to investigate a cognitivemanagement framework which adopts intelligence andcognitive approaches throughout the IoT-enabled smartcities. The framework should include reasoning and learn-ing in order to improve decisions for IoT networks.A context aware cognitive management framework ispresented in [4], which made decisions regarding IoTdevices (when, why, and how to connect) according tothe contextual background.

A. Case Study on Smart Home Networks

Smart home networks enable home owners to use energyefficiently by scheduling and managing appliances. In addition,to reduce electricity bills, smart home networks offer betterlife style, customized day to day schedule, etc. The smartgrid has provided the ability to keep the electricity demandin line with the supply during the peak time of usage. This iscalled demand-side management. DSM reduces the electricitycost by altering/shifting the system load [5]. Generally DSMis responsible for the demand response program and loadshifting. In demand response program, the customer’s loadcan be reduced in peak hours by shifting it to off-peak hours.This helps to provide more electricity for less cost.

The home appliances are becoming smart with added fea-tures of connectivity which enable the consumers to takeadvantage of the demand response program. The utility cancontact with consumers to reduce/shift their electricity con-sumption, in return for certain monetary benefits. In smarthome networks, appliance load can be further categorizedinto manageable and non-manageable loads. Here, we focuson the energy management of manageable appliance loadin smart homes since it has high energy consumption andpredictability in operations. The manageable load is furtherdivided into shiftable load e.g., washing machine, dishwasher,etc., interrupt-able load, e.g., water heater and refrigerator, andweather-based load, e.g., heating and cooling. An illustrationof the smart home network model for appliance scheduling isgiven in Fig. 2.

We consider a smart home network in which NA is the setof load types, An is the set of appliances in the n-th load typeand A is the set that is a union of all appliances. We defineT , Ct and P t

na as number of time slots in a day, tariff/cost indollars in the time slot t and P t

na power of the n-th load type’sa-th appliance in the time slot t, respectively. We formulate

5

a problem for scheduling of smart home appliances whileconsidering the tariffs and peak load. The overall objectiveis to schedule the appliances in a way such that total costis minimum, i.e., minimize the xtnaC

tP tna for whole set of

NA, A for all T time slots, where xtna is a binary variablewith value 1 when n-th load type’s a-th appliance in the timeslot t is on; otherwise 0. We consider practical constraintson time occupancy and time consecutiveness that need tosatisfy for realistic execution of appliance scheduling. Theconstraints ensure that each appliance should not occupy timeslots more than the required, and the time slots for shift-able loads are consecutive. The optimization problem hereis integer programming, such problems are generally NP-hard and require very efficient algorithms. We solved theoptimization problem using an efficient heuristic algorithm.

B. Performance Analysis

For illustration purpose, we consider only four types ofappliances namely washing machine, dryer, dishwasher, andelectric vehicle. Fig. 4 (a) shows the tariff, slot time forappliances with (thick slots) and without DSM (thin slots). Itis considered that dryer cannot be activated before washingmachine. It is evident that with DSM the appliances areactivated when the tariff is low. However, in the case of withoutDSM, there is no scheduling for appliances and they canbe activated at any time. For instance, all the appliances arescheduled at the time when the tariff is low in case of DSM.In contrast, only dryer is activated when the tariff is low in theabsence of DSM. Similarly Fig. 4 (b) shows that the total loadis less in the case of optimum energy management when thetariff is high. It is important to notice that at some times thetotal load for both optimum energy management and withoutenergy management is same. This is because of the fact thereis no shiftable load at this time.

IV. ENERGY HARVESTING IN SMART CITIES

Energy harvesting is considered as a potential solution toincrease the lifetime of IoT devices in smart cities. Energyharvesting can generally be classified into two categories:

• In ambient energy harvesting, IoT devices harvest energyfrom ambient sources such as wind, RF signals in envi-ronment, vibration, and solar. However, harvesting fromambient sources depends on their availability which isnot always guaranteed.

• In dedicated energy harvesting, the energy sources areintentionally deployed in surroundings of IoT devices.

The amount of energy harvested by each IoT device dependson the sensitivity of harvesting circuit, the distance betweenIoT device and energy source, environment, etc. Thus, thesuccess of energy harvesting for IoT devices in smart citieshas to face several challenges that are discussed below.

• Energy Harvesting Receiver Design: The harvesting cir-cuit design is the primary issue in RF-based energy har-vesting. The sensitivity required for the harvesting circuitis higher than the traditional receivers, which can resultin fluctuations in energy transfer due to the environment

and mobility (energy source and IoT devices). Therefore,efficient and reliable harvesting circuit design is requiredto maximize the harvested energy. In addition, RF-to-DCconversion is the fundamental ingredient of RF energyharvesting. Therefore, circuit designers should enhancethe efficiency of RF-to-DC conversion using advancedtechnologies.

• Energy Arrival Rate: The level of uncertainty of energyarrival rate is higher in energy harvesting from ambientsources than dedicated energy harvesting. This is becausethe former uses renewable energy sources, whereas thelatter uses dedicated energy sources for which the loca-tion is set by network designers based on the harvestingrequirements of IoT devices. The accurate and detailedmodeling of energy arrival rate is indispensable in orderto analyze the performance of energy harvesting systemsin smart cities.

• Placement of Minimum Number of Dedicated EnergySources: The IoT devices which are spatially distant fromthe energy sources can result in uneven energy harvesting.This can result in energy depletion of devices which arefar from the dedicated energy sources and thus reduce thelifetime of network. We can not do much in the case ofambient energy sources; however, optimal placement andnumber of dedicated energy sources are crucial issues indedicated energy harvesting.

• Scheduling of Energy Transmitters: Energy consumed bydedicated energy sources can be reduced by introducingtask based energy harvesting, where energy transmitterscan be scheduled for RF power transfer based on theharvesting requirements of the IoT devices. This requiresa certain level of coverage and sufficient time to har-vest. Therefore, scheduling of energy transmitters withguaranteed coverage and duration is vital for the energyefficiency of dedicated energy harvesting.

• Multi-path Energy Routing: Multi-path energy routingcollects the scattered RF energy from different sourceswith the help of RF energy routers. Then these energyrouters can transfer energy via an alternative path to IoTdevices. Multi-path energy routing is based on the ideaof multi-hop energy transfer in which relay nodes aredeployed near to the IoT devices. This will help to reducepath loss between the relay node and the IoT devices, andalso improve the RF-to-DC conversion efficiency.

A. Case Study: Scheduling of Energy Sources in DedicatedEnergy Harvesting for IoT devices

We consider a network in smart cities with dedicated RFenergy transmitters that consists of NI number of IoT devices(each device is equipped with harvesting circuit) and NE

number of energy transmitters as shown in Fig. 5. It is assumedthat energy transmitters have continuous power supply andthey can satisfy requirements of all IoT devices in the area.The IoT devices can request for power transfer to harvestingcontroller which is considered as a task k. The harvestingcontroller is considered as cloudlet controller which is acentralized resource pool with information about the location

6

Time in minutes200 400 600 800 1000 1200 1400

With DSM WMWith DSM DryerWith DSM DWWith DSM EVWithout DSM WMWithout DSM DryerWithout DSM DWWithout DSM EV

Tariff

Start time of washing machine with DSM

End time of washing machined with DSM

(a)

Time in minutes0 500 1000 1500

Tot

al L

oad

0

200

400

Tar

iff P

rice

(US

D)

0

0.005

0.01

Optimum Energy ManagmentNo Energy ManagementTariff

(b)

Fig. 4. Load pattern (a) appliances starting and ending times with and without DSM, and (b) load pattern of appliances while minimizing total electricitycost and tariff.

ControllerEnergy transmitter

IoT device

Residual energy≤ ξTh

Residual energy

Po

wer

re

que

st

AC

K

Energy transmitters activation optimization

Energy transmitter confirmation

Gra

nt

Fig. 5. A mechanism for scheduling of energy transmitters.

of IoT devices and energy transmitters. The controller canassign multiple tasks from K (K is set of tasks) to the energytransmitters. The transmit power of e-th energy transmitter isdenoted by Pe. The energy transmitter e can transfer power toa task k ∈ K if the requesting IoT device is in the harvestingrange of e. The harvesting range is denoted by φet which is 1if the task k is in the harvesting range of e and 0 otherwise. Letthe energy consumption by e-th energy transmitter in activemode be ξe,A and in sleep mode ξe,S .

We propose a scheduling scheme for energy transmittersin dedicated energy harvesting for IoT devices as shown inFig. 5. IoT devices request the controller for power transferby sending a request if their residual energy is less than apre-set threshold ξTh. The threshold is set while consideringthat the node has sufficient energy for critical operations. Therequest packet contains the requesting node’s identity (ID),controller’s ID, and energy harvesting requirements. Here, weadopt RF-MAC protocol proposed in [15]. The sensor nodewith residual energy less than a pre-set threshold can send

RFP for instant charging through access priority mechanism(for details about this mechanism, see [15]) which ensuresthat the node with residual energy ≤ ξTh gets channel accessbefore data transmission by other sensor nodes. The nodeswhich have data to transmit are forced to freeze their back-offtimers as data transmission is not possible at this time. Thecontroller receives this packet and processes it to activate theenergy transmitter(s). The harvesting controller receives thisrequest for task k and calculate φet for all energy transmitters.An energy transmitter can be activated for harvesting the targetIoT device(s) if and only if 1) task k is within the harvestingrange of e, i.e., φet = 1, and 2) task k is scheduled/ activatedon e. We define a binary variable ψe which is 1 if the energytransmitter e is scheduled / activated and 0 otherwise.

The objective here is to activate minimum number of energytransmitters to minimize the energy consumed by dedicatedenergy transmitters, i.e., ψeξe,A+(1−ψe)ξe,S . This is subjectto constraints on coverage φet, duration of energy harvestingδe, and target harvesting energy EC . One way to get an optimalsolution is to enumerate over all possible combinations ofψe, which is computationally expensive and unrealistic for alarge number of energy transmitters and tasks. Therefore, weconsider branch and bound algorithm for the scheduling ofdedicated RF energy sources. Once the activation of energysources is optimized at the controller, then a grant for powertransfer packet is sent to the energy transmitters which areselected for RF power transfer. Finally, the energy source(s)send the acknowledgment packet to IoT device(s) whichrequested the power transfer. This packet has the informationof central frequency of energy transmitter and the duration ofenergy charging.

1) Performance Analysis: We evaluate the performance ofenergy-efficient scheduling of energy transmitters. We con-sider omnidirectional energy transmitters which radiate waveswith power 46 dBm. The proposed schemes can be modifiedto use with directional energy transmitters to overcome path

7

5 10 150

500

1000

1500

2000

2500

Number of tasks (K)

Ene

rgy

cons

umed

(J)

Exhaustive search, NE=10

Exhaustive search, NE=20

Branch and bound algorithm, NE=10

Branch and bound algorithm, NE=20

Traditional WSNs, NE=10

Traditional WSNs, NE=20

(a)

10 15 20 25 300

500

1000

1500

2000

2500

3000

3500

4000

Number of energy transmitters (NE)

Ene

rgy

cons

umed

(J)

Exhaustive search, K=5Exhaustive search, K=15Branch and bound algorithm, K=5Branch and bound algorithm, K=15Traditional WSNs, K=5Traditional WSNs, K=15

(b)

Fig. 6. Impact of (a) number of tasks K on the energy consumption and (b)energy transmitters (NE ) on the energy consumption for different number oftasks K for different number of energy transmitters (NE ).

losses which can certainly help to improve the chargingefficiency. The transmit and receive energy for IoT devicesare considered from MICA2 specifications. We consider NI =200 IoT devices, which are randomly distributed in a rectan-gular field of 100m× 100 m.

Figs. 6 (a) and 6 (b) illustrate the impact of a numberof tasks and energy transmitters on energy consumption,respectively for energy-efficient scheduling scheme (branchand bound, and exhaustive search, and traditional WSN. Fig. 6(a)shows that the energy consumption is increased slowly withthe increase in a number of tasks in case of energy-efficientscheduling scheme (for given number of energy transmitters,i.e., NE = 10 and 20). This is because of the fact thatenergy transmitters are activated based on the number oftasks and their location instead of a total number of energytransmitters. We may need a different number of active energytransmitters if requesting devices are far or close from each

other. The energy consumption in case of traditional WSNsis constant regardless of the number of tasks, i.e., all theenergy transmitters are activated all the time. Thus the energyconsumption is doubled when NE = 20 compared to the casewhen NE = 10. The energy consumption in the proposedscheme is reduced at the cost of overhead and delay causeddue to the exchange of packets among IoT devices, controller,and energy transmitters. From Fig. 6 (b), it can be noticedthat the energy consumption for efficient scheduling schemesis not much affected by the increase on energy transmittersNE for given number of tasks (K = 5 and K = 15). Weconsider a small network size for which the probability thattasks are spatially nearby is high. Thus, for different numberof tasks, we may need to activate the same number of energytransmitters based on their location. Therefore, curves aresuperimposed. In contrast, traditional WSNs in IoT-enabledsmart cities activate all the energy transmitters regardlessof the number of tasks which result in a linear increasein energy consumption. Moreover, results of the branch andbound algorithm are very similar to exhaustive search casewith less complexity.

V. CONCLUSIONS AND FUTURE WORK

Energy management in smart cities is an indispensablechallenge to address due to the rapid urbanization. In thisarticle, we first presented an overview of energy managementin smart cities, and then presented a unifying framework forIoT in smart cities. Energy management has been classifiedinto two levels: energy-efficient solutions and energy har-vesting operations. We covered various directions to inves-tigate energy-efficient solutions and energy harvesting for IoTdevices in smart cities. Furthermore, two case studies havebeen presented to illustrate the significance of energy man-agement. The first case study presented appliance schedulingoptimization in smart home networks where the objectivewas to reduce the electricity cost. The second case studycovered efficient scheduling of dedicated energy sources forIoT devices in smart cities. Simulation results were presentedto show the advantage of energy management in IoT for smartcities. Possible future directions for energy management insmart cities are 1) energy-efficient mechanisms for software-defined IoT solutions which can provide scalable and context-aware data and services, 2) directional energy transmissionfrom dedicated energy sources for wireless power transfer,3) energy efficiency and complexity of security protocols arecrucial aspects for their practical implementation in IoT; thus,it is important to investigate robust security protocols forenergy constraint IoT devices, 4) fog computing can lead toenergy saving for most of the IoT applications, therefore, it isimportant to study energy consumption of fog devices for IoTapplications.

REFERENCES

[1] A. El-Mougy, M. Ibnkahla, G. Hattab, and W. Ejaz, “Reconfigurablewireless networks,” Proceedings of the IEEE, vol. 103, no. 7, pp. 1125–1158, Jul. 2015.

[2] P. Kamalinejad, C. Mahapatra, Z. Sheng, S. Mirabbasi, V. Leung, andY. L. Guan, “Wireless energy harvesting for the Internet of Things,”IEEE Communications Magazine, vol. 53, no. 6, pp. 102–108, Jun. 2015.

8

[3] W. Ejaz, G. A. Shah, H. S. Kim et al., “Energy and throughput effi-cient cooperative spectrum sensing in cognitive radio sensor networks,”Transactions on Emerging Telecommunications Technologies, vol. 26,no. 7, pp. 1019–1030, Jul. 2015.

[4] P. Vlacheas, R. Giaffreda, V. Stavroulaki, D. Kelaidonis, V. Foteinos,G. Poulios, P. Demestichas, A. Somov, A. R. Biswas, and K. Moessner,“Enabling smart cities through a cognitive management framework forthe Internet of Things,” IEEE Communications Magazine, vol. 51, no. 6,pp. 102–111, Jun. 2013.

[5] F. Qayyum, M. Naeem, A. Khwaja, A. Anpalagan, L. Guan, andB. Venkatesh, “Appliance scheduling optimization in smart home net-works,” IEEE Access, vol. 3, pp. 2176–2190, Nov. 2015.

[6] D. Mishra, S. De, S. Jana, S. Basagni, K. Chowdhury, and W. Heinzel-man, “Smart RF energy harvesting communications: challenges andopportunities,” IEEE Communications Magazine, vol. 53, no. 4, pp. 70–78, Apr. 2015.

[7] T. Ozawa, S. Kawaguchi, T. Nishijima et al., “Smart cities and energymanagement,” Fujistu Scientific and Technical Jornal, vol. 50, no. 2, pp.49–57, Apr. 2015.

[8] H. C. Pohls, V. Angelakis, S. Suppan, K. Fischer, G. Oikonomou,E. Z. Tragos, R. Diaz Rodriguez, and T. Mouroutis, “RERUM: Buildinga reliable IoT upon privacy-and security-enabled smart objects,” inIEEE Wireless Communications and Networking Conference Workshops(WCNCW). Istanbul, Turkey: IEEE, Apr. 2014, pp. 122–127.

[9] D. Bonino, M. T. D. Alizo, A. Alapetite, T. Gilbert, M. Axling,H. Udsen, J. A. C. Soto, and M. Spirito, “ALMANAC: Internet ofThings for Smart Cities,” in International Conference on Future Internetof Things and Cloud (FiCloud), Rome, Italy, Aug. 2015, pp. 309–316.

[10] D. Evans, “The Internet of Things: How the next evolution of Internetis changing everything,” Cisco, Tech. Rep., Apr. 2011.

[11] M. Erol-Kantarci and H. T. Mouftah, “Wireless sensor networks forcost-efficient residential energy management in the smart grid,” IEEETransactions on Smart Grid, vol. 2, no. 2, pp. 314–325, Jun. 2011.

[12] K. M. Tsui and S.-C. Chan, “Demand response optimization for smarthome scheduling under real-time pricing,” IEEE Transactions on SmartGrid, vol. 3, no. 4, pp. 1812–1821, Dec. 2012.

[13] M. E. Khanouche, Y. Amirat, A. Chibani, M. Kerkar, and A. Yachir,“Energy-Centered and QoS-Aware Services Selection for Internet ofThings,” IEEE Transactions on Automation Science and Engineering,vol. 13, no. 3, pp. 1256–1269, Jul. 2016.

[14] A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, andM. Ayyash, “Internet of things: A survey on enabling technologies,protocols, and applications,” IEEE Communications Surveys & Tutorials,vol. 17, no. 4, pp. 2347–2376, Nov. 2015.

[15] M. Y. Naderi, P. Nintanavongsa, and K. R. Chowdhury, “RF-MAC:A medium access control protocol for re-chargeable sensor networkspowered by wireless energy harvesting,” IEEE Transactions on WirelessCommunications, vol. 13, no. 7, pp. 3926–3937, Jul. 2014.

WALEED EJAZ (waleed.ejaz@ieee.org)[S’12, M’14,SM’16] is a Senior Research Associate at the Departmentof Electrical and Computer Engineering, Ryerson University,Toronto, Canada. Prior to this, he was a Post-doctoral fellow atQueen’s University, Kingston, Canada. He received his Ph.D.degree in Information and Communication Engineering fromSejong University, Republic of Korea in 2014. He earnedhis M.Sc. and B.Sc. degrees in Computer Engineering fromNational University of Sciences & Technology, Islamabad,Pakistan and University of Engineering & Technology, Taxila,Pakistan, respectively. His current research interests includeInternet of Things (IoT), energy harvesting, 5G cellular net-works, and mobile cloud computing.

MUHAMMAD NAEEM (muhammadnaeem@gmail.com)[SM’16] received the BS (2000) and MS (2005) degrees inElectrical Engineering from the University of Engineering andTechnology, Taxila, Pakistan. He received his PhD degree(2011) from Simon Fraser University, BC, Canada. From2012 to 2013, he was a Postdoctoral Research Associate withWINCORE Lab. at Ryerson University, Toronto, ON, Canada.Since August 2013, he has been an assistant professor with

the Department of Electrical Engineering, COMSATS Instituteof IT, Wah Campus, Pakistan and Research Associate withWINCORE Lab. at Ryerson University. His research interestsinclude optimization of wireless communication systems, non-convex optimization, resource allocation in cognitive radionetworks and approximation algorithms for mixed integerprogramming in communication systems.

ADNAN SHAHID (adnan.shahid@intec.ugent.be) receivedthe B.Eng. and the M.Eng. degrees in computer engineeringwith communication specialization from the University ofEngineering and Technology, Taxila, Pakistan in 2006 and2010, respectively, and the Ph.D degree in information andcommunication engineering from the Sejong University, SouthKorea in 2015. He is currently working as a PostdoctoralResearcher at iMinds/IBCN, Department of Information Tech-nology, University of Ghent, Belgium. From Sep 2015 a Jun2016, he was with the Department of Computer Engineering,Taif University, Saudi Arabia. From Mar 2015 a Aug 2015, heworked as a Postdoc Researcher at Yonsei University, SouthKorea. He was also the recipient of the prestigious BK 21plus Postdoc program at Yonsei University, South Korea. Hisresearch interests include device to device communication,self-organizing networks, small cells networks, cross-layeroptimization, and integration of WiFi and small cells.

ALAGAN ANPALAGAN (alagan@ee.ryerson.ca) [SM] re-ceived his B.A.Sc. M.A.Sc., and Ph.D. degrees in electricalengineering from the University of Toronto, Canada. He joinedthe Department of Electrical and Computer Engineering atRyerson University in 2001 and was promoted to full professorin 2010. He served the department as graduate program direc-tor (2004a2009) and interim electrical engineering programdirector (2009a2010). During his sabbatical (2010a2011), hewas a visiting professor at the Asian Institute of Technologyand a visiting researcher at Kyoto University. His industrial ex-perience includes working at Bell Mobility, Nortel Networks,and IBM Canada. He directs a research group working onradio resource management and radio access and networkingwithin the WINCORE Lab. His current research interestsinclude cognitive radio resource allocation and management,wireless cross-layer design and optimization, cooperative com-munication, M2M communication, small cell networks, energyharvesting, and green communications technologies. He servesas Associate Editor for IEEE Communications Surveys Tuto-rials (2012a) and Springer Wireless Personal Communications(2009a), and a past Editor for IEEE Communications Letters(2010a2013) and EURASIP Journal of Wireless Communica-tions and Networking (2004a2009). He also served as GuestEditor for two EURASIP Special Issues on Radio ResourceManagement in 3G+ Systems (2006) and Fairness in RadioResource Management for Wireless Networks (2008), and aMONET Special Issue on Green Cognitive and CooperativeCommunication and Networking (2012). He co-authored oredited three books: Design and Deployment of Small CellNetworks (Cambridge University Press, 2014), Routing inOpportunistic Networks (Springer, 2013), and Handbook onGreen Information and Communication Systems (AcademicPress, 2012). He is a registered Professional Engineer in theProvince of Ontario, Canada.

9

MINHO JO [M’07] (minhojo@korea.ac.kr) is a professorin the Department of Computer and Information Science,Korea University, Sejong Metropolitan City. He received hisPh.D in industrial and systems engineering from Lehigh Uni-versity in 1994 and his B.A. in industrial engineering fromChosun University, South Korea, in 1984. He was one of thefounding members and Founding Captain of the IT Team, LCDDivision, Samsung Electronics. He has extensive experiencein wireless communications and software development inindustry for more than 15 years. He is Founding Editor-in-Chief of KSII Transactions on Internet and InformationSystems. He serves as an Editor of IEEE Network. He is anAssociate Editor of Security and Communication Networksand Wireless Communications and Mobile Computing. He iscurrently Vice President of the Institute of Electronics andInformation Engineers (IEIE). His current research interestsinclude Internet of Things (IoT), cognitive radio, mobile cloudcomputing, network security, heterogeneous networks in 5G,cellular networks in 5G, wearable computing, optimization inwireless communications, and unmanned aerial vehicles.

10

Energy Management

Energy-efficient Solutions

Energy Harvesting

Lightweight

Protocols

Scheduling

Optimization

Low Power

Transceivers

Ambient Energy

Harvesting

Dedicated Energy

Harvesting

Energy

Arrival Rate

Cloud-based

Approach

Energy Harvesting

Receiver Design

Predictive Models for

Energy Consumption

Placement and Scheduling of

Dedicated Energy Sources

Multi-path

Energy Routing

Cognitive Management

Framework

Fig. 1. Classification of energy management for IoT in smart cities.

11

Smart meter and display

Smart appliances

Heating and cooling control

Aggregator

Utility pricing/ consumer energy selling signal

Power distribution company

Dedicated energy

transmitter

Controller for wireless power transfer

Power line

Communication Access

Smart Parking

Smart Transportation

Fig. 2. An illustration of smart cities with focus on smart homes.

12

Optimization framework for IoT in

smart cities

Cost

Interference

Non-linearEnergy

efficiency

Objective

Objectives AssignmentSchedulingSelection

Mixed integer programming

Type

Branch and bound, dynamic programming, Decomposition methods, outer approximation,

bender decomposition

Solution

ObjectiveType

Newton type algorithm, iterative

algorithms

Solution

Integer programming

TypeHeuristic, Meta heuristic,

branch and bound, dynamic programming

Solution

Cost and Interference

Selection, scheduling, assignment

Linear Simplex, interior point method

Objective

Type Solution

Fig. 3. A typical optimization framework for IoT in smart cities.

13

Time in minutes200 400 600 800 1000 1200 1400

With DSM WMWith DSM DryerWith DSM DWWith DSM EVWithout DSM WMWithout DSM DryerWithout DSM DWWithout DSM EV

Tariff

Start time of washing machine with DSM

End time of washing machined with DSM

(a)

Time in minutes0 500 1000 1500

Tota

l Loa

d

0

200

400

Tarif

f Pric

e (U

SD

)

0

0.005

0.01

Optimum Energy ManagmentNo Energy ManagementTariff

(b)

Fig. 4. Load pattern (a) appliances starting and ending times with and without DSM, and (b) load pattern of appliances while minimizing total electricitycost and tariff.

14

ControllerEnergy transmitter

IoT device

Residual energy≤ ξTh

Residual energy

Po

wer

re

que

st

AC

K

Energy transmitters activation optimization

Energy transmitter confirmation

Gra

nt

Fig. 5. A mechanism for scheduling of energy transmitters.

15

5 10 150

500

1000

1500

2000

2500

Number of tasks (K)

Ene

rgy

cons

umed

(J)

Exhaustive search, NE=10

Exhaustive search, NE=20

Branch and bound algorithm, NE=10

Branch and bound algorithm, NE=20

Traditional WSNs, NE=10

Traditional WSNs, NE=20

(a)

10 15 20 25 300

500

1000

1500

2000

2500

3000

3500

4000

Number of energy transmitters (NE)

Ene

rgy

cons

umed

(J)

Exhaustive search, K=5Exhaustive search, K=15Branch and bound algorithm, K=5Branch and bound algorithm, K=15Traditional WSNs, K=5Traditional WSNs, K=15

(b)

Fig. 6. Impact of (a) number of tasks K on the energy consumption and (b) energy transmitters (NE ) on the energy consumption for different number oftasks K for different number of energy transmitters (NE ).