Professional Documents
Culture Documents
Digital Object Identifier: Mbazingwa E. Mkiramweni, Chungang Yang, and Jiandong Li are with Xidian University; Chungang Yang is also with CETC Key Laboratory
10.1109/MWC.2017.1700250 of Data Link Technology. Zhu Han is with the University of Houston and also with Kyung Hee University.
Game Theory
Non-cooperative Cooperative
games games
Zero-sum Zero-sum
pursuit-evasion pursuit-evasion
game [5] game [9]
FIGURE 1. Classification of current applied game theoretic approaches in wireless communications with UAVs.
non-cooperative games applied in solving various optimizing UAV energy consumption such as UAV
technical problems in wireless communications circular maneuvering [10]. Such methods and oth-
with UAVs networks. ers need a central authority to control and allocate
UAVs to their optimal locations and paths, which
Applications of Game Theory in increases the energy consumption of UAVs due
to the increase of signal overhead. Game theory
Wireless Communications with UAVs approaches can allow the UAVs to operate auton-
Game-theoretic properties have been used to omously, therefore minimizing power consumption.
solve various types of challenges in wireless com- As an efficient means of power optimization,
munications with UAVs. Non-cooperative games the authors in [9] proposed the use of periodic
have been used to optimize power and energy, beaconing for UAVs acting as aerial base stations
find optimal height and coverage, and perform in a wireless communication system. The authors
coordination control of the UAVs. The cooper- modeled the beaconing period scheduling by using
ative game has been applied for task allocation a non-cooperative sub-modular game in which the
and performance efficiency problems. Here, we UAVs are considered as players who compete to
choose several typical game applications as exam- maximize their coverage probability of the mobiles
ples to characterize rational behaviors, analyze in the area of interest. Each UAV sends a beacon
the equilibrium solutions, and design distributed to the mobiles for a specific duration period. A
schemes in wireless communications with UAVs. game G = {N, {A{i∈N}}, {u{i∈N}} with N being the set
of UAVs, Ai the set of action, and ui the payoff of
Energy Saving and Power Optimization UAV i. The payoff function is formulated, which is
Energy consumption optimization is one of the the difference between the successful encounter
major challenges. With the rapid growth of appli- rate and energy consumption during the beacon-
cations with high data rate such as online video ing period. With the provided learning framework,
streaming and games, more and more energy is UAVs are allowed to reach equilibrium, then the
consumed. The challenge becomes even more existence and uniqueness of the Nash equilibri-
critical when battery-powered UAVs are applied um are checked, and simulations with different
for wireless communications. The operational, encounter rates were done.
performance, and availability duration of a UAV The sub-modular game results show that at the
are restricted by the limited energy on board. Fig- equilibrium point the UAVs efficiently optimize
ure 2 shows different ways by which a UAV can their energy consumption and at the same time
lose energy. UAVs consume power to overcome maximize the possibility of contacting the mobile
its weight so as to stay in the air, as well as for var- user on the ground.
ious kinds of motions and maneuvers. Moreover, The non-cooperative games in [11] and [12],
energy is needed to maintain inter-UAV commu- while designed to find the optimal placement for
nication links such as UAV-UAV, UAV-mobile, maximum coverage of the ground mobiles, allows
UAV-ground control station, and UAV-satellite decisions about radio resources and navigation to
communication links. be made onboard the UAV without reference to
Improvement in the battery life and energy stor- other agents. Non-cooperative games use time-
age technologies, the use of low-power compo- stamped broadcasts from the mobiles and UAVs to
nents and lightweight materials in building drones derive complete information about strategies and
have not been enough to mitigate the UAV energy payoffs. In this way, the UAVs are allowed to oper-
problem. From the operational point of view, the ate with a high degree of autonomy and without
problem can be addressed by introducing energy the need for central planning authority. Therefore,
efficiency operations to reduce UAVs’ unnecessary the power required to be assigned to the inter-
power consumption during operation. In addition, UAV links, UAV-ground control station link, and
researchers have suggested several methods for the associated communication traffic is minimized.
attacker in a future state, and if so, the IES will eject UAV
UAV
this suspected node before the attack. The perfor-
mance of this approach based on the Nash Equilib-
rium concept has shown to incur a low overhead
to detect lethal attacks with high accuracy.
To compute optimal strategies for a team of UAV
UAV
UAVs evading the attack of an aerial jammer on
the communication channel, the authors in [5] UAV
UAV
proposed a differential game theoretic approach.
The model uses three nodes, receiver, transmit-
ter and the attacker that is attempting to jam the
communication channel between the transmitter
and receiver by sending a high power noise at the
same frequency. Two problems are formulated, the
first one by considering that UAVs are initially not Task Shared Task Unallocated Task
communicating in the presence of a jammer. In
this first problem, the UAVs seeks to minimize the FIGURE 5. Different UAVs performing various tasks in a wireless network.
time for which communication remains jammed
while the jammer aims to maximize the time for
which it can jam the communication between of the UAVs’ first encounter with the mobile and
UAVs. The second problem considers a situation the cost involved. The cost is the energy consumed
in which UAVs are communicating in the presence associated with sending a beacon and switching
of a jammer, which aims to minimize the time it between states (i.e., probing/idle states). In this
can jam the communication channel between game, if a player receives a utility value less than
—
UAVs. Meanwhile, each UAV is trying to maximize Ui, will quit the cooperation. Then a two-player
the time for which they maintain the operation of NBS function is formulated as the product of the
the communication link between them. The prob- difference between the defined utilities, that is, (U1
— —
lem was formulated as a zero-sum pursuit-evasion – U1) × (U2 – U2). The optimal beaconing periods
game and the necessary conditions to arrive at the of UAVs that give maximum values of the NBS
equations governing the saddle point strategies of function at various encounter rates are calculated.
the players were derived. The cost function was Finally, we provide simulation results to illustrate
formulated as the termination time of the game. the effectiveness of the proposed NBS.
The game renders a possible way of protecting the The optimal beaconing periods obtained using
communication channel to malicious attacks from NBS for various values of encounter rates at three
aerial intruders flying in the vicinity. different cycle durations can be seen in Fig. 6a. For
different probing/idle cycling periods T, for exam-
A Cooperative Beaconing Scheduling Strategy ple when T = 20, T = 10, and T = 5, the optimal
Based on Nash Bargaining Solution: A Case Study beaconing period durations start at high values
for low values of encounter rates and decrease as
The current applied game theoretic approaches the encounter rate increases. The results demon-
have shown promising potential for solving prob- strate that when a UAV has a low probability of
lems in wireless communications with UAVs [8]. encountering a mobile, it will require longer prob-
To verify the merits of carrying UAV-aided com- ing period durations and vice versa. In addition, at
munications based on game theory, we present a Nash equilibrium, longer beaconing durations will
cooperative beaconing periods scheduling game be assigned to the UAVs with higher probing/idle
between two UAVs. Motivated by the scenario durations compared to the ones with low probing/
presented in [9], where two UAVs acting as aerial idle durations. Figure 6b presents the comparison
base stations provide network coverage to mobile of utilities that a UAV can achieve when applying
users in a geographic area. In this scenario, to three different strategies. The cooperative strat-
encounter a mobile user, the UAVs send beacons egy can achieve higher payoff compared to the
announcing their presence. Two UAVs seek to always-beaconing and non-cooperative strategies.
minimize their energy consumption by adopting For the UAV that uses an always-probing strategy,
a periodical probing strategy. Different from [9], there is a linear trade-off between probing duration
where a sub-modular, non-cooperative game was and energy consumption. As the probing duration
used, we apply cooperative strategy based on the increases, the probability of encountering a mobile
Nash bargaining solution (NBS) to find the opti- and associated energy consumption increases as
mal probing period to maximize UAVs encounter well. However, the UAVs that employ game theory
probability. More details on the NBS are given in can minimize the energy consumption involved
[14]. The NBS strategy can achieve an optimal through maximizing their payoff by obtaining the
solution while maintaining fairness among UAVs. optimal probing period durations. Moreover, it
In this case study, the two-player bargaining can be seen in Fig. 6b, at the optimal solution, the
problem is modeled. We first defined players’ utility NBS strategy receives higher reward compared
Ui, which the UAV achieves when the beaconing to the non-cooperative strategy, which indicates
period duration is allocated to it, and the minimum that the cooperative strategy is more efficient in
—
utility Ui, which a UAV obtains when no agreement saving energy. This use case shows the potential of
is reached in the bargaining process. Utility Ui is game-theoretic approaches in wireless communica-
defined as the difference between the probability tion with UAVs.
(a) (b)
FIGURE 6. a) Beaconing period durations at equilibrium for different values of encounter rates using Nash bargain solution; b) the utili-
ty achieved by the UAV using the cooperative, noncooperative and always-beaconing strategies.
• In large-scale multi-agent networks, the tradition- we will consider a massive UAV-aided wireless In finding a solution
al approaches are impractical. In these games, network in which UAVs communicate with GCS, to the interference
every player should collect the information of other UAVs and provide services to the mobiles. In problem, it is necessary
all other players, which can result in complex this model, we will focus on the interference that a to learn about the
algorithms that are considerably harder to solve UAV communicating with GCS experiences from
when dealing with massive networks. neighboring GCSs. The mean field approximation behavior of the inter-
• As the number of UAVs increases, the size of the method will be used to approach the aggregate ference. We intend to
payoff matrix increases, therefore the time required interference from other links to the UAV-GCS com- learn the characteristics
for the game to reach equilibrium also increases. munication channel and get the corresponding of interference that
Therefore, it becomes necessary to rethink the cost function. communicating UAVs
different analytical models of tackling the prob- UAV-GCS links representing communication pairs
lem in future massive, dynamic and complex UAV are considered as players. They are rational policy- experience from the
wireless networks and move toward models that fit makers whose number is arbitrarily large and even neighboring GCSs,
the characteristics of the network more appropri- goes to infinite. The actions are the possible transmit mobile devices, and
ately. Therefore, we propose to use the concepts powers corresponding to UAV-GCS pairs. Each trans- neighboring communi-
of mean field game (MFG) [15] to study the char- mitter determines the power it transmits at any time cating UAVs.
acteristics of interference and design an interfer- to minimize the cost function. The state space of the
ence-aware management scheme among UAVs in player is defined as the interference introduced by
large wireless communication with UAVs networks. other links to the UAV-GCS link and the interference
from the UAV-GCS link to other links. A control poli-
Mean Field Game for Interference Management in cy is used to minimize the average cost over the time
interval in one-dimensional states. The cost function
Massive UAV Networks for interference management in massive UAV com-
There is a growing interest from various enter- munications will be formulated by considering both
prises and government institutions in using UAVs the achieved signal-to-interference-plus-noise ratio
for different purposes. The existing and expected (SINR) performance and the transmit power of the
uses of UAVs and huge investment in the drone UAV-GCS pair.
industry indicates that a future where hundreds The interaction of players with the mean field is
if not thousands of drones will be operating in modeled by a Hamilton-Jacobi-Bellman (HJB) equa-
the air is near. Massive UAVs will also be used tion, and the movement of the mean field function
as part of the rollout of 5G networks, to enhance according to the players’ actions is defined by a
capacity, increase data rates, minimize latency, Fokker-Planck-Kolmogorov (FPK) equation. The
improve connectivity for a massive number of corresponding HJB and FPK equations are derived
devices, reduce energy and cost, and boost the and solved to obtain the mean field equilibrium
quality of experience (QoE). (MFE). The finite difference method and Lagrange
Therefore, MFG will be an appropriate method relaxation method will be used to solve the FPK
for addressing different challenges in such massive and HJB equations respectively.
UAV networks. This is because in MFG the mean field
value of space-time dynamics of context approach- Other Research and Implementation
es the real value as the number of players increases. Due to the promising development of IoT and
In addition, in the MFG framework, the concept of mobile Internet, there exist novel communication
mean field can characterize the space-time dynamics requirements of ultra-reliable low latency communica-
of context, for example, two-dimensional interference tions (uRLLC), massive machine-type communication
and energy states, which helps the generic player (eMTC), and enhanced mobile broadband (eMBB).
to make an optimal decision responding only to the Meanwhile, UAVs are also expected to be used
mean field, instead of the strategies of all the other to assist vehicular ad-hoc networks (VANETs) and
players. Another unique advantage of the introduced wireless sensor networks (WNS). It is necessary to
novel game is that MFG-based decision-making is study the characteristics and advantages of deploying
distributed with less signaling overhead. UAVs to assist these networks. Novel game-theoretic
By definition, MFGs are a special form of differ- approaches can find wider applications.
ential games, where each player has a state, a set In the future, it is also necessary to design and
of actions, and a control policy. The control policy implement a prototype of the UAV-aided wireless
maps every state to an action over a pre-defined communications network, and then evaluate the
period of time. Instead of modeling each players’ feasibility of game-theoretic approaches. Then, a
interaction with every other player, MFG models an series of experiments will be practically conducted
individual’s interaction with the effect of the collec- and the results can be collected and analyzed to
tive behavior of all the players. Thus, at each step further understand the pros and cons of game-the-
of the game, the mean field is simply the fraction oretic approaches.
of players at every state. In contrast to traditional
games, MFG can be used to model and analyze the Conclusion
interaction between individual and collective behav- Game theory offers a suitable tool that can be
iors of a large number of rational entities [15]. used to effectively model the interaction between
In finding a solution to the interference prob- autonomous wireless communication UAVs.
lem, it is necessary to learn about the behavior of Game theory always leads to sophisticated distrib-
the interference. We intend to learn the charac- uted decision processes. This article summarized
teristics of interference that communicating UAVs the typical advantages and typical challenges of
experience from the neighboring GCSs, mobile UAV communications, where we concentrated
devices, and neighboring communicating UAVs. on the recent applications of game theory for
We will use the MFG to study and develop a dis- wireless communications with UAVs. We present-
tributed interference-aware scheme. In our model, ed a classification of the existing game approach-
In the future, it is also es to various novel communication challenges [14] C. Yang et al., “Point Power Coordination for Spec-
tral-and-Energy Efficiency in Heterogeneous Small Cell
necessary to design with UAVs. For each game-theoretic application, Networks: A Bargaining Game-Theoretic Perspective,” IEEE
and implement a we characterized the rational behavior, analyzed Trans. Wireless Commun., vol. 5, no. 2, 2016, pp. 1364–76.
prototype of the equilibrium solutions, and presented the distrib- [15] C. Yang et al., “Mean Field Game-Theoretic Framework for
uted schemes of the UAVs. Finally, we simulated Interference and Energy-Aware Control in 5G Ultra-Dense
UAV-aided wireless Networks,” IEEE Wireless Commun., vol. 99, Sept. 2017,
a use case example to show the potential bene- pp. 1–8.
communications fits of game theory. We highlighted the interfer-
network, and then ence management challenge, presented the novel
mean field game-theoretic approache, and look
Biographies
evaluate the feasibili- M bazingwa E lirehema M kiramweni received his bache-
ty of game-theoretic forward to the future of game theory for wireless lor degree in electronic information engineering and masters
communications with UAVs. degree in information and telecommunication engineering at
approaches. Then, a the University of Science and Technology Beijing (USTB), Bei-
jing, China, in 2011 and 2014, respectively. He is currently pur-
series of experiments Acknowledgment suing his doctoral degree at Xidian University. He works on the
will be practically con- This work was supported in part by the National GUIDE research team, which is guided by Dr. Chungang Yang.
Science Foundation of China (61871454); by the He is also working with the Dar es Salaam Institute of Technol-
ducted and the results ogy (DIT) in Tanzania. His research interests include resource
can be collected and Fundamental Research Funds for the Central Univer- and interference management and network optimization for 5G
sities(2018); by the Natural Science Basic Research networks and UAV-assisted cellular networks and game theory
analyzed to further Plan in Shaanxi Province of China (2017JZ021); by for wireless communication.
understand the pros a special financial grant from the China Postdoctor-
Chungang Yang [S’09, M’12] is an associate professor at Xid-
and cons of game-the- al Science Foundation (2016T90894); by a special ian University, where he leads the research team of “GUIDE,
oretic approaches. financial grant from the Shaaxi Postdoctoral Science Game, Utility, Intelligent computing Design for Emerging com-
Foundation (154066); by the CETC Key Laborato- munications.” He received his bachelor and doctoral degree at
ry of Data Link Technology (CLDL-20182308); by Xidian University, Xiían, China, in 2006 and 2011, respectively.
Between September 2010 and March 2011, he held the visiting
the ISN02080001 and ISN90106180001; by the scholar position in the department of electrical and computer
111 Project under Grant B08038; and by the engineering at Michigan Technological University. Between
National Science Foundation of China under Grant March 2015 and March 2016, he held the visiting scholar posi-
61671062 and Grant 91638202. tion in the department of electrical and computer engineering at
the University of Houston. He has edited two books: Game The-
ory Framework Applied to Wireless Communication Networks
References (IGI Global, 2016), and Interference Mitigation and Energy Man-
[1] S. Morgenthaler et al., “UAVNet: A Mobile Wireless Mesh agement in 5G Heterogeneous Cellular Networks (IGI Global,
Network Using Unmanned Aerial Vehicles,” IEEE Globecom 2017). His research interests include resource and interference
Workshops, Anaheim, CA, USA, Dec. 2012, pp. 1603–08. management, network optimization, and mechanism design for
[2] Y. Zeng, R. Zhang, and T. J. Lim, “Wireless Communications cognitive radio networks, heterogeneous cellular networks, and
with Unmanned Aerial Vehicles: Opportunities and Chal- game theory for wireless communication and computing net-
lenges,” IEEE Commun. Mag., vol. 54, no. 5, May 2016, pp. works. He serves as an editor for the KSII Transactions on Inter-
36–42. net and Information Systems, and an editor of a special issue
[3] W. Saad et al., “A Selfish Approach to Coalition Formation on Green and Energy Harvesting Wireless Networks, Wireless
Among Unmanned Air Vehicles in Wireless Networks,” Communications and Mobile Computing.
Proc. Int’l. Conf. Game Theory for Networks (GameNets),
May 2009, pp. 259–67. J iandong L i [SM’05] graduated from Xidian University with
[4] P. Semasinghe, S. Maghsudi, and E. Hossain, “Game The- a bachelor degree, master degree, and Ph.D. in communica-
oretic Mechanisms for Resource Management in Massive tions and electronic system, respectively in 1982, 1985 and
Wireless IoT Systems,” IEEE Commun. Mag., vol. 55, no. 2, 1991. He has been with Xidian University since 1985, as an
Feb. 2017, pp. 121–27. associate professor from 1990 to 1994; professor from 1994;
[5] S. Bhattacharya and T. Basar, “Game-Theoretic Analysis of Ph.D. student supervisor from 1995; and Dean of the School of
an Aerial Jamming Attack on a UAV Communication Net- Telecommunication Engineering, Xidian University from 1997.
work,” Proc. American Control Conf., Baltimore, MD, USA, He also serves as Executive Vice Dean of the Graduate School,
Jun. 2010, pp. 818–23. Xidian University. He is a senior member of IEEE and the China
[6] J. Parras et al., “A New Approach for Solving Anti-jamming Institute of Electronics (CIE) and a fellow of the China Institute
Games in Stochastic Scenarios as Pursuit-evasion Games,” of Communication (CIC). He was a member of the PCN spe-
Proc. IEEE Statistical Signal Processing Workshop (SSP), Palma cialist group for the China 863 Communication high technology
de Mallorca, Palma, Spain, Jun. 2016, pp. 1–5. program between January 1993 and October 1994 and from
[7] D. Athukoralage et al., “Regret Based Learning for UAV 1999 to 2000. He is also a member of Communication Special-
Assisted LTE-U/WiFi Public Safety Networks,” Proc. IEEE ist group for The Ministry of Information Industry. His current
Global Commun. Conf. (GLOBECOM), Washington, DC, research interests and projects are funded by the 863 High Tech
USA, Dec. 2016, pp. 1–7. Project, NSFC, the National Science Fund for Distinguished
[8] H. Sedjelmaci, S. M. Senouci, and N. Ansari, “Intrusion Young Scholars, TRAPOYT, MOE and MOI
Detection and Ejection Framework Against Lethal Attacks in
UAV-Aided Networks: A Bayesian Game-Theoretic Method- Zhu Han [S’01, M’04, SM’09, F’14] received the B.S. degree
ology,” IEEE Trans. Intelligent Transportation Systems, vol. 18, in electronic engineering from Tsinghua University in 1997, and
no. 5, May 2017, pp. 1143–53. the M.S. and Ph.D. degrees in electrical and computer engineer-
[9] S. Koulali et al., “Green Strategic Activity Scheduling for UAV ing from the University of Maryland, College Park, in 1999 and
Networks: A Sub-modular Game Perspective,” IEEE Com- 2003, respectively. From 2000 to 2002 he was an R&D engineer
mun. Mag., vol. 54, no. 5, May 2016, pp. 58–64. at JDSU, Germantown, Maryland. From 2003 to 2006 he was
[10] D. H. Choi, S. H. Kim, and D. K. Sung, “Energy-Efficient a research associate at the University of Maryland. From 2006
Maneuvering and Communication of a Single UAV-based to 2008 he was an assistant professor at Boise State University,
Relay,” IEEE Trans. Aerospace and Electronic Systems, vol. Idaho. Currently, he is a professor in the Electrical and Comput-
50, no. 3, July 2014, pp. 2320–27. er Engineering Department as well as in the Computer Science
[11] P. B. Charlesworth, “A Non-Cooperative Game to Coor- Department at the University of Houston, Texas. His research
dinate the Coverage of Two Communications UAVs,” IEEE interests include wireless resource allocation and management,
Military Commun. Conf., San Diego, CA, USA, Nov. 2013, wireless communications and networking, game theory, big data
pp. 668–73. analysis, security, and smart grid. He received an NSF Career
[12] P. B. Charlesworth, “Using Non-Cooperative Games to Award in 2010, the Fred W. Ellersick Prize of the IEEE Commu-
Coordinate Communications UAV,” Proc. Workshop-Wire- nication Society in 2011, the EURASIP Best Paper Award for the
less Networking and Control for Unmanned Autonomous Journal on Advances in Signal Processing in 2015, the IEEE Leon-
Vehicles, Austin, TX, USA, Dec. 2014, pp. 1463–69. ard G. Abraham Prize in the field of Communications Systems
[13] A. Giagkos et al., “Comparing Approaches for Coordina- (best paper award in IEEE JSAC) in 2016, and several best paper
tion of Autonomous Communications UAVs,” Proc. Int’l. awards at IEEE conferences. Currently, he is an IEEE Communica-
Conf. Unmanned Aircraft Systems (ICUAS), Arlington, VA, tions Society Distinguished Lecturer. He was a 1 percent highly
USA, Jun. 2016, pp. 1131–40. cited researcher in 2017 according to the Web of Science.