Abstract: In dense traffic unmanned aerial vehicle (UAV) ad-hoc networks, traffic congestion can cause increased delay and packet loss, which limit the performance of the networks; therefore, a traffic balancing strategy is required to control the traffic. In this study, we propose TQNGPSR, a traffic-aware q-network enhanced geographic routing protocol based on greedy perimeter stateless routing (GPSR), for UAV ad-hoc networks. The protocol enforces a traffic balancing strategy using the congestion information of neighbors, and evaluates the quality of a wireless link by the q-network algorithm, which is a reinforcement learning algorithm. Based on the evaluation of each wireless link, the protocol makes routing decisions in multiple available choices to reduce delay and decrease packet loss. We simulate the performance of TQNGPSR and compare it with AODV, OLSR, GPSR, and QNGPSR. Simulation results show that TQNGPSR obtains higher packet delivery ratios and lower end-to-end delays than GPSR and QNGPSR. In high node density scenarios, it also outperforms AODV and OLSR in terms of the packet delivery ratio, end-to-end delay, and throughput.

基于GPSR和Q网络的流量感知无人机ad-hoc网络路由协议

[1]Abadi M, Barham P, Chen JM, et al., 2016. TensorFlow: a system for large-scale machine learning. Proc 12^th USENIX Conf on Operating Systems Design and Implementation, p.265-283.

[2]Basagni S, Chlamtac I, Syrotiuk VR, et al., 1998. A distance routing effect algorithm for mobility (DREAM). Proc 4^th Annual ACM/IEEE Int Conf on Mobile Computing and Networking, p.76-84.

[3]Bekmezci I, Sahingoz OK, Temel Ş, 2013. Flying ad-hoc networks (FANETs): a survey. Ad Hoc Netw, 11(3):1254-1270.

[4]Bolch G, Greiner S, de Meer H, et al., 2006. Queueing Networks and Markov Chains: Modeling and Performance Evaluation with Computer Science Applications (2^nd Ed.). John Wiley & Sons, New York, USA.

[5]Boyan JA, Littman ML, 1994. Packet routing in dynamically changing networks: a reinforcement learning approach. Proc 7^th Int Conf on Neural Information Processing Systems, p.671-678.

[6]Coutinho N, Matos R, Marques C, et al., 2015. Dynamic dual-reinforcement-learning routing strategies for quality of experience-aware wireless mesh networking. Comput Netw, 88:269-285.

[7]Farahnakian F, Ebrahimi M, Daneshtalab M, et al., 2011. Q-learning based congestion-aware routing algorithm for on-chip network. Proc 2^nd Int Conf on Networked Embedded Systems for Enterprise Applications, p.1-7.

[8]Jung WS, Yim J, Ko YB, 2017. QGeo: Q-learning-based geographic ad hoc routing protocol for unmanned robotic networks. IEEE Commun Lett, 21(10):2258-2261.

[9]Karp B, Kung HT, 2000. GPSR: greedy perimeter stateless routing for wireless networks. Proc 6^th Annual Int Conf on Mobile Computing and Networking, p.243-254.

[10]Kenta T, Takeshi M, Shinya K, et al., 2006. Experimental evaluation of an on-demand multipath routing protocol for video transmission in mobile ad hoc networks. J Zhejiang Univ-Sci A, 7(S1):145-150.

[11]Ko YB, Vaidya NH, 2000. Location-aided routing (LAR) in mobile ad hoc networks. Wirel Netw, 6(4):307-321.

[12]Li RL, Li F, Li X, et al., 2014. QGrid: Q-learning based routing protocol for vehicular ad hoc networks. Proc 33^rd Int Performance Computing and Communications Conf, p.1-8.

[13]Lin SC, Wang P, Luo M, 2016. Control traffic balancing in software defined networks. Comput Netw, 106:260-271.

[14]Lyu N, Song GH, Yang BW, et al., 2019. QNGPSR: a Q-network enhanced geographic ad-hoc routing protocol based on GPSR. Proc 88^th Vehicular Technology Conf, p.1-6.

[15]Ma X, Xu Y, Sun GQ, et al., 2013. State-chain sequential feedback reinforcement learning for path planning of autonomous mobile robots. J Zhejiang Univ-Sci C (Comput & Electron), 14(3):167-178.

[16]Mnih V, Kavukcuoglu K, Silver D, et al., 2015. Human-level control through deep reinforcement learning. Nature, 518(7540):529-533.

[17]Peng J, Williams RJ, 1996. Incremental multi-step Q-learning. Mach Learn, 22(1-3):283-290.

[18]Shi RH, Deng YY, 2008. An improved scheme for reducing the latency of AODV in ad hoc networks. Proc 9^th Int Conf for Young Computer Scientists, p.594-598.

[19]Sutton RS, Barto AG, 1998. Reinforcement Learning: an Introduction. MIT Press, Cambridge, p.1.

[20]Watkins CJCH, Dayan P, 1992. Q-learning. Mach Learn, 8(3-4):279-292.

[21]Wu C, Ohzahata S, Kato T, 2012. Routing in VANETs: a fuzzy constraint Q-learning approach. Proc Global Communications Conf, p.195-200.

[22]Wu C, Ohzahata S, Kato T, 2013. Flexible, portable, and practicable solution for routing in VANETs: a fuzzy constraint Q-learning approach. IEEE Trans Veh Technol, 62(9):4251-4263.

[23]Xu DH, Chiang M, Rexford J, 2011. Link-state routing with hop-by-hop forwarding can achieve optimal traffic engineering. IEEE/ACM Trans Netw, 19(6):1717-1730.

[24]Zhan HW, Zhou Y, 2008. Comparison and analysis AODV and OLSR routing protocols in ad hoc network. Proc 4^th Int Conf on Wireless Communications, Networking and Mobile Computing, p.1-4.

[25]Zhang JJ, Xi K, Luo M, et al., 2014. Load balancing for multiple traffic matrices using SDN hybrid routing. Proc 15^th Int Conf on High Performance Switching and Routing, p.44-49.