Research on UAV Handover Management Based on Deep Reinforcement Learning
Providing network connections for drones is a major application of future cellular network systems.When drones serve as mobile base stations or mobile user equipment in cellular networks,they need to switch between different base stations to maintain high-speed and reliable network connections.Aiming at the problems of frequent handovers and handover failures of UAVs between cellular base stations caused by high mobility of UAVs and complex flight environment,a method for optimizing handover of UAVs connected to cellular networks based on deep reinforcement learning is proposed.First of all,based on a deep reinforcement learning framework,on-line learning and decision-making for adaptive base station switching of UAVs are realized,which overcomes the shortcomings of previ-ous algorithms that result in long training time and poor generalization ability when the state space is too large.Secondly,two indicators of reference signal received power and handover times are integrated as a joint reward function to ensure that the UAV has a stable cel-lular network connection and reduces the number of invalid handovers between the UAV and the cellular base station.Experimental re-sults show that after 1 000 rounds of training,the proposed algorithm has significantly reduced the average number of handovers for UAV,effectively avoiding unnecessary handovers,reducing the probability of handover failures,and improving the receive power of UAV when connecting to cellular networks.
UAV communicationcellular networkreference signal receiving powerdeep reinforcement learning
国家科技期刊平台
NSTL
万方数据
