Trajectory Planning for Unmanned Aerial Vehicle Assisted WSN Data Collection Based on Q-Learning

Abstract

In some scenarios where Unmanned Aerial Vehicle(UAV) assists in Wireless Sensor Network(WSN) data collection,the data generation rate of each node is random and the states of sink node are inconsistent.To address the problem,this paper proposes a Q-learning-based algorithm called Q-TDUD for discontinuous UAV trajectory planning,which can improve the energy efficiency of UAV and data collection efficiency.Based on the randomness of the data generation rate of each node in the cycle,the aggregation delay model of the sink node is established.The Q-learning algorithm in reinforcement learning is used to normalize the delay time of each sink node and the uplink transmission rate of the collection link into the reward function,and the optimal discontinuous flight trajectory of the UAV is obtained through iterative calculation.Experimental results show that,compared with TSP-continues,TSP,NJS-continues and NJS algorithms,the proposed Q-TDUD algorithm can reduce the task completion time of UAV,and improve the energy efficiency and data collection efficiency of UAV.