Global Path Planning for Multi-objective UAV-Assisted Sensor Data Collection: A DRL Approach

Abstract

With the rapid development of the Internet of Things (IOT), the amount of business data carried by wireless sensor networks (WSNs) has exploded. Unmanned aerial vehicle (UAV)-assisted sensor data collection is effective and environmentally friendly. A UAV should determine which sensor nodes to visit, and the order in which to visit them according to the target requirements. Due to UAV energy and data-collection demands, a mission can have more than one goal. This study investigates the multi-objective global path planning of UAV-assisted sensor data collection (SDC). It is modeled as a multi-objective optimization problem (MOP) to maximize the amount of data collected and minimize UAV flight time. Based on a deep reinforcement learning (DRL) framework, we decompose the MOP into a series of subproblems, model them as neural networks and use an actor-critic algorithm and modified pointer network to solve each subproblem. Then the Pareto front of the path planning solution under the above constraints can be obtained through the forward propagation of the neural network. Experimental results show that the proposed method is superior to a traditional multi-objective evolutionary algorithms in terms of convergence, diversity of solutions, and time complexity. In addtion, the proposed method can be directly applied to the situation in which the number of sensors changes without retraining, and it has better robustness.

This work is supported in part by the National Key R &D Program of China under Grant 2019YFB2102000, and in part by the Collaborative Innovation Center of Novel Software Technology and Industrialization.