Deep-sarsa based multi-uav path planning and obstacle avoidance in a dynamic environment

Abstract

This study presents a Deep-Sarsa based path planning and obstacle avoidance method for unmanned aerial vehicles (UAVs). Deep-Sarsa is an on-policy reinforcement learning approach, which gains information and rewards from the environment and helps UAV to avoid moving obstacles as well as finds a path to a target based on a deep neural network. It has a significant advantage over dynamic environment compared to other algorithms. In this paper, a Deep-Sarsa model is trained in a grid environment and then deployed in an environment in ROS-Gazebo for UAVs. The experimental results show that the trained Deep-Sarsa model can guide the UAVs to the target without any collisions. This is the first time that Deep-Sarsa has been developed to achieve autonomous path planning and obstacle avoidance of UAVs in a dynamic environment.