Deep Reinforcement Learning Based on the Hindsight Experience Replay for Autonomous Driving of Mobile Robot
Minjae Park, Jin-Seok Hong, Nam-Kyu Kwon · Journal of Institute of Control Robotics and Systems · 2022
In this paper, we present a method to overcome the sparse reward problem that can occur in the autonomous driving of a mobile robot based on the deep deterministic policy gradient algorithm of reinforcement learning using simple reward engineering and the hindsight experience replay technique. The mobile robot used in the experiment is a robot operating system-based TurtleBot3, and the experimental environment was configured using Gazebo. To validate the effectiveness of the proposed technique, we present the experimental results with the application of HER.