Development of a Lunar Rover Simulator with an Interface for Reinforcement Learning

In this paper, we propose an open-source lunar rover simulator integrated with a reinforcement learning framework. We incorporate lunar environmental effects in our simulator to enhance the rover locomotion realism while employing PyBullet to simulate the dynamics of a multi-degree-of-freedom planetary rover. We also extend our lunar rover simulator to interface with a reinforcement learning framework using OpenAI Gym environment and a compatible training workflow. We demonstrate the success of this integration through an autonomous rover navigation task on the lunar environment in our simulator using two reinforcement learning algorithms, DDPG and TD3. The lunar rover simulator is available at https://github.com/assawayut/LunarRoverSim and an accompanying video can be accessed at https://www.youtube.com/watch?v=8HMqAkDbpqI.