High maneuverability control of Single-track Two-wheeled Robot in Narrow Terrain based on Reinforcement Learning

Abstract

The single-track two-wheeled (STTW) robot has the advantages of small size and flexibility, and it is designed to travel in narrow terrain such as mountains or jungles. In this article, a reinforcement learning-based control method is proposed to solve the problem that STTW robots drive fast in narrow terrain with limited visibility. This control method integrates STTW robot path planning, trajectory tracking, and balance control in a single framework. Based on this framework, we define state, action, and reward function for narrow terrain passing tasks. At the same time, we design the actor network and the critic network structures and use the Twin Delayed Deep Deterministic Policy Gradient (TD3) to train these neural networks to construct a controller. Next, we verify the proposed control method on a simulation platform. The simulation results show that the obtained controller allows the STTW robot to effectively pass the training terrain. In addition, this article conducts a simulation comparison to prove advantages of training with the TD3 algorithm and the effectiveness of the reward function.