Step climbing method for crawler type rescue robot using reinforcement learning with Proximal Policy Optimization
Mifu Totani, Noritaka Sato, Yoshifumi Morita · 2019
It is a huge burden on an operator when he/she tele-operates a rescue robot traveling on a rough terrain. Therefore, the purpose of this study is to reduce this burden by controlling the robot autonomously. As a first step, we propose a step climbing method for a crawler type rescue robot by using reinforcement learning with Proximal Policy Optimization (PPO). The input data are the image of a camera on the robot and a posture image of the robot. We verified the effectiveness of the proposed method using a dynamics simulator.