Goal-driven Motion Control of Snake Robots with Onboard Cameras via Policy Improvement with Path Integrals

Abstract

Due to the redundant degrees of freedom and complex gaits, the onboard visual localization and motion control of a snake robot's goal-driven motion are very challenging. Firstly, to overcome the localization difficulties caused by the swing of the robot's head, we propose a periodic visual localization strategy, which divides the motion control into several control periods. After each control period, the snake robot stops moving and rotates the onboard camera to estimate its position. Subsequently, a stable and robust motion control is required to enable the snake robot to reach the goal. We propose a two-stage control framework based on the Policy Improvement with Path Integrals (PI <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ) and gait equation. We use the PI <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> optimization method to generate and optimize a stable and robust control policy. The optimized control policy can guide the gait equation to generate suitable gaits according to the position, ensuring that the robot always moves towards the goal. The two-stage control framework enables the snake robot with an onboard camera to perform precise goal-driven motion. The simulation and experiment demonstrate that the proposed method is accurate and efficient.