Non-Periodic Gait Planning Based on Salient Region Detection for a Planetary Cave Exploration Robot

Abstract

A limbed climbing robot can traverse uneven and steep terrain, such as Lunar/Martian caves. Towards the autonomous operation of the robot, we first present a method to detect topographically salient regions in 3D point cloud as the robot ’s graspable targets, and next, we introduce a strategy to plan a non-periodic gait for the robot from the detected discrete graspable options. The proposed gait planner is implemented and validated in our open dynamic climbing robot simulation platform assuming the 3 kg class four-limbed climbing robot testbed moving over steep and uneven Lunar terrain.