사족 보행 로봇의 보행을 위한 마찰 제약 조건을 고려한 지각 기반 동작 계획

Abstract

The legged robots, such as quadruped robots, have the advantage of being able to traverse through complex, challenging environments. For this to be possible, the robots must be aware of the complex area to traverse and plan the desired footsteps according to the desired walking direction considering the environment. In this paper, we describe an end-to-end framework pipeline from environment perception to footstep planning to let the quadruped robots traverse complex environments with friction constraints. The environment perception extracts planar regions that the legged robots can step on from the 2.5-dimension height map with the collected point cloud data. To overcome locomotion over rough terrain, footstep planning optimizes the desired footstep by considering the friction coefficients, and body pose planning determines the pose using robot’s states on stairs. The entire computation process of the environment perception completes to detect the planar regions within 80 Hz, and the motion planner generates the optimal foot trajectory and body trajectory to get a specified location. We experimentally validate our approach with the quadruped robot Canine by autonomously navigating stairs, stepping stones, and gaps over simulation environments.