Motion planning of multi-limbed robots subject to equilibrium constraints: The free-climbing robot problem

摘要

This paper addresses the problem of planning the motion of a multi-limbed robot in order to "free-climb" vertical rock surfaces. Free-climbing only relies on frictional contact with the surfaces rather than on special fixtures or tools like pitons. It requires strength, but more importantly it requires deliberate reasoning: not only must the robot decide how to adjust its posture to reach the next feature without falling, it must plan an entire sequence of steps, where each one might have future consequences. In this paper this process of reasoning is broken into manageable pieces by decomposing a free-climbing robot's configuration space into manifolds associated with each state of contact between the robot and its environment. A multistep planning framework is presented that decides which manifolds to explore by generating a candidate sequence of hand and foot placements first. A one-step planning algorithm is then described that explores individual manifolds quickly. This algorithm extends the probabilistic roadmap approach to better handle the interaction between static equilibrium and the topology of closed kinematic chains. It is assumed throughout this paper that a set of potential contact points has been presurveyed. Validation with real hardware was done with a four-limbed robot called LEMUR (developed by the Mechanical and Robotic Technologies Group at NASA-JPL). Using the planner presented in this paper; LEMUR free-climbed all indoor near-vertical surface covered with artificial rock features.