Motion Planning for Legged Robots on Irregular and Steep Terrain

Dr. Bretl is from Stanford University
A key constraint on the motion of a legged robot is maintenance of equilibrium. The robot must apply contact forces with hands and feet on the terrain that exactly compensate for gravity without causing slip. On regular and flat terrain, this constraint can often be satisfied by a pre-computed, fixed gait. But on irregular and steep terrain, such as rocky outdoor slopes or broken urban environments, useful contacts may be sparse and arbitrarily distributed. Not only must the robot decide how to adjust its posture to reach the next contact without falling (reasoning about contacts, friction, equilibrium, collision, and torque limits), it must plan an entire sequence of steps, where each one might have future consequences.

This talk will present a motion planning approach developed for legged robots. It will introduce the computational issues raised by such robots (for example, searching geometrically complex and high-dimensional configuration spaces, sampling feasible configurations in such spaces, and allocating time among many search queries) and will describe specific algorithmic techniques to handle these issues. In particular, it will show how these techniques have been applied to a four-limbed "free-climbing" robot called LEMUR (developed by the Jet Propulsion Laboratory), enabling it to climb an indoor, near-vertical surface covered with artificial rock features. Ongoing work also extends these techniques to a humanoid robot (HRP-2) and a six-legged lunar vehicle (ATHLETE).