Skip to content

Optimization and Learning for Rough Terrain Legged Locomotion

Authors: Matthew Zucker, Nathan Ratliff, Martin Stolle, Joel Chestnutt, J. Andrew Bagnell, Christopher G. Atkeson, James Kuffner

Published: 2011 (Journal Paper)

Source: International Journal of Robotics Research

DOI: 10.1177/0278364910392608

Summary

Proposes a hierarchical optimization framework for quadruped locomotion over rough terrain, combining anytime footstep planning with dynamic body motion optimization. Uses inverse optimal control to learn cost functions from demonstrations and real-time re-planning for robustness, demonstrated on the LittleDog robot.

Abstract

This paper presents a novel approach to legged locomotion over rough terrain rooted in optimization. The approach relies on a hierarchy of fast, anytime algorithms to plan a set of footholds and the dynamic body motions required to execute them. Components within the planning framework coordinate to exchange plans, cost-to-go estimates, and certificates that ensure the output of an abstract high-level planner can be realized by lower layers of the hierarchy. The burden of careful engineering of cost functions is mitigated by a simple inverse optimal control technique, and robustness is achieved by real-time re-planning of the full trajectory, augmented by reflexes and feedback control. We demonstrate the approach guiding the LittleDog quadruped robot over various rough terrains.

Tags

  • Legged locomotion

  • Rough terrain

  • Footstep planning

  • Trajectory optimization

  • Inverse optimal control

  • Quadruped

  • Anytime planning