Colloquia Archive
Motion Planning for Legged Locomotion
April 15, 2008
AbstractLegged vehicles have attracted interest for applications that require high mobility. They could provide military troop support and logistics in rocky, steep, and forested terrain. Scientific applications include exploration of cliffs and volcanoes, and even other planets. They have been considered for industrial uses, such as logging and construction. They are also attractive for navigating human environments for search and rescue and building inspection, as well as humanoid robots navigating homes and offices as personal assistants. However, legged vehicles are difficult to control, because many joints must be coordinated to maintain balance while executing a task. Control is particularly difficult in extremely uneven, steep, or cluttered terrain -- precisely the situations in which legs are an advantage. In these situations, a legged robot must sense the environment, and plan ahead.
I present a motion planner for legged locomotion that reasons directly in the vehicle's high dimensional configuration space, producing motions that are guaranteed to avoid collision and satisfy balance constraints. I focus on work on the ATHLETE (All-Terrain Hex-Legged Extra-Terrestrial Explorer), a six-legged vehicle with 36 joints, which is being developed by NASA for lunar exploration. However, the same methods work for mechanisms with an arbitrary number of joints and legs and a variety of terrains from flat ground to near-vertical faces. The planner takes a two-step approach. First it produces a sequence of candidate steps using graph search techniques. To plan single steps, it uses probabilistic roadmaps (PRMs), which have been established as state-of-the-art for motion planning in high-dimensional spaces. I demonstrate the planner on simulated examples of ATHLETE and other robots on rocky and steep terrain, where simpler techniques like gaits would fail. I conclude with extensions of this research to other problems, such as a manipulation planning system that enables the Honda ASIMO humanoid robot to push objects on a table.
BiographyKris Hauser is a PhD candidate at Stanford University, where he is a member of the Artificial Intelligence Lab in the Department of Computer Science. His research focuses on motion planning algorithms for complex mechanical systems, such as legged locomotion and manipulation. Kris earned his undergraduate degrees in Computer Science and Mathematics at the University of California at Berkeley.
