The MIT Leg Laboratory explores active balance and dynamics in legged systems, robots and animals alike. Systems that balance actively can use footholds that are widely separated, they can move along narrow paths with a narrow base of support, and they can use their kinetic energy as a bridge to increase their effective size. Dynamics is a key ingredient in the behavior of animals and it will weigh heavily in the development of useful legged vehicles.

A dynamic treatment need not be an intractable treatment. We have found that simple algorithms can provide balance and control for a variety of dynamic legged systems. A single set of control algorithms, modified in various ways, has successfully controlled numerous running machines. Several simple algorithms currently under development have had promising results on walking machines. Long term goals for the lab include development of robot control strategies useful for both walking and running.

The techniques used to control each of the running machines derive from a single simple set of control algorithms. They focus on support, posture, and propulsion. These algorithms have been adapted for hopping, pronking, biped running, fast running, trotting, pacing, bounding, and simple gymnastic maneuvers. The ability of simple algorithms to operate under these diverse circumstances suggests their fundamental nature.

Current walking experiments explore fundamental issues in walking. In keeping with the spirit of the lab, we exploit the dynamics of our walking robots to perform a task.

The Leg Lab also applies techniques from robotics to the control of computer simulated creatures. This includes task-level control of behavior and automatic tuning of control parameters.

Back to the Leg Lab Home Page