Agile and Dynamic Bipedal Locomotion: Taking on the Wave Field

Jessy Grizzle, University of Michigan

Is it great fortune or a curse to do legged robotics on a University campus that has Maya Lin’s earthen sculpture, The Wave Field? Come to the talk and find out! We will discuss our work on model-based feedback control and optimization for highly dynamic locomotion in bipedal robots. The presentation will evolve in stages. We’ll start with feedback controllers that induce locally exponentially stable periodic walking gaits and illustrate a range of aperiodic behavior that can be achieved with such controllers. Then, robustness to terrain variation will be explicitly brought into the design problem, leading to gaits that can handle walking outdoors on grass, asphalt and 2 to 4 cm thick planks tossed randomly about the laboratory. Next, it is argued that under more extreme perturbations, these gaits typically fail due to ground reaction force constraint violations, and a means to address this with a library of gaits and interpolation will be discussed and illustrated. At this stage, the simpler parts of the University of Michigan’s Wave Field can finally be tackled. The last part of the talk will discuss preliminary work on using Supervised Machine Learning to approximate an explicit MPC controller. When combined with robot vision, we hope this will allow the full complexity of the Wave Field to be conquered. In any case Jovanotti assures us, "Non c'è scommessa più persa di quella che non giocherò". I for one will keep trying!