Event Location: GHC 8102

Abstract: We expect legged robots to be highly mobile. Human walking and running can execute quick changes in speed and direction, even on non-flat ground. Indeed, analysis of simplified models shows that these quantities can be tightly controlled by adjusting the leg placement between steps and that the leg placement can also compensate for disturbances including changes in the ground height. However, to date, legged robots do not exhibit this level of agility or robustness, nor is it well understood what prevents them from attaining this performance.

In this thesis, we aim to implement highly mobile running on our bipedal robot ATRIAS while maintaining a consistent handle for comparing theoretical expectations, simulation results, and experimental results. We want to systematically understand how the properties of the hardware platform limit our application of theoretically derived control and to design the control accordingly.

Specifically, we build our control strategy around the spring mass model. We extend the analysis of the simplified model to 3D to derive a leg placement strategy that provides near deadbeat tracking of both steering angles and running speeds on unobserved terrain. Based on this nominal motion from the lower order model, we define a full-order reference trajectory for the robot. We then use optimal control techniques and inverse dynamics to track the reference by modulating the ground reaction forces. We will apply Lyapunov arguments in conjunction with the emergent cost function and well established frequency space techniques to quantify the expected performance given the limitations of the hardware. In the end, we expect to demonstrate an implementation that matches our understanding of how well the agility and robustness of the simplified model carries over to a real system.

Committee:Hartmut Geyer, Chair

Christopher G. Atkeson

Koushil Sreenath

Jerry Pratt, Institute for Human and Machine Cognition