An Approach to Model-Based Control of Frictionally Constrained Robots

Aaron Greenfield

Abstract

I will be talking about an approach to generating controls for robotic systems subject to frictional contacts with the environment. The target applications for this approach are a flipping task for the RHex robot and a climbing task for a snake robot. We start by presenting a dynamics model based on the assumptions of system rigidity and coulomb friction. This dynamics model is solved instantaneously for its "dynamic response", which is the unknown system accelerations and reaction forces written as a function of the control torque. The dynamic response is then optimized over to select the controls which produce the desired instantaneous behavior, in particular for the two target applications.

Presentation Slides

Ppt (2.3Mb)

 

Speaker Appointments

For appointments, please contact Aaron Greenfield (agreenfi@andrew.cmu.edu).