Control of Robots in Uncertain Environments

How should robots imbued with complex dynamics be controlled in order to maximize task performance and robustness while also preserving beneficial aspects of their dynamics? This question motivates out current focus within control, and our lab explores it through a variety of different approaches. While traditional approaches (e.g. impedance control) get us pretty far, we are exploring optimization-based control which can encode the unique physical attributes of our new classes of integrated multimaterial robots, allowing us to solve more complex control problems and directly encode notions of safety.

A major motivator for our current work is the apparent contradiction between precision of traditional feedback control and the closed loop compliance of a robotic system. Essentially, as we turn up our proportional control knob to decrease the control error, we reduce the compliance of the system, as seen in the following simple demo. preview Had I not removed my finger, the robot would have either crushed my finger or the motors would have burnt out. This means that our cuddly, “safe” soft robot is not so safe when operated with pure feedback control. Thus, for this work we are seeking alternatives.

Control Barrier Functions for Soft Robots

preview

Sensorless Fofce Control

preview

Impedance Control

preview

Safe Control with Actuator Saturation

preview

Publications