Faculty Advisor

Berenson, Dmitry

Abstract

Many compliant robots suffer from actuation uncertainty. When executing paths, contact with obstacles can lead to the robot becoming unable to continue execution. This paper presents a robustness metric for compliant robots that captures the probability of the robot completing a path, i.e. the probability of avoiding stuck configurations. Our approach constructs a set of reachable C-space volumes between the waypoints of a path. We can then identify stuck configurations within these volumes and approximate their joint pre-image, which we then use to compute the probability of successfully reaching a way-point and subsequently the probability of reaching the path’s goal. We find that our method computes similar robustness predictions to forward simulation more efficiently.

Publisher

Worcester Polytechnic Institute

Date Accepted

March 2016

Major

Computer Science

Major

Robotics Engineering

Project Type

Major Qualifying Project

Accessibility

Unrestricted

Advisor Department

Computer Science

Share

COinS