Introspective Visuomotor Control: Exploiting Uncertainty in Deep Visuomotor Control for Failure Recovery

TL;DR

This paper introduces an uncertainty-based method for failure detection and recovery in deep visuomotor control, significantly improving task success rates by monitoring and exploring uncertain states.

Contribution

It proposes a novel approach leveraging policy uncertainty to detect failures and recover in visuomotor tasks, without relying on tactile feedback or explicit failure detection.

Findings

Improved success rates: 12% in pushing, 15% in pick-and-reach, 22% in pick-and-place.

Uncertainty effectively indicates potential failures in visuomotor control.

Trajectory monitoring and exploration based on uncertainty enhance task robustness.

Abstract

End-to-end visuomotor control is emerging as a compelling solution for robot manipulation tasks. However, imitation learning-based visuomotor control approaches tend to suffer from a common limitation, lacking the ability to recover from an out-of-distribution state caused by compounding errors. In this paper, instead of using tactile feedback or explicitly detecting the failure through vision, we investigate using the uncertainty of a policy neural network. We propose a novel uncertainty-based approach to detect and recover from failure cases. Our hypothesis is that policy uncertainties can implicitly indicate the potential failures in the visuomotor control task and that robot states with minimum uncertainty are more likely to lead to task success. To recover from high uncertainty cases, the robot monitors its uncertainty along a trajectory and explores possible actions in the state-action space to bring itself to a more certain state. Our experiments verify this hypothesis and show a significant improvement on task success rate: 12% in pushing, 15% in pick-and-reach and 22% in pick-and-place.