Introspective Visuomotor Control: Exploiting Uncertainty in Deep Visuomotor Control for Failure Recovery
This addresses a common limitation in robot manipulation tasks by enabling recovery from errors without tactile feedback or explicit vision-based failure detection, though it is incremental as it builds on existing uncertainty methods.
The paper tackled the problem of failure recovery in imitation learning-based visuomotor control by using policy network uncertainty to detect and recover from out-of-distribution states, resulting in significant improvements in task success rates: 12% in pushing, 15% in pick-and-reach, and 22% in pick-and-place.
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.