Introspective visuomotor control: exploiting uncertainty in deep visuomotor control for failure recovery
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...
| Main Authors: | , , , , |
|---|---|
| Format: | Conference item |
| Language: | English |
| Published: |
Institute of Electrical and Electronics Engineers
2021
|
| _version_ | 1826281212476391424 |
|---|---|
| author | Hung, C-M Sun, L Wu, Y Havoutis, I Posner, I |
| author_facet | Hung, C-M Sun, L Wu, Y Havoutis, I Posner, I |
| author_sort | Hung, C-M |
| collection | OXFORD |
| description | 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.
|
| first_indexed | 2024-03-07T00:25:24Z |
| format | Conference item |
| id | oxford-uuid:7df969e7-7302-44dc-9994-5fca582aac20 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T00:25:24Z |
| publishDate | 2021 |
| publisher | Institute of Electrical and Electronics Engineers |
| record_format | dspace |
| spelling | oxford-uuid:7df969e7-7302-44dc-9994-5fca582aac202022-03-26T21:07:11ZIntrospective visuomotor control: exploiting uncertainty in deep visuomotor control for failure recoveryConference itemhttp://purl.org/coar/resource_type/c_5794uuid:7df969e7-7302-44dc-9994-5fca582aac20EnglishSymplectic ElementsInstitute of Electrical and Electronics Engineers2021Hung, C-MSun, LWu, YHavoutis, IPosner, IEnd-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. |
| spellingShingle | Hung, C-M Sun, L Wu, Y Havoutis, I Posner, I Introspective visuomotor control: exploiting uncertainty in deep visuomotor control for failure recovery |
| title | Introspective visuomotor control: exploiting uncertainty in deep visuomotor control for failure recovery |
| title_full | Introspective visuomotor control: exploiting uncertainty in deep visuomotor control for failure recovery |
| title_fullStr | Introspective visuomotor control: exploiting uncertainty in deep visuomotor control for failure recovery |
| title_full_unstemmed | Introspective visuomotor control: exploiting uncertainty in deep visuomotor control for failure recovery |
| title_short | Introspective visuomotor control: exploiting uncertainty in deep visuomotor control for failure recovery |
| title_sort | introspective visuomotor control exploiting uncertainty in deep visuomotor control for failure recovery |
| work_keys_str_mv | AT hungcm introspectivevisuomotorcontrolexploitinguncertaintyindeepvisuomotorcontrolforfailurerecovery AT sunl introspectivevisuomotorcontrolexploitinguncertaintyindeepvisuomotorcontrolforfailurerecovery AT wuy introspectivevisuomotorcontrolexploitinguncertaintyindeepvisuomotorcontrolforfailurerecovery AT havoutisi introspectivevisuomotorcontrolexploitinguncertaintyindeepvisuomotorcontrolforfailurerecovery AT posneri introspectivevisuomotorcontrolexploitinguncertaintyindeepvisuomotorcontrolforfailurerecovery |