Research paper
Safer task learning · Remembering causes of irrecoverable errors
The learner reasons about dead ends to avoid repeating actions such as breaking objects. When no safe alternative is found, it queries a teacher; evaluation includes a robot task and a planning benchmark.
Reviewed 2026-11-05 · Source and bibliography checked
- Original title
- Safe robot execution in model-based reinforcement learning
- Publisher
- IEEE
- Authors
- David Martinez, Guillem Alenya, Carme Torras
- DOI
- 10.1109/iros.2015.7354295
- Publication date
- 2015-09
- Source language
- en
- Edition
- 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS); 2015-09; DOI 10.1109/iros.2015.7354295
- Access
- Publisher metadata and open author manuscript
- Publication status
- Published
Related resources and engineering notes
Source overview
The learner reasons about dead ends to avoid repeating actions such as breaking objects. When no safe alternative is found, it queries a teacher; evaluation includes a robot task and a planning benchmark.
Use and review
Institutional abstract and bibliography checked. Task-level error avoidance must be assessed separately from physical-contact and injury safety.
Sources & editions
Safe robot execution in model-based reinforcement learning ↗
IEEE · 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS); 2015-09; DOI 10.1109/iros.2015.7354295
IROS2015 author abstract https://www.iri.upc.edu/publications/show/1671; robot task and planning benchmark are distinct evaluations.