Research paper
Pedestrian navigation · Reinforcement learning with uncertainty estimates
MC-Dropout and bootstrapping estimate neural-model uncertainty for a pedestrian-avoidance policy. Simulation compares its response to unfamiliar pedestrian behaviour against a policy that does not account for uncertainty.
Reviewed 2026-11-05 · Source and bibliography checked
- Original title
- Safe Reinforcement Learning With Model Uncertainty Estimates
- Publisher
- IEEE
- Authors
- Bjorn Lutjens, Michael Everett, Jonathan P. How
- DOI
- 10.1109/icra.2019.8793611
- Publication date
- 2019-05
- Source language
- en
- Edition
- 2019 International Conference on Robotics and Automation (ICRA); 2019; DOI 10.1109/icra.2019.8793611
- Access
- Publisher metadata and open author manuscript
- Publication status
- Published
Related resources and engineering notes
Source overview
MC-Dropout and bootstrapping estimate neural-model uncertainty for a pedestrian-avoidance policy. Simulation compares its response to unfamiliar pedestrian behaviour against a policy that does not account for uncertainty.
Use and review
Checked archive abstract and publication metadata. Relative improvement in simulation is not calibrated risk probability or a safety guarantee around real pedestrians.
Sources & editions
Safe Reinforcement Learning With Model Uncertainty Estimates ↗
IEEE · 2019 International Conference on Robotics and Automation (ICRA); 2019; DOI 10.1109/icra.2019.8793611
ICRA 2019 bibliography; author archive abstract https://arxiv.org/abs/1810.08700v2. Conference date retained instead of earlier archive posting.