Research paper
safe-control-gym · Repeatable safety comparisons for robot learning
The benchmark provides constrained cart-pole and quadrotor simulation tasks, with controlled disturbances. It supports comparisons of controller performance, data use and constraint violations under a shared experimental setup.
Reviewed 2026-11-04 · Source and bibliography checked
- Original title
- safe-control-gym: a Unified Benchmark Suite for Safe Learning-based Control and Reinforcement Learning in Robotics
- Publisher
- arXiv
- Authors
- Zhaocong Yuan, Adam W. Hall, Siqi Zhou, Lukas Brunke, Melissa Greeff, Jacopo Panerati, Angela P. Schoellig
- DOI
- 10.48550/arXiv.2109.06325
- Publication date
- 2021-09-13
- Source language
- en
- Source country
- US
- Edition
- arXiv:2109.06325v4
- Access
- open_archive_metadata_and_manuscript
- Publication status
- Preprint
Source overview
The benchmark provides constrained cart-pole and quadrotor simulation tasks, with controlled disturbances. It supports comparisons of controller performance, data use and constraint violations under a shared experimental setup.
Use and review
This is a research benchmark using selected simulated dynamics. Benchmark results leave physical robot integration, sensing, timing and safety-function validation to the application. The date is the first arXiv posting.
Sources & editions
safe-control-gym: a Unified Benchmark Suite for Safe Learning-based Control and Reinforcement Learning in Robotics ↗
arXiv · arXiv:2109.06325v4
arXiv abstract, author metadata and submission history. First posted 2021-09-13; v4 dated 2022-07-26.