Event
RobustRLlib Released: Unifying 17 Robust Reinforcement Learning Algorithms
1 reports · 1 sources
Xi Chen and other authors have released RobustRLlib, an algorithm-centric robust reinforcement learning library and benchmark that integrates 17 representative online, offline, and safe reinforcement learning algorithms under a unified interface, accompanied by an evaluation platform covering various real-world shifts and humanoid robot, vision-language-action tasks. The authors state that the platform conducts systematic evaluation across 273 perturbation configurations, providing a capability distribution of current robust reinforcement learning algorithms and showing which scenarios remain effective and which still have notable robustness gaps. The project is open-sourced, providing code, a project page, and tutorials for reproduction and extension. The above evaluation conclusions come from the authors' own account and have not yet been independently verified.
Report timeline
- RobustRLlib released: unifies 17 robust reinforcement learning algorithms and provides evaluation across 273 perturbation configurations
RobustRLlib is an algorithm-centric library and benchmark for robust reinforcement learning that integrates 17 representative online, offline, and safe reinforcement learning algorithms under a unified interface. Its evaluation platform covers a variety of real-world shifts as well as humanoid robot and vision-language-action tasks, and it presents the capability distribution of current robust reinforcement learning algorithms across 273 perturbation configurations, identifying which scenarios remain effective and where substantial robustness gaps persist. The project provides open-source code, a project page, and tutorials to facilitate reproduction and extension.
arXiv cs.RO ·