🤖 AI Summary
This work addresses the challenge of balancing performance and safety constraints in model-free reinforcement learning when an accurate dynamics model is unavailable. The authors propose a robust Koopman-CBF framework that constructs affine control barrier functions (CBFs) in a lifted space using a finite-dimensional Koopman operator and employs a quadratic programming-based safety layer to correct policy actions in real time. To mitigate errors from Koopman approximation, they introduce a projection residual margin derived from historical trajectories and design an actor regularization mechanism to reduce reliance on the safety filter. Experiments demonstrate that the method achieves zero constraint violations in the CartPole task while matching the return of unconstrained SAC, and significantly reduces violations in high-dimensional Safety Gymnasium tasks. The study also reveals limitations of first-order velocity barriers and linear EDMD models.
📝 Abstract
Safe reinforcement learning (RL) for robotic systems requires policies that improve task performance while satisfying state and input constraints during both training and deployment. Control barrier functions (CBFs) provide a principled mechanism for enforcing forward invariance through minimally invasive safety filters, but their use in model-free RL is limited by the need for accurate dynamics and hand-designed barrier certificates. We propose Robust Koopman-CBF SAC, a safety-filtered actor--critic framework that learns a finite-dimensional Koopman predictor from data, constructs affine CBF constraints in the lifted space, and enforces them through a quadratic-program safety layer. To account for finite-dimensional Koopman approximation error, the CBF condition is tightened using a projected residual margin estimated from held-out rollout data. The critic is trained on the executed safe action, while the actor is regularized toward the Koopman-CBF feasible set, reducing dependence on the filter over training. Across safe-control benchmarks, the method achieves zero constraint violations on CartPole stabilization and tracking while matching or exceeding unconstrained SAC returns. On high-dimensional Safety Gymnasium locomotion tasks, the method reduces violations in some settings but also exposes important limitations of first-order velocity barriers and linear EDMD models, motivating high-order and multi-step Koopman-CBF extensions. These results suggest that robust Koopman-CBF filters are a promising bridge between model-free RL and certifiable safety, while clarifying the structural conditions under which such filters remain effective. All code is available at \href{https://github.com/DhruvKushwaha/Koopman-CBF-Soft-Actor-Critic}{Github Repository}.