Provably Safe Reinforcement Learning via Action Projection using Reachability Analysis and Polynomial Zonotopes

Abstract

While reinforcement learning produces very promising results for many applications, its main disadvantage is the lack of safety guarantees, which prevents its use in safety-critical systems. In this work, we address this issue by a safety shield for nonlinear continuous systems that solve reach-avoid tasks. Our safety shield prevents applying potentially unsafe actions from a reinforcement learning agent by projecting the proposed action to the closest safe action. This approach is called action projection and is implemented via mixed-integer optimization. The safety constraints for action projection are obtained by applying parameterized reachability analysis using polynomial zonotopes, which enables to accurately capture the nonlinear effects of the actions on the system. In contrast to other state-of-the-art approaches for action projection, our safety shield can efficiently handle input constraints and dynamic obstacles, eases incorporation of the spatial robot dimensions into the safety constraints, guarantees robust safety despite process noise and measurement errors, and is well suited for high-dimensional systems, as we demonstrate on several challenging benchmark systems.

References in corpus (2)

• Safe Reinforcement Learning Using Black-Box Reachability Analysis
• Fully-Automated Verification of Linear Systems Using Inner- and Outer-Approximations of Reachable Sets

Cited by in corpus (4)

• Safety Filtering While Training: Improving the Performance and Sample Efficiency of Reinforcement Learning Agents
• Provable Traffic Rule Compliance in Safe Reinforcement Learning on the Open Sea
• Safe Autonomous Environmental Contact for Soft Robots using Control Barrier Functions
• Leveraging Analytic Gradients in Provably Safe Reinforcement Learning