Boundary Sampling to Learn Predictive Safety Filters via Pontryagin's Maximum Principle

TL;DR

This paper introduces a boundary sampling method based on Pontryagin's Maximum Principle to efficiently learn safety filters for autonomous systems, enhancing safety and learning speed.

Contribution

It proposes a novel boundary trajectory characterization to guide data collection for Hamilton-Jacobi Reachability, improving safety filter learning in high-dimensional systems.

Findings

PMP sampling improves learning efficiency and convergence speed.

The method reduces failure rates in safety-critical scenarios.

Achieves real-time safety filtering with around 3ms computation time.

Abstract

Safety filters provide a practical approach for enforcing safety constraints in autonomous systems. While learning-based tools scale to high-dimensional systems, their performance depends on informative data that includes states likely to lead to constraint violation, which can be difficult to efficiently sample in complex, high-dimensional systems. In this work, we characterize trajectories that barely avoid safety violations using the Pontryagin Maximum Principle. These boundary trajectories are used to guide data collection for learned Hamilton-Jacobi Reachability, concentrating learning efforts near safety-critical states to improve efficiency. The learned Control Barrier Value Function is then used directly for safety filtering. Simulations and experimental validation on a shared-control automotive racing application demonstrate PMP sampling improves learning efficiency, yielding faster convergence, reduced failure rates, and improved safe set reconstruction, with wall times around 3ms.