Interacting safely with cyclists using Hamilton-Jacobi reachability and reinforcement learning

2026-02-20

AI Summary

This work addresses the safety–efficiency trade-off faced by autonomous vehicles when interacting with cyclists. The authors propose a novel approach that integrates Hamilton–Jacobi reachability analysis with deep Q-learning: by solving a time-varying Hamilton–Jacobi–Bellman inequality, they construct a state-dependent safety metric and embed it as a structured reward within a reinforcement learning framework. Furthermore, a model of cyclist responses to vehicle behavior is incorporated to enhance interaction naturalness. This study is the first to embed the rigorous safety guarantees provided by Hamilton–Jacobi reachability directly into the reward design of reinforcement learning, while also integrating a human behavioral feedback mechanism to improve the realism and comfort of human–autonomy interaction. Simulations demonstrate that the method significantly improves traffic efficiency without compromising safety, outperforming state-of-the-art approaches and yielding driving behaviors more aligned with human norms.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Natural Language Processing: Safety and Robustness
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Research challenges in human and human-AI computation
• Security and Privacy: Large-scale security measurements

Abstract

In this paper, we present a framework for enabling autonomous vehicles to interact with cyclists in a manner that balances safety and optimality. The approach integrates Hamilton-Jacobi reachability analysis with deep Q-learning to jointly address safety guarantees and time-efficient navigation. A value function is computed as the solution to a time-dependent Hamilton-Jacobi-Bellman inequality, providing a quantitative measure of safety for each system state. This safety metric is incorporated as a structured reward signal within a reinforcement learning framework. The method further models the cyclist's latent response to the vehicle, allowing disturbance inputs to reflect human comfort and behavioral adaptation. The proposed framework is evaluated through simulation and comparison with human driving behavior and an existing state-of-the-art method.

Problem

• autonomous vehicles
• cyclist interaction
• safety
• optimal navigation
• human-vehicle interaction

Innovation

• Hamilton-Jacobi reachability
• reinforcement learning
• autonomous driving
• human-robot interaction
• safety-critical control