Red-Team Multi-Agent Reinforcement Learning for Emergency Braking Scenario

TL;DR

This paper introduces a red-team multi-agent reinforcement learning framework that actively generates corner cases in safety-critical scenarios, enhancing the robustness testing of autonomous vehicle decision-making.

Contribution

It proposes a novel red-team multi-agent RL framework with a constraint graph and threat zone model to uncover corner cases outside the data distribution.

Findings

Significantly impacts AV decision-making safety.

Generates diverse corner cases.

Provides a new approach for safety-critical scenario testing.

Abstract

Current research on decision-making in safety-critical scenarios often relies on inefficient data-driven scenario generation or specific modeling approaches, which fail to capture corner cases in real-world contexts. To address this issue, we propose a Red-Team Multi-Agent Reinforcement Learning framework, where background vehicles with interference capabilities are treated as red-team agents. Through active interference and exploration, red-team vehicles can uncover corner cases outside the data distribution. The framework uses a Constraint Graph Representation Markov Decision Process, ensuring that red-team vehicles comply with safety rules while continuously disrupting the autonomous vehicles (AVs). A policy threat zone model is constructed to quantify the threat posed by red-team vehicles to AVs, inducing more extreme actions to increase the danger level of the scenario. Experimental results show that the proposed framework significantly impacts AVs decision-making safety and generates various corner cases. This method also offers a novel direction for research in safety-critical scenarios.