Adversarial Agent Behavior Learning in Autonomous Driving Using Deep Reinforcement Learning

TL;DR

This paper introduces a deep reinforcement learning approach to generate adversarial behaviors in autonomous driving, aiming to identify failure scenarios by challenging rule-based surrounding agents.

Contribution

It presents a novel learning-based method to derive adversarial behaviors that cause failures in autonomous driving simulations, improving safety testing.

Findings

Adversarial agents decrease the cumulative reward of rule-based agents.

The method successfully identifies failure scenarios in autonomous driving.

Demonstrates effectiveness across multiple rule-based agent models.

Abstract

Existing approaches in reinforcement learning train an agent to learn desired optimal behavior in an environment with rule based surrounding agents. In safety critical applications such as autonomous driving it is crucial that the rule based agents are modelled properly. Several behavior modelling strategies and IDM models are used currently to model the surrounding agents. We present a learning based method to derive the adversarial behavior for the rule based agents to cause failure scenarios. We evaluate our adversarial agent against all the rule based agents and show the decrease in cumulative reward.