Adversarial Agent Behavior Learning in Autonomous Driving Using Deep Reinforcement Learning
This work addresses safety-critical issues in autonomous driving by improving adversarial testing, though it appears incremental as it builds on existing rule-based agent models.
The paper tackles the problem of modeling adversarial behavior for rule-based agents in autonomous driving by introducing a learning-based method to generate failure scenarios, resulting in a demonstrated decrease in cumulative reward.
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.