Diverse Critical Interaction Generation for Planning and Planner Evaluation

TL;DR

This paper introduces RouteGAN, a generative model that creates diverse, style-controlled vehicle interactions for testing autonomous vehicle planners, improving the evaluation of safety and robustness in traffic scenarios.

Contribution

The paper presents RouteGAN, a novel styled generative model that produces diverse and critical vehicle interactions, addressing limitations of existing methods in naturalness and variety.

Findings

RouteGAN generates diverse interactions across various traffic scenarios.

The model can produce interactions with different safety levels by adjusting style coefficients.

Testing with RouteGAN reveals differences in collision rates among planners.

Abstract

Generating diverse and comprehensive interacting agents to evaluate the decision-making modules is essential for the safe and robust planning of autonomous vehicles~(AV). Due to efficiency and safety concerns, most researchers choose to train interactive adversary~(competitive or weakly competitive) agents in simulators and generate test cases to interact with evaluated AVs. However, most existing methods fail to provide both natural and critical interaction behaviors in various traffic scenarios. To tackle this problem, we propose a styled generative model RouteGAN that generates diverse interactions by controlling the vehicles separately with desired styles. By altering its style coefficients, the model can generate trajectories with different safety levels serve as an online planner. Experiments show that our model can generate diverse interactions in various scenarios. We evaluate different planners with our model by testing their collision rate in interaction with RouteGAN planners of multiple critical levels.