TrafficBots: Towards World Models for Autonomous Driving Simulation and Motion Prediction
This work addresses the need for configurable and scalable traffic simulators for autonomous vehicle planning, representing an incremental improvement over existing data-driven methods.
The authors tackled the problem of data-driven traffic simulation for autonomous driving by formulating it as a world model, resulting in TrafficBots, which simulates realistic multi-agent behaviors and achieves good performance on motion prediction tasks as shown in experiments on the Waymo open motion dataset.
Data-driven simulation has become a favorable way to train and test autonomous driving algorithms. The idea of replacing the actual environment with a learned simulator has also been explored in model-based reinforcement learning in the context of world models. In this work, we show data-driven traffic simulation can be formulated as a world model. We present TrafficBots, a multi-agent policy built upon motion prediction and end-to-end driving, and based on TrafficBots we obtain a world model tailored for the planning module of autonomous vehicles. Existing data-driven traffic simulators are lacking configurability and scalability. To generate configurable behaviors, for each agent we introduce a destination as navigational information, and a time-invariant latent personality that specifies the behavioral style. To improve the scalability, we present a new scheme of positional encoding for angles, allowing all agents to share the same vectorized context and the use of an architecture based on dot-product attention. As a result, we can simulate all traffic participants seen in dense urban scenarios. Experiments on the Waymo open motion dataset show TrafficBots can simulate realistic multi-agent behaviors and achieve good performance on the motion prediction task.