TSDiT: Traffic Scene Diffusion Models With Transformers

TL;DR

This paper presents TSDiT, a novel traffic scene diffusion model with transformers that generates diverse, realistic vehicle trajectories for autonomous driving, improving scene understanding and prediction accuracy.

Contribution

Introducing a diffusion-transformer hybrid model for trajectory generation that enhances diversity and realism in traffic scene predictions.

Findings

Outperforms existing methods in trajectory smoothness and accuracy

Produces diverse and realistic agent trajectories

Effective in complex traffic scenarios

Abstract

In this paper, we introduce a novel approach to trajectory generation for autonomous driving, combining the strengths of Diffusion models and Transformers. First, we use the historical trajectory data for efficient preprocessing and generate action latent using a diffusion model with DiT(Diffusion with Transformers) Blocks to increase scene diversity and stochasticity of agent actions. Then, we combine action latent, historical trajectories and HD Map features and put them into different transformer blocks. Finally, we use a trajectory decoder to generate future trajectories of agents in the traffic scene. The method exhibits superior performance in generating smooth turning trajectories, enhancing the model's capability to fit complex steering patterns. The experimental results demonstrate the effectiveness of our method in producing realistic and diverse trajectories, showcasing its potential for application in autonomous vehicle navigation systems.