arXiv 2411.15139
DiffusionDrive: Truncated Diffusion Model for End-to-End Autonomous Driving
By Bencheng Liao, Shaoyu Chen, et al.
Published 2024-11-22
Citation lineage
Review the prior work and downstream research connected to this paper.
Recently, the diffusion model has emerged as a powerful generative technique for robotic policy learning, capable of modeling multi-mode action distributions. Leveraging its capability for end-to-end autonomous driving is a promising direction. However, the numerous denoising steps in the robotic diffusion policy and the more dynamic, open-world nature of traffic scenes pose substantial challenges for generating div…