Diffusion-Planner and Hyper-Diffusion-Planner
These are ecosystem siblings where Hyper-Diffusion-Planner builds upon and extends the foundational diffusion-based planning approach of Diffusion-Planner, progressing from flexible guidance mechanisms to end-to-end autonomous driving capabilities.
About Diffusion-Planner
ZhengYinan-AIR/Diffusion-Planner
[ICLR 2025 Oral] The official implementation of "Diffusion-Based Planning for Autonomous Driving with Flexible Guidance"
Diffusion Planner employs a DiT-based architecture that jointly models vehicle trajectories and environmental context as a unified future trajectory generation problem, eliminating heavy reliance on iterative refinement. It achieves ~20Hz real-time inference through fast diffusion sampling and supports flexible guidance mechanisms (e.g., classifier guidance) for controllable planning. The framework is evaluated on nuPlan's closed-loop benchmark and integrates with nuplan-devkit for end-to-end autonomous driving validation.
About Hyper-Diffusion-Planner
ZhengYinan-AIR/Hyper-Diffusion-Planner
The official implementation of "Unleashing the Potential of Diffusion Models for End-to-End Autonomous Driving"
Scores updated daily from GitHub, PyPI, and npm data. How scores work