SparseDrive: End-to-End Autonomous Driving via Sparse Scene Representation

vis_sparsedrive.mp4

News

• 24 June, 2024: We reorganize code for better readability. Code & Models are released.
• 31 May, 2024: We release the SparseDrive paper on arXiv. Code & Models will be released in June, 2024. Please stay tuned!

Introduction

SparseDrive is a Sparse-Centric paradigm for end-to-end autonomous driving.

• We explore the sparse scene representation for end-to-end autonomous driving and propose a Sparse-Centric paradigm named SparseDrive, which unifies multiple tasks with sparse instance representation.
• We revise the great similarity shared between motion prediction and planning, correspondingly leading to a parallel design for motion planner. We further propose a hierarchical planning selection strategy incorporating a collision-aware rescore module to boost the planning performance.
• On the challenging nuScenes benchmark, SparseDrive surpasses previous SOTA methods in terms of all metrics, especially the safety-critical metric collision rate, while keeping much higher training and inference efficiency.

Overview of SparseDrive. SparseDrive first encodes multi-view images into feature maps, then learns sparse scene representation through symmetric sparse perception, and finally perform motion prediction and planning in a parallel manner. An instance memory queue is devised for temporal modeling.

Model architecture of symmetric sparse perception, which unifies detection, tracking and online mapping in a symmetric structure.

Model structure of parallel motion planner, which performs motion prediction and planning simultaneously and outputs safe planning trajectory.

Results in paper

• Comprehensive results for all tasks on nuScenes.

Method	NDS	AMOTA	minADE (m)	L2 (m) Avg	Col. (%) Avg	Training Time (h)	FPS
UniAD	0.498	0.359	0.71	0.73	0.61	144	1.8
SparseDrive-S	0.525	0.386	0.62	0.61	0.08	20	9.0
SparseDrive-B	0.588	0.501	0.60	0.58	0.06	30	7.3

• Open-loop planning results on nuScenes.

Method	L2 (m) 1s	L2 (m) 2s	L2 (m) 3s	L2 (m) Avg	Col. (%) 1s	Col. (%) 2s	Col. (%) 3s	Col. (%) Avg	FPS
UniAD	0.45	0.70	1.04	0.73	0.62	0.58	0.63	0.61	1.8
VAD	0.41	0.70	1.05	0.72	0.03	0.19	0.43	0.21	4.5
SparseDrive-S	0.29	0.58	0.96	0.61	0.01	0.05	0.18	0.08	9.0
SparseDrive-B	0.29	0.55	0.91	0.58	0.01	0.02	0.13	0.06	7.3

Results of released checkpoint

We found that some collision cases were not taken into consideration in our previous code, so we re-implement the evaluation metric for collision rate in released code and provide updated results.

Main results

Model	config	ckpt	log	det: NDS	mapping: mAP	track: AMOTA	track: AMOTP	motion: EPA_car	motion: minADE_car	motion: minFDE_car	motion: MissRate_car	planning: CR	planning: L2
Stage1	cfg	ckpt	log	0.5260	0.5689	0.385	1.260
Stage2	cfg	ckpt	log	0.5257	0.5656	0.372	1.248	0.492	0.61	0.95	0.133	0.097%	0.61

Detailed results for planning

Method	L2 (m) 1s	L2 (m) 2s	L2 (m) 3s	L2 (m) Avg	Col. (%) 1s	Col. (%) 2s	Col. (%) 3s	Col. (%) Avg
UniAD	0.45	0.70	1.04	0.73	0.66	0.66	0.72	0.68
UniAD-wo-post-optim	0.32	0.58	0.94	0.61	0.17	0.27	0.42	0.29
VAD	0.41	0.70	1.05	0.72	0.03	0.21	0.49	0.24
SparseDrive-S	0.30	0.58	0.95	0.61	0.01	0.05	0.23	0.10

Quick Start

Quick Start

Citation

If you find SparseDrive useful in your research or applications, please consider giving us a star 🌟 and citing it by the following BibTeX entry.

@article{sun2024sparsedrive,
  title={SparseDrive: End-to-End Autonomous Driving via Sparse Scene Representation},
  author={Sun, Wenchao and Lin, Xuewu and Shi, Yining and Zhang, Chuang and Wu, Haoran and Zheng, Sifa},
  journal={arXiv preprint arXiv:2405.19620},
  year={2024}
}

Acknowledgement

• Sparse4D
• UniAD
• VAD
• StreamPETR
• StreamMapNet
• mmdet3d