EgoFSD: Ego-Centric Fully Sparse Paradigm with Uncertainty Denoising and Iterative Refinement for End-To-End Self-Driving
Haisheng Su, Wei Wu, Zhenjie Yang, Isabel Guan
AI summary
Problem
Current end-to-end autonomous driving systems rely on dense Bird’s Eye View features or exhaustively model all surrounding agents, leading to computational redundancy, information loss, and inferior efficiency.
Approach
EgoFSD introduces an ego-centric fully sparse paradigm that hierarchically selects only the closest in-path vehicles and stationary objects, then jointly predicts motion and iteratively refines trajectories using position-level diffusion and trajectory-level denoising.
Key results
- 59% reduction in average L2 error
- 92% decrease in collision rate
- 6.9× faster running efficiency
- Superior performance on nuScenes and Bench2Drive datasets
Why it matters
Provides a highly efficient and reliable blueprint for real-time end-to-end autonomous driving by eliminating dense feature bottlenecks and focusing resources on ego-relevant interactions.
Abstract
Current End-to-End Autonomous Driving (E2E- AD) methods resort to unifying modular designs for various tasks (e.g. perception, prediction and planning). Although optimized with a fully differentiable framework in a planning- oriented manner, existing end-to-end driving systems lacking ego-centric designs still suffer from unsatisfactory performance and inferior efficiency, due to rasterized scene representation learning and redundant information transmission. In this pa- per, we propose an ego-centric fully sparse paradigm, named EgoFSD, for end-to-end self-driving. Specifically, EgoFSD con- sists of sparse perception, hierarchical interaction and iterative motion planner. The sparse perception module performs detec- tion and online mapping based on sparse representation of the driving scene. The hierarchical interaction module aims to select the Closest In-Path Vehicle / Stationary (CIPV / CIPS) from coarse to fine, benefiting from an additional geometric prior. As for the iterative motion planner, both selected interactive agents and ego-vehicle are considered for joint motion prediction, where the output multi-modal ego-trajectories are optimized in an iterative fashion. In addition, position-level motion diffusion and trajectory-level planning denoising are introduced for uncertainty modeling, thereby enhancing the training stability and convergence speed. Extensive experiments are conducted on nuScenes and Bench2Drive datasets, which significantly reduces the average L2 error by 59% and collision rate by 92% than UniAD while achieves 6.9× faster running efficiency.