This research focuses on 2D image perception and multi-sensor 3D surround perception methods, aiming to develop end-to-end solutions that integrate perception and decision-making. The goal is to build a general framework for unified autonomous driving perception and decision-making, driving the advancement of autonomous driving technology into a new stage.

Representative Work:

Unified BEV (Bird’s-Eye View) Perception Framework, feature construction and model optimization in bird’s-eye view representation

• BEVFormer: Learning Bird's-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers

  [6th Most Influential Paper at ECCV 2022]

  [Top 100 Most Cited AI Papers of 2022]

  [Winner of the Waymo Open Dataset Challenge 2022]

• BEVFormer v2: Adapting Modern Image Backbones to Bird's-Eye-View Recognition via Perspective Supervision

  [CVPR 2023 highlight paper]

End-to-End Autonomous Driving Solution Integrating Perception and Decision-Making

• Planning-oriented Autonomous Driving

  [CVPR 2023 best paper]