Sensor Fusion and Efficient Data Processing in Autonomous Driving

The recent advancements in autonomous driving research have significantly focused on enhancing sensor fusion techniques and improving the efficiency of data processing. A notable trend is the integration of radar and camera data for 3D object detection, addressing the limitations of each sensor by leveraging their complementary strengths. Innovations in transformer-based models and query-based frameworks have shown promising results in accurately detecting objects in complex environments. Additionally, there is a growing emphasis on developing efficient LiDAR data alignment methods for precise mapping and localization, with continuous-time trajectory optimization emerging as a key technique. Furthermore, the field is witnessing the development of high-precision on-device depth perception systems, which combine multiple data sources to achieve superior depth mapping accuracy. These developments collectively push the boundaries of autonomous driving technology, aiming for safer and more reliable systems.

Sources

SEGT: A General Spatial Expansion Group Transformer for nuScenes Lidar-based Object Detection Task

Timealign: A multi-modal object detection method for time misalignment fusing in autonomous driving

HGSFusion: Radar-Camera Fusion with Hybrid Generation and Synchronization for 3D Object Detection

Efficient LiDAR Bundle Adjustment for Multi-Scan Alignment Utilizing Continuous-Time Trajectories

RaCFormer: Towards High-Quality 3D Object Detection via Query-based Radar-Camera Fusion

RCTrans: Radar-Camera Transformer via Radar Densifier and Sequential Decoder for 3D Object Detection

4D Radar-Inertial Odometry based on Gaussian Modeling and Multi-Hypothesis Scan Matching

MobiFuse: A High-Precision On-device Depth Perception System with Multi-Data Fusion

SCKD: Semi-Supervised Cross-Modality Knowledge Distillation for 4D Radar Object Detection

PC-BEV: An Efficient Polar-Cartesian BEV Fusion Framework for LiDAR Semantic Segmentation

Built with on top of