The field of autonomous driving is witnessing significant developments in perception and planning, with a focus on enhancing the accuracy and reliability of end-to-end driving systems. Researchers are exploring innovative approaches to integrate bird's-eye view (BEV) perception, spatial-aware representations, and gated fusion mechanisms to improve the performance of autonomous vehicles in complex scenarios. Noteworthy papers in this area include ME$^3$-BEV, which proposes a novel approach to autonomous driving using deep reinforcement learning and BEV perception, and GMF-Drive, which introduces a hierarchical gated mamba fusion architecture for end-to-end autonomous driving. Other notable works include Risk Map as Middleware, which provides an interpretable cooperative end-to-end driving framework, and CBDES MoE, which proposes a hierarchically decoupled Mixture-of-Experts architecture for functional modules in autonomous driving.
Advancements in Autonomous Driving Perception and Planning
Sources
ME$^3$-BEV: Mamba-Enhanced Deep Reinforcement Learning for End-to-End Autonomous Driving with BEV-Perception
FMCE-Net++: Feature Map Convergence Evaluation and Training
GMF-Drive: Gated Mamba Fusion with Spatial-Aware BEV Representation for End-to-End Autonomous Driving
Progressive Bird's Eye View Perception for Safety-Critical Autonomous Driving: A Comprehensive Survey
Decoupled Functional Evaluation of Autonomous Driving Models via Feature Map Quality Scoring
Risk Map As Middleware: Towards Interpretable Cooperative End-to-end Autonomous Driving for Risk-Aware Planning
CBDES MoE: Hierarchically Decoupled Mixture-of-Experts for Functional Modules in Autonomous Driving
The SET Perceptual Factors Framework: Towards Assured Perception for Autonomous Systems