BEVDriver combines the reasoning capabilities of Large Language Models with Bird's-Eye View (BEV) maps to create a more reliable autonomous driving system that can handle complex scenarios while maintaining transparency.

• Integrates LLMs as generalized decision-makers for vehicle motion planning
• Leverages BEV representations for improved spatial awareness and navigation
• Demonstrates robust closed-loop driving performance in challenging environments
• Enhances human-AI interaction through natural language capabilities

This research advances autonomous driving engineering by addressing key challenges in reliability, safety, and transparency—critical factors for building public trust in self-driving technology and accelerating industry adoption.

BEVDriver: Leveraging BEV Maps in LLMs for Robust Closed-Loop Driving