MapAgent: An Industrial-Grade Agentic Framework for City-scale Lane-level Map Generation
Researchers have developed MapAgent, an agentic framework designed to automate the generation of lane-level maps for urban environments. This system integrates explicit specification verification and constraint-aware reasoning with a vectorization backbone to ensure compliance with mapping standards and traffic regulations. MapAgent selectively applies its verification process to areas with lower confidence from the initial mapping, maintaining high throughput while improving accuracy, especially in complex scenarios. The framework has been successfully integrated into Baidu Maps, supporting map generation for over 360 cities and achieving over 95% automation. AI
IMPACT Automates a critical infrastructure task for autonomous driving, potentially accelerating deployment and reducing costs.