Search papers, labs, and topics across Lattice.
This paper introduces MapAgent, an advanced framework designed to enhance lane-level map generation for autonomous driving by integrating explicit specification verification and constraint-aware reasoning into the mapping process. By employing a Judge-Planner-Worker architecture, MapAgent effectively diagnoses mapping errors and generates minimal corrective edits, significantly improving the accuracy of lane configurations in complex urban environments. Experimental results demonstrate that MapAgent outperforms existing production baselines, achieving over 95% automation in lane-level map production across 360 cities in China.
MapAgent achieves over 95% automation in lane-level map generation, transforming how cities can maintain accurate navigation infrastructure.
Lane-level maps are critical infrastructure for autonomous driving and lane-level navigation, yet constructing and maintaining standardized lane networks for hundreds of cities remains highly labor-intensive. Recent end-to-end vectorized mapping methods can predict lane geometry and topology directly from sensor data, but they typically treat mapping specifications and traffic regulations as implicit, dataset-dependent supervision. Moreover, in complex scenes (e.g., worn or missing markings and occlusions), correct lane configurations are often under-determined by visual evidence alone, making specification violations a major source of human post-editing. We propose MapAgent, an industrial-grade agentic architecture that augments a vectorization backbone for specification-compliant lane-map production. Rather than merely adding an agent loop to map prediction, MapAgent couples backbone perception with explicit specification verification, constraint-aware reasoning, and deterministic map editing under a bounded, verification-driven Judge-Planner-Worker loop. A vision-language Judge diagnoses errors by jointly inspecting visual evidence and draft vectors, while a tool-calling Planner generates minimal corrective edits with post-edit re-validation. To remain scalable for city-scale production, MapAgent is selectively triggered only on tiles with low backbone confidence, adding modest overhead while preserving throughput. Experiments on real-world datasets show consistent gains over strong production baselines, especially in complex and long-tail scenarios. Additionally, MapAgent has been integrated into Baidu Maps, supporting lane-level map generation for over 360 cities nationwide and elevating the overall production automation to over 95%, demonstrating MapAgent's practicality and effectiveness for large-scale lane-level map generation.