RoadWeaver: Large-Scale Lane-Level HD Map Generation from Scratch for Autonomous Driving Simulation

2026-08-11

AI Summary

This work addresses the lack of scalable and diverse lane-level high-definition (HD) maps in autonomous driving simulation, a limitation exacerbated by conventional approaches that rely on manual or localized generation and struggle to produce large-scale, well-connected road networks. To overcome this, the authors propose a coarse-to-fine, end-to-end generative framework that synthesizes complete lane-level HD maps from scratch for the first time. The method first generates a global road layout, then refines network connectivity, and finally constructs geometrically accurate lane structures consistent with topological constraints. Achieving 99.8% map reachability and only 10.7% dead-end ratio, the approach reduces endpoint alignment error to 0.24 meters—94.4% lower than existing methods—while generating each map in just 1.39–3.50 seconds, substantially improving both quality and efficiency.

Abstract

Autonomous driving simulation requires diverse and scalable lane-level HD maps to support long-horizon evaluation across complex road networks. Existing approaches either rely on handcrafted or reconstructed real-world maps, which limits scalability, or generate only local road structures rather than complete HD maps. We present RoadWeaver, a coarse-to-fine framework for from-scratch generation of diverse, large-scale HD maps. RoadWeaver first synthesizes a global road layout, expands it into a connected road network, and then constructs lane-level geometry with topologically consistent lane connectivity. Experimental results show that RoadWeaver achieves a 99.8\% reachability, a 10.7\% dead-end ratio, and an endpoint alignment error of 0.24 m. Compared with SOTA generation methods, it reduces endpoint alignment error by 94.4\% while generating complete HD maps in 1.39--3.50 s. The generated maps can be directly deployed in driving simulators, providing scalable simulation environments for future closed-loop evaluation of autonomous driving systems. The training code and an out-of-the-box implementation of RoadWeaver will be released upon acceptance.

Problem

• HD map generation
• autonomous driving simulation
• large-scale road networks
• lane-level mapping
• map scalability

Innovation

• HD map generation
• autonomous driving simulation
• coarse-to-fine framework
• lane-level geometry
• topological consistency