FASTopoWM: Fast-Slow Lane Segment Topology Reasoning with Latent World Models

2025-07-31

AI Summary

Existing lane topology reasoning methods suffer from weak temporal modeling: they over-rely on historical queries, are sensitive to pose estimation errors, and lack sufficient temporal propagation. This paper proposes a novel fast-slow dual-path spatiotemporal co-modeling framework operating in the Bird’s Eye View (BEV) space, jointly performing lane segment detection and topology reasoning. We introduce an action-driven latent query mechanism and a lightweight BEV world model to enable robust historical state propagation. Additionally, we design a streaming temporal propagation scheme coupled with parallel supervision at both the query level and the BEV feature level to enhance spatiotemporal consistency. Evaluated on OpenLane-V2, our method achieves significant improvements: 37.4% mAP for lane segment detection and 46.3% OLS for centerline prediction—setting new state-of-the-art performance. The proposed approach delivers a more reliable and temporally robust solution for end-to-end autonomous driving systems.

Technology Category

• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Computer Vision: Large Vision Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Lane segment topology reasoning provides comprehensive bird's-eye view (BEV) road scene understanding, which can serve as a key perception module in planning-oriented end-to-end autonomous driving systems. Existing lane topology reasoning methods often fall short in effectively leveraging temporal information to enhance detection and reasoning performance. Recently, stream-based temporal propagation method has demonstrated promising results by incorporating temporal cues at both the query and BEV levels. However, it remains limited by over-reliance on historical queries, vulnerability to pose estimation failures, and insufficient temporal propagation. To overcome these limitations, we propose FASTopoWM, a novel fast-slow lane segment topology reasoning framework augmented with latent world models. To reduce the impact of pose estimation failures, this unified framework enables parallel supervision of both historical and newly initialized queries, facilitating mutual reinforcement between the fast and slow systems. Furthermore, we introduce latent query and BEV world models conditioned on the action latent to propagate the state representations from past observations to the current timestep. This design substantially improves the performance of temporal perception within the slow pipeline. Extensive experiments on the OpenLane-V2 benchmark demonstrate that FASTopoWM outperforms state-of-the-art methods in both lane segment detection (37.4% v.s. 33.6% on mAP) and centerline perception (46.3% v.s. 41.5% on OLS).

Problem

• Improving lane topology reasoning with temporal information
• Reducing reliance on historical queries and pose estimation
• Enhancing temporal perception via latent world models

Innovation

• Parallel supervision of historical and new queries
• Latent query and BEV world models
• Fast-slow lane segment topology reasoning