Fu Yanping ; Liao Wenbin ; Liu Xinyuan ; xu Hang ; Ma Yike ; Dai Feng ; Zhang Yucheng

Abstract

As an emerging task that integrates perception and reasoning, topologyreasoning in autonomous driving scenes has recently garnered widespreadattention. However, existing work often emphasizes "perception over reasoning":they typically boost reasoning performance by enhancing the perception of lanesand directly adopt MLP to learn lane topology from lane query. This paradigmoverlooks the geometric features intrinsic to the lanes themselves and areprone to being influenced by inherent endpoint shifts in lane detection. To tackle this issue, we propose an interpretable method for lane topologyreasoning based on lane geometric distance and lane query similarity, namedTopoLogic. This method mitigates the impact of endpoint shifts in geometric space, andintroduces explicit similarity calculation in semantic space as a complement.By integrating results from both spaces, our methods provides morecomprehensive information for lane topology. Ultimately, our approach significantly outperforms the existingstate-of-the-art methods on the mainstream benchmark OpenLane-V2 (23.9 v.s.10.9 in TOP$_{ll}$ and 44.1 v.s. 39.8 in OLS on subset_A. Additionally, ourproposed geometric distance topology reasoning method can be incorporated intowell-trained models without re-training, significantly boost the performance oflane topology reasoning. The code is released athttps://github.com/Franpin/TopoLogic.