RATopo: Improving Lane Topology Reasoning via Redundancy Assignment

2025-08-21

AI Summary

Existing lane topology inference methods predominantly adopt a “detection + one-to-one matching” paradigm, suffering from sparse supervision and insufficient geometric diversity, which severely limits their capacity for modeling topological connections. To address this, we propose RATopo—a Redundant-Assignment Topology Inference framework. RATopo reconstructs the Transformer decoder by swapping the order of self-attention and cross-attention layers and introduces a multi-branch parallel cross-attention module. Moreover, it employs a one-to-many label assignment strategy to retain redundant predictions, thereby enriching both the quantity and geometric diversity of topological supervision. This design breaks the conventional matching bottleneck and exhibits model-agnosticism and plug-and-play compatibility. Evaluated on OpenLane-V2, RATopo significantly improves accuracy in inferring lane–lane and lane–traffic-element topological relationships, while seamlessly integrating with mainstream end-to-end architectures.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Computer Vision: Large Vision Models
• Machine Learning: Mixture of Experts (MoE)

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in 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 topology reasoning plays a critical role in autonomous driving by modeling the connections among lanes and the topological relationships between lanes and traffic elements. Most existing methods adopt a first-detect-then-reason paradigm, where topological relationships are supervised based on the one-to-one assignment results obtained during the detection stage. This supervision strategy results in suboptimal topology reasoning performance due to the limited range of valid supervision. In this paper, we propose RATopo, a Redundancy Assignment strategy for lane Topology reasoning that enables quantity-rich and geometry-diverse topology supervision. Specifically, we restructure the Transformer decoder by swapping the cross-attention and self-attention layers. This allows redundant lane predictions to be retained before suppression, enabling effective one-to-many assignment. We also instantiate multiple parallel cross-attention blocks with independent parameters, which further enhances the diversity of detected lanes. Extensive experiments on OpenLane-V2 demonstrate that our RATopo strategy is model-agnostic and can be seamlessly integrated into existing topology reasoning frameworks, consistently improving both lane-lane and lane-traffic topology performance.

Problem

• Improving lane topology reasoning in autonomous driving systems
• Addressing suboptimal supervision in lane connectivity modeling
• Enhancing diversity and quantity of topological relationship assignments

Innovation

• Swapping cross-attention and self-attention layers
• Using multiple parallel cross-attention blocks
• Enabling one-to-many assignment for supervision