ARETE: Attention-based Rasterized Encoding for Topology Estimation using HSV-transformed Crowdsourced Vehicle Fleet Data

2026-04-27

AI Summary

This work addresses the challenge of generating road vectors for real-time high-definition map updates by proposing an end-to-end method based on Detection Transformer (DETR). Leveraging crowdsourced vehicle trajectories, the approach efficiently produces high-precision, direction-aware road centerlines and lane dividers. The key innovation lies in aggregating trajectories into local rasterized patches and encoding both presence and directional information using the HSV color space, which is then integrated with geometric constraints within the DETR framework to ensure geometrically consistent lane topology estimation. Experiments on an internal dataset as well as public benchmarks—including nuScenes and nuPlan—demonstrate the method’s effectiveness, achieving significant improvements in both the accuracy and directional plausibility of vectorized lanes.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Computer Vision: Large Vision Models
• Intelligent Robots: Localization, Mapping, and Navigation

Application Category

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

Abstract

The continuous advancement of autonomous driving (AD) introduces challenges across multiple disciplines to ensure safe and efficient driving. One such challenge is the generation of High-Definition (HD) maps, which must remain up to date and highly accurate for downstream automotive tasks. One promising approach is the use of crowdsourced data from a vehicle fleet, representing road topology and lane-level features. This work focuses on the generation of centerlines and lane dividers from crowdsourced vehicle trajectories. We adopt a Detection Transformer (DETR)-based approach, where a rasterized representation of vehicle trajectories is used as input to predict vectorized lane representations. Each lane consists of a centerline with an associated direction and corresponding lane dividers that are geometrically constrained by the centerline. Our method includes the extraction of local tiles, from which crowdsourced vehicle trajectories are aggregated. Each tile undergoes a transformation into a rasterized representation encoding both the presence and direction of each trajectory, enabling the prediction of vectorized directed lanes. Experiments are conducted on an internal dataset as well as on the public datasets nuScenes and nuPlan.

Problem

• HD maps
• crowdsourced vehicle trajectories
• lane topology estimation
• centerline extraction
• autonomous driving

Innovation

• Attention-based
• Rasterized Encoding
• Topology Estimation
• Crowdsourced Trajectories
• Vectorized Lane Prediction