Road Selection Using Multicriteria Fusion for the Road-Matching Problem

Najjar, Maan El Badaoui El; Bonnifait, Philippe · 2007 · OpenAlex-citations

DOI: 10.1109/tits.2007.895312

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the road-matching problem in Advanced Driving-Assistance Systems (ADAS), specifically focusing on selecting the most likely road segment from a digital map database given an estimated vehicle position. The authors argue that traditional arc-matching methods, which project vehicle positions onto road segments, introduce geometric distortions and fail to account for lateral variations or ambiguous situations. Instead, they propose an absolute localization approach where the vehicle’s pose is estimated in the map frame, and the key challenge becomes selecting the correct road segment without immediate projection. This selection is critical for real-time navigation and safety applications, as it reduces processing load and ensures robust localization despite sensor imprecision and map inaccuracies. The methodology combines sensor fusion with a multi-criteria decision strategy based on Belief Theory. First, the vehicle’s pose (position and heading) is continuously estimated using an Extended Kalman Filter (EKF) that fuses data from a differential GPS receiver and an odometric system derived from Anti-Lock Braking System (ABS) sensors on the rear wheels. This fusion provides high-availability positioning, compensating for GPS outages in urban environments or tunnels. The EKF also quantifies the estimation error covariance, which is essential for the subsequent selection process. The road selection strategy employs Belief Theory to merge two primary criteria: proximity and heading alignment. The proximity criterion evaluates the Euclidean distance between the estimated position and candidate road segments, weighted by the position error ellipse. The heading criterion assesses the angular difference between the vehicle’s estimated heading and the orientation of the road segments, considering the heading error bounds and vehicle speed. Experimental results were obtained using data from an experimental vehicle equipped with ABS sensors and a differential GPS receiver, tested on a 4.5 km route using an accurate digital roadmap (Géoroute V2). The study demonstrates that the EKF effectively fuses GPS and odometry data, providing consistent pose estimates with quantified uncertainty. The Belief Theory-based selection strategy successfully identifies the correct road segments, particularly in ambiguous situations where multiple roads are close to the estimated position. The approach handles uncertainties from sensor noise, map errors, and coordinate transformations by assigning mass functions to hypotheses (Yes, No, Perhaps) regarding segment validity. The local strategy for mass assignment, which treats each segment separately, proved more effective for real-time applications than global strategies. The significance of this work lies in its robust handling of uncertainty in vehicle localization. By using Belief Theory, the method can quantify the degree of belief in selected roads and detect off-road situations when all segments have low belief values. This approach is particularly valuable for safety-critical ADAS applications, such as curve warning systems, where reliable road matching is essential. The integration of low-cost ABS sensors with GPS provides a cost-effective solution for continuous localization, while the multi-criteria fusion enhances accuracy and reliability compared to single-criterion methods. The paper contributes to the field by offering a formal framework for road selection that accounts for both sensor and map uncertainties, improving the performance of map-matching algorithms in complex driving environments.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• gaze based attention detection