Driving with Context: Online Map Matching for Complex Roads Using Lane Markings and Scenario Recognition

TL;DR

This work tackles online map matching in complex, multilevel road networks where GNSS and SD-map limitations cause frequent errors. It introduces an HMM-based method that enriches SD maps with lane markings and incorporates driving scenario recognition to generate multiple emission factors, alongside standard pose-based emissions and connectivity-driven transitions. The lane-marking association forms a lane-level enriched SD map, while ICP-based re-localization and scenario-aware emissions reduce misattribution on road splits and elevated-underneath road pairs; the approach is integrated via a multi-factor HMM and solved with Viterbi. Large-scale experiments on the Zenseact Open Dataset and Shanghai roads show substantial gains in $F_1$ score and overall accuracy, demonstrating strong practical potential for robust online map matching in challenging urban environments.

Abstract

Accurate online map matching is fundamental to vehicle navigation and the activation of intelligent driving functions. Current online map matching methods are prone to errors in complex road networks, especially in multilevel road area. To address this challenge, we propose an online Standard Definition (SD) map matching method by constructing a Hidden Markov Model (HMM) with multiple probability factors. Our proposed method can achieve accurate map matching even in complex road networks by carefully leveraging lane markings and scenario recognition in the designing of the probability factors. First, the lane markings are generated by a multi-lane tracking method and associated with the SD map using HMM to build an enriched SD map. In areas covered by the enriched SD map, the vehicle can re-localize itself by performing Iterative Closest Point (ICP) registration for the lane markings. Then, the probability factor accounting for the lane marking detection can be obtained using the association probability between adjacent lanes and roads. Second, the driving scenario recognition model is applied to generate the emission probability factor of scenario recognition, which improves the performance of map matching on elevated roads and ordinary urban roads underneath them. We validate our method through extensive road tests in Europe and China, and the experimental results show that our proposed method effectively improves the online map matching accuracy as compared to other existing methods, especially in multilevel road area. Specifically, the experiments show that our proposed method achieves $F_1$ scores of 98.04% and 94.60% on the Zenseact Open Dataset and test data of multilevel road areas in Shanghai respectively, significantly outperforming benchmark methods. The implementation is available at https://github.com/TRV-Lab/LMSR-OMM.

Figures (12)

• Figure 1: A challenging scenario in a multilevel road area. In this scenario, the vehicle is driving on the ramp of an elevated road, but the positioning drifts to either the elevated road or the adjacent surface road, making it difficult to achieve accurate map matching.
• Figure 2: The overall framework of online map matching in complex road networks using lane markings and scenario recognition.
• Figure 3: Schematic diagram of calculating the distance from the vehicle to the road and the road heading angle. In the diagram, $p_N$ represent vehicle positioning points at time $N$, which form the vehicle trajectory together with $p_{N - 1}$, $p_{N - 2}$, and $p_{N - 3}$. $\theta_N$ denotes the vehicle heading angle at time $N$. Specifically, $d_1$, $d_2$, and $d_3$ are the distances from the vehicle's positioning point $p_N$ to roads $r_1$, $r_2$, and $r_3$, respectively. When a projection point exists on the road, the perpendicular distance is taken; otherwise, the distance to the road endpoint is used. Additionally, $\theta_1$, $\theta_2$, and $\theta_3$ represent the road heading angles at the nearest points from the vehicle to roads $r_1$, $r_2$, and $r_3$, respectively.
• Figure 4: An example of the lane markings generated by the multi-lane tracking method. Each red curve represents a tracked lane marking, and the black dashed curve represents the vehicle trajectory. The base map for reference is from Google Satellite Map.
• Figure 5: Schematic diagram of enriched SD map fusion. The thick polylineswith arrows are the roads of SD map, and the thin solid and dashed polylines represent the solid and dashed lane markings, respectively. In this figure, (a) shows the overlay of lane markings and the SD map in the same coordinate system, and note that there is no correlation between the lane markings and the SD map before performing the association. (b) shows the association result of multiple lane markings with multiple roads. The associated roads and lane markings are marked with the same color In this figure. One road can be associated with multiple lane marking instances, and different parts of one lane marking instance can also be associated with multiple roads with different probabilities.