OpenNovelty.org

Online Navigation Refinement: Achieving Lane-Level Guidance by Associating Standard-Definition and Online Perception Maps

ICLR 2026 Conference SubmissionAnonymous Authors
OpenReview Score: 6.0Download Report PDF
online navigation refinementgeographic information systemsnavigationstandard definition maponline perception map
Abstract:

Lane-level navigation is critical for geographic information systems and navigation-based tasks, offering finer-grained guidance than road-level navigation by standard definition (SD) maps. However, it currently relies on expansive global HD maps that cannot adapt to dynamic road conditions. Recently, online perception (OP) maps have become research hotspots, providing real-time geometry as an alternative, but lack the global topology needed for navigation. To address these issues, Online Navigation Refinement (ONR), a new mission is introduced that refines SD-map-based road-level routes into accurate lane-level navigation by associating SD maps with OP maps. The map-to-map association to handle many-to-one lane-to-road mappings under two key challenges: (1) no public dataset provides lane-to-road correspondences; (2) severe misalignment from spatial fluctuations, semantic disparities, and OP map noise invalidates traditional map matching. For these challenges, We contribute: (1) Online map association dataset (OMA), the first ONR benchmark with 30K scenarios and 2.6M annotated lane vectors; (2) MAT, a transformer with path-aware attention to aligns topology despite spatial fluctuations and semantic disparities and spatial attention for integrates noisy OP features via global context; and (3) NR P-R, a metric evaluating geometric and semantic alignment. Experiments show that MAT outperforms existing methods at 34 ms latency, enabling low-cost and up-to-date lane-level navigation.

Disclaimer
This report is AI-GENERATED using Large Language Models and WisPaper (A scholar search engine). It analyzes academic papers' tasks and contributions against retrieved prior work. While this system identifies POTENTIAL overlaps and novel directions, ITS COVERAGE IS NOT EXHAUSTIVE AND JUDGMENTS ARE APPROXIMATE. These results are intended to assist human reviewers and SHOULD NOT be relied upon as a definitive verdict on novelty.
NOTE that some papers exist in multiple, slightly different versions (e.g., with different titles or URLs). The system may retrieve several versions of the same underlying work. The current automated pipeline does not reliably align or distinguish these cases, so human reviewers will need to disambiguate them manually.
If you have any questions, please contact: mingzhang23@m.fudan.edu.cn

Overview

Overall Novelty Assessment

The paper introduces Online Navigation Refinement (ONR), a framework that associates standard-definition maps with online perception maps to refine road-level routes into lane-level navigation. It resides in the 'SD-to-Online Map Association for Hybrid Navigation' leaf, which contains only two papers including this work. This represents a relatively sparse research direction within the broader taxonomy of fourteen papers, suggesting the specific problem of many-to-one lane-to-road association under spatial misalignment has received limited prior attention in the examined literature.

The taxonomy reveals that neighboring research directions focus on map-enhanced topology reasoning and sensor fusion for lane-level positioning. While sibling topics like 'Map-Enhanced Lane Topology Reasoning' leverage map priors to improve connectivity prediction, they do not explicitly address the SD-to-online association challenge under severe spatial fluctuations and semantic disparities. The broader 'Sensor Fusion and Localization' branch emphasizes multi-modal positioning accuracy but typically assumes either HD maps or operates without map priors, creating a distinct boundary from this work's hybrid SD-online approach.

Among thirty candidates examined, none clearly refute the three core contributions: the OMA dataset (ten candidates examined, zero refutable), the MAT transformer architecture (ten candidates, zero refutable), and the NR P-R metric (ten candidates, zero refutable). This limited search scope suggests that within the top-thirty semantically similar papers, no prior work directly provides the same combination of lane-to-road correspondence annotations, path-aware attention mechanisms for topology alignment, or navigation-specific evaluation metrics. The absence of refutable candidates across all contributions indicates potential novelty, though the search scale leaves open the possibility of relevant work beyond these thirty papers.

Based on the examined literature, the work appears to occupy a distinct position addressing a specific gap: associating coarse SD maps with noisy online perception under many-to-one mappings. The taxonomy structure and contribution-level statistics suggest novelty in both the problem formulation and the proposed solutions, though the analysis is constrained by the top-thirty semantic search scope and does not claim exhaustive coverage of all related map association or lane-level navigation research.

Taxonomy

Core-task Taxonomy Papers
14
3
Claimed Contributions
30
Contribution Candidate Papers Compared
0
Refutable Paper

Research Landscape Overview

Core task: Associating standard-definition maps with online perception maps for lane-level navigation. The field addresses how autonomous vehicles can achieve precise lane-level positioning by combining coarse prior maps with real-time sensor observations. The taxonomy reveals four main branches: Map Association and Integration Frameworks focus on methods that fuse standard-definition (SD) map data with dynamically constructed perception maps, often exploring hybrid strategies that balance prior knowledge and online updates. Sensor Fusion and Localization for Lane-Level Positioning emphasizes multi-modal sensor integration—such as camera, GPS, and radar—to refine vehicle positioning within lane boundaries. Perception Algorithms and Map Construction deals with the computational techniques for detecting lanes and building local representations from sensor streams, including works like Lane Detection GPS[8] and Multiple Lanes Vision[12]. Autonomous Navigation Systems and Map Reliability examines end-to-end navigation pipelines and the robustness of map-based guidance, with efforts like Navigation Without HD Maps[7] exploring alternatives to high-definition priors. A particularly active line of work investigates how to augment or replace expensive high-definition maps with lighter SD maps enriched by online perception. Hybrid Navigation Framework[2] and SMART Map Priors[3] exemplify strategies that leverage prior topology while adapting to real-time observations, balancing computational cost and accuracy. Online Navigation Refinement[0] sits within this SD-to-online association cluster, emphasizing the integration of coarse map priors with dynamically perceived lane structures. Compared to Hybrid Navigation Framework[2], which may focus on broader multi-sensor fusion architectures, Online Navigation Refinement[0] appears to concentrate more directly on the association mechanism itself—how to align and update SD map elements with live perception outputs. Meanwhile, works like Augmenting Lane Perception[1] and Camera Lane Integration[4] highlight the role of vision-based methods in constructing reliable online maps, underscoring ongoing questions about sensor modality trade-offs and the minimal map fidelity required for safe lane-level navigation.

Claimed Contributions

Online Map Association Dataset (OMA)

The authors introduce OMA, the first public benchmark dataset for online navigation refinement that provides structured lane-to-road correspondences between standard definition maps and online perception maps, derived from nuScenes and OpenStreetMap with manual annotations.

10 retrieved papers
Map Association Transformer (MAT)

The authors propose MAT, a lightweight transformer-based model that uses path-aware attention to handle topological alignment and spatial attention to integrate noisy online perception map features through global context modeling, enabling real-time map association.

10 retrieved papers
Navigation Refinement Precision-Recall (NR P-R) Metric

The authors develop NR P-R, a new evaluation metric that measures both geometric similarity and association accuracy for map alignment tasks, designed to evaluate any map generation method using only ground-truth perception map annotations.

10 retrieved papers

Core Task Comparisons

Comparisons with papers in the same taxonomy category

Contribution Analysis

Detailed comparisons for each claimed contribution

Contribution

Online Map Association Dataset (OMA)

The authors introduce OMA, the first public benchmark dataset for online navigation refinement that provides structured lane-to-road correspondences between standard definition maps and online perception maps, derived from nuScenes and OpenStreetMap with manual annotations.

Contribution

Map Association Transformer (MAT)

The authors propose MAT, a lightweight transformer-based model that uses path-aware attention to handle topological alignment and spatial attention to integrate noisy online perception map features through global context modeling, enabling real-time map association.

Contribution

Navigation Refinement Precision-Recall (NR P-R) Metric

The authors develop NR P-R, a new evaluation metric that measures both geometric similarity and association accuracy for map alignment tasks, designed to evaluate any map generation method using only ground-truth perception map annotations.

Table of Contents