Lane Network Graph Updates Using Probe Data and Bayesian Matching

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing navigation systems lack accurate lane-level information necessary for self-driving vehicles and advanced driver assistance systems, as current methods struggle to efficiently update lane network models using probe data, particularly in areas with spatial variance and concentration on lane centers.

Innovation Solution

A lane network graph (LNG) model is generated and updated using time-ordered sequences of location data from probe vehicles, incorporating Bayesian updates and map matching techniques to add new lanes and refine node positions, headings, and uncertainty measures, linked to a localization layer for enhanced accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If probe data is used to update lane network models, then lane-level navigation accuracy is improved, but the complexity of data processing and model updating increases

Engineering Contradiction:
Improvelane-level navigation accuracyVSAvoiddata processing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system automatically processes probe data and updates the lane network graph model without manual intervention. The model self-updates by integrating new drive data, matching locations to existing nodes, and adjusting lane representations based on accumulated probe information, eliminating the need for manual map updates.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system continuously collects probe data from vehicles and uses it to refine and update the lane network graph model. The updated model then provides improved navigation guidance back to vehicles, creating a closed-loop feedback system where real-world driving data continuously improves the accuracy of the navigation infrastructure.

Inventive Principle:
Principle #23Feedback

2Adaptability or versatility

If the lane network model is continuously updated with new drives, then the adaptability to changing road conditions is improved, but the computational resources and time required for updating increase

Engineering Contradiction:
Improveadaptability to changing road conditionsVSAvoidmodel updating time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The system performs map matching by comparing probe locations against the existing lane network graph model before full integration. This preliminary matching step identifies which probe data points correspond to known lanes and intersections, allowing the system to efficiently determine what updates are actually needed rather than processing all data comprehensively.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system updates only the portions of the lane network graph model that are affected by new probe data. Rather than reprocessing the entire road network, it selectively updates specific lanes, intersections, or road segments where probe data indicates changes or improvements, significantly reducing computational overhead.

Inventive Principle:
Principle #16Partial or excessive action

3Manufacturing precision

If Bayesian updates are applied to refine node positions and headings, then the precision of lane network data is improved, but the computational complexity of each update operation increases

Engineering Contradiction:
Improvenode position precisionVSAvoidupdate operation complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The system uses probabilistic uncertainty estimates that can be quickly calculated and discarded after each update cycle. Rather than maintaining complex, persistent models of every possible error source, it uses simplified uncertainty representations that are computationally inexpensive to calculate and update, enabling frequent refinements without excessive computational burden.

Inventive Principle:
Principle #27Cheap short-living objects (Disposable)

Data Source

PatentEP3769296B1Generation and update of a lane network graph model
Publication Date: 2026.01.28 HERE GLOBAL BV
  • EP3769296B1 patent drawingFigure 1
  • EP3769296B1 patent drawingFigure 2A~2B
  • EP3769296B1 patent drawingFigure 3

AI summary

A new drive includes a plurality of instances of location data that are received. The location data was captured by a probe apparatus onboard a vehicle as the vehicle traversed a lane network. The plurality of instances of location data are map matched using a lane level map matcher. The matched nodes (e.g., nodes of the LNG model representing the lane network to which an instance of location data was map matched) are identified and updated based on the new drive. Unmatched instances of location data (e.g., instances of location data that cannot be satisfactorily matched to a node of the LNG model) may be used to generate new nodes of the LNG model. A portion of the LNG model may be updated based on an update to a linked localization layer.