Lane-level geometry from probe data deconvolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current GPS systems lack the accuracy to determine the specific lane of a road segment, making it difficult to gather lane-level traffic information and perform lane-level navigation due to the narrow width of lanes and inherent location estimation errors.
Innovation Solution
A system that uses probe data from multiple vehicles to generate a probe density histogram, applies deconvolution methods like the Maximum Entropy Method to identify statistically significant peaks representing lanes, and computes lane-level properties to provide navigational assistance and semi-autonomous vehicle control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS location data is used to determine vehicle position, then real-time location information is obtained, but the accuracy is insufficient to identify specific lanes due to 7.8 meter confidence interval versus 2.5-4 meter lane width
Solution Approach 1:
The patent combines location data from multiple probe vehicles into a collective histogram representation, merging individual uncertain measurements into a group pattern that reveals lane-level information through statistical aggregation
Solution Approach 2:
The system performs preliminary deconvolution processing on the histogram data to remove location error effects before identifying lane positions, preparing the data in advance to enable accurate lane detection despite GPS inaccuracies
2Measurement precision
If probe data from multiple vehicles is collected and processed using deconvolution methods, then lane-level geometry and traffic information accuracy is improved, but data processing complexity and computational requirements increase
Solution Approach 1:
The patent introduces a histogram as an intermediary representation that transforms raw probe data into a processed format suitable for deconvolution analysis, serving as a mediator between data collection and lane identification
Solution Approach 2:
The system changes the parameter representation from individual vehicle coordinates to histogram frequency distributions, transforming the data structure to enable effective deconvolution and peak detection for lane identification
Data Source
AI summary
Provided herein is a method for establishing lane-level data from probe data. Methods may include receiving probe data points associated with a plurality of vehicles; determining, for each of the probe data points, a location and road segment corresponding to the location; generating, from the probe data points associated with a first road segment, a cell-density image of the first road segment, where the cell-density image represents a volume of probe data points at each of a plurality of cells of a grid overlaid on the first road segment; applying a deconvolution method to the cell-density image to obtain a refined cell-density image having a lower degree of data point spread; determining, from the refined cell-density image, a number of paths along the first road segment, where each path represents a lane; and computing, from the refined cell-density image, lane-level properties of the probe data of the first road segment.