Graph-Based Map Matching for Complex Intersections
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing map matching algorithms struggle with complex intersections, incomplete road networks, and poor GPS trace quality, often failing to deliver accurate results, especially with uni-directional and bi-directional networks, and traces with point clouds, loops, or zig-zags.
Innovation Solution
A method that generates a matching graph using probe data trace points and line segments, where each trace point's candidates are determined within a specified distance, forming a graph with levels representing trace points, and using graph theoretic methods to find the best path, simplifying and scoring to select the best match results, incorporating hysteresis and distance criteria.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional map matching algorithms are used, then processing speed may be maintained, but matching quality deteriorates on complex intersections and incomplete road networks
Solution Approach 1:
The algorithm segments the road network into line segments and organizes them in a hierarchical structure with levels and clusters. Each trace point is matched against specific candidate line segments rather than the entire network, dividing the complex matching problem into manageable sub-problems that improve both accuracy and efficiency.
Solution Approach 2:
The algorithm applies different matching criteria and candidate selection strategies based on local characteristics of the road network and trace data. For example, it adjusts candidate line segment selection based on whether the trace point is near an intersection, junction, or along a straight road segment, allowing optimized matching quality for each local context.
2Reliability
If complete map matching is applied to ensure all trace points are allocated, then coverage is improved, but accuracy deteriorates when digital vector map is incomplete
Solution Approach 1:
The algorithm allows partial matching where only some trace points need to be matched to line segments. Rather than forcing every trace point to find a match (which would reduce accuracy on incomplete networks), it permits unmatched points and focuses on achieving high-quality matches for the subset of points that can be reliably matched.
Solution Approach 2:
The algorithm introduces intermediate structures such as candidate line segments and matching graphs that mediate between trace points and the digital map. These intermediaries allow the system to handle incomplete networks gracefully by providing multiple potential match paths and selecting the most reliable ones.
3Measurement precision
If graph based matching with multiple candidates is used, then matching quality is improved, but computational complexity increases
Solution Approach 1:
The candidate line segments are organized into levels and clusters that segment the search space. The algorithm processes candidates in a hierarchical manner, eliminating unlikely candidates at each level before reaching the final matching graph, which significantly reduces the number of candidates that need full evaluation.
Solution Approach 2:
The algorithm performs preliminary filtering and candidate selection before constructing the full matching graph. It pre-identifies candidate line segments based on spatial proximity and topological relationships, and pre-evaluates some matching criteria, so that the main graph-based optimization works with a reduced and pre-filtered candidate set.
Data Source
AI summary
An improved method for matching traces derived from probe data to one or more line segments in a digital vector map. Points in a probe trace are provisionally matched one-by-one to line segments in the digital vector map to identify all possible matching candidates. A graph of the matching candidates is created having one or more paths. The graph has a plurality of sequential levels corresponding to the points in the probe trace. Each matching candidate is assigned to a level of the graph corresponding with trace point to which it relates. Edges are established between matching candidates in adjacent levels provided they are topologically related to one another. The graph is simplified and scored. The best paths deliver the matching results. The invention allows use of graph theoretic methods to find the best path through the graph, which in turn represents an efficient map matching algorithm. The concepts of this invention may be used in conjunction with longitudinal distance as matching criterion.