Map Matching
An Introduction
Prof. David Bernstein
James Madison University
Computer Science Department
bernstdh@jmu.edu
The Map-Matching Problem
The Situation:
A person/vehicle is moving along a finite set of streets \(\overline{\mathcal{N}}\)
We are provided with an estimate of this person's location at a finite number of points in time denoted \(\{0, 1, ..., T \}\)
The person's actual location at time \(t\) is denoted by \(\overline{P}^{t}\) and the estimate is denoted by \(P^{t}\)
The Problem:
Find the street in \(\overline{\mathcal{N}}\) that contains \(\overline{P}^{t}\)
The Map-Matching Problem (cont.)
A Complication:
We do not know the street system, \(\overline{\mathcal{N}}\) exactly
We have a
network (or graph)
representation \(\mathcal{N}\) and a set of piecewise linear curves (one curve associated with each arc/link)
The Actual Problem(s):
Match the estimated location \(P^{t}\) with a curve, \(A\) in the "map"
Then determine the street \(\overline{\mathcal{A}} \in \overline{\mathcal{N}}\) that corresponds to the actual location \(\overline{P}^{t}\)
A secondary concern is to determine the position on the curve \(A\) that best corresponds to \(\overline{P}^{t}\)
The Map-Matching Problem (cont.)
Point-to-Point Matching
The Idea:
Match
P
t
to the "closest" breakpoint in the piecewise linear curve
A Difficulty:
Point-to-Curve Matching
The Idea:
Identify the arc that is closest to \(P^{t}\)
One Approach:
Point-to-Curve Matching (cont.)
One Difficulty:
Point-to-Curve Matching (cont.)
Another Difficulty:
Curve-to-Curve Matching
The Idea:
Find the arc that is closest to the piecewise linear curve, \(P\) defined by the points \(P^{0}, p^{1}, ..., P^{m}\)
A Difficulty:
Curve-to-Curve Matching (cont.)
Another Approach:
Use the "total" distance
A Difficulty:
Curve-to-Curve Matching (cont.)
Another Approach:
"Correct" for differences in length
A Difficulty:
Using Topological Information
Getting Started
Using Topology (cont.)
Using Connectivity/Topology
There's Always More to Learn