Processing Improvement of Map-Matching for Travel Time Prediction Model

Abstract

For trajectory-based travel time prediction model, map matching shows its excellence in terms of GPS data processing by providing an efficient technique to generate the vehicle trajectory on the digital map. The transit vehicle trajectory contains essential information about arrival time at bus stops and delay at major intersections. An understanding of reliable map-matching method is necessary for the development of the accuracy of real-time prediction result accuracys. This thesis provides an enhanced map-matching method, which has better performance in terms of accuracy ofinferred path inference and link identification accuracy, compared with Spatial-temporal matching method, a well-recognized map-matching method used in previous literature. Compared with the existing map-matching method, a reference point file is added to originalthe digital map, converting the point-to-curve match to point-to-point match. The map is also divided into equal digital grids by latitude and longitude to narrow down the matching scale. The feasibility and the accuracy of the method are evaluated in different traffic environment using real field geometric information and GPS data. The last part of the thesis will beis the comparison analysis of between single transit trajectory predictionprediction results using the matching resultsderived from from both map-matching methods,. The field test which is conducted on 23rd Avenue corridor from Legar transit center to Century Park transit center in Edmonton.