A study of travel time prediction using universal kriging

Abstract

This study describes an approach for predicting travel time by kriging. The kriging method, which is a means of spatial prediction, is used as prediction measure for car travel time in an imaginary four-dimensional space. Each point in the space represents a single journey: the coordinates of a point are the coordinates of the origin and destination on a plane. The travel time can then be viewed as a function over this four-dimensional space. The prediction relies on the feature that nearby points (in the four-dimensional space) will have almost the same travel time. In this approach, it is not necessary to break down travel times from origin to destination into link travel times. The approach will also allow us to use information from “probe vehicles” for travel time prediction in the near future. The feasibility of this approach is demonstrated through a case study in London and its environs. The case study uses 200 observations for verification. The multiple correlation coefficient of estimated travel time and the verification data is 0.9045. The results indicate that 95% prediction limits are between ±10 minutes and ±30 minutes for travel between two arbitrary points. This prediction method is effective for urban districts with links having changeable travel time owing to congestion.