A Novel Loglinear Model for Freeway Travel Time Prediction

Abstract

As traffic congestion continues to grow worldwide, freeway travel time prediction is becoming increasingly important. During the past decade, numerous research projects have been carried out in travel time estimation. A variety of algorithms and techniques have been developed, primarily for predicting short-term travel time (less than 30 minutes ahead). However, these travel time prediction methods cannot be applied for long-term travel planning. In this paper, a loglinear travel time prediction model is proposed to estimate the travel time that begins at a long-term future moment of departure. Instantaneous and historical traffic data from loop sensors on difference freeways are collected and analyzed. Coefficients in the model are obtained using these training data. By using the proposed loglinear algorithm, the travel time for each segment of the freeways is predicted. The travel time prediction is performed in real-time based on the travel time of each segment. This model is scalable to freeway networks with arbitrary travel routes. It is unique in that it considers various traffic patterns during different days in one week. It is also simple, stable, and computationally efficiency, with low storage cost requirements. Real world data are used to evaluate the proposed loglinear predictor. The performance of our model is compared with the results of the commonly used predictors.