Real-Time Freeway-Experienced Travel Time Prediction Using N-Curve and k Nearest Neighbor Methods

Abstract

This study presents a methodology for freeway travel time prediction that uses only count data. The proposed models include the generalized N-curve method in conjunction with the k nearest neighbor (kNN) method so that the travel time predicted for traversing a defined freeway segment at a certain departure time is similar to what a driver actually experiences. A real-world traffic network and demand are replicated in a traffic simulation model in which several scenarios are produced to serve as the test bed for evaluation and validation of the proposed algorithms. The proposed single-NN algorithm best predicts travel times for light, free-flow traffic conditions, and the multiple-NN algorithm best predicts travel times for congested traffic conditions. The hybrid-NN algorithm merges the single-NN and multiple-NN algorithms, exploiting each one where most suitable. A numerical analysis concludes the potential of the proposed models.