Abstract:
Variable message signs on motorways give travellers estimates of expected travel time. Knowledge of these times can be useful for a driver in deciding which routes to take in order to avoid congestion. If effective, this can also help reduce overall congestion. It is therefore important that these estimates are accurate. Several methods proposed in literature for estimating travel time are explored in this thesis. The methods from literature can broadly be grouped as trajectory based, flow based and data driven. These methods use detected traffic data such as vehicle counts and speeds in order to estimate travel time. However, all methods have their limitations in being able to estimate travel time accurately. A selection of methods from literature are implemented and compared. A new method for estimating travel time called the Average Vehicle Tracking (AVT) method is also proposed. The method uses linear regression and a least squares approach to find the number of vehicles in a section at a particular point in time. Travel time can then be estimated by finding when a particular vehicle enters and leaves the section, based on vehicle counts at either end of the section. Variations of this method, and hybrids of the method and some of the methods from literature are also proposed. The drawback of the proposed AVT method is that it depends greatly on perfect detector measurements, which is usually unattainable in the real world. The selected methods are tested on a simulation of a section of Auckland's motorway, and are compared under various conditions. The AVT method outperforms the others in most conditions, with a data driven regression approach also performing well. An extension to this project is to make travel time predictions, which take into account changing road conditions. Approaches for prediction are outlined, that build on the ideas presented in this thesis. Overall the AVT and regression methods are found to provide very good results. These methods show significant improvements on the travel times estimated by a method similar to the one used in practice on Auckland's motorways.