Heuristic network generator :an expert systems approach for selection of alternative routes during incident conditions
MetadataShow full item record
Congestion on the freeways of the U.S. has increased multifold over the past few years. A significant portion of this congestion is caused by non-recurring events such as incidents. Diversion has been accepted as a method that can reduce delays during incidents.
The process of diversion involves the selection of the alternate routes, which is currently done off-line and is not responsive to each incident case. The volumes on these preselected routes on that particular day are also ignored. The preselected routes, in most cases, serve only to bypass the link on which the incident occurs. Considering the volumes that flow on the freeways, this leads to considerable delays in terms of lost time and productivity. Another important issue that is currently neglected is user compliance.
The network generator is used to reduce the delays in selection of these alternate routes. It uses characteristics such as the congestion levels and available capacities in selection of alternate routes in real-time. Also, used in selecting alternate routes are feasibility criteria, that significantly affect the available capacities on the links. These include presence of trip generators (schools, offices, etc.) or safety factors (icy bridges, height restrictions, etc.). The model thus generates a reduced network and a set of alternate routes to divert the traffic upstream of the incident. Disutilities that drivers associate with route-choice, such as the number of left-turns and signals, the relative time spent on the freeway and arterials are attached to each route. The routes with the minimum disutilities are displayed to the user. A user-equilibrium assignment module to predict traffic flows in the future is also incorporated into the framework. As a precursor to the network generator, there is a module which calculates the clearance time for an incident. It uses other characteristics of the incident such as the weather and time of occurrence in order to predict if the delays are significant to initiate diversion.
Numerous tests were conducted in order to validate the rules and functions developed. The tests were based on varying incident and traffic conditions.
The results showed that the model, was able to select better routes for off-peak conditions rather than peak conditions. There is a threshold value of the delay caused by the incident, beyond which the model is very effective.
- Masters Theses