Data and Modeling Support of the Management of Diversion Routes During Freeway Incidents

Abstract

A promising traffic management strategy is the application of special signal timing plans on alternative routes during freeway incidents. The development of such plans requires the estimation of the route diversion during incident conditions. This study utilizes a data analytic approach to support the estimation of the diversion rate during incidents and to use this information as an input to the development of special signal timing plans during freeway incidents. First, a method is developed to predict the rate of traffic diversion caused by incidents based on the freeway mainline detector data combined with incident data. The diversion prediction method utilizes a combination of cumulative volume analysis, clustering analysis, and predictive data analytics. Three predictive data analytic methods: linear regression, multilayer perceptron, and support vector machine models, are investigated to predict diversion as a function of incident attributes. Next, a methodology is proposed to develop special signal plans to manage the demand surge on the diversion routes without deteriorating the intersection’s overall performance. The evaluation of the developed methodology indicates that it can significantly reduce the delays on the alternative routes.