An analysis of the multiple paths used in traffic assignment algorithms

Abstract

Historically, the four-step planning process has been used to estimate future roadway volumes and identify future roadway needs. Recently, the focus has shifted to the use of planning models to make system operations decisions. These decisions require the analysis of route-based metrics in addition to roadway volumes. The traffic assignment techniques typically used in planning models are link-based and do not explicitly enumerate or evaluate routes. To date, no research has been conducted to assess the traffic assignment techniques with respect to the routes identified in large-scale networks. An evaluation of routes identified between origin/destination (O/D) pairs using the convex combinations traffic assignment algorithm (CCA) was presented. The analysis procedures were illustrated using a simple network. Routes were analyzed in two large-scale networks at a desegregate level with respect to the assumptions of user equilibrium and several attributes that are commonly cited to influence drivers' route choice behavior. According to the principles of user equilibrium, drivers are assigned to the network in a manner such that the travel times on all used routes are equal. The results of this research revealed that the travel times among the identified routes were equal in only a small percentage of the O/D pairs. Infeasible routes were identified in several O/D pairs that had travel times that significantly exceeded that of the minimum path. The infeasible routes were typically identified in the first three iterations of the assignment. It was hypothesized that the presence of these infeasible routes affects the ability of the algorithm to converge. A heuristic strategy was developed and evaluated solution. As part of this technique, n iterations were performed using the CCA. At the nth iteration, the first k routes were removed from the solution and the assignment weights for the k+ I through n iterations were re-calculated. A new CCA assignment was performed for n+l through N iterations and the final link flows were calculated. The implementation of this strategy resulted in an improvement of the assignment results with respect to several route-based metrics.