Method of Successive Averages Versus Evans Algorithm: Iterating a Regional Travel Simulation Model to the User Equilibrium Solution

The results of comparative tests of two methods for iterating a regional travel demand model system are presented. Model iteration is necessary to ensure consistency between model input and output speeds, as required by current federal legislation. Two methods were tested: the Evans algorithm and the method of successive averages. A series of tests using alternative assignment techniques was conducted for each method. Criteria for evaluating the iteration methods included convergence error, average highway speeds compared with observations, highway vehicle miles traveled compared with Highway Performance Monitoring System estimates, transit boardings compared with observations, and computer running time. It was concluded that the Evans algorithm performed the best, primarily on the basis of superior computational efficiency, although good results were obtained by using the method of successive averages. Use of the Evans algorithm is recommended, embedded within a formal assignment restart, for iterating the model system. Multiple iterations of highway assignment should be used in the initial model loop and all-or-nothing assignments in subsequent iterations of the modeling chain.