A Mixed Evolutionary Algorithm to Solve the O-D Matrix Estimation Problem

The entropy maximizing (EM) model is a main approach to solve the O-D matrix estimation problem, but the current algorithms solving the EM model is limited by the difficulty of choosing a proper initial solution for searching procedure, such as the Newton´s method and the Levenberg-Marquardt algorithm (LMA). A mixed evolutionary algorithm (MEA) integrating genetic algorithm (GA) and LMA is presented here to solve the EM model. In MEA, GA is used to obtain a solution which is in the neighbourhood of the optimal solution which satisfies the precision first. Then considering that GA is slow to converge, LMA is applied to quickly converge into the optimal solution which satisfies the precision with GA´s result as its initial solution. The numerical results show that MEA performs quite better than GA and LMA in solving the EM model