Combined Genetic Computation of Microscopic Trip Demand in Urban Networks

Abstract

This chapter describes a combined genetic computation approach for estimating time-varying Origin-Destination (O-D) trip demand matrices from traffic counts in urban networks. The estimation procedure combines a microscopic model simulating traffic flow conditions with a genetic algorithm to synthesize the network O-D trip matrix, through determining the turning flow proportions at each intersection. The proposed approach avoids the restrictions involved in employing a user-optimal Dynamic Traffic Assignment (DTA) procedure and carries out a stochastic global search of the optimal O-D trip and turning flow distributions. The multi-objective, single-level optimization formulation of the problem provides a mutually consistent solution between the resulting O-D matrix and path/link flow pattern, which minimizes the difference between estimated and observed link flows. The model implementation into a real arterial sub-network demonstrates its ability to microscopically estimate trip demand with satisfactory accuracy and fast computing speeds which allow its usage in dynamic urban traffic operations.