A Dual Genetic Algorithm using Central Composite Design to Enhance Speed of Optimization for High Wo

The calibration of car-following models has received much attention in effort to reflect more realistic microscopic traffic simulations. A genetic algorithm (GA), in spite of an absence of the definitive agreements, has been widely employed to find optimum set of parameters in car-following models. Although the GA can solve the unstandardized problems, the efficiency of algorithm, i.e., the number of iterations and population size required to obtain a point near the optimum, is a challenge for high workload optimization problems, like the microscopic traffic simulation. The objective of this paper is to provide an approach to enhance the optimization speed of GA by using a central composite design (CCD). The proposed approach uses a dual GA, with an additional step to find “better” initial population of GA. The optimization goes through mainly three sequential steps: (1) experimental design using CCD for a quadratic response surface model (RSM) estimation, (2) 1st GA procedure using the RSM with CCD to find a near-optimal initial population for a next step, and (3) 2nd GA procedure to find a final solution. The proposed method was applied in calibrating the parameters of Gipps car-following model with respect to maximize likelihood of the spacing distribution between a lead and following vehicle. To validate the improvements, a conventional approach using a “single” GA was compared, under both simulated and real vehicle trajectory data. The result evidently shows that the proposed approach aids to start the optimization with an initial population that have better fitness values, and thereby empowers the efficiency of GA.