Font Size: a A A

Automatic Calibration Method Research For Mixed Traffic Microscopic Simulation Models

Posted on:2013-04-01Degree:MasterType:Thesis
Country:ChinaCandidate:J J HeFull Text:PDF
GTID:2232330374490639Subject:Traffic and Transportation Engineering
Abstract/Summary:PDF Full Text Request
Promoted by computer simulation technology, microscopic traffic simulation hasbeen developed vigorously. With the advantages of economic, convenient, strongtargeted, reusable, and etc, traffic simulation technique can be used in many trafficfields, such as traffic organization and the impact analysis, traffic wave analysis, ITSevaluation, and operation condition analysis. However, a traffic simulation modelwithout effective calibration and validation will be regarded as “black box” operation,which means it is impossible to judge whether the simulation model could operate asreal condition. Therefore, simulation model calibration has been valued and regardedas a popular research topic. Through all the domestic and foreign research, calibrationis mostly conducted in homogeneous traffic simulation models, which ignore thephenomenon of mixed traffic. In order to make up the defect, a calibration methodconsidering the mixed traffic was presented in this paper, which takes both theaggregate and disaggregate traffic calibration into account. Besides, an automaticcalibration program has been developed to settle micro-simulation model calibrationissues of mixed traffic flow efficiently.Micro-simulation model calibration of mixed traffic flow corresponds to acomplex multi-variable and multi-objective optimization problem. Genetic algorithm(GA) is a popular and mature method to settle these problems. Though GA hasadvantages of strong global search ability and robustness, its disadvantages ofbecoming premature easily when facing large search problems cannot be ignored.Therefore, improvements and new processing techniques such as simulated annealingalgorithm, niche technique, ranking selection method and new cross point generatemethod for parameter sets have been implanted into the traditional GA to form a newalgorithm named improved simulated annealing genetic algorithm (ISAGA).Atomizing the model calibration procedure need, in addition to reasonable andfeasible calibration methods, data chains between compiled calibration procedure andtraffic simulation software to realize the data transfer and simulation control. In thispaper, MATLAB has been used to compile the automatic calibration procedure namedISAGA program. This program can be divided into three parts: simulation module,control module and ISAGA module, which are work together to achieve automaticcalibration. Finally, the CORSIM simulation model of an expressway section in Guangxi has been chosen as a test site. By the ISAGA program calibration, the aimfitness of the test site has been controlled into target threshold (≤15%) while theprogram iterate to17thgeneration, which indicate that this ISAGA program isefficient; Also the travel time fitness of aggregate and disaggregate calibration areseparately within3%and13%, which all prove that this ISAGA automatic calibrationprogram feasible and the disaggregate calibration is necessary.
Keywords/Search Tags:Microscopic Simulation, Mixed Traffic, Parameter Calibration, GeneticAlgorithm, Simulated Annealing Algorithm
PDF Full Text Request
Related items