| This paper has investigated performance of several microscopic traffic flow models using experimental data setstaken from car following experiments conducted in a test track. Speed and headway data measured by RTK GPSreceivers for ten passenger cars participated in the experiment were used for performance evaluation、 on the basis ofhow good the models fit with the data sets. The optimum values for models’ parameters were determined usinggenecop algorithm、 an optimization technique、 based on two different approaches、 first using speed data、 and secondusing headway data. The optimized performance of each model was analyzed for different driving conditionsintroduced by different level of disturbances in the speed of lead car that includes half wave、 one wave、 two wave、three wave、 random and constant speed patterns. The analysis based on speed data produced relatively lowerpercentile error in estimation of speed. First five models have given very close performance with average percentileerror ranging from 3.87% to 4.71% and standard deviation of 1.09% to 1.64%. The analysis based on headway dataproduced relatively higher percentile error in estimation of headway with first three models performing better thanother three models. The average percentile error for first three models ranges from 12.04% to 12.91% with standarddeviation of 4.53% to 5.13%. The interpersonal variations in percentile error have been found higher than intermodelvariations in most of the cases、 indicating influence of individual drivers’ behavior in car followingphenomena. |
