Parameter Calibration For Microscopic Traffic Simulation Model Of VISSIM

Since Microscopic Traffic Simulation Model is cost-effective, risk-free, and user-friendly, it is widely used in traffic engineering fields. Microscopic Traffic Simulation Model uses many independent parameters to describe traffic system functioning, traffic flow characteristics, driver behavior patterns. Since the value of parameters wield influence on simulation results, it is vital to perform Parameter Calibration. Previous studies mainly focus on the selection and reformation of Parameter Calibration algorithm, while this dissertation goes a further step by performing feasibility analysis and sensitivity analysis on default parameters in earlier stage of Parameter Calibration.The dissertation reviews previous studies on Microscopic Traffic Simulation Model and Parameter Calibration, summaries simulation platform, simulation object and parameter calibration procedure, based on which the dissertation is developed as follows:This article summarized basic data and evaluation data required for parameter calibration at first, and chose travel time of the main-side road as the evaluation index. Then both parts of the data were collected real-time, summarized and sorted. A reasonable simulation model based on VISSIM was built on basis of survey data, and data collector were set.Then, after model parameters were introduced in detail, this article found it’s impossible for the research object through histogram method and confidence interval method. Since there are lots of default parameters, calibration for all parameters was a waste of time and unnecessary. This paper made use of the nonlinear mapping ability of BP neural network, combined with the sensitivity analysis. Then the parameter sensitivity analysis method based on BP neural network was proposed and the set of parameters was determined, including Observed vehicles, Additive Part of Safety Distance, Multiple Part of Safety Distance, Maximum Deceleration,-lm/s2per distance, Accepted deceleration, Waiting Time Before Diffusion, Safety distance reduction factor and Maximum Deceleration for cooperative braking.Finally, it established a calibration model which the square sum of the relative error of the simulation output values and the measured values used as the objective function. SPSA algorithm is chosen as optimal algorithm for this article by comparing the algorithms. Calibrating the parameters Based on improving the parameter calibration process of SPSA algorithm, fresh data validation results show the effectiveness of calibration.