CALIBRATION AND VALIDATION OF VISSIM MODEL OF AN INTERSECTION WITH MODIFIED DRIVING BEHAVIOR PARAMETERS

A. C. Dey, S. Roy and M. A. Uddin. Student, Dept. of Civil Engineering, Bangladesh University of Engineering and Technology, Dhaka-1000, Bangladesh. ...................................................................................................................... Manuscript Info Abstract ......................... ........................................................................ Manuscript History Received: 01 October 2018 Final Accepted: 03 November 2018 Published: December 2018

Microscopic simulation models have been widely used in both transportation operations and management analysis because simulation is safer, less expensive and faster than field implementation and testing. The usefulness of these models in making design and traffic control decisions will mainly depend on their accuracy and reliability. This paper describes the detail procedure for the calibration and validation of a microscopic model of highly congested intersection Mirpur-10 in Dhaka. Legs of this intersection is composed of both motorized vehicles (Bus, Passenger car etc.) and non-motorized vehicles (Rickshaws, Bicycles). Most cases, drivers are rarely concern about the lane based traffic operation. Addressing this phenomena, a microsimulation VISSIM model with modified driving behavior parameters helps to create a virtual environment representing the traffic scenario, optimize the problems and visualize the outputs that is important to face the challenges of transportation system at present and future.

…………………………………………………………………………………………………….... Introduction:-
Advances in computational technology along with the increased complexity of roadway design and management have created an environment in which microscopic simulation models have become useful tools for transportation engineers. Microscopic simulations can be utilized in several different transportation areas. They can be used to evaluate alternate timing plans and geometric changes before implementing the design in the field. They are also appealing in the estimation of certain quantities that are not easily estimated or observed from the field. Microscopic simulation models contain numerous independent parameters to describe traffic control operation, traffic flow characteristics, and drivers' behavior. These models contain default values for each variable, but they also allow users to input a range of values for the parameters. Changes to these parameters during calibration should be based on field measured or observed conditions and should be justified and defensible by the user.
Dhaka is a large and densely populated metropolitan area which has one of the most diverse road transportation systems in the world. This system consists of both motorized vehicles (viz. bus, mini-bus, car, cng, baby taxi, motorcycle etc.) and non-motorized vehicles (viz. rickshaw, rickshaw van, bicycle etc.) modes. This mixed flow of vehicles leads to many problems, like conflicts at intersections, when number of non-motorized vehicle increases which adversely affects the speed and flow of other vehicles. Another feature of this traffic is the absence of lane marking and lane discipline. The lane widths are also not constant. Analytical modeling of such traffic is in nascent ISSN: 2320-5407 Int. J. Adv. Res. 6(12), 107-112 108 stage. Micro-simulation is favored to study and model heterogeneous traffic (Mathew et al., 2010). It is a useful tool to effectively analyze and evaluate proposed improvements and alternatives. For example, an intersection can be simulated for different signal timing plans and its effect found before implementing it. VISSIM which is used in this study is a microscopic, behavior-based multi-purpose traffic simulation to analyze and optimize traffic flows (VISSIM 5.30 User manual). VISSIM is better in terms of ease of use and does not require cumbersome coding (Park et al., 2002) VISSIM has the ability to model the interaction between the various modes of transit with automobile traffic, ability to generate vehicles randomly and flexibility in modeling complex geometries (Moen et al., 2000). The components or parameters of simulation model requiring calibration include traffic control operations, traffic flow characteristics, and drivers' behavior. Model Validation tests the accuracy of the model by comparing traffic flow data generated by the model with that collected from the field.

Test Site and Simulation Model: -Study Area: -
The study area is the intersection of Mirpur-10 roundabout, located in Dhaka metropolitan area. The study area includes four link roads connected with various infrastructures, cantonment and shopping mall. Dhaka cantonment, Mirpur DOHS is located in north direction, Sher-e-Bangla National Stadium in on west, Mirpur-13, 14 residential areas are on east and lots of government offices, hospital, institutes are located in the south direction which makes it one of the busiest intersection in Dhaka. Figure 1. Shows Mirpur-10 intersection.

Determination of Measures of Effectiveness: -
In this step one has to determine a performance measure, identify uncontrollable input parameters and controllable input parameters. Uncontrollable input parameters may include existing geometry, traffic counts, current signal timing plans, etc. Controllable input parameters in the simulation program may include lane changing distances, waiting times before diffusion, minimum headways, minimum and maximum look ahead distances, etc.

Data Collection: -
Once the measures of effectiveness have been identified, the next step in the calibration and validation process is data collection from the field. Performance measures and uncontrollable input parameters should be collected from the field.
VISSIM Network Model Coding: -As described before simulation via VISSIM starts with the introduction of network coding. It should be done by creating links, connecting them, inputting vehicles composition, creating route decision, setting signal controls, data collection points and configuration.

109
Identification of Calibration Parameters: -All calibration parameters within the microscopic simulation model must be identified. Examples of the controllable calibration parameters are lane change distance, desired speed, and minimum headway distances in the simulation model. Acceptable ranges for each of the calibration parameter should be determined.

Evaluation of Parameter Sets: -
In this step multiple runs will be conducted to verify whether the parameter sets identified in the previous step generate statistically significant results. For each parameter set, a distribution of performance measure will be developed and compared with the field measure.

Collection of New Data Set for Validation: -
In order to perform validation of the microscopic simulation model a new set of field data under untried conditions should be collected. One way of collecting validation data would be collecting data for different time periods or conditions.

Implementation of the Proposed Procedures: -Data Collection: -
Data of heterogeneous traffic flow such as traffic volume, vehicles composition, speed, and signal timing of the chosen road along with geometric data are collected. The vehicle flow input is given in start of leg near first foot over bridge and the outflow is found by placing 4K digital camera at selected 4 locations at peak time 9.00AM-11.00AM and 5.00PM-7.00PM. The research area is manually signalized intersection. It is intersected by two roads of two-way two lanes. East-west direction road and north-south direction road both are assumed major road. The width of each lane is not fixed but average 3.25m. Two hours of data in the morning peak and evening peak are collected. Figure 2. And Figure 3. show data collection and vehicles composition respectively.   Model calibration: -Calibration is the process in which various parameters of the simulation model are adjusted till the model accurately represents field conditions. VISSIM has a COM interface which can be used to calibrate VISSIM externally through a code. Algorithm, a random search and optimization technique is used to generate random sets for parameters within specified bounds and the calibration code is run till it finds the least mean absolute percentage error value between the actual and simulated measure (Siddharth et al., 2013).  Table 1. The calibrated models are then evaluated with a new set of data under untried conditions, including the input volumes, traffic composition, and other required data. This study adopts the Geoffrey E. Heaver (GEH) statistic to compare field traffic volumes with those obtained from simulation data. As a general guideline for model validation, GEH values less than 5 indicate good fit (UK highway agency). Several simulations run with different parameter for confirmation. Table 2. shows that GEH value of the microscopic model is 2.863 which indicates a well calibrated model and represents the field traffic condition with remarkable accuracy. Figure 6. also shows the least variation between actual flow and validated flow.   The proposed procedure appears to be effective in the calibration and validation for VISSIM for signalized intersections. Two important issues were encountered during the implementation of calibration and validation procedure. The first issue dealt with statistical testing when claiming the calibrated model was equal to the field data. The second issue was the importance of visualization in the calibrations process. The study only utilized a single day of data collection and two measures of performance. It is recommended to collect multiple days of field data, if possible, in order to consider variability of field data. A range of flows and geometries would give more general conclusions. It is also recommended to utilize other performance measures such as number of stops, delays, fuel consumptions, or emissions to see if they produce different variability.