Simulation Model Configuration and Calibration. We set traffic volumes for the approaches based on actual volumes observed by traffic sensors. Some traffic volume data were double-checked by ground-truth video tapes recorded at the test intersections to enhance the reliability of the model calibration process. Traffic flows of intersection approaches generated by the simulation program were reasonably distributed in the range of 50 vehicle-per-hour-per-lane (vphpl) to 1250 vphpl and matched field observed volumes very well. We estimated the passenger ridership on buses on the basis of annual CT ridership (National Transit Database 2004). In our model we selected 12 passengers per vehicle (ppv) as the ridership. The average vehicle occupancy for general-purpose vehicles was estimated to be 1.2 occupants per vehicle, as determined by King County Metro on the basis of field observations (King County Department of Transportation 2002). Additionally, the generation rate of passengers was set as 10 persons per hour (pph) on the basis of the number of boardings at each stop (Community Transit 2005). Other parameters, such as bus headways, locations of bus stops, and so on, were calibrated according to the real values. Figure 7-1 shows a snapshot of the simulation model for the intersection of 164th Street and 36th Avenue. We also calibrated the traffic control settings of the simulation model by using actual traffic operation parameters and control plans. Internal parameters for the simulation model were properly adjusted to ensure the model’s appropriateness to the corresponding application. After the simulation model was properly calibrated, we conducted a six-hour simulation test: three hours for TSP on and the other three hours for TSP off.
Simulation Model Configuration and Calibration. Traffic volumes in each approach were collected from mid-block virtual loops by using the VIPs. Directional volumes were manually extracted from video tapes recorded at the test intersections on typical weekdays. These volume data were used to configure the simulation model for traffic generation. Traffic volumes generated by our VISSIM simulation model were reasonably distributed in the range of 30 vphpl to 980 vphpl, which matched our field observations very well. The traffic control parameters used by the VISSIM model were calibrated by using the actual control plans and timing parameters. We estimated the passenger ridership on buses on the basis of CT’s annual ridership data (National Transit Database 2004). Consequently, we used 12 ppv as the ridership for our simulation model. The average vehicle occupancy for general-purpose vehicles was configured to be 1.2 occupants per vehicle on the basis of field observations by King County Metro (King County Department of Transportation 2002). Additionally, the generation rate of passengers was set at 20 pph in our simulation model according to CT’s study on the number of boardings at each stop (Community Transit 2005). The other parameters, such as bus headways, bus stop locations, and so on, were calibrated according to the real values. Because the corridor was very long, we show only a snapshot of the simulation model at one example intersection of 196th Street and SR 99 in Figure 7-2. Figure 7-3 provides a three-dimensional view of the simulation model at the intersection of 200th Street SW and SR 99. Because of the stochastic features of the simulation models, multiple simulation iterations were essential to enhance the reliability of the simulation results. By changing the VISSIM simulation random seeds, the random vehicle generation could be realized. In this analysis, a total of 20 iterations were conducted, ten scenarios with TSP functions and ten without TSP functions. The test period was three hours for each scenario. Figure 7-2 A Snapshot of the Phase Two Simulation Model Figure 7-3 A Three-Dimensional Snapshot of the Phase Two Simulation Model CHAPTER 8 PHASE ONE RESULTS AND DISCUSSION
