Pedestrian-vehicle Conflict Analysis at Signalized Intersections Using Micro-simulation Jiawei Wu, Essam Radwan, Hatem Abou-Senna Department of Civil, Environmental & Construction Engineering University of Central Florida, Orlando, Florida Abstract Field Data Collection The MAPE value for the total conflicts varied from 12.7% to 73.2% for different maixmum TTC and PET thresholds. It is found that when the TTC ranges from 2.6 to 2.8 seconds and PET threshold ranges from 8 to 9, the best goodness-of-fit between the observed and the simulated conflict of mean PET is achieved with the MAPE value under 13%. Therefore, the suitable maximum TTC and PET thresholds for pedestrian-vehicle conflicts are 2.7 and 8, respectively. The main goal of this study is to explore whether the VISSIM simulation model and Surrogate Safety Assessment Model (SSAM) can be used to provide the reasonable estimates for pedestrian-vehicle conflicts at signalized intersections. A total of 42 hours videos were recorded at seven signalized intersections for field data collection. The calibrated and validated VISSIM model was used to generate pedestrian-vehicle conflicts and SSAM software was used to extract these conflicts by processing the vehicle trajectory file from the VISSIM model. The results showed that there was a best goodness-of-fit between simulated conflicts and observed conflicts when the maximum TTC threshold was set to be 2.7 and the maximum PET threshold was set to be 8. The result also indicated that there was a significant statistical relationship between the simulated conflicts and the observed conflicts. The data collection in the field was used to develop, calibrate, and validate the VISSIM and SSAM simulation models. Seven intersections were selected from urban areas in Florida, the United States. No. Intersection Name 5-year Ped Crashes Location County 1 Primrose Dr & Colonial Dr 9 Orlando Orange 2 Silver Star & Hiawassee Rd 20 Pine Hills 3 Sand Lake Rd & I-Drive 6 4 Kirkman Rd & Conroy Rd 13 5 Martin Luther King & US 92 7 Daytona Beach Volusia Orange Ave & Kaley St 8 Semoran Blvd & Pershing Ave Introduction VISSIM Calibration and Validation Relationship between Simulated Conflicts and Observed Conflicts In the United States, nearly 76,000 pedestrians were reported injured in 2012. Although the number only accounts for 3% percent of all the people injured in traffic crashes, the number of pedestrian fatalities is still around 14% of total traffic fatalities (National Highway Traffic Safety Administration, 2014). In recent years, microscopic traffic simulation techniques have been widely used in transportation to evaluate different traffic safety strategies. A software application was called “Surrogate Safety Assessment Model (SSAM)” designed by Federal Highway Administration (FHWA) to perform statistical analysis of vehicle trajectory data output from microscopic traffic simulation models. Several studies attempted to demonstrate that the vehicle conflicts from SSAM could reflect the safety assessment in the real world. However, few researches discussed the pedestrian-vehicle conflict by using SSAM, Therefore, the main objective of this study is to examine if the VISSIM simulation model and the SSAM could estimate pedestrian-vehicle conflicts at signalized intersections. The VISSIM version 7 was used to develop the vehicle/pedestrian simulation model at signalized intersections. The VISSIM simulation model was calibrated to reproduce the performance measures for both traffic and pedestrians, such as traffic volume, pedestrian volume, queue length and pedestrian crossing time. By applying the Chi- square tests, it was found out the difference in these measures between the field and the simulation model were not statistically significant. A total of seven intersections were separated into two groups, including a calibration dataset with five intersections, and a validation dataset with two intersections. The five intersections were used to develop and calibrate the VISSIM models, while the other two intersections were used to testify the effectiveness of simulation model calibration. It is found that the p-value of independent variable is 0.00, indicating that number of simulated conflicts is significantly associated with the number of observed conflicts. The R2 value for the model was 0.8825, which means that 88.25% of the variability in the observed conflicts can be explained by the variation in the simulated conflicts. Consequently, there is a correlation between simulated conflicts from VISSIM and obseved conflict in the field. Conclusions SSAM Calibration There were two major findings in this study: First, the suitable maximum TTC and PET thresholds for pedestrian-vehicle conflicts were found through measuring the differences between the mean PET observed in the field and the mean PET simulated in VISSIM and SSAM by using the MAPE. Second, the number of simulated conflicts was significantly related to the number of observed conflicts according to the linear regression results. Two threshold values for surrogate measures of safety were used in SSAM to detect the conflicts, which are maximum TTC and maximum PET. Since the pedestrian-vehicle conflict is totally different from the vehicle- vehicle conflicts, the maximum TTC and PET thresholds need to be adjusted for pedestrian-vehicle conflicts. The mean absolute percent error (MAPE) was used to measure the differences between the mean PET observed in the field and the mean PET simulated in VISSIM and SSAM to calibrate the SSAM. MAPE= 1 𝑛 𝑖=1 𝑛 | 𝑐 𝑠 𝑖 − 𝑐 𝑜 𝑖 𝑐 𝑜 𝑖 | Where n represents the number of intersections, 𝑐 𝑠 𝑖 represents the mean PET of the simulated conflicts for one intersection, and 𝑐 𝑜 𝑖 represents the mean PET of the observed conflicts for one intersection. Acknowledgements The work reported in this paper is part of an ongoing research project under contract which is sponsored by the Southeast Region Transportation Center (STC) and the Florida Department of Transportation (FDOT).