Development of a Surrogate Safety Assessment Framework Incorporating Traffic Simulator, Vehicle Dynamics, Driver Warning, GPS/INU, and V2V/V2I

This dissertation developed an integrated surrogate safety assessment framework to proactively assess traffic safety using realistic vehicle trajectories as well as potential positioning errors and V2V/V2I communication delays on the vehicle safety applications. To this end, the following simulators were developed and integrated into the surrogate safety assessment framework: 1) vehicle dynamics model-integrated traffic safety simulation environment, 2) V2V/V2I communication delays simulator, 3) GPS/INU positioning error simulator, and 4) driver warning generator.
First, a vehicle dynamics model (i.e., CarSim) was integrated with a microscopic traffic simulation model (i.e., VISSIM) for a surrogate safety assessment, based on more realistic vehicle trajectories. This idea was initiated from the fact that the microscopic traffic simulation model can generate various traffic situations and the vehicle dynamics model has an extensive capability of modeling the vehicle dynamics including pitch, yaw, and roll and generating realistic vehicle trajectories. To take advantage of these capabilities, the two simulation models (i.e., VISSIM and CarSim) were integrated and used to estimate the number of traffic conflicts. In addition, a driver aggressiveness model derived from the Next Generation Simulation (NGSIM) project’s lane change vehicle trajectories was incorporated to the lane change vehicles in VISSIM. The resulting VISSIM vehicle trajectories were processed through CarSim to account for the vehicle dynamics and the traffic conflicts were identified through the Surrogate Safety Assessment Model (SSAM). The VISSIM-CarSim integrated simulation environment resulted in 9.5% fewer traffic conflicts compared with the existing VISSIM-only approach.
Second, the results of the two conflict estimation approaches, that is, from the proposed approach (i.e., VISSIM-CarSim) and the existing approach (i.e., VISSIM-only), were analyzed to estimate their correlation with the actual traffic crashes. These correlations were then used to assess and compare the effectiveness of these two approaches for assessing traffic safety. This correlation analysis was based on the number of traffic crashes and traffic conflicts from two freeway corridors (i.e., I-495 and SR-267) during a peak hour (i.e., from 5 P.M. to 6 P.M.). This analysis showed that the traffic conflicts obtained from the proposed approach exhibits a stronger correlation (i.e., 0.72 of correlation coefficient) with traffic crashes than the existing approach did (i.e., 0.61 of correlation coefficient). Both traffic conflicts computed for both approaches showed a statistically significant relationship with the actual traffic crashes. In addition, a cross-validation test on the confidence intervals of the correlation coefficients showed that the correlation coefficients have very tight confidence intervals (i.e., 0.02 for both cases). This indicates that traffic conflict can be used as a traffic safety estimator but also the newly developed vehicle dynamics model-integrated traffic safety simulation environment was found to be a superior, valid alternative for assessing the surrogate safety.
In addition, the V2V/V2I communication connection probability model reflecting that communication performance can be degraded according to the number of transceivers (i.e., vehicles) in a specific area and the distance between transceivers was developed. The GPS/INU simulator, which simulates the VISSIM X and Y coordinates according to the assumed positioning system corresponding to the positioning accuracy, was also developed. When the driver warnings are triggered in VISSIM, the V2V/V2I communication delays simulator potentially delays the warnings, and the GPS/INU simulator provides GPS/INU erroneous vehicle trajectories on the fly. A perception-reaction time (PIEV) was adopted in the middle of driver warnings and actual vehicle response (i.e., deceleration) to reflect a realistic driver response.
Consequently, although the driver warnings reduced 28% to 35% of dangerous conditions under no-communication delays and positioning errors, the communication delays degraded the effect of driver warnings ranging from 8% to 15%. In addition, the effectiveness of driver warnings based on various GPS/INU technologies improves as the accuracy level of the GPS/INU devices increases. Therefore, two important findings are highlighted: 1) the probability of false alarm would decrease as the high-accuracy positioning system is deployed in the vehicle safety applications, and 2) the traffic safety estimation result can be different according to the accuracy level of the positioning systems assumed. Accordingly, this dissertation research conveys to the traffic safety research community that potential positioning errors need to be considered when the traffic safety is estimated under advanced vehicle safety applications scenarios (e.g., Connected Vehicles applications)