|
Abstract:
|
Driver related factors such as speeding and driver inattention contribute greatly to motor vehicle crashes. In this study, we propose a method to quantify the riskiness of driver’s behavior for various roadway scenarios and environmental conditions. A risk rating score is created for each driver based on the probability of performing a risky behavior given a specific environmental condition. Given the randomness in real-world traffic data, a Bayesian approach was used to improve the model. The risk rating score, as a function of posterior distributions, reveals the riskiness of a driver’s behavior and the credibility associated with that rating based on prior beliefs and the observed data. This approach is examined using data from a naturalistic driving study called Safety Pilot Model Deployment (SPMD). The specific road segment being examined for this case was a crosswalk with a Rectangular Rapid Flash Beacon (RRFB) installed. The model demonstrates a way to design in-vehicle alert algorithms that are better customized for individual drivers.
|
