Driver behavior assessment based on the belief theory in the driver-vehicle-environment system

This paper proposes to evaluate the driver safety by integrating the information of the Driver, the Vehicle and the Environment (DVE). The adopted strategy is composed of two fusion levels (local and global fusion levels) and use the Demptster-Shafer Theory (DST). The development of the Basic Probability Assignment (BPA) is based on the developed Fuzzy Inference System (FIS) (in the vehicle entity) and models. To reduce the complexity of the fusion problem in the Vehicle entity, the output membership function is fixed to the two developed FISs and the rule table is tuned relatively to this assumption. Application to the Canadian drivers in the city of Montreal shows the validity of the developed risk modes in case of one sample data and a driving mission. Comparison results between the Dempster-Shafer (DS) and the sixth version of the Proportional Conflict Redistribution (PCR6) combination rules show that the relevance of the PCR6 in case of high conflict between sources. Data related to the experimental test is recorded using the developed on-board data collector device that integrates an Inertial Navigation System (INS) and a Global Positioning System (GPS). Our approach assumes that the driving behavior is evaluated over the defined referential subsets which are the Low Risk (LR), Medium Risk (MR) and High Risk (HR), LE u MR and HR and the driving scores are the masses over these subsets.