Driving style assessment based on the GPS data and fuzzy inference systems

Car insurance can be computed according to the client´s driving behaviour. This option is based on algorithms which use data from the black box. To carry out the monitoring function, the black box integrates an Inertial Navigation System (INS) sensors, a Global Positioning System (GPS), which are Micro Electromechanical Systems (MEMS) based and a flash memory. Researchers were interested in driver behaviour modelling by means of Kalman Filtering (KF) and Hidden Markov Modelling (HMM) and most of them are based only on the driver´s acceleration profile during a period of time. This research paper expands the number of parameters, by adding the jerk, to estimate the driver aggressiveness. It proposes two approaches for driving behaviour supervision based on the vehicle velocity signal which is acquired from the Global Positioning System (GPS). The first approach presents three driving behaviour indicators based on the vehicle acceleration and jerk. Simulation results show some weaknesses in terms of driver aggressiveness estimation. For this purpose, the second approach is based on a developed Fuzzy Inference System (FIS) model. The inputs of the FIS model are the vehicle acceleration and jerk and the output remains the aggressiveness score. Experimental results show that the driver aggressiveness is better estimated by the proposed FIS model since the two input parameters are taken into account simultaneously. Furthermore, it will be shown that the denazification techniques and the GPS signal noise have an impact on the driver behaviour estimation.