Development and Testing of a Wearable Sensor based Driving Confidence Monitoring System for a Prototype Electric Vehicle

Abstract

This study aims to detect the driving confidence level by measuring skin electrical conductance and traction motor torque. Measurement data of 38 drivers were collected, feature vectors (mean, standard deviation, kurtosis, skewness) were extracted, and the classes of the drivers were determined while comparing to the safety expert driver data. The best classification method for the galvanic skin response sensor and the current sensor of the electric traction motor is determined. Fine kNN (fine k-nearest neighbors) is the most successful classification method for the galvanic skin response sensor data and the artificial neural network is the most successful method for the current sensor data. In the second part of the study, the driving confidence of the drivers is elaborated around the road junctions where the driver has to detect and prevent possible hazards. Logged data of the galvanic skin sensor, current sensor and velocity measurement is analyzed to monitor the confidence level while approaching and passing two junctions on the pre-specified route. In the third part of the study, based on the GSR sensor measurement, electric motor torque is intervened when a deviation between the average value and measured value. An experimental setup including the RFID card containing the average skin conductivity of each driver is presented.