Modelling and recognition of a driver’s starting intentions

Abstract

This paper explores the classification, modelling, and recognition of a driver’s starting intentions of a passenger car equipped with an automatic mechanical transmission or a dual-clutch transmission. A driver’s intentions can be affected by a number of sources such as the driver’s current control actions, his or her visual scanning behaviour, and the traffic environment. The control actions, such as acceleration actions, are used to infer the driver’s starting intentions. Basically, the driver kinesthetically senses the acceleration, but not the speed. Therefore, a dynamic model for the driveline is proposed according to engine test data and the general parameters of the vehicle in order to analyse the influences of the pedal opening, the first derivative of the pedal opening, and the second derivative of the pedal opening on the starting intentions. An orthogonal test design is applied to treat the variables and their combinations for the driver’s decision making during launching the vehicle. It is demonstrated by the numerical results that the sequences of the factors exert different effects on the starting intentions, but the effects of their interactions can be ignored. According to these factors, a hidden Markov model is used to model and recognize the starting intentions. The results show that the system is able to recognize the driver’s starting intentions accurately.