| Autonomous vehicles are the development direction of automotive technology,and hands off/on detection(HOD)is one of the essential systems for autonomous vehicles.In this paper,we built an intelligent driving platform for HOD based on haptic sensing and driver-in-the-loop simulation driving technology for the demand of HOD involved in the field of human-computer interaction and risk control of smart networked vehicles.further studied on the law between vehicle driving conditions and driver’s grip,and established fuzzy inference and BP neural network to identify the typical driver’s grip.The main research contents are as follows:1.We built an intelligent driving platform for HOD based on the flexible pressure sensor array hardware system and software systems such as LabVIEW and Prescan,and developed a real-time driver grip information acquisition system.The HOD system can effectively acquire the driver grip characteristics information,integrate signal acquisition,processing,calculation,result display and driving simulation,which lays the foundation for the subsequent research.2.The laws between vehicle driving conditions and driver grip were studied.The test working conditions such as uniform straight line,double button line and free lane change were built on HOD system.The laws of vehicle speed and steering wheel turning angle on grip force were investigated respectively.The research result shows that the driver’s psychological state and grip posture will change and the magnitude of grip force on the steering wheel will change with the change of vehicle speed and steering wheel turning angle.3.We used a fuzzy inference to combine vehicle driving conditions with driver grip states based on the results of the above-mentioned study on the effect of vehicle operating conditions on driver grip.We take two parameters,vehicle speed and steering wheel angle,as inputs and quantified steering wheel pressure thresholds as outputs.Expert experience and recognition rate of HOD are introduced to build a fuzzy inference model to compare the thresholds with steering wheel grip for grip posture algorithm recognition.4.The data with coordinate attributes of 144 pressure sensor points were imported into the normalization procedure by output of the fuzzy inference the threshold value.The label matrix of different driver grip states were established respectively.Then we used the BP neural network to train the test samples.Finally,the results show that is ability to recognize typical driver grip by BP neural network. |
