Driver Activity Recognition And Co-driving Strategy Research Under Human-machine Collaborative Driving

Human-machine collaborative driving,which is jointly controlled by the driver and the automatic system,is a major trend in the future development of smart vehicles,and the dynamic adjustment and allocation mechanism of human-machine authorities is one of the key issues.On the one hand,with the intelligence of the cockpit and the diversification of driving scenes,it is necessary to consider the influence of various "human-vehicle-road" factors and conduct a comprehensive quantitative analysis of the driver state under co-driving.On the other hand,to improve vehicle performance and driving experience,how to provide timely automatic assistance according to dynamically changing driver states is also worth further discussion.Based on the theory of the human-machine co-driving system framework,control interaction strategy and driver feature identification,a U-shaped dynamic authority assignment strategy based on driver activity recognition is proposed.Specific research methods are as follows:(1)Theoretical proposal and parameter modeling of the driver activity model.Aiming to comprehensively characterize the driver’s state and obtain the quantitative basis of humanmachine authorities’ assignment,a driver activity model is proposed based on the driver’s cognitive theory and ergonomics characteristics,which take driver’s attention distribution and perceived situational pressure as the influencing factors.A driving behavior experiment under large curvature road condition is designed.Feature extraction and principal component analysis modeling are carried out for steering control data under the gradient of experimental tasks.The result shows that the modeling parameters are consistent with the theoretical hypothesis.After extracting the driving performance index,the driving performance-driver activity curve is fitted by the least square method,and the U-shape variation law is verified.(2)Driver activity recognition model based on driving state monitoring.To solve the problem that the driver activity model depends on controller data input and is difficult to be applied to the human-machine collaborative driving environment,the direct mapping relationship between driving state and driver activity is established.Indicators such as eye-movement fixation time,scan frequency and driving perceived pressure are selected as driver state monitoring methods.Designing typical scenarios and corresponding experimental tasks of driver distraction and stress respectively,driving experiments with cross-variables are conducted,and the driving state indicators and driver activity values are calculated according to a certain time window.An artificial neural network model is established to learn and train the driver activity tags in different driving states,and the effectiveness and universality of the model are verified by the comparison test of the human-machine co-driving system.(3)U-shaped co-driving strategy design and prototype testing.To explore a rational mechanism of human-machine authorities’ assignment,the driving behavior mechanism under humanmachine collaborative driving is analyzed by introducing the Flow Theory,and a U-shaped dynamic authorities assignment strategy is proposed based on the driver activity recognition model.Prescan/Simulink co-simulation is applied to simulate the prototype of the system,and the human-machine cooperation degree,driving stability,comfort and subjective driving experience are evaluated by combining objective driving indexes with subjective driving scales.Based on the human-machine interaction characteristics of the U-shaped co-driving strategy in various driver activity scenarios,a user interface application scheme to enhance driver trust is designed.Through the parametric modeling and experimental research on driver activity,driving state recognition and human-machine co-driving strategy,the paper puts forward a U-shaped dynamic authorities allocation mechanism based on the real-time monitoring of the driver’s comprehensive state.It can provide appropriate automated auxiliary according to different driving scenarios,which has a high degree of flexibility and humanity.The study provides further direction and parameter basis for human-machine cooperation research under co-driving.