The Integrated Cognitive Architecture Of Driver Take-over Process Of Human-computer Driving Vehicle

The situation that both driver and in-vehicle intelligent system control vehicles together will exist for a long time before the realization of fully automatic driving,and the resulted human-computer mismatch,driver distraction and cognitive limitations become important factors that affect the traffic safety.At present,the cognitive research of driver's take-over on autonomous vehicles mainly focuses on local cognitive behavior in the field of domestic traffic safety,such as operation characteristics,situational awareness,and attention allocation.Research methods are limited to behavioral experiments and lack of computable modeling.Considering the complexity of cognitive system,the development of cognitive architecture modeling is slow,and there is no related research on integrated modeling and simulation for integration of cognitive behavior of driver's take-over on autonomous vehicles.The specific work is shown as below:Firstly,driver's take-over modes for the human-computer driving vehicle were classified based on domestic and foreign literature,take-over conditions and typical scenes were defined.After analyzing the factors of driver's take-over process of human-computer driving vehicle and the non-driving tasks in the vehicle,the cognitive architecture models of non-driving tasks for human-computer driving were established,including the standard visual Su RT(Surrogate reference task),the standard auditory 2-back task,and the DRT(Detection response task)task.According to the non-interrupt processing mechanism of QN-ACTR,the single task model components were combined together to establish a cognitive architecture model for the driver's take-over process of human-computer driving,and the situation that the model production rules need queuing processing was analyzed.Based on the simulation driving experiments,the cognitive architecture model of human-computer driving vehicle take-over process was verified from non-driving tasks,take-over warning time and traffic risk level.The established model does not adjust any cognitive neuropsychology parameters and it was better adapted to driving data.Based on the above research,the cognitive appraisal system of human-computer driving was used to evaluate the driver's take-over eye movement characteristics,driving load and vehicle lateral longitudinal stability in the human-computer driving task.The horizontal gaze dispersion was evaluated by the eye movement behavior records of the driver model.In this paper cognitive resource bottleneck and performance impairment mechanism were studied under driving load by resource demand decomposition analysis.According to checking the production matching of the driver model,the human factor mechanism of driving performance was analyzed.The cognitive appraisal system provided a theoretical basis for improving the traffic safety problem of the human-computer driving,and also provided a new idea for later development and application of the cognitive architecture model for the driver's take-over process of human-computer driving vehicle.