Research On Driving Behavior Characteristics And Its Impacts On Mixed Traffic Flow Based On Metacognitive Theory

The emergence of heterogeneous mixed driving between manual and autonomous driving has brought about changes in driving methods and behavioral characteristics.In complex heterogeneous mixed driving environments,the driving subject has gradually transitioned from human driving to human-machine driving.This dissertation focuses on"driving behavior characteristics"and"mixed traffic flow",drawing on the metacognitive theory in the field of psychology.It relies on the national key research and development plan"Vehicle Road Collaborative System Element Coupling Mechanism and Collaborative Optimization Method".From two aspects of"human driving behavior characteristics"and"human-machine driving behavior characteristics",driving behavior characteristics at three levels of"knowledge analysis,emotional experience,action monitoring"are deeply characterized,and their impacts on mixed traffic flow are systematically studied.Main research contents are as follows:（1）A driving rule extraction algorithm based on metacognitive knowledge analysis is proposed to address insufficient utilization of human common driving experience and dynamic characteristics of drivers in mixed driving environments.A cognitive driving behavior model is established based on knowledge analysis.Besides,a rule extraction algorithm based on artificial neural network integration（ANNI-REA）is designed to explain the complexity and discreteness of driving behavior.Experimental results show that the algorithm improves accuracy by about 0.4%and reduces complexity by about10%compared to common algorithms,verifying the effectiveness of cognitive rule extraction and the correctness of the extracted driving rules.This study provides driving rules’support for anthropomorphic driving of autonomous vehicles in mixed driving environments.（2）The research on human driving emotion mostly uses a single model,which is prone to the problem of low recognition accuracy.A driving emotion recognition model based on metacognitive emotional experience is established.Firstly,extract facial expression features of drivers,and then fuse facial expression and driving behavior based on support vector machine（SVM-FE&DB）for multimodal emotion recognition.Experimental results show that the model recognition accuracy is 92.57%,and the recall rate is 92.76%.This study reveals the internal micro mechanism of driver external behavior decision-making,providing a guarantee for improving the safety of manual driving takeover in mixed traffic environments.（3）At present,there are problems with single evaluation factors and low environmental adaptability in the study of driving style.As a complex behavioral characteristic that includes free lane changing and free following,free driving style lacks comprehensive quantitative evaluation methods.Therefore,a method for evaluating and predicting free driving style based on metacognitive action monitoring has been proposed.This method utilizes fuzzy comprehensive support vector machine（FC-SVM）to quantitatively evaluate and qualitatively predict free driving style from safety,economy,comfort,and other aspects,taking into account both the influence of personal behavior and environmental factors.Experimental results show that the stability of FC-SVM evaluation results is the best,and compared with other commonly used methods,its macro average accuracy is 89.2%.This study provides a guarantee for improving road traffic efficiency in mixed traffic environments and also provides an important basis for studying the impact of different driving styles on mixed traffic flow during mode switching.（4）At present,research on manual takeover of L3 level autonomous driving mainly focuses on impacts of human-machine interaction,and fails to fully consider the potential risks of driver road rage during mode switching.In order to accurately identify the risk state of road rage emotion in the takeover,first input the eigenvector R and other parameters of driver’s emotion change,and then judge whether the driver is in road rage emotion through the classification function.The road rage risk identification model is constructed by fusing vehicle status and environmental information,and the improved random forest algorithm based on spark（SPA-RF）is applied.Experimental results show that the accuracy of SPA-RF algorithm in identifying road rage risk reaches 97.9%,improving the safety and stability of manual takeover by drivers with road rage emotions in mixed traffic environments.（5）Considering that existing take-over control（To C）models do not fully consider the factors that affect driver reaction time,a take-over control model of response based on cognitive analysis（CAR-To C）based on ACT-R cognitive architecture is constructed.The extracted driver steering wheel control rule X₃（LC）is applied.We study the impact of different driver response times when taking over the steering wheel in mixed traffic flow on typical merging ramps.A quantification method for the uncertainty of driver situational cognition is also proposed,describing the cognitive effects of drivers with different cognitive characteristics on vehicle clusters.Experimental results indicate that when the driver takes over the steering wheel with a reaction time of 4.2 seconds in To C,the traffic state is optimal.The model constructed in this study provides a research foundation for other takeover models in mixed traffic flow,and the designed dynamic takeover strategy improves the safety and traffic efficiency of mixed traffic flow.（6）Considering the current lack of a vehicle evolution model for personalized driver takeover,a take-over control based on cellular automata（PCA-To C）model for drivers with different driving styles is constructed,taking into account the impact of different driving styles on mixed single lane highways and typical merging ramps with multiple lanes during human-machine mode switching.Experimental results indicate that an increase in the proportion of mild vehicles has a moderating effect on cautious driving vehicles,while an increase in the proportion of cautious vehicles has an inhibitory effect on aggressive driving vehicles.When the proportions of aggressive,moderate,and cautious types are 1:2:1,the traffic state is optimal when vehicles are manually taken over under the same conditions.This study not only enriches the theoretical system of traffic flow,but also provides technical support for improving the comprehensive performance of mixed traffic.