| With the rapid development of intelligent vehicles,the development and application of the autonomous driving algorithm have become an important condition for the successful commercialization of intelligent vehicles.At present,it has become the mainstream research direction to develop,test and iterate the autonomous driving algorithm by constructing the virtual vehicle driving environment,which effectively reduces the cost of the autonomous driving algorithm verification and improves the development efficiency of the autonomous driving algorithm in the industry.Therefore,it is an important means to improve the authenticity of the virtual test environment by studying the characteristics and mechanism of human-driven vehicles and making the virtual test environments have the characteristics of traffic vehicles in the real world.Since people play a very important role in the closed-loop system of human-vehicle road,and will bring great uncertainty to the driving process,especially the anger emotion of people will have a profound effect on driving,it is necessary to study the performance of anger in the simulation scene of human driving process.On the one hand,it can contruct the angry driving behavior at any time and enhance the authenticity of the virtual driving test environment.On the other hand,the differences of driving behaviors under different emotions are further explored to enrich the consideration dimensions of the driver model.This paper successfully collected relevant driving data of drivers in anger by inducing drivers to produce angry behaviors,and combined with the third-party NGSIM（Next Generation Simulation）data after analysis and processing to train the identification model of the driver’s angry driving behavior.In addition,the identification model was successfully used to establish the data set of anger and normal driving,and the probability density function was formed statistically,so as to improve the output function of the existing driver model of our research group and obtain the random driving behavior data of angry drivers.Based on this technical route,the research contents of this topic mainly include the following two aspects:（1）Identification model of the driver’s angry driving behavior: The inducing factors that produce anger emotion can be screened out by using the PAD（pleasure,arousal,dominance）emotion theory model.The relevant driving data of drivers in anger were collected successfully by inducing drivers to produce angry behaviors,and combined with the third-party NGSIM data after analysis and processing to establish the vehicle kinematics data set under anger and non-anger.The identification model of the driver’s angry driving behavior is obtained through training according to the Gaussian Mixed Model – Hidden Markov Model（GMM-HMM）.（2）Construction of the driver emotion model under anger: NGSIM data are classified by using the above identification model of the driver’s angry driving behavior and all data are divided into two types of driving behavior data sets of angry driving and non-angry driving,thus forming a data set of angry driving behavior based on NGSIM.Based on the classified NGSIM anger driving behavior data set,the probability distribution function of the dominant eigenvector was built on the basis of the driver model constructed previously by the group,which will serve as one of the references of the driver model movement decision,thus forming the driver emotion model with anger emotion style and generating random driving beh avior data of angry drivers.Finally,the effectiveness of the driver emotion model under anger emotion is verified to some extent with the help of MATLAB – based simulator simulation,and the implementation path of emotion – induced driving experiment based on PAD emotion theory model is successfully explored.This study improves the authenticity of the virtual vehicle driving environment,further enrichs the consideration dimensions of the driver model,and indirectly supports the improvement of the development efficiency of the autonomous driving algorithm. |
