| In recent years,autonomous vehicles have become a hot topic in the field of transportation.It can sense the external environment in real time,realize intelligent decisions,avoid obstacles in advance,and effectively reduce the occurrence of traffic accidents.However,in the urban environment,due to the complex and changeable driving scenarios.the behavior of traffic participants is challenging to predict.At the same time,due to the influence of bad weather,equipment failure and other factors,autonomous vehicles will inevitably encounter collisions,which is called moral dilemma.The moral dilemma caused by autonomous vehicles is a worldwide problem.Due to the lack of theories and methods of quantitative and qualitative analysis of ethics,there is little research on engineering application.In this paper,the virtual driving experiment is used to restore the typical moral dilemma scene,and the influencing factors of driving decision in the scene are analyzed.An improved PSO-LS-SVM driving decision-making model is established,which includes input indicators such as moral factors,legal factors and emotional intensity,to predict and verify the effectiveness of the model for drivers’ decision-making under the moral dilemmas.The main research work in this paper is mainly focused on the following aspects:(1)In this paper,combined with China’s traffic regulations and accident liability cases,we establish and then design 12 typical "ghost probe" moral dilemma driving scenes.Specifically,the moral factors in driving decision are represented by the number and type of collision objects,and the legal factor is represented by the right of way.Then we carry out virtual driving experiments,collect driving performance parameters and driving decisions,and record specific decisions to form a database.Finally,the gray correlation entropy model is used to extract and rank the influence factors of driving decision from human-vehicle-road-environment.It provides specific input indicators for the following driving decision prediction model.(2)In this paper,twelve typical moral dilemmas are selected as the experimental paradigm of fear emotion.The Ergolab human-computer environment interactive platform is used to record the subjects ’heart rate variability signals(HRV)and respiratory signals(RESP).At the same time,we improve the hormone regulation module and emotion generation module in the artificial endocrine model,and add the two variables of distance to the collision target and the brake pedal to the formula to make the model more in line with the characteristics of the intensity of fear under moral dilemmas.Finally,the characteristic values of the physiological indexes are substituted into the hormone concentration in the endocrine model to calculate the driver’s fear intensity value for the target ahead in the driving simulation experiment to represent human instincts in moral dilemmas,which is as an important part of the factors affecting driving decisions.(3)In this paper,the improved PSO algorithm is used to optimize the parameters of the traditional LS-SVM model.Seven indicators such as driving performance index,moral index,legal index and fear intensity of the target ahead are selected as inputs,and braking and left turning+braking are taken as inputs to establish a driving decision prediction model based on improved PSO-LS-SVM in moral dilemma scenarios.The experimental samples are brought into the model for training and verification.The results showed that the proposed prediction model had higher prediction accuracy than the traditional model.At the same time,it was found that adding the intensity of fear emotion to the decision-making model could improve the fitness of the sample and make the model have faster iterative speed. |
