Automatic Trust Management Of High-Speed Train Drivers Based On Multi-Modal Physiological Information

With the continuous acceleration of automation,road traffic,rail transit,aviation fields have appeared different degrees of automated driving.Autonomous driving has become an important way to solve road congestion and safety problems,changing the way people travel.High-speed rail autonomous driving technology has realized semi-automated operation(GOA2 level)on the Beijing-Zhangjiakou railway and Beijing-Xiongan railway in China,a milestone in the development of bullet train automation.Autonomous driving in urban rail transit offers many benefits,such as mitigating negative impacts on train drivers’ health,improving efficiency and safety during normal operation.At the same time,many countries have started high-speed trains with 400km/h speed class.Such rapid development makes it more urgent for autonomous driving function to play a larger proportion.If the driver’s job is fully automated,human error such as distraction or misoperation will be reduced,ensuring the safe operation of trains.However,these benefits can only be realized if autonomous driving systems are successfully applied to high-speed trains,and automated trust is an important factor in achieving this goal.Trust is a key factor influencing operators’ acceptance of automation and willingness to use autonomous driving systems.Operators are more likely to use automation systems they trust and less likely to use automation systems they don’t trust.However,if too much trust is placed in automation,it will not be able to respond quickly in emergency situations and will cause problems such as degraded driving ability over a long period of time.If you do not trust the automatic system,you will not be able to play the effect of automation,can not effectively slow down the workload of people.Therefore,in the field of high-speed trains,adjusting the automatic trust level of train drivers is the key problem to avoid accidents and ensure the safety of high-speed trains.This paper takes the automatic trust of high-speed train drivers as the research object,and firstly explores how to evaluate the automatic trust in real time and accurately.Then,the confirmation method of the optimal trust level is proposed,and the optimal trust level is determined as the calibration target to carry out trust calibration.Finally,a calibration method based on system transparency real-time adjustment is proposed to stabilize the automatic trust level of train drivers to the optimal level.The reasonable interaction between the train driver and the automatic system is realized,so that the human-automatic interaction system can exert its maximum effect and reduce the accidents caused by trust deviation.The research in this paper is carried out according to the idea of "discover the problem → analyze the reason →solve the problem",and the main contents are as follows:(1)Automatic trust assessment model of high-speed train drivers based on multi-modal physiological information.Automated trust evaluation is the premise of automated trust management.The high-speed train driving scene is constructed,the fault judgment task is simulated,the physiological data of train drivers(electrodermal activity/EDA;electrocardiograms/ECG;respiration/RSP;and functional near-infrared spectroscopy/f NIRS)under different automatic trust levels is collected,the classification model of physiological information and automatic trust level is established by machine learning algorithm,and the physiological data of different modes and their combinations are compared as input.Classification accuracy under different classification algorithms.The results show that the accuracy of input physiological data of four modes can reach 90% under random forest classification model.The automatic trust level of train drivers can be accurately output by inputting physiological data of train drivers.The automatic trust level can be represented by the real-time monitoring of physiological information,which ensures the objectivity and real-time performance of automatic trust assessment.(2)Automatic trust calibration model of high-speed train drivers based on system transparency regulation.Automated trust management aims to stabilize automated trust level to an appropriate level and avoid security accidents caused by trust imbalance.Based on the experimental data,this paper takes the optimal performance as the goal(shortest fault judgment time,highest judgment accuracy)to construct the objective function to determine the optimal level of automatic trust,and takes this as the calibration goal.Then,the driver fault judgment experiment was set up under different system transparency to explore the relationship between system transparency and automatic trust level.The results show that when the transparency of the system is improved,the automatic trust level of train drivers is gradually enhanced and the driving state is gradually relaxed.However,due to the limitation of initial trust,the influence degree is limited and the automatic trust level can only be adjusted in a small range.Based on the empirical results,an automatic trust calibration measure based on system transparency adjustment is proposed to realize automatic trust calibration of high-speed train drivers.In this study,the automatic trust evaluation model of high-speed train drivers based on multi-modal physiological information was constructed to realize real-time monitoring of automatic trust.The objective of automated trust calibration is to determine the optimal trust level of a specific system with the goal of optimal performance.Then,based on the automated trust evaluation model,the influence of different system transparency on automated trust of train drivers was explored,and automatic trust calibration measures based on transparency regulation were proposed to realize automated trust management.Enrich the theory of automated trust management and consolidate the foundation for the development of intelligent transportation.