Research Of ATC's Fatigue State Model Based On Multi Physiological Parameter Fusion

In recent years,with the rapid development of civil aviation,the shortage of control personnel and the delay of flight arise.The fatigue problem faced by ATC(air traffic controllers)is increasing.If the controller is in the fatigue state but still on duty,it will directly threaten aviation safety.So how to determine the fatigue state of the controller becomes a critical issue.This paper draws on a variety of other areas of the method to determine the fatigue,with multiple physiological parameters and fatigue degree as input and output,based on artificial intelligence learning algorithm framework,constructs the evaluation model of fatigue according to information fusion technology,the point of the research as follows:Research on the current domestic and foreign's ATC fatigue were summarized,according to the related literature,this paper analyzes of the causes of fatigue,the fatigue measurement method of other existing areas.The three most practical fatigue judging methods were compared,this paper analyzed their advantages and disadvantages,finally decide to use information fusion model to determine the ATC fatigue.The experimental simulation software was based on the tower control,we recruited 20 students to participate in the test,using MP150 multi channel physiological record instrument and eye tracker for recording ATC of normal group and sleep deprivation group.After filtering,we observe the time domain and frequency domains indexes,we use SPSS software and analyze significant between normal group and fatigue group by four kinds of physiological indexes.Data analysis based on SPSS,because of the physiological parameters is not reasonable calibration of fatigue level,this paper analyzes the KSS results and ATC performance results,and extended them to the whole point of fatigue experiments as the fatigue judgment benchmark.To avoid the curse of dimensionality,partial correlation analysis was used to determine the best five physiological indexes,the physiological indexes of low correlation were eliminated.So the input and output of the model were determined.Model building based on MATLAB,this paper compares the most widely used artificial intelligence learning algorithms.Finally using the BP neural network,CART decision tree and Gauss kernel support vector machine algorithm to construct three kinds of fatigue prediction model.This paper compares the gap of the accuracy of between these three algorithms,among them,the prediction accuracy of support vector machine model is highest.Thus,the current fatigue can be determined according to the current physiological parameters,and an alarm can be provided to the controller who enters the fatigue state.