Study Of Pilot-Induced Delay Modeling Based On Hidden Markov Model

Nowadays,aviation safety has become one of the most important topics in both industry and people's daily life.With the fulfillment of aviation mechatronics system,the accident rate caused by the aircraft system has been greatly reduced;however,comparatively,human factors has become the key element which obstacles the improvement of aviation safety.As digital simulation plays an important role in evaluation of human factors,the purpose of this thesis is to explore the techniques of modeling and simulating the pilot-induced delay inside a pilot model so as to provide theoretic foundation and technical support to the improvement of aviation safety.The contents of this thesis are as follows.The pilot's cognitive structure is analyzed and the multi-level model of pilot-induced delay is discussed by introducing the theory of hidden Markov model.A pilot-induced delay model is proposed based on the extended hidden Markov model,the layers of which,including the flight state,cognitive state,and time delay,are defined and the measuring methods of these states are given.The pilot's physiological data,namely heart rate(HR),respiration depth(RD),respiration rate(RR),saccade rate(SR),and pupil diameter(PD),and flight data are acquired through design and implementation of the full flight simulation of a left-engine failure scenario.The datasets of flight state,cognitive state,and time delay are set up,and the similarity and difference of which are analyzed among different flight crew.Furthermore,the proposed pilot-induced delay model is instantiated with the acquired datasets and the experimental results show that the proposed method is feasible and applicable.Through cross validation,the model is verified for its reliability and accuracy.The results also suggest that the selection of model structure and parameters should be based on the negotiation of both the descriptive ability and prediction accuracy of the model.