Research On Cognitive Process Monitoring And Competence Of Ship Piloting Operation

Accurate evaluation of ship piloting competency is a basic premise for preventing piloting human error,and in-depth study of cognitive competency is a key scientific issue in establishing a more advanced and better classification selection and training system for piloting competency.The three-level structure of situation awareness(SA),as a classical model of behavioral cognitive processes,is an important theoretical foundation to systematically structure pilots’ behavioral cognitive mechanisms and achieve monitoring of cognitive processes.In the current risk assessment of ship pilotage situations,SA is usually considered only as a single risk factor,and there are few relevant continuous and in-depth systematic studies on its specific structure,content and influencing factors,as well as how to identify its intrinsic cognitive processes by changing the physiological state of individuals and the external environment.In view of this,this paper explores the mechanism of the role of SA in cognitive processes of piloting unsafe behaviors and monitors the low SA levels in pilots’cognitive competency processes,which has important scientific significance and practical application value for reducing humancaused accidents in ship pilotage.To address these issues of cognitive process monitoring and competence of ship piloting operation,based on the identification of SA as an important cognitive element that plays a key moderating role in behavioral cognitive mechanisms,this paper found that eye movement and electroencephalogram(EEG)data can effectively characterize the cognitive hierarchical processes of perception,comprehension,and prediction in the pilots’SA model.Random forest(RF),support vector machine(SVM)and convolutional neural network(CNN)algorithms are good at mining the complex multidimensional correlation effects of pilot’s multimodal physiological data.Therefore,the identification models of perceptual levels in SA cognitive progresses based on pilots’ eye movement data using RF-SVM algorithm,and the identification model of situational understanding levels in SA cognitive progress based on EEG data using RF-CNN algorithm were constructed,and the identification accuracy was found to be maintained above 84.8%.The SA level identification method with D-S(Dempster/Shafer)decision-level fusion based on the model output of eye movement and EEG data has optimal identification results(more than 89.3%),which realizes the dynamic monitoring of pilots’SA complete cognitive competency processes in pilotage continuous task situations,and promotes the effective application of multimodal physiological data-based cognitive competency identification methods in the pilots’ competency classification screening,training and testing situations.The results of this paper provide key technical support for the development of an adaptive evaluation system of pilots’ cognitive competency based on intelligent technology and lay the foundation and framework for monitoring the cognitive process and competency of ship piloting operation in China.The main research contents and progresses are as follows:(1)The Motivation-Cognition-Action-Feedback(MCAF)mechanism was designed to reveal and quantify the correlations of cognitive elements in piloting unsafe behavior,and it provides an important theoretical basis for identifying the key moderating role of SA in the piloting cognitive mechanisms.This study is the first to introduce a systematic SA cognitive model to accurately reflect the pilots’ cognitive processes,i.e.,to effectively identify the behavioral cognitive mechanisms based on SA and theory of planned behavior(TPB)variables using structural equation model(SEM),which quantifies the"black box" of cognitive processes into explicit and concrete influence paths,and provides the essential theoretical foundations for further cognitive process monitoring and competency research.(2)A new way of S A perceptual competency identification applicable to ship piloting situations,i.e.,SA levels identification model based on eye movement metrics using RFSVM algorithm,is proposed to solve the problem of reasonable evaluating the pilots’cognitive competency by their perceptual ability of situational elements and realize the dynamic monitoring of perceptual competency in SA cognitive processes.This study reveals the correlation between pilot’s eye movement metrics and SA levels based on the permutation test method and achieves effective classification of pilot’s multi-noise eye movement data based on RF-SVM algorithm.The results show that the identification accuracy in crossing encounter scenes(94.3%)has better performance than traditional RF(86.8%)and SVM algorithms(86.2%),and the accuracy in poor visibility scenes increases from 86.9%for traditional RF algorithm to 93.4%for RF-SVM.(3)A pilots’SA comprehension identification based on combined EEG multi-band metrics using RF-CNN algorithm is constructed to solve the problem of reasonable evaluating pilots’ cognitive competency by their situational comprehension and realize the dynamic monitoring of comprehension competency in SA cognitive processes.This study reveals the correlation between pilots’ EEG combined features of alpha,beta,and theta frequency bands in the frontal lobe(F)and central area(C)and SA levels based on the permutation test method and achieves effective classification of pilots’ complex multidimensional EEG data based on RF-CNN algorithm.The results show that the identification accuracy can be improved from 78.1%in the traditional RF algorithm to 84.8%in RF-CNN in unberthing scenes,and the accuracy(88.7%)in crossing encounter scenes is also improved by 8.9%compared with the traditional RF algorithm(79.8%).(4)A new method of S A cognitive competency identification based on D-S evidence theory to fuse model results of eye movement and EEG data in decision level is proposed,which solves the problem that a single physiological metric is difficult to completely represent the SA cognitive process because of its limited information characteristics and failure to integrate the information complement and correlation between multi-modal physiological data,and achieves dynamic monitoring of the SA complete cognitive competency process in continuous pilotage tasks.The results show that the optimal identification accuracy of SA cognitive competency(average 95.7%)can be obtained by using the model outputs of RF-SVM based on eye movement data and RF-CNN based on EEG data for D-S decision level fusion,which improves 7.8%and 6.3%,respectively,compared with the former two input level fusion models.