| The behavior of dispatching personnel plays a decisive role in the safe operation of the train,especially in the emergency scenario of abnormal operation.The improvement of train scheduling automation,on the one hand,reduces the daily workload of dispatchers,while they need to focus on the dynamic changes of the monitoring screen as long as possible.On the other hand,when an emergency occurs,dispatchers need to deal with the emergency quickly and correctly.Facing the emergency scene of abnormal operation,this paper takes the cognitive-decision process of high-speed rail dispatchers as the research object.Aiming at the deep relationship between physiological data,operational data and cognitive-decision process,a method of constructing cognitive decision model based on structural equation model is proposed.The specific research contents include the following aspects.(1)For the construction of cognitive-decision model,this paper analyzed the existing research results at home and abroad.The theoretical method of structural equation model,fuzzy cognitive graph theory,mathematical model of support vector regression and the basic principle of Particle Swarm Optimization were introduced.The research content of this paper was clarified,and the theoretical basis of this research was expounded.(2)The structure of the cognitive-decision mechanism model and the method of data collection for the measurement model were designed.Based on Wickens information processing model,the brain load element was added,so that the cognitive-decision mechanism model for high-speed rail dispatchers was established combining with perceptual load theory and situational awareness concept.Objective measurement methods and subjective evaluation method were both used to measure the observed variables such as perception,comprehension,decision-making,working memory and attention,using 3D-SART scale,1-back test and attention network test.(3)A particle swarm optimization algorithm combined features selection with SVR parameters selection was designed and employed for the assessment of brain load,which was uesd to optimize the training process of support vector regression models based on multi-source physiological signals.Using a combination of signal processing theory and mathematical methods,three independent signal processing and feature extraction channels were constructed for three types of physiological signals,namely,EEG,ECG and eyes movements,which can be well adapted to the respective data characteristics of multimodal physiological parameters.Under the condition that better feature subsets and SVR parameters were used as the optimization-seeking dimensions of the PSO,a mathematical model with the objective of minimizing the error metric MSE was established to optimize the training process of the support vector regression model for brain load.The performance differences between features selection or SVR parameters alone and joint optimization,between MAE orR~2 fitness function and MSE fitness function were compared from the perspectives of particle search dimension and algorithm fitness function.It has been proved that the joint optimization algorithm with MSE as the objective function works best.(4)The structural equation model was fitted,and the model was modified according to the rationality of the path coefficients,significance analysis and the test of the fitting index.The exact structure of the cognitive-decision mechanism model and the corresponding quantified path coefficients were obtained.On this basis,the inference evolution process of the fuzzy cognitive graph was used to analyze the operational principle of each element in the model on cognitive-decision process. |
PDF Full Text Request