基于多源生理数据与模糊建模方法的操作员功能状态预测与调节

发布时间：2018-04-21 01:38

本文选题：操作员功能状态 + 适应性自动化　；参考：《华东理工大学》2014年博士论文

【摘要】：从技术可行性、经济性和安全性的角度出发,人类已经意识到,以彻底取代人类操作员为目的的完全自动化的实现正变得越来越困难,人类操作员仍将继续长期存在于各种系统中。因此,对人机交互系统的研究成为了自动化技术发展的另一个分支。在高安全性要求的人机交互系统中,微小的事故往往可能造成巨大的损失。其中,由操作员功能状态(Operator Functional State, OFS)失效造成的其所承担的任务无法有效完成是导致各种事故发生的重要原因。为此,一些学者提出了适应性自动化(Adaptive Automation, AA)的概念。在AA系统中,通过对OFS进行估计和预测,一旦发现操作员出现高风险状态,立刻对其任务负荷进行调整或提醒操作员采取一定措施,以使操作员所承担的任务要求与其当前的状态两者相匹配。在AA系统的构建中,建立可以对OFS进行精确估计和预测的模型是一个关键问题。本文通过对操作员电生理数据进行采集和分析,使用模糊建模方法建立了基于操作员电生理数据的OFS估计和预测模型,完成了如下的研究工作： (1)使用aCAMS (automation-enhanced Cabin Air Management System)软件对5名操作员被试进行了多任务负荷仿真实验,并采集了被试在不同任务负荷下的电生理数据及其任务性能数据。对电生理数据进行滤波、功率谱分析、数据平滑等预处理,通过相关性分析,得到了3个EEG特征作为OFS模型的输入,使用被试的任务性能数据作为OFS的量化指标和模型的输出,为后续的模糊建模工作提供了数据集； (2)使用粒子群优化(Particle Swarm Optimization, PSO)算法对OFS模糊模型的参数进行估计。在该过程中,比较了PSO算法与增量型PID控制器的内在联系,将二者结合提出了一种新的搜索方式,开发了一种增量型PID控制的PSO算法(IPID-PSO)。为了检验该算法的有效性,首先在7个基准函数的优化问题中进行了测试,发现对于多峰函数,IPID-PSO算法在优化效果上有优于其它3种PSO算法的表现。接着,将IPID-PSO算法应用于OFS模糊模型的参数估计中,所得的模糊模型实现了对OFS的良好估计； (3)使用Wang-Mendel (WM)方法进行OFS模糊建模。在基于WM方法进行模糊模型设计时,分析了高斯隶属函数的宽度参数σ对模糊模型抗噪声能力的影响。为了确定最优的σ值,在使用聚类法进行论域划分时,设计了一种混合高斯隶属函数,将对σ的确定转化为对相邻隶属函数重叠度δ值的确定。为了得到最优的δ值,首先比较了使用不同δ值的模糊模型在4个数据集预测中的性能,得到了适用于不同含噪水平数据的最优δ值,说明了进行最优δ值选取的普遍意义。接着,将同样的比较应用于OFS模糊建模中,取得了类似的结论,并实现了对OFS的良好估计。同时,比较结果显示,使用聚类方法加混合高斯隶属函数的论域划分形式在OFS模糊建模中表现出了优于传统均匀论域划分形式的性能； (4)为了实现AA系统的功能,即对操作员高风险状态的预防,使用了OFS预测的概念,据此建立了OFS动态预测模型,并进行了仿真验证。对OFS预测模型的结构进行了估计,结果显示,采用WM方法的一阶模糊模型可以获得最优性能。为了提高对高风险OFS的有效预测率,用多模型策略代替了单模型策略,并建立了多个WM模型用于OFS预测。为了验证该预测模型的有效性,设计了一种自适应任务分配策略,对基于该自适应任务分配策略的人机交互控制系统进行了仿真。仿真结果显示,在该人机交互系统中,OFS得到了有效的调节,操作员的任务性能水平得到了显著改善,同时,操作员高风险状态出现的次数大大减少,从而大幅度提高了人机系统的安全性。
[Abstract]:From the perspective of technical feasibility, economy and security, human beings have realized that the realization of complete automation for the purpose of completely replacing human operators is becoming more and more difficult, and human operators will continue to exist in various systems for a long time. Therefore, the research of human-computer interaction system has become the development of automation technology. Another branch. In high security man-machine interaction systems, small accidents can often cause huge losses. Among them, the failure of the operator's functional state (Operator Functional State, OFS) failure to be effectively completed is an important cause of the occurrence of various accidents. The concept of Adaptive Automation (AA). In the AA system, by estimating and predicting the OFS, once the operator finds a high risk state, it immediately adjusts its task load or reminds the operator to take certain measures to match the operator's task requirements with the current state. In AA In the construction of the system, it is a key problem to establish a model that can accurately estimate and predict the OFS. By collecting and analyzing the operator's electrophysiological data, a OFS estimation and prediction model based on the operator's electrophysiological data is established by using the fuzzy modeling method, and the following research work is completed.
(1) using aCAMS (automation-enhanced Cabin Air Management System) software to carry out a multi task load simulation experiment on 5 operators, and collect the electrophysiological data and the task performance data of the subjects under different task loads. The electrophysiological data are filtered, power spectrum analysis, data smoothing and other preprocessing, through correlation. In sex analysis, 3 EEG features are obtained as the input of the OFS model. Using the task performance data of the subjects as the quantization index of OFS and the output of the model, the data set is provided for the following fuzzy modeling work.
(2) the Particle Swarm Optimization (PSO) algorithm is used to estimate the parameters of the OFS fuzzy model. In this process, the internal relation between the PSO algorithm and the incremental PID controller is compared. A new search method is put forward by combining the two parties, and a PSO algorithm (IPID-PSO) for incremental PID control is developed. The effectiveness of the algorithm is tested first in the optimization of 7 benchmark functions. It is found that for multi peak function, the IPID-PSO algorithm is superior to the other 3 PSO algorithms in the optimization effect. Then, the IPID-PSO algorithm is applied to the parameter estimation of the OFS fuzzy model, and the fuzzy model has achieved a good estimation of the OFS.
(3) using the Wang-Mendel (WM) method to make OFS fuzzy modeling. In the design of the fuzzy model based on the WM method, the influence of the width parameter sigma of the Gauss membership function on the anti noise ability of the fuzzy model is analyzed. In order to determine the optimal value of the sigma, a hybrid Gauss membership function is designed when the clustering method is used to divide the domain. In order to obtain the optimal delta value, the performance of the fuzzy model using different delta values in the prediction of 4 data sets is compared. The optimal delta value suitable for different noise level data is obtained, and the universal significance for the optimum selection of the delta value is explained. Then, the same ratio will be compared. Compared with the OFS fuzzy modeling, a similar conclusion is obtained and a good estimation of OFS is achieved. At the same time, the comparison results show that the clustering method and the domain division of the mixed Gauss membership function are better than the traditional uniform domain classification in the OFS fuzzy modeling.
(4) in order to realize the function of the AA system, that is to prevent the high risk state of the operator and use the concept of OFS prediction, the OFS dynamic prediction model is established, and the simulation verification is carried out. The structure of the OFS prediction model is estimated. The result shows that the first order fuzzy model of the WM method can obtain the optimal performance. The effective prediction rate of risk OFS is replaced by a multi model strategy and multiple WM models are used for OFS prediction. In order to verify the effectiveness of the prediction model, an adaptive task allocation strategy is designed, and the simulation of the man-machine cross control system based on the adaptive task allocation strategy is simulated. The simulation results show that In the human-computer interaction system, the OFS has been effectively adjusted. The performance level of the operator has been greatly improved. At the same time, the number of high risk states of the operator is greatly reduced, which greatly improves the security of the man-machine system.

【学位授予单位】：华东理工大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：X912.9;TP18

【参考文献】