Speech Title: Prediction and Regulation of Operator Functional State Using Electrophysiological Data and Fuzzy Rule-based Models

Abstract：From the viewpoint of technical feasibility, economy and safety, one has realized that the implementation of fully-automated system to completely replace humans is impractical and human factors always exist in many complex sociotechnical systems. Therefore, the study of human-machine (HM) integrated systems has become an emerging research topic in the field of control systems and automation technology. In safety-critical HM systems, even a seemingly small/trivial accident may likely cause huge loss of lives or properties. The operator performance degradation or breakdown due to the deteriorated operator functional state (OFS) has been found to be one of main reasons of such accidents. To prevent accidents, the notion of adaptive automation (AA) was proposed. An AA system is based on accurate assessment and prediction of the OFS. Once a high-risk OFS is detected, the tasks would be reallocated between human and machine in order to make them match well. In the design and implementation of AA systems, building an accurate model for OFS evaluation and prediction is crucial. Based on the electrophysiological data measured under an aCAMS task environment from five volunteer participants, in this work we employed fuzzy modeling method to elicit the OFS predictive model. The main contributions of this work include:
1) Identification of fuzzy model structure: Physiological data and task performance data were collected while operators were working under different levels of task difficulty/complexity. From the correlation analysis of the pre-processed physiological data, we selected 3 EEG-related features as the input variables of the OFS model. The task performance was treated as the model output.
2) PSO-tuned fuzzy model: A new incremental-PID-controlled Particle Swarm Optimization (IPID-PSO) algorithm was employed to tune the parameters of fuzzy OFS model.
3) Data-driven fuzzy model based on WM method: The benchmark Wang-Mendel (WM) method was employed for fuzzy OFS modeling. The relationship between the width parameter of Gaussian membership function and the noise rejection property of fuzzy model was analyzed. The clustering method was employed for fuzzy partition of the I/O space. By using a two-sided shape of Gaussian membership function, the determination of σ was converted to the determination of overlapping degree δ between two adjacent membership functions. The comparative results demonstrated that the use of clustering-based fuzzy partition method and two-sided Gaussian membership functions outperforms traditional grid partition method for fuzzy OFS modeling.
4) Design of adaptive HM system with regulated OFS: In order to prevent high-risk OFS in an AA system, OFS has to be accurately predicted. The OFS predictive model was constructed and validated. It was shown that the WM-method-based 1st-order I/O model led to the best performance for OFS prediction. To improve the prediction precision of high-risk OFS, a multi-model strategy was used to build multiple sub-models for OFS prediction. Based on the predictive model elicited, an adaptive function allocator was designed to form an adaptive HM system. Simulation results showed that in the designed adaptive HM system: i) the OFS can be effectively regulated with highly reduced number of risky or vulnerable OFS; ii) the operator performance can be significantly improved; and iii) the overall performance and safety of HM system can be substantially enhanced.