New methods of mathematical modeling of human behavior in the manual tracking task

Modeling human behavior is a difficult task. The modeling of physical systems (e.g. aircraft, automobiles) and processes based on physics are very mature and well founded. Currently the human behavior models are not as sophisticated as these models of physical systems. This is in part due to the fact that human behavior represents highly complex nonlinear, sometimes non-deterministic and adaptable systems. Conventional approaches of state machines, optimal control theory, general control theory and expert systems have been applied to human behavior modeling. The result in many cases has been human behavior portrayals that are not robust, are unrealistic and do not support the full man-machine systems modeling requirements for analysis and design of complex human-in-the-loop systems. The various time constants and parameter values of these models must be carefully tuned for each application. This time consuming and costly task needs to be streamlined. This dissertation explores applying data driven model development to obtain human behavior models for the manual tracking task. Data driven models use actual experimental data in a systematic manner to develop models. A new mathematical approach known as soft computing (SC) is emerging that shows promise in dealing with the inherent complexity of modeling human behavior.;One such SC technique known as adaptive neuro-fuzzy inference systems (ANFIS) offers a new and dynamic method to develop human behavior models. These techniques use the power of artificial neural networks to learn and classify data, advanced cluster algorithms to partition data, optimized learning algorithms and fuzzy logic.;This research focuses on the application of such new mathematical methods to model human behavior and performance. In particular this dissertation discusses the application of an SC mathematical modeling technique to model human behavior. In this dissertation the neuro-fuzzy technique is applied to a specific human task, the manual tracking problem It is desirable to have models of the human operator as an integral part of a dynamical system. Models of this sort can be used in systems design and trade studies of manual control systems.