Current control system analysis and design techniques are almost entirely based on the use of parametric equations. As an alternative to this, the thesis aims to propose and develop an approach or methodology for linear control systems which is exclusively non-parametric. That is, at no stage of the modeling, analysis, and design process are parametric equations to be used. The proposed approach is based on the system representation by impulse response (IR) models. Such models, which can also be parametric, are assumed here to be available only in nonparametric or graphical form.

In the thesis, mathematical bases for nonparametric analysis and design are found to be the theories of linear Volterra integral equations and of Mikusinski's convolution quotients. The unique capability of Volterra integral equations to characterize the feedback system structure is recognized. However, the classical analytical solution approach using resolvent kernels is found to be inappropriate for nonparametric analysis.

The general methodology developed in the thesis is a combination of an operational solution method based on convolution quotient theory and a numerical evaluation method based on convolution and deconvolution algorithms. This nonparametric approach is applied to feedback control systems. Various types of nonparametric analysis are performed and demonstrated by numerical examples.

The nonparametric method is also applied to control system design. A basic nonparametric design method is proposed and developed first and is then extended to systems with time-delay, actuator saturation, and nonminimum-phase (NMP) characteristics. The feasibility of extension of nonparametric design to multi-input multi-output (MIMO) systems is demonstrated by design of a two-input two-output control system. The nonparametric controller IR models are realized and used in the form of finite-impulse response (FIR) digital filters.

An application of the nonparametric method to an actual manipulator control system is completed. The real-time control results of the experiments confirm the effectiveness of the method.