DEVELOPMENTS IN THE CONTROL OF MULTIVARIABLE SERVOMECHANISMS

Abstract

This thesis presents the results of some new investigations and developments in various different aspects of the control of multivariable servomechanisms. In such servomechanisms, the operating conditions and performance requirements typically vary with different applications. For example, in some classes of applications, a fairly complex control system (requiring, say, on-line real-time estimation of the system dynamics and on-line real-time performance optimization) may be needed to achieve the desired specifications. Thus, the first piece of work in the thesis presents an investigative study in multiprocessing for real-time adaptive controllers with high computational requirements. The work includes real-time experimental results demonstrating the effectiveness of the approach adopted. Following this, the development of an iterative algorithm for solving the multivariable "pole-assignment by output feedback" problem is presented. This output feedback problem is of interest in those multivariable servomechanism applications where the traditional, computationally straight-forward, time-invariant output feedback controllers for multivariable dynamical systems are adequate and may be used. The development here effectively provides, for the first time, a suitable iterative procedure for computing the output feedback gain matrix without requiring any special restrictions on the matrix structure. Next, the development of a direct adaptive controller using a feedforward paradigm for multivariable servomechanisms is presented. This method is applicable to classes of servomechanisms where the structure of the system dynamics permits the feedforward paradigm to be used, and where it is not necessary to have the complexity of on-line real-time optimization. Finally, initial developments in realizing an autonomous controller for a typical multivariable servomechanism (in our case, a gyro-mirror targetting system) are presented; the work here is a preliminary study of the autonomous control concept and descriptions of the functional components of this autonomous controller, the hard ware implementation platform, features of the software,' and an experimental session in employing the proposed controller are included.