Analysis of linear time invariant systems with time delay. (English) Zbl 0769.93051
Summary: Time Delay Control has recently been suggested as an alternative scheme for control of systems with unknown dynamics and unpredictable disturbances. The proposed control algorithm does not require an explicit plant model nor does it depend on the estimation of specific plant parameters. Rather, it uses information in the recent past to directly estimate the unknown dynamics at any given instant, through time delay. In earlier papers, analysis and implementation of Time Delay Controller for nonlinear systems were discussed. This paper analyzes the continuous Time Delay Controller for a class of linear systems and presents necessary and sufficient conditions for control system stability. A necessary condition for stability is derived using the properties of linear time-delayed systems. This condition involves only a few of the system and controller parameters and facilitates design of the Time Delay Controller.
It is proved that the necessary condition is also sufficient if the delay time is chosen to be infinitesimally small. The convergence of closed loop system error to zero for certain classes of inputs and disturbances when the system is stable is also established. It is also shown that certain approximations in the control algorithm and certain additional unmodeled dynamics render the closed loop system under continuous Time Delay Control to be not exponentially stable due to the controller poles on the imaginary axis at infinitely high frequencies. However, in digital implementation, all the signals are prefiltered by anti-aliasing filters prior to sampling. Hence, the highest frequency component is automatically limited and the issue of exponential instability is not encountered. A discussion is presented comparing Time Delay Control with Repetitive Control. It is indicated that the Time Delay Controller can perform the functions of a repetitive controller with the delay time replaced by the period of the reference input while the repetitive controller can perform the functions of Time Delay Controller for sufficiently small “period” for a certain class of linear systems. Furthermore, examples are included to illustrate the results.
It is proved that the necessary condition is also sufficient if the delay time is chosen to be infinitesimally small. The convergence of closed loop system error to zero for certain classes of inputs and disturbances when the system is stable is also established. It is also shown that certain approximations in the control algorithm and certain additional unmodeled dynamics render the closed loop system under continuous Time Delay Control to be not exponentially stable due to the controller poles on the imaginary axis at infinitely high frequencies. However, in digital implementation, all the signals are prefiltered by anti-aliasing filters prior to sampling. Hence, the highest frequency component is automatically limited and the issue of exponential instability is not encountered. A discussion is presented comparing Time Delay Control with Repetitive Control. It is indicated that the Time Delay Controller can perform the functions of a repetitive controller with the delay time replaced by the period of the reference input while the repetitive controller can perform the functions of Time Delay Controller for sufficiently small “period” for a certain class of linear systems. Furthermore, examples are included to illustrate the results.
