Adaptive control of a continuous-type system with time-varying input delay (Q1119231)

An adaptive algorithm is constructed for controlling a linear unstable first order differential equation having a known time-varying input delay and an unknown time parameter. The goal in the adaptive control is to get the output of the closed-loop system to follow asymptotically the output of a given stable first order input delay model. The controller belongs to a class of model reference adaptive controllers (MRAC). The adaptive control algorithm is formed by two parts: a finite dimensional recursive parameter estimator of a functional differential equation type, and the control law based on the estimated parameter value. The control algorithm is based on the finite spectrum assignment (FSA) principle developed originally for known constant delay systems. In applying the FSA principle a prediction of the state of the system is constructed by using the differential equation to be controlled, i.e. the system model. In predicting the state the system input up to the time-to-go is used. This gives a causal predictor. The state prediction method developed here needs the inversion of the time argument function t-h(t) in which h is the delay function. Consequently, some restrictions are posed for h. It must not pick up those past values of the control already used, and it must be bounded and differentiable. The parameter estimator is obtained by applying Liapunov's method in a function quadratic in the parameter and a state error. The adaptive control algorithm is proved to be globally stabilizing and to result in output convergence. Parameter convergence is not obtained. Only finite differences, i.e. terms \(da(t,k)=a(t)-a(t-k)\) for all finite k, converge. In the proof an essential step is to prove stability and convergence of the solution to a functional differential equation which is essentially driven by the difference da(t,h(t)). An example is given to illustrate the inversion of the delay function needed in the algorithm.