Velocity gradient algorithms for delay systems (Q1097855)

Algorithms are developed for adjusting the parameters of a controller in plants in which there is delay in the state and the control. The systems are of the form \[ \dot x(t)=F_ 1(\dot x(t-\tau),\zeta)+F_ 2(x(t),x(t-\tau),u(t),u(t-h),\zeta)+f(t) \] where \(\zeta\) is a vector of parameters. The controller is given by \[ u(t)=u(x(t),x(t- \tau),c(t),r(t)) \] where c(t) is a vector of adjustable parameters and r(t) is an input. Adjustment by the velocity-gradient method \[ \dot c(t)=-P \text{grad}_ c B(x_ t(s),u_ t(\theta), c(t), r(t), \zeta) \] is to be chosen so that \[ \lim_{t\to \infty} J(x_ t(s), z(t), u_ t(\theta), r(t))=0 \] where J is a positive definite estimating functional. Here P is a positive definite symmetric matrix, B is the total time derivative of J in an appropriate sense, \(z(t)=x(t)-F_ 1(x(t-\tau),\zeta)\), and \(x_ t\) and \(u_ t\) are the usual symbols for x and u, respectively, over their delay intervals. Using a Liapunov- Krasovskii functional, conditions are found that ensure convergence of the adaptive scheme for various cases.