Adaptive control of a static multiple input multiple output system

This paper introduces a control scheme that incorporates adaptive learning and optimization in a closed loop system. An estimated Jacobian that identifies a plants characteristics is obtained using Kalman filter estimation. The obtained Jacobian is used to solve a constrained optimization problem to determine the optimal change in control parameters; this approach is compared to a previously presented approach that utilizes the pseudo inverse of the Jacobian to calculate the input control. Adaptive control schemes of complex, nonlinear systems are in general sensitive to initial conditions, and may not converge for all initial conditions. To overcome this problem a rule-base of "good" initial conditions are stored and used whenever the control starts to diverge.