Suboptimal finite-dimensional controller for a distributed process. I: A deterministic problem of analytical design (Q1062656)

The problem is considered of a finite dimensional approximation of an optimal controller for a distributed process which is described by a boundary-value problem of parabolic type. The proposed approximation is proved to converge to a control optimal in terms of a quadratic criterion as the dimension of the controller increases. The error is estimated in terms of a quadratic functional computed on trajectories of the distributed process. The ''finite dimensional'' description of distributed processes is closely related to the aggregation of the dynamic description of a finite but highly dimensional system. The latter approach is itself equivalent to the Galerkin method, so the problem of the effectiveness of a finite dimensional suboptimal controller is the traditional problem of estimating the accuracy of Galerkin approximations. An example is considered of controlling the temperature field in an infinite plate of fixed thickness by varying the temperature of the surrounding medium.