Constrained linear quadratic Gaussian control with process applications (Q791497)

In this paper linear quadratic Gaussian control problems are examined with additional: (a) inequality constraints on the variance of some state and input variable; (b) prescribed form of the controller, which must be a linear instantaneous transformation of some output variable. Algorithms for solving both problems are given, which are based on the optimization of the Lagrangian. For the first problem the algorithm solves at each step a LQG problem with fixed values of the Lagrange multipliers, which are subsequently updated by a quasi-Newton formula. The same ideas are applied for the second problem with a gradient algorithm. No convergence proofs are given. For the adaptive extension of such algorithms the reader is addressed to previous papers of the third author [Automatika 19, 415-418 (1983; Zbl 0509.93043) and Int. J. Control 38, 229-235 (1983; Zbl 0513.93041)]. Finally, three industrial applications are lengthy discussed.