Linear quadratic feasible predictive control (Q1295004)

A linear quadratic feasible predictive control (LQFPC) algorithm is proposed. The LQFPC gives comparable performance to the recently developed linear quadratic constrained predictive control (LQCPC) and linear quadratic infinite horizon predictive control (LQIHPC) algorithms. The proposed LQFPC algorithm ``centres'' predictions around a ``mean level'' control strategy and thus overcomes (under same weak assumptions) the feasibility difficulties associated with LQCPC. In addition, LQFPC uses a prediction class which converges to the ``feasible unconstrained optimal'' and offers thus an alternative to the set-point conditioning associated with the LQIHPC algorithm. Numerical examples illustrate the advantages of LQFPC at the expense of a modest, often insignificant increase in the final value of the cost function.