Optimal tuning of H∞ fixed-structure robust controller against multiple high-level requirements using evolutionary computation

This work deals with the optimal controller synthesis against high-level multiple requirements using evolutionary computation. Indeed, such stochastic algorithms are interesting to solve problems based on complex industrial specifications and so seem to be particularly well suited to optimal robust controller synthesis. Using the H-infinity loop-shaping framework, the optimal weights/controller tuning without any structural assumption (in terms of poles/zeros/damping) on the searched filters (except of course their order) is investigated. The absence of such any structural assumption is important to avoid affecting the quality of the solution toward a complex specification and allows reducing the synthesis problem to a simple one with static scalings in place of frequency weights. Using a version of differential evolution algorithm well adapted for high dimensional control problems, computing directly a fixed-structure controller for complex industrial specifications toward a generic nonconstraint fitness with quite reasonable computing time is achieved. The illustrating example deals with the line-of-sight stabilization problem.