Integrated multiple-objective control design methodologies: With applications to flight control

The research in this thesis is motivated by the integrated flight/propulsion control problem, which is almost one-way coupled and has multiple performance specifications. This thesis consists of four parts, representing a logical progression for multiple objective controller design from one-way coupled systems to fully coupled systems: (1) First, block triangular, or hierarchical, systems are considered. Choosing a block triangular controller, leads to a block triangular closed-loop system, allowing each subsystem to be dealt with separately. A Q-parameterization and a direct state-space method are presented for performing the design. A direct procedure for designing the off-diagonal term is presented, which if certain conditions are met, decouple the subsystems exactly. A limitation of the use of a block triangular controller is that inherent in its formulation, the aim is to eliminate subsystem interaction rather than exploit it. The block triangular approach is thus extended, to maintain the benefits of the sequential controller design process, and to also make use of the useful interaction. (2) The procedure for exact block triangular system decoupling has a natural extension to fully coupled systems. It is shown that under certain conditions (for fully coupled systems), the subsystem controllers may be designed separately such that when they are embedded in a centralized controller, the closed-loop system is decoupled into its subsystems. (3) A parameter optimization method is presented for controller design without any restriction on plant or controller structure. A series of cost functions are defined in such a way that they and all of the gradients have analytical expressions. The use of LQ cost functions as barrier functions to enforce stability is introduced. By use of multiplier methods it is shown how to obtain initial stabilizing controllers that can stabilize n specified plants simultaneously as well as how to impose Hinfinity and mu type constraints. The overall optimization can be solved using any unconstrained method provided the line-search is restricted to the feasible region. (4) The methods are applied to the integrated flight/propulsion control problem.