Control engineering is concerned with modifying the behavior of dynamical systems to achieve desired goals. Aerospace systems, such as advanced high performance tactical fighter aircraft, large flexible space structures, and variable-cycle gas turbine aeroengines are becoming increasingly complex. It is imperative that engineers develop an integrated control-system design methodology for high performance controllers, which satisfy multiple design criteria and real-world hardware constraints. Controls research in the School concentrates on three control approaches: fixed architecture control, robust control and adaptive and neuro-fuzzy control.
Fixed-architecture control theory provides a framework for methodologies
in multivariable control-system design. This approach was
originally developed to address the problem of fixed-order linear-quadratic
dynamic compensator design to obtain low-order, high-performance controllers.
The motivation for restricting the controller order arises from limitations
on computer processing capability typical of aerospace applications such
as vibration suppression in large space structures, high performance tactical
fighter aircraft, and variable-cycle gas turbine aeroengines. Subsequently,
considerable effort has extended the fixed-structure approach to address
a broad range of issues in multivariable control system design, including
distributed-parameter systems; discrete-time and sampled-data systems;
decentralized control; model reduction; multiobjective control; parameter-robust
control; and pole and eigenstructure placement. The key feature of fixed-structure
control theory is that all of the preceding issues are addressed within
a common mathematical framework permitting simultaneous treatment of multiple
design goals.
Unavoidable discrepancies between mathematical models and real-world systems can result in degradation of control-system performance including instability. Ideally, feedback control systems are designed to be robust with respect to uncertainties, or perturbations, in the plant characteristics. Such uncertainties arise due to limitations in performing system identification prior to control system implementation, or because of unpredictable system changes that occur during operation. Thus, robustness analysis and synthesis must play a key role in control-system design. Recent research has led to the development of new robustness analysis and synthesis tools for linear and nonlinear control. These are significantly less conservative than previous methods because of their inclusion of phase information regarding the system uncertainty. Specifically, a general framework for robustness analysis and synthesis is being developed that is flexible in addressing a large class of uncertainty structures and restrictive in excluding uncertainties that are not physically meaningful for aerospace systems.
Adaptive and neuro-fuzzy control provide a framework for identifying
and controlling highly uncertain aerospace systems. High levels of redundancy
facilitate adaptation and fault tolerance. Moreover, control processing
is massively parallel and decentralized. Adaptive control is capable of
working around some levels of sensor/motor failure and of recovery from
damage to the neural processing required for control. New techniques in
adaptive and neuro-fuzzy control are being explored in the areas of flight
control of fighter aircraft, helicopter and missile autopilot design,
and integrated flight/propulsion control.