Michael J. Acheson
(Advisor: Prof. J.V.R. Prasad]

will propose a doctoral thesis entitled,

Optimal Methods for Control Allocation

On

Tuesday, November 27th at 10:30 a.m.
Montgomery Knight Building 317
 

Abstract
Control Allocation continues to rise in prominence in modern flight control.  Typical aircraft configurations have more control effectors than desired outcomes (e.g. moments or accelerations) and the problem is to find the “best” way to allocate the control effectors to achieve these control outcomes.  Control allocation is therefore routinely cast as solving the underdetermined linear optimal control problem.  The potential benefits for a real-time control allocation flight control system are substantial.  Control allocation allows for achievement of desired control outcomes while simultaneously achieving an optimal secondary objective such as minimum drag or minimum control power.  Control allocation also allows for achieving control outcomes that were previously unavailable using the legacy flight control method of grouping or “ganging” flight control effectors.  Additionally, control allocation inherently accommodates control effector failure modes that has the potential to substantially improve safety and the linear matrix algorithms are tractable with linear stability analysis and certification methods.

An in-depth review of existing control allocation algorithms is presented with an emphasis on their limitations in meeting the needs of an “idealized” control allocation algorithm.  The novel Prediction Method and the Affine Generalized Inverse algorithms are presented which guarentee weighted optimal control allocation solutions across the entire Attainable Moment Set and do so in numerically robust real-time algorithms which are adaptable to variable size controls effectiveness matrices.  Also, a novel control effector unsaturation identification and location method is presented which has been applied to the Prediction Method throughout the Attainable Moment Set.  This unsaturation method, when applied to the legacy Cascading Generalized Inverse method resolves the inability of this method to guarantee optimal solutions throughout the Attainable Moment Set.  Finally, the groundwork for the offline generation of a family of weighted generalized inverses is presented which enables direct computation of optimal control allocation solutions without being restricted to a strict subset of the Attainable Moment Set.  Numerical studies and comparisons using an autocoded Simulink ® s-function show the ease and advantages of the new algorithms applied to existing control applications.

Committee

• Prof. J.V.R Prasad – School of Aerospace Engineering (advisor)
• (Pending) Prof. Eric Feron – School of Aerospace Engineering
• Dr. Irene Gregory – Senior Control Researcher, NASA Langley Research Center