Description
Adaptive Control Research and Flight Testing
by Buddy Michini
For non-conventional flight vehicles, it is difficult to develop
accurate dynamic and aerodynamic physical models. Some examples include flapping, indoor
and miniature flight vehicles that typically operate at low Reynolds numbers and transition
through distinct phases of flight. This variation in flight dynamics makes adequate control
of the vehicle difficult or impossible using classical control techniques. One possible solution
is to use adaptive techniques whereby a control algorithm adapts to changes in plant dynamics
and attempts to correct the system's response to some desired reference model (model reference adaptive
control, or MRAC).
A variant of standard MRAC, known as L1 adaptive control, includes a low-pass filter
in the loop to limit the bandwidth of the control signal. While this algorithm shows good
stability properties and transient performance guarantees, selection of the parameters to generate
some desired reference performance is not immediately obvious. Current research is focused on
developing optimization algorithms to select these parameters based on a multi-objective cost function
incorporating stability margins, closed-loop performance, and hardware limitations. Using RAVEN, controllers
can eventually be implemented and flight tested on vehicles such as the three-wing tailsitter below.
