Research in adaptive control theory is motivated by the presence of uncertainties. Uncertainties may be due to a lack of accurate modeling data combined with modeling approximations that result in unmodeled dynamics. They may also be due to external disturbances, failures in actuation and airframe damage. Adaptive control is also motived by the desire to reduce control system development time for systems that undergo frequent evolutionary design changes, or that have multiple configurations or environments in which they are operated. Model reference adaptive control (MRAC) is a leading methodology intended to guarantee stability and performance in the presence of high levels of uncertainties.
This course will present a review of a number of well-established methods in MRAC. Starting with MRAC problem formulation and an overview of classical robustness and stability modifications, this course will continue to introduce the adaptive loop recovery approach that allows the approximate retention of reference model loop properties such as relative stability margins. The course will also present Kalman filtering in adaptive control, in which a Kalman Filter framework is used to update adaptation gains that enables meeting a given performance criteria without excessive tuning.
Who Should Attend
Two novel adaptive control laws are also presented: concurrent learning adaptive control and derivative-free adaptive control. Concurrent learning is a memory-enabled adaptive control method that uses selected recorded data concurrently with instantaneous measurements for adaptation. Concurrent learning guarantees exponential tracking combined with parameter identification for a wide class of adaptive control problems, without requiring persistency of excitation. Derivative-free adaptive control is particularly well suited for systems with sudden (and possibly discontinuous) change in uncertain dynamics, such as those induced through reconfiguration, payload deployment, docking, or structural damage. It provides superior adaptation and disturbance rejection properties, and computable transient and steady-state performance bounds.
The course will also discuss emerging results in connecting machine learning with adaptive control. A special section will be devoted to implementation and flight testing of adaptive control methods, including discussion of the pseudo control hedging methods for handling actuator dynamics and saturation. The course will conclude with discussing extensions to decentralized adaptive control, output feedback adaptive control, unmodeled dynamics, and unmatched uncertainties.
This course will enable individuals from industry and academia to learn more about recent advances in adaptive control theory. The course will provide the tools needed for real-world adaptive control applications, and will be relevant to practicing professionals from electrical, mechanical, and aerospace industries.