Ph.D. Proposal Presentation by
Thursday, May 19, 2005
(Dr. Nader Sadegh, Chair)
"Discrete-Time Neural Network Based Adaptive Control for Nonlinear Dynamic Regimes"
A discrete-time adaptive feedback control design using the neural networks (NN) is investigated. Unlike its compatible continuous-time approaches, the proposed discrete-time NN-based feedback control design is applicable to both non-minimum and minimum phase system of arbitrary relative degree. Stability of the proposed adaptive control design with discrete-time NN is analyzed and proved using Lyapunov and/or passivity-based approach. Simultaneously discrete-time adaptation laws of NN are investigated. Discrete-time investigation is performed to obtain the results of the compatible state-of-the-art continuous-time ones through different approach in discrete-time domain.
Dynamic inversion methodologies are implemented for the control design of nonlinear systems. The discrete-time NN plays a key role as the principal element of the adaptation to eliminate the inversion errors, which subsequently shows robustness against the parameter uncertainties and the unmodeled dynamics of the system at highly nonlinear operational regimes. Pseudo control hedging (PCH) technique is introduced to prevent the NN adaptive element from attempting to adapt to various actuation nonlinearities such as actuator position and rate saturation. Control allocation is also introduced for the case of redundant control effectors.
The proposed discrete-time NN based adaptive c ontrol design demonstrate s its adaptive p erformance for two highly maneuverable aerial vehicles operated with various uncertainties at nonlinear dynamic regimes : F-15 ACTIVE and FQM-117B UAV.