Multilayer Neural Control of Aircrafts With Disturbance Compensation

Abstract

In this work, a high-performance multilayer neurocontroller is integrated for unmanned aircraft systems (UASs) with unknown modeling uncertainties. Specially, large endogenous disturbances and exogenous disturbances can be online compensated feed forwardly. Notably, the online learning ability of the multilayer neural networks is strengthened via the weight updating laws composed of different errors. By introducing the second-order filter based backstepping procedure, the integrated algorithm not only protects from the adverse influences generated by tedious derivation iteration, but also possesses a simple scheme for implementation. Additionally, the practicability of the integrated control algorithm is validated on UASs which may suffer from largely unknown endogenous uncertainties and exogenous disturbances via comparative application results.