Abstract
This research aims at proposing an intelligent flight control system which makes unmanned aerial vehicles tolerant of actuator faults. To make the unmanned system downsizing and low-cost, this study focuses on keeping the redundancy by software approach. The purpose of this study is to establish the systematic learning-based design method of control system with the evaluation function based on control purpose. The proposed technique adds to normal flight control system with detection, identification, and accommodation mechanisms. Each mechanism consists of neural network. In this study, the availability of the proposed system is verified by six-degree-of-freedom nonlinear simulation.
