Indirect model reference adaptive control of quadrotor UAVs using neural networks

In this paper we propose an Artificial Neural Network (ANN) based controller for a quadrotor UAV. We use an Indirect Model Reference Adaptive Control scheme to show trajectory tracking in the presence of dynamically modeled thrust and drag coefficients. A plant emulator ANN continuously and accurately predicts the next plant output in addition to back propagating to compute the errors in the control inputs. The control inputs are generated based on the state error and its derivative by the controller ANN. Unknown or changing plant parameters are handled effectively in this manner.