Neural network control with neuro-sliding mode observer applied to quadrotor helicopter

Abstract

An adaptive neural control scheme based on a new observer applied to quadrotors Helicopter is proposed in this paper. This technique is realized by using two parallel feedforward Artificial Neural Networks (ANN) for each subsystem of the quadrotor. The first one estimates on line the equivalent control term and the second ANN generates observer's corrective term. The main purpose in our work is to reduce the amplification of measurement noise caused by a conventional sliding mode observer by using a new observer. The proposed observer has the same structure as the sliding mode observer. But the discontinuous function in the corrective term is replaced by an adequate ANN to minimize the undesirable phenomenons. The learning algorithms of the two ANNs (controller and observer) are obtained using the Lyapunov stability method. The simulation results are given to highlight the performances of the proposed control scheme.