Adaptive Neural Network-Based Dynamic Surface Control for a 3-DoF Helicopter

This paper presents a new robust control structure to control a 3-degree-of-freedom (3-DoF) helicopter in the presence of uncertainties, disturbance and actuators saturation. The 3-DoF helicopter is an underactuated system where only two DoFs can be fully controlled, so elevation and pitching motions are considered in this paper. Robust control methods conventionally generate control signals with high amplitude to suppress model uncertainties. Nonetheless, the practical actuators might not be able to supply such control efforts, causing actuator saturation. The actuator saturation phenomenon highly affects the motion control performance of the system and even can cause instability. To address this issue in this paper, firstly a smooth prevention function is utilized as an approximation of the practical actuator saturation, limiting the input torque/force in a prescribed range. However, to avoid violating the stability of the system by a prevention function, it should be considered in the dynamic model of the system, leading to a non-affine structure. Hence, the problem of controlling a non-affine system is addressed by proposing a new third-order dynamic model, as well as a robust control scheme. In this approach, a backstepping control is utilized, equipped with an adaptive radial basis function neural network to estimate and compensate for the adverse terms. Finally, the effectiveness of the proposed scheme is evaluated through a detailed simulation study on a 3-DoF helicopter.