A learning based approach for reduction of the pendulum behavior of ducted-fan helicopter

Pendulum-like behavior of the helicopter is an effect occurs in the case where the movement of the body of the helicopter lags with respect to its rotors, leading to undesired oscillations when the helicopter is required to hover at desired positions. In this paper, an online learning based controller is designed to dampen the oscillation by coupling a two layers forward neural network with a classical PD control method. The neural network is able to approximate the nonlinear dynamics of the pendulum in an online manner, thus improving the overall performance of the linear PD controller. Simulations results have shown that the proposed technique is capable of suppressing the undesired oscillation in an automated manner and alleviating the needs for adjusting the tuning parameters of the controller.