Stabilization of a quadrotor system using an optimal neural network controller

This paper presents a new and practical neural network-based optimal control method for a quadrotor system. Due to the difficulty in obtaining the exact model of the quadrotor system and also to deal with its nonlinear characteristic in both simulation and practice, choosing an effective intelligent controller possessing less complexity can increase system stability. For this purpose, the kinematic relations and the 6-DOF dynamic model of the quadrotor are first extracted. Subsequently, a neural network control method is proposed as the main controller to overcome the system nonlinearities as well as the corresponding unstable dynamics. Accordingly, an enumerative learning method is used to optimize the neural network's weights. At last, the performance of the proposed control method is evaluated and then compared to a PID control method using simulation and experiment. The outcomes of this paper clearly reveal that the optimal neural network controller has a good performance in achieving minimum error.

Automated summary

This paper presents a new and practical neural network-based optimal control method for a quadrotor system. By extracting the kinematic relations and dynamic model of the quadrotor, a neural network controller is used to overcome the system's nonlinearities and unstable dynamics, achieving minimum error through weight optimization.