Neural Network Control and Wireless Sensor Network-based Localization of Quadrotor UAV Formations

Abstract

In recent years, quadrotor helicopters have become a popular unmanned aerial vehicle (UAV) platform, and their control has been undertaken by many researchers However, a team of UAV's working together is often more effective than a single UAV in scenarios like surveillance, search and rescue, and perimeter security. Therefore, the formation control of UAV's has been proposed in the literature. Saffarian and Fahimi present a modified leader-follower framework and propose a model predictive nonlinear control algorithm to achieve the formation Although the approach is verified via numerical simulations, proof of convergence and stability is not provided. In the work of Fierro et al., cylindrical coordinates and contributions from wheeled mobile robot formation control The work by Gu et al. proposes a solution to the leader-follower formation control problem involving a linear inner loop and nonlinear outer-loop control structure, and experimental results are provided The associated drawbacks are the need for a dynamic model and the measured position and velocity of the leader has to be communicated to its followers. Xie et al. present two nonlinear robust formation controllers for UAV's where the UAV's are assumed to be flying at a constant altitude. The first approach assumes that the velocities and accelerations of the leader UAV are known while the second approach relaxes this assumption In both the designs, the dynamics of the UAV's are assumed to be available. Then, Galzi and Shtessel propose a robust formation controller based on higher order sliding mode controllers in the presence of bounded disturbances In this work, we propose a new leader-follower formation control framework for quadrotor UAV's based on spherical coordinates where the desired position of a follower UAV is specified using a desired separation, d s , and a desired-angle of incidence, d and bearing, d . Then, a new control law for leader-follower formation control is derived using neural networks (NN) to learn the complete dynamics of the UAV online, including unmodeled dynamics like aerodynamic friction in the presence of bounded disturbances. Although a www.intechopen.com