A multilayer neural dynamic controller design method of quadrotor UAV for completing time-varying tasks

Abstract

In order to track the time-varying trajectory efficiently and accurately, a multilayer neural dynamic controller is proposed and exploited for quadrotor unmanned aerial vehicles (UAVs). Previous UAV controllers based on neural dynamic methods used forces and torques as control variables. However, in fact, the real control variables of a UAV are the PWM waves for controlling motors. What is more, the motors are always modeled as one-order systems since the existence of inertia and friction, which means that using forces and torques as control variables is not proper and may lead to low precision and inefficient tracking for UAVs. To solve this problem, a new UAV controller which uses exactly what inputs to motors as control variables is designed, so that the designed controller is more practical for UAVs. In the design process, a multilayer neural dynamic controller is obtained by applying the neural dynamic method to position layer, velocity, torque layer and motor layer successively. Both theoretical analysis and computer simulation results verify the effectiveness, convergence, stability and accuracy of the proposed multilayer neural dynamic controller for tracking time-varying trajectories.