Event-Sampled Control of Quadrotor Unmanned Aerial Vehicle Using Neural Networks

In this paper, an event-sampled output-feedback neural network (NN) controller for a quadrotor unmanned aerial vehicle (UAV) is considered. First an observer design is presented, allowing the need for a full knowledge of the state-vector to be avoided. Next, a kinematic controller is designed in order to find a desired translational velocity; the information provided by the kinematic controller will be used in the design of a virtual controller wherein a desired rotational velocity will be determined such that the UAV's orientation converges to its desired value. Finally, the information from the observer, the kinematic controller, and the virtual controller are used in the design of a dynamic controller where NNs will be implemented to approximate uncertainties in the UAV's dynamics; the signals generated by the dynamic controller will ensure that the desired lift velocity and the desired rotational velocities are tracked. In all these designs, the effects of sampling errors are highlighted. Next, by designing an appropriate event-execution law, the sampling errors are shown to be bounded during the inter-event period. Finally, the effectiveness of the proposed event-sampled controller will be demonstrated with simulation results.