Adaptive Neural Network-Based Sliding Mode Tracking Control for Quadrotor Flights with a Payload

Abstract

PurposeUnmanned Aerial Vehicles (UAVs) have emerged as an innovative technology with diverse applications,including medical delivery. Despite the advantages of integrating UAVs into medical systems, certain products may losetheir therapeutic effectiveness when exposed to high-intensity or continuous vibrations. Such payload sensitivitydemands robust and adaptive control strategies to ensure accurate trajectory tracking with minimal oscillations.MethodsIn this sense, an adaptive sliding mode control (SMC) technique, based on an artificial neural network, isdesigned for the quadrotor with its payload. Different combinations of payload weight (1\% to 20\% of the quadrotor mass)and stiffness nominal values (21.49 N/m to 361.88 N/m) are chosen to create a robust dataset by considering therectangular trajectory configuration. Once trained, this adaptive coefficient is included in the control design to adjust in realtime to the undesired dynamics of the payload under various scenarios, including different trajectories, time-varyingpayload, and external disturbances.ResultsNumerical results show that the use of adaptive sliding mode control can positively compensate for theundesirable effects of the payload and consequently increase the stability of the system. A performance index, the RootMean Square (RMS), is also used to evaluate the nominal attenuation of the states of the quadrotor and its payload.ConclusionTherefore, the Adaptive SMC technique allows compensation for the payload's dynamic effects, regardlessof the trajectory type, nominal payload value, or boundary uncertainties such as external disturbances. This enables thebroader use of UAVs in the medical field, ensuring safe transportation with minimal vibration.