Networked Adaptation and Digital Twin Concept Applied to Control of Quadrotors

Abstract

This paper aims to investigate how sampling time variation affects the performance of a control system consisting of online adaptive and local full state-feedback controllers in a digital twin context. To address potential parameter variation and plant uncertainties, a Model Reference Adaptive Controller (MRAC) is designed. To reduce the computational burden, Digital Twin and cloud computing techniques can be used to process the adaptation law, while an adjustable local controller implements the control action. Considering a non-ideal communicating channel or the need to reduce the transmitted data size, a decrease in sampling time for adaptation could be needed, requiring an evaluation of the situation based on the relation efficiency/security. Moreover, a full-state feedback controller with integral action (FSFIA) is designed to be compared with the adaptive controller based on performance metrics. Computational experiments are conducted where the quadrotor nonlinear dynamic is exposed to sampling time variation and non-ideal operating conditions, such as measurement noise, bounded external disturbances, and loss of actuator effectiveness. By defining T1 as the sampling time used in transmission and reception and T2 as the sampling time for update control actions and measurements, the results indicate that values around T1 = 6 T2 could be potential optimal values for the design of the proposed system and reveal that the benefits of MRAC do not justify its application in the range of 30 T2 ≤ T1 ≤ 50 T2.