Modelagem matemática da curva de descarga da bateria de um quadrirotor rolling spider utilizando a teoria de identificação de sistemas

Abstract

This work presents the application of systems identification techniques in the mathematical modeling of the battery discharge curve of a Rolling Spider quadrotor, in order to identify the model that represents the highest accuracy. In the mathematical modeling, the structures of linear parametric models ARX, ARMAX, and Box-Jenkins were used. The non-recursive least squares(MQNR) method was also used to estimate a transfer function that represents the quadrotor battery discharge curve. The parameters and validation of the ARX, ARMAX, and Box-Jenkins frameworks were obtained with the aid of the systemIdentification toolbox present in Matlab / Simulink software. The parameters and validation of non-recursive least squares were obtained from an algorithm implemented in Matlab software. The database used in the estimation and validation of the mathematical models was obtained experimentally from a platform of tests and acquisition of data. From the results of the validation of the estimated mathematical models, the discrete model MQNR was selected to represent the discharge curve of the Rolling Spider quadrotor, because it presented better result when compared to the other models. The discrete model MQNR presented an accuracy of 97.5914%, while the discrete models ARMAX and Box-Jenkins presented an accuracy of 65.55% and 66.95% respectively. The ARX discrete model presented the worst result, with an accuracy of 56.86%.