Online Parameter Estimation of the B-Matrix of a Quadcopter in Time- and Frequency-Domain

The performance of control algorithms for eVTOL, like incremental nonlinear dynamic inversion (INDI), relies on precise models of the system. Instead of using pre-identified models, we aim to identify the model online. In this paper we compare sequential least squares in frequency-domain to extended Kalman filtering for identification of the B-matrix, used in INDI. We demonstrated the online identification of the B-matrix in mini-quadcopter flight tests. Both methods can identify constant and time-varying parameters, but only with sufficient excitation. Otherwise, the parameters cannot be estimated precisely and drift away. In this study, we lay the basis for integrating online parameter estimation into INDI.     «

The performance of control algorithms for eVTOL, like incremental nonlinear dynamic inversion (INDI), relies on precise models of the system. Instead of using pre-identified models, we aim to identify the model online. In this paper we compare sequential least squares in frequency-domain to extended Kalman filtering for identification of the B-matrix, used in INDI. We demonstrated the online identification of the B-matrix in mini-quadcopter flight tests. Both methods can identify constant and ti...     »