Rank one update-based efficient adaptive control allocation for multicopter

This paper proposes an efficient rank one update-based adaptive control allocation (CA) that shows satisfactory computational efficiency. A typical adaptive CAmodule includes parameter estimation and CA parts. In this paper, the parameter estimation part uses a recursive least square (RLS) filter to obtain an estimated CA matrix, and the CA module is designed based on the pseudoinverse method. Because of using the pseudo-inverse method to allocate pseudo commands from the upstream controller, the singular value decomposition (SVD) of the CA matrix is required in every time step. Completely decomposing the matrix by using SVD methods will waste too much computational resources. Noticing that the estimated matrix is only perturbed by a rank one matrix in the parameter update step of RLS, a rank one update method is integrated into the RLS filter in this paper for updating SVD results of the estimated matrix without using SVD methods. Benefited from the low computational complexity of the rank one update method, the computational efficiency of the whole adaptive CA part is therefore enhanced. In the meantime, the control performance will not be deteriorated by using the rank one update method. The proposed rank one update-based adaptive CA is validated on a complex high-fidelity multicopter model and shows satisfactory computation and control performance.     «

This paper proposes an efficient rank one update-based adaptive control allocation (CA) that shows satisfactory computational efficiency. A typical adaptive CAmodule includes parameter estimation and CA parts. In this paper, the parameter estimation part uses a recursive least square (RLS) filter to obtain an estimated CA matrix, and the CA module is designed based on the pseudoinverse method. Because of using the pseudo-inverse method to allocate pseudo commands from the upstream controller, th...     »