Hybrid-Estimation-Based Adaptive Control Allocation Under High Multicollinearity in Sustained Steady Flight

A hybrid-estimation-based adaptive control allocation (ACA) is proposed in this paper that incorporates a hybrid estimator and a cascaded control allocation (CA) scheme. The hybrid estimator assesses the multicollinearity of the correlation matrix and blends ordinary least squares and principal component regression solutions using a multicollinearity-dependent blending factor. When an electric vertical takeoff and landing (eVTOL) aircraft remains in hover during a steady flight, the absence of maneuver commands leads to low persistent excitation, which in turn aggravates the multicollinearity of the regressor matrix. In such a case, the hybrid estimator provides biased but low-variance parameter estimates with reduced sensitivity to noise. These parameter estimates, along with multicollinearity information, are fed into a proposed cascaded CA for allocating desired pseudo-control-rate commands. In the first stage of the cascaded CA, the desired commands are fulfilled by intermediate control inputs. In the second stage, when high multicollinearity is detected, excitation signals along the weak directions of the correlation matrix can be injected promptly into the system, which facilitates regulation of the correlation matrix and further reduces the impact of multicollinearity. The effectiveness of the proposed method is validated on a high-fidelity eVTOL aircraft model in hover mode. By considering the multicollinearity within the proposed hybrid-estimation-based ACA, the tracking performance in the roll, pitch, and heave channels is significantly improved, despite limited enhancement in the yaw channel.