Abstract: For most of the adaptive multi-beamforming algorithms, it is difficult to keep the robustness of the beams of the strong and weak desired signals at the same time when the power of the desired signals differs greatly and the number of snapshots is not enough. In this paper, a robust adaptive multi-beamforming algorithm based on the covariance matrix reconstruction is proposed. The covariance matrix is reconstructed by combining the subspace transformation and diagonal loading technique. The algorithm can not only keep each beam stable when the power of the desired signals differs greatly and the number of snapshots is not enough, but also keep the robustness of the nulling of each beam. Finally, beam pattern simulations are carried out to verify the improvement of the reconstructed covariance matrix for the robustness of the main lobe and nulling of each beam. The output signal-to-interference plus noise ratio (SINR) comparisons show that the anti-interference performance of the proposed algorithm in this paper is better.