In this paper, we introduce a robust beamforming technique using array antennas. The proposed solution constitutes two stages, the first stage exploits the space-alternating generalized expectation-maximization (SAGE) algorithm to decompose the received GNSS signal into its constituent signals, i.e., the direct and the reflected signals. The SAGE algorithm estimates, for the direct and reflected signals, the Angle of Arrival (AoA) and the received covariance matrix. The second stage, on the other hand, utilizes the Minimum Variance Distortionless Response (MVDR) algorithm to produce the weight vector that steers the main beam towards the satellite’s direction and the nulls towards the multipath effect. The MVDR uses the AoA of the direct path and the covariance matrix of the reflected path in order to minimize the multipath effect only. The experimental results reveals that the proposed technique outperforms the traditional MVDR beamforming technique in a complex canyon-like environment.