Channel Estimation for mmWave Satellite Communications with Reconfigurable Intelligent Surface

We consider an mmWave satellite communication system with a reconfigurable intelligent surface (RIS) to enhance the signal coverage, where both the satellite and the served users are equipped with phased arrays. Different from the existing methods that separately estimate the uplink channel from the user to the RIS and that from the RIS to the satellite, we directly estimate the cascaded channel by proposing two schemes. In the first scheme with two stages, we power off the last antennas of the satellite, user and the RIS in the first stage and then transceive some pilot symbols, while in the second stage we power off the first antennas of the satellite, user and the RIS and then transceive the same pilot symbols. Then we perform the channel estimation based on the estimating-signal-parameter-viarotational-invariance-techniques (ESPRIT) method. In the second scheme that does not power off any antenna and needs only one stage, we propose a null space projection (NSP) algorithm, where the equivalent channel matrix is estimated through projecting the dictionary steering vectors to the null space of the received signal covariance matrices. Simulation results show that the NSP scheme needs much fewer pilots and much lower hardware complexity than the ESPRIT scheme but with some sacrifice in channel estimation performance.