HOOC-EM: Fast Beam Sweeping for LEO Mega-Constellation Customer Terminals

The popularity of low Earth orbit (LEO) satellite broadband networks has grown significantly in the past decade. These LEO networks transmit at high carrier frequencies due to spectral availability, resulting in narrow beamforming at the satellite and customer terminal (CT). However, existing methods for beam sweeping have only been developed for terrestrial mmWave networks (e.g., IEEE 802.11ad, IEEE 802.11ay, IEEE 802.15.3c, and the 3GPP’s 5G NR). These standards call for exhaustive beam sweeping procedures, selecting the best candidate beam pair after each azimuth and elevation is searched. Such methods are unsuitable for LEO network CTs due to satellite dynamics and frequent satellite handoff. This paper proposes a maximum likelihood (ML) technique for CT beam sweeping in LEO broadband networks. The proposed method, called hexagonal odds-based orbital comb and entrainment method (HOOC-EM), first divides the celestial sphere into hexagonal and pentagonal sectors, according to CT beamwidth. A probability density function (PDF) is applied to the sectors and each sector is searched in a ML fashion (i.e., the most likely sector is searched first and so on) stopping once a signal is detected. An empirical a priori PDF and a Markov model are developed using orbital simulations of the constellation. The a priori PDF is dynamically updated according to the Markov model to account for satellite dynamics. This paper compares HOOC-EM to other ML methods informed by two other a priori PDFs, one uniform, and one linear with respect to elevation. These latter methods are not updated dynamically according to the Markov model. HOOC-EM is also compared to a traditional exhaustive raster scan beam sweep. The performance of the proposed method is analyzed by comparing the average time to detection as a function of receiver beamwidth. Numerical results indicate that HOOC-EM is an effective method for LEO broadband beam sweeping and superior to other techniques in cases where the Markov model is recent.