Joint pilot optimization and channel estimation using deep learning in massive MIMO systems

Nasser Sadeghi, Masoumeh Azghani, Seyed Amir Mortazavi,
Joint pilot optimization and channel estimation using deep learning in massive MIMO systems,
Digital Signal Processing,
Volume 165,
2025,
105287,
ISSN 1051-2004,
https://doi.org/10.1016/j.dsp.2025.105287.
(https://www.sciencedirect.com/science/article/pii/S1051200425003094)
Abstract: In order to leverage the potential benefits of the massive Multiple-input multiple-output (MIMO) systems, it is crucial to have the accurate channel state information at the transmitter side (CSIT). This paper focuses on the joint pilot optimization and time varying channel estimation in multiuser massive MIMO systems using Deep Learning. The proposed method consists of two off line and on line stages. In the offline mode, a channel estimation network is trained and an offline pilot matrix is optimized. In the online mode, the joint pilot design and channel estimation is conducted using the deep learning scheme. A deep learning layer has been designed inspired by the sparse recovery schemes. The designed layer is used both in the pilot optimization network and in the online channel estimation and pilot optimization network. In the offline pilot optimization network, the inherent sparsity property of the channel has been exploited with the application of several designed layers. The proposed method is capable of tracking the channel variations over time with a reduced number of pilots. The performance of the proposed method has been evaluated in various simulation scenarios using two different channel models. The results confirm the superiority of the suggested scheme in offering a high precision channel estimation with much lower pilot overhead compared to its counterparts.
Keywords: Massive MIMO; Channel estimation; Deep learning; Pilot optimization; Joint pilot optimization and channel estimation