A Hybrid Scheme for Reconfigurable Intelligent Surfaces: How Many Elements Should be Estimated?

In this work, a low-complexity hybrid scheme is presented for a wireless network assisted by a reconfigurable intelligent surface (RIS), where channel estimation is required for only a subset of the elements. Specifically, in order to reduce the channel training overhead and boost the performance of the RIS-aided network, the RIS is partitioned in two sub-surfaces, which are sequentially activated to assist the communication. The elements of the first sub-surface align their phase shifts, based on the acquired channel state information (CSI) from a channel training period, whereas the elements of the second sub-surface randomly rotate the phase of the incident signals. The performance of the proposed scheme is investigated under the effect of imperfect CSI acquisition at the RIS. Analytical expressions for the outage probability are derived and useful insights on the optimal configuration of the RIS are provided. We show that, by optimizing the number of elements that need to be estimated, the proposed scheme provides significant performance gains and overcomes the limitations caused by the imperfect CSI acquisition.