Feasibility of the promising large intelligent surface (LIS) concept, as well as its scalability, relies on the use of low-cost hardware components, raising concerns about the effects of hardware distortion. We analyze LIS systems with receive-chain (RX-chain) hardware distortion, showing how it may limit performance gains when scaling up these systems. In particular, using the memory-less polynomial model, analytical expressions are derived for the signal to noise plus distortion ratio (SNDR) after applying maximum ratio combining (MRC). We also study the effect of back-off and automatic gain control on the RX-chains. The derived expressions enable us to evaluate the scalability of LIS when hardware impairments are present. The cost of assuming ideal hardware is further analyzed by quantifying the minimum scaling required to achieve the same performance with non-ideal hardware. The analytical expressions derived in this work are also used to propose practical antenna selection schemes for LIS, and we show that such schemes can improve the performance significantly leading to increased energy efficiency. Specifically, by turning off RX-chains with lower contribution to the post-MRC SNDR, we can reduce the energy consumption while maintaining performance. We also consider a more practical scenario where the LIS is deployed as a grid of multi-antenna panels, and we propose panel selection schemes to optimize the complexity-performance trade-offs and improve the system overall efficiency.

Large intelligent surface (LIS) has gained momentum as a potential 6G-enabling technology that expands the benefits of massive multiple-input multiple-output (MIMO). On the other hand, orthogonal space-division multiplexing (OSDM) may give a promising direction for efficient exploitation of the spatial resources, analogous as what is achieved with orthogonal frequency-division multiplexing (OFDM) in the frequency domain. To this end, we study how to enforce channel orthogonality in a panel-based LIS (P-LIS) scenario. Our proposed method consists of having a subset of active LIS-panels coherently serving a set of users, and another subset of LIS-panels operating in a novel low-power mode by implementing a receive and re-transmit (RRTx) process. This results in an inter-symbol interference (ISI) channel, where we characterize the RRTx processing required to achieve simultaneous orthogonality in time and space. We then employ the remaining degrees of freedom (DoFs) from the orthogonality constraint to minimize the RRTx processing power, where we derive a closed-form global minimizer, allowing for efficient implementation of the proposed scheme.

We envision a future in which multi-antenna technology effectively exploits the spatial domain as a set of non-interfering orthogonal resources, allowing for flexible resource allocation and efficient modulation/demodulation. We may refer to this paradigm as orthogonal space-division multiplexing (OSDM). On the other hand, reconfigurable intelligent surface (RIS) has emerged as a promising technology which allows shaping the propagation environment for improved performance. This paper studies the ability of three extended types of reconfigurable surface (RS), including the recently proposed beyond diagonal RIS (BD-RIS), to achieve perfectly orthogonal channels in a general multi-user multiple-input multiple-output (MU-MIMO) scenario. We consider practical implementations for the three types of RS consisting of passive components, and obtain the corresponding restrictions on their reconfigurability. We then use these restrictions to derive closed-form conditions and explicit expressions for achieving arbitrary (orthogonal) channels. We also study the problem of exploiting the degrees of freedom (DoFs) from the channel orthogonality constraint to maximize the channel gain while maintaining the passive RS constraints, and we propose some initial methods with satisfying performance. Finally, we provide some channel estimation and RS configuration techniques within this framework, where the computations are assumed to be performed at the BS, and we derive some limits on the amount of overhead required to achieve channel orthogonalization with RSs. The numerical results confirm the theoretical findings, showing that channel orthogonality with passive RSs can be effectively achieved in practical environments as long as the direct channel is not significant with respect to the RS cascaded channel. We thus take some important steps towards realizing OSDM.

Reconfigurable surface (RS) has been shown to be an effective solution for improving wireless communication links in general multi-user multiple-input multiple-output (MU-MIMO) setting. Current research efforts have been largely directed towards the study of reconfigurable intelligent surface (RIS), which corresponds to an RS made of passive reconfigurable elements with only phase shifting capabilities. RIS constitutes a cost- and energy- efficient solution for increased beamforming gain since it allows to generate constructive interference towards desired directions, e.g., towards a base station (BS). However, in many situations, multiplexing gain may have greater impact on the achievable transmission rates and number of simultaneously connected devices, while RIS has only been able to achieve minor improvements in this aspect. Recent work has proposed the use of alternative RS technologies, namely amplitude-reconfigurable intelligent surface (ARIS) and fully-reconfigurable intelligent surface (FRIS), to achieve perfect orthogonalization of MU-MIMO channels, thus allowing for maximum multiplexing gain at reduced complexity. In this work we consider the use of ARIS and FRIS for simultaneously orthogonalizing a MU-MIMO channel, while embedding extra information in the orthogonalized channel. We show that the resulting achievable rates allow for full exploitation of the degrees of freedom in a MU-MIMO system with excess of BS antennas.

Cell-free massive multiple-input multiple-output (MIMO) consists of a large set of distributed access points (APs) serving a number of users. The APs can be far from each other, and they can also have a big number of antennas. Thus, decentralized architectures have to be considered so as to reduce the interconnection bandwidth to a central processing unit (CPU) and make the system scalable. On the other hand, the APs in a heterogeneous network might have limited processing capabilities and fully-decentralized processing may not be available. In a recent paper, a trade-off between level of decentralization and decentralized processing complexity has been identified. Furthermore, a novel matrix decomposition – the WAX decomposition – which, if applicable to the channel matrix, allows for exploitation of said trade-off without loss of information. The results on WAX decomposition are only available for random channel matrices with-out specific structures, while in a cell-free massive MIMO scenario the channel can have sparse structures. In this work, we study the applicability of WAX decomposition to cell-free massive MIMO with its implications to the above-mentioned trade-off.

Large intelligent surface (LIS) is a technology that extends massive MIMO by considering an even greater number of antennas distributed throughout vast areas. In order to be able to implement this technology, it is crucial to consider decentralized architectures so as to make the whole system scalable. We consider a LIS divided into several LIS panels of smaller size, which can be located far away from each other. We present a modular architecture that allows combining different LIS panels using a binary tree. This architecture is also valid in a cell-free massive MIMO scenario. We make use of a newly defined matrix decomposition, the WAX decomposition, to define the modules that are used within our architecture. We also study the lossless dimensionality reduction in the data to be processed, which can be achieved using our proposed architecture.

Algorithms for Massive MIMO uplink detection are typically based on a centralized approach, by which baseband data from all antenna modules need to be routed to a central node for further processing. In the case of Massive MIMO, where hundreds or thousands of antennas are expected in the base-station, such architecture requires high interconnection bandwidth between antennas and the central node. Recently, decentralized architectures have been proposed to maintain low interconnection bandwidth, where channel-state-information (CSI) is obtained locally in each antenna node and not shared. Further, Massive MIMO performance is sensitive to CSI quality. However, in the literature, ideal CSI is typically assumed in decentralized systems, which is not only far from reality but also limits the generality of the analysis.This paper proposes a decentralized (a term that will be defined in the main body of the paper) architecture with the following main features: (i) the channel matrix is not made available at any single node, (ii) there is no inter-communication among antennas, (iii) the architecture used during the payload data phase, is reused to provide a certain statistic to a processing node, (iv) A non-standard channel estimation problem based on said statistic arises, (v) a matrix inversion is needed (in case of zero-forcing) at said processing node.A hefty share of the paper is devoted to (iv).

In this paper we present fully decentralized methods for calculating an approximate zero-forzing (ZF) equalizer in a large intelligent surface (LIS). A LIS is intended for wireless communication and facilitates unprecedented MU- MIMO performance, far superior to that of Massive MIMO. Antenna modules in the grid connect to their neighbors to exchange messages of information needed for interference cancellation in a fully-decentralized fashion, making the system scalable. By a careful design of how the messages are routed, we show that the proposed method is able to cancel inter-user interference sufficiently well without any centralized coordination, opening the door for the realization of this type of structures.