Massive multiple-input multiple-output (M-MIMO) is a key technology for 5G networks which consists on equipping the base station (BS) with hundreds or thousands of antennas. Increasing the antenna separation is paramount in order to make real advantage from an array dimension of the order of thousands of antennas. One potential approach is the integration of the antenna array into large structures, which is referred to as extra-large-scale MIMO (XL-MIMO). However, when employing extremely large arrays, centralized processing architectures face a challenging complexity. A promising solution is to divide the antenna array into disjoint units, referred to as subarrays, with individual processing units. In this paper, we modify the M-MIMO channel model aiming to take two implications of the extreme array dimensions into account: the spherical wavefronts, which is called near-field propagation, and the concentration of the majority of energy received from a specific user on a small portion of the array, namely spatial non-stationarity. Considering the latter, it is not efficient from an energy perspective to activate the antennas with lower channel gains to transmit/receive the signal to/from a given user. Thus, antenna selection methods are quite important in the considered scenario. Two antenna selection (AS) methods for XL-MIMO are proposed in this paper: the adjustable-flexible antenna selection (FAS), and the fixed subarray selection (FSS). Numerical results demonstrate that, by judiciously selecting the antennas subset used to detect the signal of each user in the uplink (UL), the number of active antennas can be reduced considerably in a scenario with up to 64 active users, reducing the power consumption without compromising the throughput. The EE is hugely increased in such scenario. Finally, the FSS scheme achieved the same performance of FAS, while having a simpler hardware implementation.

大规模多输入多输出（M-MIMO）是5G网络的一项关键技术，该技术包括为基站（BS）配备数百或数千个天线。为了从成千上万个天线的阵列尺寸中获得真正的优势，增加天线间距至关重要。一种潜在的方法是将天线阵列集成到大型结构中，这被称为超大规模MIMO（XL-MIMO）。但是，当采用超大型阵列时，集中式处理体系结构将面临复杂的挑战。一种有前途的解决方案是将天线阵列分为具有单个处理单元的不相交的单元，称为子阵列。在本文中，我们修改M-MIMO信道模型，以考虑极端阵列尺寸的两个含义：球形波前（称为近场传播）和从特定用户接收的大部分能量集中在阵列的一小部分上，即空间非平稳性。考虑到后者，从能量的角度来看，激活具有较低信道增益的天线来向给定用户发送信号或从给定用户接收信号不是有效的。因此，在所考虑的情况下，天线选择方法非常重要。本文提出了两种用于XL-MIMO的天线选择（AS）方法：可调-柔性天线选择（FAS）和固定子阵列选择（FSS）。数值结果表明，通过明智地选择用于检测上行链路（UL）中每个用户信号的天线子集，在多达64个活动用户的情况下，活动天线的数量可以大大减少，从而在不影响吞吐量的情况下减少了功耗。在这种情况下，EE大大增加了。最后，FSS方案具有与FAS相同的性能，同时具有更简单的硬件实现。