Page 65 - Kaleidoscope Academic Conference Proceedings 2021

Connecting physical and virtual worlds

of CPUs and accommodations of APs. The principle of CF- and improve the throughput, reliability, and energy
mMIMO signal processing is that one CPU processes signals efficiency per user. In CF-mMIMO, a large number of
from all APs. However, the nationwide mobile network antenna sites communicate with a small number of UEs in a
cannot be realized because the CPU has a limitation of coverage area, and each antenna site; whereas, baseband
processing capacity. To realize the network, it requires the signal processing functions are concentrated in a Centralized
placement of CPUs and accommodations of APs. The other Processing Unit (CPU) that controls them and processes the
type of RAN management is an assignment of resources in a signals that are connected via a fronthaul. Various
RAN, such as computing for CPUs, radio, and transport links. technologies have been considered depending on the
Since the total RAN resources are limited, the RAN capacity of the fronthaul link and the functions to be realized
resources should be used effectively. This management is by the CPU and antenna sites, respectively.
based on the user’s situations in the RAN, such as the link
quality of user terminals or devices, user’s terminal Bjornson et al. categorized the degree of coordination
movement, occasions of mobile communication services, between each antenna site and CPU in CF-mMIMO into four
and the RAN resource usage. levels and characterized them [7]. The network structure for
each level is as follows. Level 4 is completely centrally
controlled, and the signals received by each antenna site are
transmitted to the CPU for channel estimation. Level 3 is
based on Large Scale Fading Decoding (LSFD) and consists
of two stages. In the first stage, each antenna site performs
the channel estimation, and only the channel statistics are
transmitted to the CPU. In the second stage, combining and
detection are performed at the CPU. Level 2 is a simplified
version of Level 3, where only the average value is
calculated at the CPU in the second stage. In this way, the
amount of information transmitted in the fronthaul network
can be reduced. Level 1 is a small cell network in which each
Access Point (AP) operates completely independently and
there is no information transmission with the CPU. The
conclusion is that central control provides better
characteristics than distributed processing for each antenna
Figure 2 – An overview of user-centric RAN architecture site. On the other hand, Ngo et al. state that the Channel State
Information (CSI) is not shared among antenna sites, but is
To enhance the uplink capacity, increasing the number of only used in the uplink and downlink processing of each
MIMO multiplexes on the user terminal is an effective way antenna site [8]. As described above, various configurations
even if there is almost no space to install additional antennas. and processing methods have been proposed on the AP and
Here, we propose a new concept of “virtualized terminal”, a CPU sides to realize CF-mMIMO. However, in order to
user terminal that can treat the antennas of its peripheral realize a user-centric RAN architecture in the real world, it
devices, such as glasses and watches equipped with radio is necessary to find a configuration of CF-mMIMO that can
interface, as antennas of the user terminal itself. The flexibly meet the various communication quality
capability of the user terminal can be expanded. Here, the requirements of users and can be realized with a realistic
connection with the peripheral devices and user terminal scale of equipment, signal processing, and the number of
should be constructed by wireless links, which are ultra-high APs. However, the above-mentioned work has not yet
speed, large capacity, and ultra-low latency even if the sufficiently investigated the system design from the
communication range is very short. That is, the use of ultra- perspective of realizing CF-mMIMO in the real world.
wide bandwidth, e.g., Terahertz, is very suitable for such
requirements. However, to utilize the terahertz band, Recently, a measurement-based evaluation revealed that the
fundamental research, propagation model, development of performance of a semi-distributed deployment is comparable
RF front-end devices and antennas and baseband signal to that of a fully-distributed deployment in terms of coverage
processing chips, and so forth are still required. in an indoor environment while reducing the number of APs
[9]. Therefore, we are also investigating the effectiveness of
3. KEY TECHNOLOGIES OF USER-CENTRIC semi-distributed deployment for a wider range of conditions
RAN ARCHITECTURE using computer simulations. Our simulation results also
verify that semi-distributed deployment has the potential to
3.1 Key technology 1 “Cell-free massive MIMO” significantly reduce the number of antenna sites without
changing the frequency utilization efficiency [10]. Here, we
Cell-free massive MIMO (CF-mMIMO) is a technology that show one of the simulation results under the simulation
has been actively researched since around 2015 and is a conditions shown in Table 1. Here, we assume that the APs
combination of distributed MIMO and massive MIMO. By are regularly placed in an area of 1 square kilometer as the
extending distributed MIMO to massive antennas, it is CF-mMIMO environment. In addition, the number of APs
expected to increase the number of simultaneous connections and the number of antennas per AP are varied within the

– 3 –