Cell-free Communication

Cell-free (CF) communication is a cutting-edge technology expected to play an important role in developing 6G wireless networks. This technology extends the traditional massive MIMO technology, which has been widely deployed in 5G networks. CF communication is anticipated to deliver numerous advantages to 6G networks, such as boosting capacity, enhancing reliability, minimizing latency, and improving energy efficiency. It could also help support new applications and use cases, such as ultra-reliable low-latency communications (URLLC), massive machine-type communications (mMTC), and enhanced mobile broadband (eMBB).

Introduction to Cell-Free Communication

As the name suggests, CF communication refers to communication in a network that doesn’t have cell boundaries (i.e., the cell boundary concept disappears in CF communication).

In a traditional cellular network (i.e., centralized radio access network technology), a geographical service area is divided into cells. One BS serves each cell. These traditional cellular networks can’t meet the expectation of next-generation wireless networks (i.e., they can’t serve a large number of UEs as required in machine-type communications such as IoT, uniform coverage, and higher data rate) since the cellular concept leads to Inter-cell Interference (ICI) to UEs at the cell boundary.

**Basic components of CF communication network** (*Source*)

A 4G wireless networks attempt to resolve the ICI issue within the cellular paradigm using distributed coordinated multipoint (CoMP) technology. However, the UE at the cell edge is still affected by ICI from neighboring cell clusters (more discussion is provided in subsequent sections).

CF architecture is based on a user-centric model. CF communication networks eliminate cell boundaries, thus avoiding the ICI in the cellular paradigm. In cell-free architecture, each UE gets served by all APs in its area of significance [1].

One way to implement CF communication is by using radio stripes (as shown in the figure below), which can be easily deployed in dense areas, viz., stadiums, malls, etc. Radio stripes consist of multiple antennas embedded in cables or adhesive tapes that can be easily installed anywhere [2]. Depending on the frequency range, different antennas can be used. These antennas are generally referred to as APs, and these APs form a single massive MIMO BS, managed by a central processing unit (CPU) through the fronthaul link. CPU is located at the end of the stripe, and these are connected to antennas to serve all users in a wide area.

CF Massive MIMO

CF mMIMO is a technology that essentially combines the best features of different technologies, including ultra-dense networks (UDNs), cellular mMIMO networks, CoMP-JT, and CF architectures.

In particular, the large number of distributed antennas in the UDN and the ultra-density of the network bring the antennas closer to the UE and thus provide macro-diversity. In cellular mMIMO networks, many antennas serve all the UEs in the serving area. Likewise, in cell-free mMIMO networks, the total number of antennas (serving the UEs) is much higher than the number of UEs. The CoMP-JT feature

In CF mMIMO, there are more APs than UEs, and the distance between APs and UEs is small. APs cooperate to serve the UEs via coherent joint transmission and reception. C-RAN technology enables cooperation between APs. Specifically, APs are connected to edge-cloud processors through front haul connections, the so-called CPU in the CF mMIMO literature.

A CF mMIMO network is a way to implement a user-centric network since each UE is served by coherent joint transmission from its all-closest surrounding APs [4].

From the UE perspective, there are no cell boundaries in CF mMIMO, and all antennas serve the UEs simultaneously. In particular, many distributed single or multiple antennas simultaneously serve all UEs through a fronthaul network, CPU at the same time-frequency resource, and by exploiting local CSI and performing joint transmission.

**Different technologies associated with CF mMIMO**

Cellular mMIMO vs. Distributed mMIMO vs. CF mMIMO

In a conventional cellular mMIMO network, the service area (i.e., area to be served) is divided into sub-areas called cells. Each cell consists of one BS with co-located large antenna arrays and multiple UEs, where a BS serves all the UEs in the cell. Since co-located antennas process each signal at one BS, spectral efficiency is not so high [6].

Higher spectral efficiency can be achieved by employing distributed mMIMO and co-processing each signal at multiple APs. It can be implemented in either a network-centric manner conventional way) or a user-centric manner.

In the conventional network-centric implementation of distributed mMIMO networks, co-processing is performed in a network-centric manner. In particular, the APs are divided into disjoint clusters, forming a cellular network with distributed antennas within each cell cluster. APs in a cluster cooperate to serve the UEs (i.e., APs in a cluster transmit jointly to the UEs) residing in their common coverage area. It is equivalent to a conventional cellular network with coordinated clusters called CoMP. The CoMP technology reduces the SNR variations within each cluster but keeps the cellular structure. However, interference from neighbouring cell clusters impacts UEs at the cell edge of a cell cluster. Further, 3GPP LTE standardization of CoMP-JT has not achieved many practical gains.

In the user-centric implementation of distributed mMIMO network, co-processing is performed in a user-centric manner. User-specific clusters are formed, i.e., the formation of dynamic (possibly partially overlapped) clusters of APs based on the needs of each served UE. Each UE is served by coherent joint transmission from its selected subset of APs. This approach eliminates the cell boundaries resulting in no inter-cell interference.

A CF mMIMO network is a way to implement a user-centric network.

**The critical differences between Cellular mMIMO, Distributed mMIMO, and CF mMIMO**

. Illustration of Cellular mMIMO vs. Distributed mMIMO vs. CF mMIMO — **Illustration of Cellular mMIMO vs. Distributed mMIMO vs. CF mMIMO**

Advantages of CF Massive MIMO

Uniform service for everyone: By placing antennas everywhere and using user-centric architecture.
Flexibility: In terms of many use cases.
Fewer components: Only small components, viz., radio stripes, are required. Also, it’s possible to manufacture radio stripes using printed electronics methods.
CF mMIMO systems require comparatively small antennas and analog circuitry for the RF modules to allow the AP to be placed in any geographic location [8].
ICI mitigation: The signal co-processing feature in CF mMIMO networks converts the ICI signal into a sound signal. In CF mMIMO networks, all the APs provide service to all UEs in their area of significance; thereby, CF mMIMO networks eliminate the cell boundaries resulting in no ICI. Further, there is no overhead loss from cell handovers [9].
Easy installation: As discussed, CF mMIMO communication can be implemented by using radio stripes, which can be easily installed anywhere.
Invisible: The radio stripes provide an excellent look at the surroundings, unlike the BS antennas on the tower.
Scalability and reduced cost: Easy to scalable and cost is reduced compared to regular BSs, which are very expensive.
Strong macro diversity: The distributed antenna network topology and the ultra-densification feature in CF mMIMO networks provides strong macro-diversity gain, which indicates a more reliable communication link and a high channel gain.
Beamforming and Multiplexing Gains: Inherited from mMIMO since many antennas are distributed throughout the geographical area and serve the UEs coherently [10].

Ongoing Research, Challenges, and Future Research Directions

mmWave CF mMIMO systems

Both mmWave and CF mMIMO techniques are best suited for short-range communications. In particular, the CF mMIMO systems significantly reduce the distance between the APs and UEs while it provides macro diversity gains. This path diversity gain can be used to avoid the blockage effects of mmWave. These features of CF mMIMO resemble the basic requirements of mmWave, making mmWave CF mMIMO a suitable technology for future wireless networks. The research is already going on in the direction of mmWave-aided CF mMIMO systems [11]-[14]. However, there are several open research problems in designing mmWave-aided CF mMIMO systems, which may be considered future research directions.

Non-orthogonal Multiple Access (NOMA) CF mMIMO systems

NOMA and CF mMIMO techniques can further improve the spectral efficiency of the network further. Having these capabilities, a combination of NOMA and CF mMIMO techniques has drawn significant interest. Further, in CF mMIMO, the distances between UEs and APs are more diverse than those in centralized mMIMO systems. This gives further motivation to combine the NOMA and CF mMIMO techniques. The existing literature suggests that the performance of cell-free mMIMO-NOMA systems was initially investigated in [15]-[16]. However, cell-free mMIMO-NOMA systems' performance in uplink scenarios has yet to be adequately investigated in the literature. Therefore, the performance analysis of mMIMO-NOMA systems in uplink scenarios is a possible future research direction.

Compliance with existing standards

Cell-free data transmission is likely to be implemented in 5G. However, further work is required for standardization and conceptual development [17].

Prototype development

Developing a practical network from its theoretical concept requires a prototyping phase. The existing literature suggests that the initial cell-free mMIMO-based prototype developed was pCell [18]. However, further research can be carried out to develop a more efficient and straightforward prototype using simple and low-cost components. Specifically, one can start by demonstrating synchronization and joint processing capabilities with fewer APs in a limited area and then carry on with more APs and a larger coverage area.

Channel Modeling

In the literature, the performance of CF networks is mainly analyzed for Rayleigh fading channels [19]-[22]. However, the practical channel modeling of CF networks will likely involve a mixture of both LoS and NLoS paths and will vary depending on the carrier frequency. Therefore, modeling a practical channel for CF networks is a possible research direction.

Resource allocation and broadcasting

Resource allocation and broadcasting functionalities (viz., scheduling, paging, pilot allocation, system information broadcast, random access, etc.) traditionally rely on a cellular architecture. These functionalities in CF networks require new algorithms and protocols [23].

Machine learning (ML) for CF mMIMO

ML tool has been extensively employed for physical layer communications (viz., channel state information feedback, channel estimation, power control, etc.). Furthermore, the existing literature considers using ML tools to improve the system performance of CF mMIMO [24]-[31]. The research is still ongoing in this direction and is a possible future research direction.

Supporting IoT with CF mMIMO

Internet-of-things (IoT) is an evolving technology for next-generation wireless systems. It has been comprehensively investigated for co-located mMIMO systems. Most recently, CF mMIMO has been investigated to support IoT networks [30], [32]-[34]. The benefits offered by CF mMIMO (viz., macro-diversity gain, better coverage, etc.) may allow CF mMIMO to perform better in supporting IoT systems than those supported by co-located mMIMO systems. Therefore, it is interesting to study the use of CF mMIMO further to support IoT.

**Cell-Free mMIMO with RIS**

RIS is an emerging technology for beyond 5G applications capable of significantly boosting the performance of wireless networks. The existing literature shows that CF mMIMO with RIS enhances the system performance, mainly when the direct links between APs and UEs need more quality. However, the key challenges and promising research directions associated with RIS-aided CF mMIMO systems are indicated in [35].

Synchronization

Synchronization between APs is critical to facilitate coherent joint transmission [10]. The impact of imperfect synchronization on network performance has yet to be studied in depth.

Technical issues and challenges with CF mMIMO deployment:

o The main challenge with CF mMIMO is deployment cost reduction. It is challenging for operators to install many distributed antennas because of the time, effort, and cost required in the deployment process.

o Another challenge with CF mMIMO is its scalability to the broader service area. In a typical ideal CF mMIMO, all the antennas are connected to the CPU and are expected to serve all UEs. It seems reasonable to meet these expectations in a small area (i.e., sports stadium, shopping mall, building, etc.). However, deploying CF MMIMO in a wide service area (i.e., outdoor area) is quite challenging, as the number of APs and UEs increases with the increase in the service area. Therefore, the CPU complexity becomes prohibitively high for cooperative signal processing of many antennas for all UEs.

o Another concern with CF mMIMO deployment is its smooth transition from cellular to CF system.

Applications

The practical deployment of radio stripes under different application scenarios is illustrated in the figure below [36]. The radio strips are labeled with white, facilitating invisible installation into existing building elements.

Conclusion

CF mMIMO is undoubtedly a key enabler for next-generation wireless technologies. In cellular mMIMO systems, only the lucky UEs close to BS experience a good quality of service. However, CF mMIMO changed that perception by ensuring that all UEs are equally important; each UE is located centrally and is served by multiple APs. In the past few years, CF mMIMO has attracted much attention from academic research whose main objective is to make the concept practical. So far, most aspects have been addressed theoretically, and the potential is widely recognized; the industry appears ready to pick up and invest in this paradigm shift.