5G mobile communication system has commercialized deployment, and its continuous evolution will be integrated with the deep economic economy and form a good 5G industry ecology. In this context, international organizations and governments have plans to carry out 6G mobile communication systems. At present, although 6G has not uniform definitions, there are some initiatives for application scenarios, technical trends and key indicators .On June 6, 2021, my country's Ministry of Industry and Information Technology IMT-2030 (6G) promotion group officially released "6G Overall Vision and Potential Key Technology" White Paper , combing 6G overall vision and eight business application scenarios and corresponding indicator requirements Some key technical indicators of 6G, including: The peak transmission rate of the system will reach the Tbit / S level, the user experience rate reaches 10 Gbit / s, and the time to increase to 100 μs levels simultaneously, the reliability reaches 99.999 99%, etc. And put forward ten key technologies, pointing that 6G will continue to minat the potential of low frequency sections in 5G, improve the spectrum efficiency of the system; the deep tillage millimeter wave frequency band improve the transmission rate and system robustness, and to Taihaz The optical band, expands the spectrum resources of wireless communication, providing ultra-high capacity, large-scale mobile communication services.
Empty port wireless transmission technology has always been an embodiment of the core capabilities of the mobile communication system in the past, and is also the main technical pathway to achieving 6G key indicators. In the spectrum used in the existing 5G (including millimeters and sub-6 GHz bands), the shortage of spectrum resources is still very prominent, and the spectral efficiency needs to be further improved.
Multi-antenna technology and intensive networks as the main method of improving spectral efficiency is widely used in 3g to 5g. The number of transmission antennas of the base station increases from 2 to 5G applications to 5G applications 64 or even 128, and the cell division is also from macro. , Microcell to picocell. However, the problem of physical implementation problems encountered in the centralized system and the interference problem encountered in the cell split, making the spectral efficiency of the 5G system not sustainable. Therefore, it is necessary to break the traditional honeycomb structure and the small-sized way of thinking, using a new type of cellular network and corresponding large-scale collaborative MIMO transmission technology .
Due to the low-band spectrum resource, the expansion spectrum resources are the most straightforward way to increase the peak rate, and thus the terahertz from 5G millimeter wave to the higher frequency band is a major solution to 6G peak rate. However, the high-band wide bandwidth, near-optics and easy to be obscribed, so that it faces numerous technical challenges in mobile communication applications. Under the cellular architecture, by collaborative transmission, it is possible to effectively solve the problem of high frequency bands easy to be blocked, and the robustness of the link can be improved.
This paper first introduces the relationship between noncommitted large-scale MIMO and technology, and then introduces key technologies with large-scale wireless transmission without cells, respectively, and discusses the future research direction of cellular large-scale MIMO.
1 Evolution of multi-antenna technology and technical principles of cellular system
1.1 Multi-antenna technology evolution
Multi-antenna technology is an effective way to improve the spectral efficiency of wireless communication systems. From 2G to 5G, the number of base station antennas receives 64 issued 64 issued by 1 to 16, the data stream of the parallel transmission from 1 to 16, and the system's spectrum efficiency is also large. Lifting. As shown in Figures 1 (a) ~ (c), from 3G to 5G, multi-antenna technology has also passed the point MIMO, point-to-point multi-user MIMO and multi-point multi-user distributed MIMO. Commercial 5G uses a small-to-point MIMO technology in the indoor scenario, a large-scale MIMO in the outdoor macro honeycomb in the outdoor macro. The 5G small cellular network enhances coverage and transmission rates by intensive deployment of low-power stations, but its interference problem is difficult to further increase. The 5G large-scale MIMO can greatly increase the spectral efficiency, but at the same time its power consumption, weight and cost is large, and the bottleneck will be encountered by further increasing the performance of the single-station antenna.
In the deployment of the cellular mobile communication system, the remote wireless unit (RRU) of the fiber optic lattice can be enhanced. A simple way is that the baseband unit (BBU) of the base station is allocated different time-frequency resources for different users. When the uplink receives the baseband signals of the plurality of RRUs, they are sent to BBUs, and multiple RRUs are sent when they are sent. The same signal. This common community implementation is still widely used in 5G small honeycombroom deployment, and also the concept of early distributed antenna systems. When the received signals of the plurality of RRUs are transparent to the BBU, multi-user distributed MIMO (as shown in FIG. 1 (c), as shown in FIG. 1 (c)) is used to obtain a spatial multiplex gain and macro set. Different from a common community implementation, multiple users of distributed MIMO can share the same time-frequency resources, and then significantly improve the spectral efficiency of the system .
1.2 COMP, C-RAN, distributed MIMO and non-cellular system
The application of the RRU and the application of Cloud Wireless Access Network (Cloud-Ran) provide a support for distributed collaboration transmission. C-RAN introduces the concept of baseband pool, gathering the baseband signals of multiple RRUs to baseband pools, thereby increasing the flexibility of the system and reduces the cost of deployment. C-RAN is a deployment and implementation of a wireless access network that supports collaborative transmission or supports non-cooperative transmission. Currently in the 4G and 5G C-RAN commercial deployment, the cooperative transmission of joint processing is not used.
4G introduced collaborative multi-point transmission (CoMP) technology. The CoMP allows collaboration between multiple access points in a cell and a collaboration of multiple sites in the cell. COMP collaboration transport technology includes joint processing, interference coordination, collaborative beamforming, collaborative scheduling. However, 4G introduced CoMP technology is still based on cellular implementation, and due to the limited capacity of interactions between sites, the number of collaborative nodes and antennas is limited, and the advantage of CoMP has not been played.
The infrastructure of non-honeycomb large-scale MIMO is still dependent on distributed RRU deployment, theoretically, a multi-user distributed MIMO. There is a cellular system that can be used in centralized processing and distributed processing. Centralized processing can be deployed by C-RAN, a baseband signal of multiple RRUs to aggregate to a centralized BBU pool, which is jointly processed in the BBU pool. In theory, this centralized implementation can achieve optimal performance . However, it is difficult to achieve an unlimited expansion of the "non-cellular" scale due to the implementation of the bottleneck in the signal processing capability of the BBU pool.
1.3 Scalable noncommitted large-scale MIMO
Figure 2 shows an extensible implementation without cellular uplink transmission method . There are K users in the system, with n single antenna RRU. For uplink transmission, at each RRU, its receive signal Yn, advance row, the initial detection result of K user signal SK can be obtained, and after quantify the detection result, the next level baseband processing unit is transmitted as needed. . In the baseband processing unit, the detection result of a particular user sent by the plurality of RRUs can be merged, and the final detection result of the user can be obtained.
The above implemented manner has the following advantages:
（1）Distributed coherence reception in RRU implementation, no need to interact with other RRUs;
（2）In theory, even if the conjugated multiplication of simple channels, the number N n tends to be infinite, and the user interference can still be eliminated;
（3）Under the support of the forward network, the user can be implemented in different baseband cells, and an arbitrary expansion of the RRU size and user size can be realized. Therefore, the above-mentioned non-cellular implementation method is scalable.
For downlink, we can still use the scalable implementations shown in Figure 2. It can be seen that distributed implementation, its core thinking is that the transceiver is divided into coherent reception / send, signal merge / distribute two entity modules. In theory, the two modules can be distributed, and the system scale can be expanded. However, there is also the following problems compared to centralized distribution implementation:
（1）Centralized implementation can adopt better receivers and precoding, so concentrated, centrally, can achieve better performance than distributed.
（2）Distributed implementation may initiate an increase in the predecessor, as shown in FIG. Each RRU needs to send each user's detection output to the next level, because the pre-transmission overhead is greatly increased.
As described above, it can be seen that there is a honeycomb large-scale MIMO is a distributed MIMO, a realization of the implementation architecture of COMP, has a certain difference from C-RAN. Below, we introduce the challenges and key technologies that have no honeycomb large-scale MIMO in high-frequency sections and low frequency bands.
2 Low-band no honeycomb large-scale MIMO key technology
In the low frequency band of SUB-6 GHz, the channels of large-scale distributed MIMO have the following characteristics:
（1）Multi users are different from multiple nodes, resulting in greater changes in the frequency domain;
（2）Multiple users to multiple nodes of Doppler are not the same, and when the user moves, the channel is changed in the time domain;
（3）The user and node are large, resulting in a large channel matrix dimension. The above three features lead to challenges, transmission method design, etc. in the three characteristics of the three characteristics.
2.1 Channel information acquisition technology
The time division duplex mode can be used to utilize the empty channel channel, and the downlink channel information is obtained according to the uplink detection, thereby reducing the difficulty of the downlink channel information acquisition. Therefore, there is an important role in the intermodular calibration of the null. With the advancement of radio frequency chip technology, the consistency of multiple channels within a single RRU is more mature. However, the null port calibration between multiple RRUs is required because there is a cellular system. Considering that the distance between the RRUs is large, there is a need to study high-performance calibration algorithms, such as the iterative coordinate decline detection of the iterative coordinates proposed by . In addition, in the actual deployment, since the plurality of RRUs are difficult to do, downlink combined precoding needs to consider clock deviations between the RRUs. Considering the clock deviation between the RRUs, the vaccination signal estimation and tracking can be used, and the vaccination signal between the design RRU is required, and the clock synchronization and the transcordability calibration are realized. Fortunately, due to the flexible frame structure of 5G NR, the empty port signal between the RRU can be transparent to the terminal.
When the user is more than a large number of users, the pilot overhead of the uplink detection channel is an important issue. With the sparseness of the channel power domain, the pilot reuse technique can be used to reduce the pilot overhead . For uplink data channels, we need to estimate the delay between multiple users to multiple RRUs, Doppler and other statistics, using parameterized channel estimation methods, and obtain a more accurate demodulation reference signal channel estimate.
The downlink channel status information reference signal (CSI-RS) has an important aid for channel estimation of downlink shared channels. Tracking reference signals (TRS) using terminal transparent can achieve statistical characteristics of multiple RRU composite channels. However, when there is no cellular large-scale MIMO system with user-centric transmission method , only partial RRUs are user service, using traditional TRS, which can cause statistical characteristic measurements. Matching. Therefore, in the user-centered non-cellular large-scale MIMO system, it is necessary to study the CSI-RS configuration and design.
2.2 Distributed transmission method
In order to achieve unlimited expansion of non-honeycomb, it is necessary to consider distributed collaborative receivers and precoding. For uplink receivers, independent multi-user detection can be separated by independent multi-user detection on the RRU side, and multi-user detection can take maximum than merge, zero, minimum mean square error, maximum likelihood and other receivers. The user signal after multi-user detection is quantified to the next level to perform the merge of user signals. For downlink precoding, the RRU can be used to transmit, force zero-pre-encoding or regularization of zero precoding. Considering that the independent receiver or precoding pre-transfer overhead is large, the performance is poor, and the performance of the part RRU combined with a large receiver or the partial RRU combined with precoding is required.
As mentioned earlier, the overall channel has changed in the time frequency. The difficulty of using joint precoding and receivers is the complexity of implementation. For example, when multiple subbands use the same precoding, the subband is not able to be too wide. When the interference of  is used to suppress the receiver, the subband width of the same interference suppression matrix cannot be too wide.
There are also power control and downlink multi-user power allocation associated with the upper downlink transmission. Unlike traditional centralized MIMOs, the upstream power control is achieved for the QoS requirements of the terminal for the cellular system. There are more research on the downstream power distribution of collaborative MIMO. However, for cellular systems, the scalability of the algorithm is needed. In addition, when the multi-RRU is combined with precoding, the power allocation needs to consider the power constraint of each RRU. Document  proposed a scalable power distribution method implemented by greedy algorithm.
When using the user-centric non-cellular system, it is also necessary to study the association of the user and the RRU. Due to the multi-node collaboration capability, user location information can be obtained using an uplink detection channel and a received signal strength. Depending on the user location information, the association of the user can be implemented, and the reuse of the reference signal can be assisted.
3 High-frequency section without cellular large-scale MIMO key technology
Mmmm is a new technology introduced in 5G. Because of the near-optical, easy to be blocked characteristics, the robustness of the link is one of its main challenges. Therefore, the current 5G millimeter wave does not have a large-scale commercial. In addition, since the symbol duration of the millimeter wave system is short, it is also a technique for realizing low-delay. The collaborative transmission technology introduces the millimeter wave system, on the one hand, its robust problem can be solved, achieving ultra-low time delay, high reliability transmission, on the other hand, can improve the spectral efficiency of the system, thereby increasing the total system throughput. Therefore, millimeter wave large-scale collaboration MIMO combined with non-cellular realization architecture will be one of the key technologies that meet 6G high peak rate, high spectrum efficiency, and low-time delay.
However, millimeter waveless nest large-scale MIMO will face more challenges, including:
（1）Affected by the impact of the phase noise and the consistency of the millimeter wave RF front-end channel, the overall upper and downlink channels have tutaneous and the timeliness of calibration still needs to be studied.
（2）Since millimeter wave systems typically use mixed precoding, multiple nodes and multi-user beam scans require further research in non-cellular systems.
（3）The upstream joint reception of millimeter waves has strong implementability, but is different from the low frequency band, the receiver needs to design analog reception beam. Depending on the uplink detection channel, you can solve analog reception beam. After simulating the receiving beam, the multi-user interference of the upstream can be solved in combination with a receiver similar to a low frequency band.
（4）Deciding multi-user collaboration is a systematic problem, especially how to obtain downlink channel information, and achieve mixing precoding. When the empty mouth is available, a collaborative mixed precoding design can be used . When the empty mouth is unavailable, the terminal feedback downlink channel is required. The use of manual intelligence to realize the channel compression feedback is a recent research hotspot , and the feedback overhead is expected to reduce the degree of acceptable degree using the sparseness of the millimeter wave system channel.
4 Network assisted full duplex technology based on non-honeycomb large-scale MIMO
Dual-way way is also a hot spot for mobile communication standards. 5G uses flexible duplex. As the same frequency full-duplex (CCFD) technology is gradually maturing, its application in 6 g is further concerned. However, 5G introduced flexible duplex and CCFD are in the networking, inevitably face cross-link interference issues , ie, the RRU in the transmit state is in the reception state, the interference of the RRU in the reception state, and the terminal transmitted by the uplink. Down the interference of the terminal receiving terminal. Collaborative transmission capabilities without cellular large-scale MIMO provide strong support for more free duplex.
Figure 3 shows a schematic diagram of a network assisted full duplex (NAFD) based on a cell-free frame, which realizes the flexible duplex mode . Its main working principle includes: the upper and down line wireless links are simultaneously performed on the same frequency resources; each RRU is connected to the base station baseband processing unit (BBU) through the predecessor link, and implements a combined baseband processing by the BBU; each RRU is Transceivers to implement transmission or reception or simultaneously send and receive, and determine the appropriate duplex mode by the BBU based on the traffic load of the entire network.For the CCFD RRU, the RRU's transmission and reception self-interference can be eliminated on an analog domain, so we can see it as two RRUs, one for uplink, and the other for downlink. On the other hand, for the transmission of the RRU and the interference between the RRU, the channel matrix between the links can be obtained at a very low overhead estimate, and the centralized processing of the BBU allows it to get all terminals in advance. The downlink signal can be eliminated on the digital domain. Therefore, under the condition of cellular framework, multiple semi-duplex RRUs can be used to achieve full duplex, which is why we call this two-way way NAFD.
The NAFD system still has interference from the uplink user to the downlink user. The main way to eliminate the interference includes the following two:
1）When the down user can estimate the channel of interference users, the interference of the uplink user can be eliminated by interference cancellation techniques;
2）This interference is used in BBU using joint up-down user scheduling and packet pairing or uplink power control.
Compared with existing duplex technologies, there is a difference between the following differences. First, compared to traditional time division duplex, NAFD provides low-delayed services; NAFDs can support non-symmetrical services without reducing spectral utilization than conventional frequency divisions. Second, compared to 5G flexible duplex technologies, for the non-cellular architectural NAFD, the RRU can be half-duplex or CCFD, through joint processing, can reduce the intersection of flexible duplex, mixed half-duplex and CCFD network. interference. In addition, the NAFD based on the cellular architecture can support 5G NR flexible time division duplex: When all RRUs operate in half-duplex mode, different RRUs are different, at the same time, part RRU transmission, part of RRU receive The use of NAFD can reduce the intersection interference caused by this scene. In theory, the performance comparison of NAFD and CCFD is similar to the contrast of distributed MIMO and centralized MIMO, and distributed MIMO can obtain additional power gain and macro psex . Due to the increase in RRU density, NAFD can achieve better performance than CCFD.
NAFD is a free duplex method based on a cellular architecture. At present, it is still facing more problems, including:
（1）In the actual 5G NR system, due to the advance reception of the upstream requirements, the RRU is not aligned in time, how to solve this asynchronous interference problem, need to consider when standardized design.
（2）Eliminating cross-link interference depends on the collaboration between the RRU, using a centralized BBU scheme, it can better eliminate interference. When using distributed transmission and reception, the interference elimination capability requires further research.
（3）With full dynamic RRU transmission and reception control, you need to go to the global angle to study the transmission and receiving mode selection  to reduce interference, improve system capacity.
None of the honeycomb large-scale MIMO is an effective way to break the traditional honeycomb structure and achieve large-scale collaboration. Its basic theory is inherited in multi-user distributed MIMO, which has been widely proved to have significant performance gain. With the advancement of the RF device, the air-port calibration can support non-honeycomb MIMO collaborative transmission, which is the spectrum efficiency of the SUB-6GHz system and improves reliability. After nearly 20 years of research and continuous experimental verification , there is no such thing as a large-scale MIMO to play an important role in 6G systems. The ideological application of the honeycomb large-scale MIMO is in the millimeter wave system, which will have an important support for super upward, and is an important technical approach to further deep millimeter wave bands.However, it is necessary to see that there is still a lot of problems in millimeter waves, and it needs further research and verifies its achievability through experiments. There is a honeycomb large-scale MIMO is an important means of solving the interference problem facing the CCFD network, but how to solve more free and flexible dual-way ways to interfere, there are still many work that requires further study.