KR102589838B1 - Efficient user-serving method in user-centric cell-free massive MIMO model - Google Patents

Abstract

The present invention relates to a user support method that can efficiently eliminate unsupported users when there are a large number of users in an area in a downlink user-centric cell-free massive MIMO model. More specifically, it relates to a user support method that can effectively eliminate unsupported users in an area while preventing unsupported users in the area. It is about methods to ensure user connectivity and throughput.
According to the present invention, the proposed method can operate appropriately under various environments and can effectively reduce the number of unsupported users in an area while properly maintaining the throughput and calculation amount of the entire system.

Description

Efficient user-serving method in user-centric cell-free massive MIMO model {Efficient user-serving method in user-centric cell-free massive MIMO model}

The present invention relates to a method for supporting a large number of users in a user-centric cell-free massive MIMO model, and more specifically, to significantly reduce the number of unsupported users in a downlink user-centric cell-free massive MIMO system. At the same time, it is about an efficient user support method to achieve an appropriate sum rate.

With the advent of 5G, high data rate support has become a feature, but it has been accompanied by limitations in coverage due to the propagation characteristics of the mmWave band. As a way to overcome this, an innovative network model, cell-free massive MIMO, was proposed. Unlike the existing cellular model, the Cell-free Massive MIMO model is a model in which geographically distributed access points (APs) support all users together.

The Cell-free Massive MIMO model maintains the advantages of existing Massive MIMO, such as channel hardening and favorable propagation, and is resistant to the influence of the surrounding environment and shadow fading, so it can provide uniform spectral efficiency to all users in the area. However, the cell-free massive MIMO model has the limitation of having too high a complexity to operate realistically. Accordingly, user-centric cell-free massive MIMO is receiving attention as an alternative for more realistic operation.

The user-centric model is a feasible implementation of the cell-free model and has feasible complexity, but has the limitation that the number of users that can be supported for each AP is limited. Accordingly, in order to create connections with unsupported users in the area, a method has been proposed to connect unsupported users to the AP with the best channel gain.

The above method can effectively eliminate unsupported users when there are a small number of users, but when applied to an environment with a large number of users, it causes the problem of disconnecting existing users to create new connections.

Korean registered patent 10-1161195

The purpose of the present invention is to provide an efficient user support method that can provide services to all users even when there are a large number of users within a user-centric cell-free massive MIMO model. Additionally, while reducing the number of unsupported users in the area, it is possible to guarantee connection to existing users and maintain an appropriate sum rate.

The present invention proposes a user allocation method to support services to all users when there are a large number of users in an area in a downlink user-centric cell-free massive MIMO model.

Observations are made on the distribution characteristics of per-user throughput in this downlink user-centric cell-free massive MIMO model, and the observation results are used as the basis for performing appropriate user allocation. Through the distribution characteristics, it can be seen that among all APs providing services to any user, only the top few APs are responsible for most of the user's throughput.

Accordingly, higher-ranking APs are fixed to existing users, and lower-ranking APs are selected as candidates for providing services to unsupported users. In addition, the candidates assigned to non-supported users are repeatedly assigned from APs within a certain distance in the order of channel gain on the non-supported user side, and the data rates supported for the non-supported users are cumulatively added.

If the throughput exceeds the average of users existing on the entire network, the repetition is stopped and search for the next unsupported user begins. The method presented in the present invention places priority on ensuring connectivity with existing users and an appropriate level of data rate, and to maintain realistic complexity, the maximum number of users that can be supported per AP is limited to the number of antennas mounted on each AP. do.

According to the present invention, in a user-centric cell-free massive MIMO model where a large number of users exist, the connection and data rate of existing users can be guaranteed and the number of unsupported users in the area can be greatly reduced. The user support method built according to the present invention enables realistic implementation, unlike existing cell-free massive MIMO, and can solve user support limitations in the user-centric cell-free massive MIMO model. In addition, CDF performance for sum rate performance and per-user throughput under various environments within the system can also be properly maintained.

[Figure 1] is a diagram showing the overall configuration of a wireless communication system with multiple users in a user-centric cell-free massive MIMO system according to the present invention.
[Figure 2] is a diagram showing the distribution of per-user throughput supported per user based on the environment [Figure 1] according to the present invention.
[Figure 3] shows the sum of the total per-user throughput per user for [Figure 2] according to the present invention and the total number of APs when sorting the total APs supporting that user in descending order based on the data rate supported. This diagram shows a comparison of the per-user throughput sum after deactivating support for APs below the median.
[Figure 4] is a diagram showing the actual algorithm of the AP allocation process for guaranteeing existing user support and unsupported users according to the present invention.
[Figure 5] is a diagram showing simulation results of the number of unsupported users occurring per user-centric cell-free massive MIMO model and comparison model according to the present invention.
[Figure 6] is a diagram showing the comparison results of the sum rate performance per user-centric cell-free massive MIMO model and comparison model according to the present invention according to changes in the number of users in the area under various environments.
[Figure 7] is a diagram showing CDF simulation results for per-user throughput per user-centric cell-free massive MIMO model and comparison model according to the present invention.
[Figure 8] is a bar graph showing the simulation results of the number of unsupported users remaining as a result of the increase in the number of users per user-centric cell-free massive MIMO model and comparison model according to the present invention.

1. Prior art

In the cell-free massive MIMO model, it is common for all APs to support all users within the area. number of APs and the number of users in the area is When expressing it as Because the value of is set to a rather high number (ex.64, 128, 256, ...) as in cellular massive MIMO, it entails high computational complexity. In addition, the cell-free massive MIMO model is a technology that is difficult to implement realistically even if the number of users increases slightly, because the complexity of the system also increases proportionally as the number of users increases.

On the other hand, in the user-centric massive MIMO model, not all APs support all users, and the number of users that can be supported per AP is limited or can vary selectively depending on the environment. Therefore, even if the number of users increases compared to the existing cell-free massive MIMO model, the complexity within the overall model can be maintained at a certain level and thus corresponds to a feasible cell-free model.

Figure 1 is a diagram showing a downlink user-centric cell-free massive MIMO model. Same as the existing cell-free massive MIMO model, the APs (100) in the model are connected directly or indirectly to the central processing unit (CPU) (110), which is a logical unit, through the fronthaul (120), and the connection between CPUs is optical. Assume that cable(130) is used. Additionally, although not shown in the drawing, each AP uses very simple conjugate beamforming (CB) precoding when transmitting downlink data.

In the present invention, as shown in FIG. 1, the maximum number of users that can be supported by each AP (100) is limited. Figure 1 shows a situation assuming that the maximum number of users that can be supported per AP is 2. As explained previously, in the user-centric cell-free massive MIMO model, each AP (100) does not support all users (140) for realism in implementation, so unsupported users (150) may occur within the model, which may occur within the area. This happens more frequently as the number of users increases.

Accordingly, the present invention proposes a method for efficiently providing services to unsupported users 150 that occur within the model shown in FIG. 1.

2. Overall system model and Spectral Efficiency according to the present invention

2.1 Network model and channel model

The present invention considers a user-centric cell-free massive MIMO model in which the number of users that can be supported for each AP (100) is limited to the number of antennas installed per AP (100). Within that model equipped with two antennas There are APs Assume that there are single antenna users. Additionally, the operation between the AP and the user is assumed to be time division duplexing (TDD). Each coherence block of the corresponding network model is expressed in three phase sections as shown in Equation 1 below.

is a channel estimation section (hereinafter referred to as channel estimation) through uplink pilot transmission, is the downlink data transmission section, means the uplink data transmission section.

In the present invention, since a flat fading channel is assumed, AP and user the channel between Expressed as It is defined as here represents large scale fading. is small scale fading (Complex Normal Distribution) is satisfied.

Below, the channel estimation section and downlink data transmission section will be described. Since the present invention deals with a downlink user-centric cell-free massive MIMO model, it does not include a description of the uplink data transmission section.

2.2 Channel estimation through uplink pilot transmission

In the present invention, channel estimation using the MMSE estimator is applied. About the pilot as follows: When defined as, means the pilot transmission length and is the user represents the pilot sequence for and the power is It is defined as At this time, the AP and the receiving user The pilot signal transmitted by is expressed in Equation 2 below.

here refers to the pilot's power and Corresponds to the noise generated when receiving from AP _l and follows independent and identically distribution (iid) and N(0,1) (Complex Normal Distribution).

AP _l calculates the estimate locally through the signal received in Equation 2 above. Equation 3 below means the estimated signal of AP _l .

The MMSE estimator used in the present invention applies the estimate obtained in Equation 3 above. Accordingly, the channel estimate between AP _l and user k finally obtained is expressed as Equation 4 below.

In the following 2.3, downlink data transmission is explained using the channel estimate obtained through Equation 4.

2.3 Downlink data transmission

The present invention deals with downlink data transmission in the user-centric cell-free massive MIMO model. All APs apply conjugate beamforming (CB) as precoding when transmitting downlink data. Equation 5 below is the signal transmitted by AP _l to user k. Express.

In equation 5 above, means the downlink power of AP _l indicates that CB precoding is applied using the channel estimate obtained in 2.2 above. This refers to the signal being transmitted. is the user It refers to the data symbol transmitted to There are restrictions on raw power. In equation 5 above, is AP _l and user It corresponds to the power control coefficient between and is assumed to satisfy the conditions shown in Equation 6 below.

In downlink data transmission through the above equations, the user The signal that is finally received can be expressed in Equation 7 below.

In the present invention, it is assumed that all APs do not support all users, so the connection matrix as shown in Equation 8 below Define.

Equation 8 above is a random AP _l and user As a connection matrix that determines whether there is a connection between If the value of is 0, the corresponding AP and the user are in an unconnected state, and if the value is 1, they are in a connected state. Also, in equation 8 is the user refers to a set of APs that support . In the present invention, the connection matrix Apply to Equation 9 that follows.

The final downlink spectral efficiency in the user-centric cell-free massive MIMO model can be expressed as Equation 9 above. This has the same reformulation as the existing cell-free massive MIMO, but has a connection matrix It has the characteristic of determining whether or not to support a specific user.

3. User connection procedure of the present invention

The present invention proposes an efficient user support method based on the throughput distribution as shown in FIG. 2. FIG. 2 is a diagram showing the distribution of per-user throughput supported per user based on the model environment of FIG. 1.

Figure 2 shows the distribution of throughput supported by APs for each user when four users are randomly selected. Figure 2 shows the distribution by throughput of APs supported for four users. Taking User 2 as an example, it receives throughput support of 0.3355 BPS from all supported APs, and supports throughput greater than the median throughput. It shows that the throughput of APs is 0.3275 BPS.

That is, through the distribution results of the four users in Figure 2, it can be confirmed that most of the throughput supported for a specific user is composed of a small number of APs with high throughput support. Additionally, it can be seen that the average value of the total throughput of users is much larger than the median value of the amount of service provided by each AP. This ultimately means that multiple low-throughput APs have little impact on the overall throughput of the user.

Figure 3 shows a comparison of the change in total throughput that appears when disconnecting from APs below the median based on the throughput provided among the APs connected to each user based on the observation results in Figure 2.

Figure 3 shows simulation results for a total of 10 random users. For each user, the graph L on the left represents the total throughput when supported by all APs initially connected, and the graph R on the right represents the total throughput after disconnecting from APs that provide throughput below the median. Since the amplitude change of graph L and graph R is not large for each user, consistent with the observation results shown in FIG. 2, it is confirmed in FIG. 3 that high-ranking APs with high throughput account for most of the total throughput.

Hereinafter, the efficient user support method proposed by the present invention is presented in the flowchart of FIG. 4. Hereafter, the description of the procedure according to the present invention follows the order indicated by the symbols in FIG. 4.

Number 200 in the flowchart indicates information about the initial connection state. channel matrix corresponds to the channel defined in 2.1 is the user Represents a set of initial APs that support . and means the power value and connection matrix of the channel estimate obtained in 2.3, respectively, and the threshold represents the throughput value corresponding to the median among the throughput of users in the entire network.

above threshold The selection standard is selected as the median of the throughput for each supported user within the existing network. This is used to determine the level of service provision for unsupported users in line with the throughput level of the entire network to set a reasonable standard. It's a method.

Afterwards, it is sequentially checked whether support is not available for the user according to steps (201)-(202) of FIG. 4. If it is determined that a specific user is an unsupported user, that is, if there is no element of the set of APs supporting the user according to (203), the process proceeds to STEP 1 (204). In (204), the power value of the channel estimate for the corresponding unsupported user As a set of candidate APs with high Define. The range of the candidate group is within a certain distance based on the location of the unsupported user. dog or Can be set to APs. In the present invention The candidate group was selected from APs that correspond to the following. This is a set of AP candidates. This is called the selection stage. In this way, by selecting a set of AP candidates based on the physical distance between the user and the AP, it is possible to guarantee basic reception sensitivity and service quality.

Candidate set After selection, the user connection AP confirmation step (205) is performed to prevent the occurrence of another unsupported user. In the user connection AP confirmation step (205), Check whether there is a user who receives service only from a specific AP within the network. If such a user exists, the APs supporting that user are aggregated. A step 206 of removing the connection AP candidates excluded from is performed. This is to prevent disconnection of existing users in order to support unsupported users.

Afterwards, it moves on to STEP 2 (207), where the set Gather existing users who are receiving service from APs within Perform the user subset setting step defined by . This (207) process corresponds to the process for later STEP 4.

User set defined in STEP 2 (207) The AP candidate group channel gain extraction process, which extracts the index of the users within the network and the channel gain value of the APs (or APs) connected to each user, is performed in STEP 3 (208). This is a subset of the above users This is the process of extracting the channel gain between each subset of users included in and the APs supporting these subset users, and this is used in the priority assignment step below.

In the following STEP 4 (209), a user set is created based on the channel gain values extracted in STEP 3 (208). A priority assignment step is performed in which priority is given in order of users receiving support from the most APs among users within the system. As seen in the observation results of FIGS. 2 and 3 described above, priority is given based on the channel gain value extracted in the channel gain value extraction process (STEP 3), and the more users supported by multiple APs, the higher the overall This applies to the high probability that a small number of APs at the top will be responsible for most of the throughput. In other words, this is a step where priority is given in the order of users with the highest throughput supported by the AP, and then to users with the highest side channel gain value.

Afterwards, in STEP 5 (210), a connection AP selection step is performed to extract APs that support less than the median throughput of each AP for that user in the order of the users assigned priority in STEP 4 (209). By going through the process of (210), it is possible to ensure the connection of existing users and at the same time properly maintain the throughput of existing users.

Afterwards, in STEP 6 (212), the AP connection step for unsupported users is performed in which the APs extracted in STEP 5 (210) are released from existing users and sequentially connected to non-supported users. At this time, release from existing users is performed from users with a large number of connected APs. This is to avoid damaging throughput for existing users as much as possible.

When connecting to the release AP for an unsupported user, the threshold value obtained in the previous step (200) is the standard, so if the cumulative throughput provided to non-supported users is the threshold If it is less than or equal to the cumulative throughput and the threshold value comparison step (211) and STEP 6 (212), the process is repeated to sequentially assign the extracted AP to the unsupported user.

Afterwards, the cumulative sum δ of the throughput of unsupported users obtained through STEP 6 (212) is calculated, and the cumulative sum δ is the threshold value. If it exceeds , stop STEP 6 (212), move to step 213 in Figure 4 and repeat the above-described process for another unsupported user (214), or complete the entire process if there are no more users to search. (215).

In STEP 6 above, the threshold This is a method of adjusting the throughput for unsupported users to an appropriate level through repeated size comparison with the threshold value before connecting all extracted APs. If it exceeds , no further connections are made. Through this, it is possible to ensure appropriate throughput for unsupported users and minimize throughput damage to existing users. In addition, even if all extracted APs are connected, the threshold If it does not exceed , the search for the next unsupported user proceeds without any additional connections. This is also because priority is given to minimizing throughput damage to existing users.

Figure 5 shows the simulation results of the user-centric cell-free massive MIMO model according to the present invention and the number of unsupported users per comparison model through a color map.

Number of APs in Figure 5 is set to 100 and is the number of antennas on the AP. is 4, the number of users in the area was set to 100. “Minimum-user” in FIG. 5 simply connects to the AP closest to a specific unsupported user without any additional consideration under the user-centric cell-free massive MIMO model, and connects to the user who receives the weakest throughput from the AP. This corresponds to the method of disconnecting. Additionally, “Full CF” refers to the existing cell-free massive MIMO model, and “UC” refers to the user-centric cell-free massive MIMO model. On the color map of Figure 5, the blue portion represents users who are not supported, and the yellow portion represents supported users. Through the simulation results of the drawing, it was confirmed that applying the user support method presented can effectively eliminate unsupported users in the area.

Figure 6 is a diagram showing the same comparison for each model as in Figure 5, but showing the sum rate performance comparison for each model as the number of users in the area increases under various environments. The number of APs and the number of installed antennas used in the simulation in FIG. 6 were set the same as in FIG. 5. In the simulation of Figure 6, a total of three different environments were applied, and their descriptions are as follows.

A. Incomplete channel information & sufficient orthogonal pilots exist for the number of users

B. Incomplete channel information & 60 orthogonal pilots exist

C. Perfect channel information & sufficient orthogonal pilots for the number of users

As a result of the simulation, it was confirmed that if the proposed method is applied, the number of unsupported users can be drastically reduced and performance similar to that of not applying a separate user support method can be achieved as long as the number of users does not increase beyond a certain level.

Figure 7 shows the CDF (Cumulated Distribution Function) simulation results for per-user throughput according to the present invention for each model. As can be seen in Figure 7, applying the proposed user support method can significantly reduce the number of unsupported users in the area compared to the existing user-centric cell-free massive MIMO model, and does not significantly damage the throughput of users.

The following Figure 8 shows simulation results of the final number of unsupported users as the number of users increases. Even if the method presented in FIG. 8 is applied, it can be seen that if the number of users increases beyond a certain level, unsupported users will exist, which is set at 204 in FIG. 4. This is because the range is not set to the entire range, and at the same time, it guarantees the connection of existing users at (205). This is because as the number of users increases, the number of users supported only by a single AP within the area also increases. This is because the number of elements also decreases. In addition, in the model used in the present invention, the number of users that can be supported for each AP is also calculated by considering realistic calculation amount and complexity. Setting slightly less also affects the results in Figure 8. However, it was confirmed that applying the proposed method can significantly reduce the number of unsupported users in the area compared to the case where a separate user support method is not applied.

100: AP (Access Point)
110: CPU (Central Processing Unit)
120: Fronthaul
130: Optical cable
140: User receiving support
150: Unsupported user
200: Initial value for existing connections in the model
201: Start of checking whether an unsupported user exists
202: Part where the process of checking whether unsupported users exist is carried out for all users
203: Process of checking whether support exists for a specific user
204: STEP 1. Candidate AP set P selection process
205: User confirmation process supported by a single AP
206: Process of ensuring connection of existing users
207: User set S selection process connected to set P
208: Information extraction process for set S
209: Prioritization process for elements in set S
210: AP extraction process according to the priority given to set S
211: threshold The process of comparing the size and size with
212: Sequential connection and cumulative sum for APs extracted from 210 and unsupported users
Calculation progress
213: Decision process for further exploration
214: Another unsupported user search process
215: Entire course ends

Claims (8)

As a method of allocating multiple APs to users in the downlink user-centric cell-free massive MIMO model,
An AP candidate set selection step of selecting a set P , which is a subset of candidate APs for supporting unsupported users, as an AP candidate set;
A user connection AP confirmation step of checking whether there is a user who receives service only from APs in the selected AP candidate group P ;
A user subset setting step of defining existing users receiving service from APs in the selected AP candidate set P as a user subset S ;
An AP candidate group channel gain extraction step of extracting channel gains between each subset of users included in the user subset S and APs supporting these subset users;
Based on the extracted channel gain, a priority assignment step of assigning priority to the subset users included in the user subset S in the order of the largest number of supported APs;
A connection AP selection step of extracting APs supporting a throughput lower than the median among the APs connected to the user, starting from the highest ranking user, as connection candidate APs according to the assigned priorities of the users;
A non-supported user AP connection step of connecting the extracted connection candidate APs to non-supported users in order;
AP user allocation method comprising: According to clause 1,
The AP candidate set P selection step is,
An AP user allocation method characterized by selecting APs existing within a predetermined distance as a set of AP candidates based on the physical distance between unsupported users and APs. According to claim 1 or 2,
After the user connection AP confirmation step,
If there is a user supported only by a single AP in the set P , the connection AP candidate group removal step of preventing disconnection of the existing user by excluding the AP from the set P :
An AP user allocation method further comprising: According to clause 1,
The AP candidate channel gain extraction step is,
An AP user allocation method characterized by extracting a channel gain based on the throughput supported by each user included in the user subset S from the AP. According to clause 1,
The step of assigning priority to the users included in the user subset S in the order of the highest number of supported APs,
Prioritizing users included in the subset S and APs supporting these users in order of highest channel gain;
AP user allocation method characterized by: According to clause 1,
The unsupported user AP connection step is,
In order to not damage the throughput of existing users as much as possible, disconnect from the user with the largest number of connected APs and connect to unsupported users.
AP user allocation method characterized by: According to clause 1,
The unsupported user AP connection step is,
Connect the extracted APs to non-supported users in order and calculate the cumulative sum of the throughput of the non-supported users,
The calculated cumulative sum is set to a predetermined threshold value. Compared to , the threshold value before connecting all extracted APs If it exceeds the threshold, no further connections are made to unsupported users, and even if all extracted APs are connected, the threshold If it is not possible to exceed , proceed to search for the next unsupported user without any further connection.
AP user allocation method characterized by: In clause 7,
above threshold Is,
It is the median throughput for each user supported within the existing network.
AP user allocation method characterized by:

Images