A Downlink Multi-Point Cooperative Scheduling Method for a Fully Decoupled Network

Abstract

A downlink multi-point cooperative scheduling method for a fully decoupled network. 1) The user estimates the channel information value based on the pilot sequence in the downlink frame, and then returns the locally estimated channel information to the edge baseband signal processing center of the base station. 2) According to the measured channel information, considering that the power allocation between the base station and the user is evenly distributed, the matching relationship between the user and the base station is modeled as a many-to-many graph matching problem, that is, modeled as a many-to-many matching model with the maximization of the total capacity as the criterion, and a many-to-many initialization and exchange matching algorithm is designed. 3) According to the many-to-many matching scheme between the user and the base station and the received channel information, considering the interference of coherent transmission between users, an optimization problem with the sum of user weighted rates as the objective is established. The fractional programming is used to transform the cooperative scheduling problem into a non-convex problem with multiple variables, and the block coordinate descent method is used to obtain the beamforming matrix of the base station for the user.

Description

Technical Field

[0001] The present invention belongs to the new generation of communication technologies, relates to the field of access network cooperative beamforming, and designs a matching and cooperative beamforming method for multiple base stations and multiple users. Background Art

[0002] With the development of access network technologies, the cellular access network has evolved from the distributed 4G network, where each base station is equipped with an independent baseband processing unit and transmitting antenna, to the centralized base station deployment of the 5G centralized processing unit (CU) and distributed processing unit (DU). The centralized base station deployment has many advantages: cost reduction. As the frequency increases, base station deployment will become more and more intensive. By deploying a centralized baseband processing center and connecting signal processing units through high-speed interfaces, there is an advantage in cost reduction; enhancing cooperation ability. The interaction between 4G base stations is distributed, and the stations communicate through the standard X2 interface. Since cooperation between base stations requires real-time signal processing or signaling interaction, the rate of the distributed interface represented by X2 obviously cannot support it. The centralized network architecture can perform centralized data processing through a centralized signal processing unit and rely on a low-latency high-speed data transmission interface between stations for fast signaling interaction and data transfer.

[0003] However, with the introduction of the goals of "carbon peak and carbon neutrality", cost and network energy consumption issues must be considered when designing 6G. The next-generation access network needs to solve not only the access services for mobile users, but also ensure the access of large-scale machine communication. If the traditional intensive base station deployment method is used, even with a centralized architecture, the cost of base station deployment is unacceptable. The main costs at the base station transmitter are the duplexer and the operational amplifier module. Considering the scenario where large-scale machine communication and mobile users coexist, the demand for uplink receiving nodes is much greater than that for downlink nodes. Therefore, Quan Yu et al. proposed a base station deployment method with independent uplink and downlink in the article "A Fully-Decoupled RAN Architecture for 6G Inspired by Neurotransmission" published in the Journal of Communications and Information Networks in 2019. The downlink transmitter and the uplink receiver are completely physically independent. Multiple uplink and downlink transmitters are deployed in one scenario, and there is a macro base station with wide coverage for centralized control. The edge baseband signal processing center is responsible for processing the baseband signals and signaling transmission services between the uplink base stations, downlink base stations, and control base stations.

[0004] After searching the existing literature, it is found that the acquisition of channel state information (CSI) is crucial in downlink beamforming. Since the uplink receiving antennas and downlink transmitting antennas are physically completely independent, methods for estimating downlink channel information based on uplink channel information, such as the channel estimation mode of time-division duplex (TDD), cannot be directly applied to fully decoupled downlink beamforming. Asmaa Abdallah et al. proposed a channel estimation algorithm for frequency-division duplex (FDD) cell-free massive MIMO systems in "Efficient Angle-Domain Processing for FDD-Based Cell-Free Massive MIMO Systems" published in 《IEEE Transactions on Communications》, which uses the angle of arrival of uplink signals to estimate the angle of departure of signals. However, this method still cannot be directly applied. Therefore, we send traditional downlink pilot signals, the user terminal estimates the channel, and transmits the estimated channel to the edge baseband signal processing center in an analog or digital manner.

[0005] In downlink multi-node multi-user coherent transmission, signal interference among multiple users is a very serious problem, which depends on the connection relationship between users and the base station and the power allocation scheme of the base station. Considering the connection guarantee characteristics of user services, the traditional single-connection scheme between users and the base station is not as good as the multi-connection scheme in terms of user service guarantee. However, this scheme requires obtaining the optimal matching relationship between users and the base station, and the index of the quality of the user-base station matching depends on the size of the network throughput, which is a very complex non-convex objective, an optimization problem with mixed integer constraints. Therefore, new user-base station matching algorithms and base station power allocation schemes need to be designed.

[0006] In summary, the present invention solves the problems existing in the prior art: (1) Traditional channel estimation algorithms based on channel reciprocity in TDD or angle reciprocity in academic FDD cannot be directly applied to channel estimation in fully decoupled networks. (2) Different from the single connection between users and the base station, the matching problem of multi-users and multi-base stations is a complex many-to-many matching problem, and the performance of algorithms based on traditional downlink signal received power (RSRP) cannot be guaranteed. (3) The matching and power allocation between users and the base station are subject to mixed integer constraints, and the optimization objective is the weighted value of the user achievable rate. The signal-to-interference-plus-noise ratio of the objective constraint is a non-convex expression. Therefore, this problem is a complex mixed integer programming problem, and a power allocation scheme considering user-base station matching and inter-station interference needs to be redesigned. Compared with the traditional access algorithm based on downlink reference signal received power, the flexible multi-point cooperation and efficient beamforming algorithm of the present invention can significantly improve the downlink throughput of the access network and enhance the network performance. Summary of the Invention

[0007] Aiming at the problems existing in the prior art, the object of the present invention is to propose a joint multi-connection and beamforming scheme based on downlink multi-point cooperation in a fully decoupled network.

[0008] The present invention is implemented as follows. A downlink multi-point cooperation scheduling method for a fully decoupled network: The multi-connection user collaborative beamforming method includes the following steps:

[0009] Step 1: The user estimates the channel information value based on the pilot sequence in the downlink frame, and then returns the locally estimated channel information to the edge baseband signal processing center of the base station.

[0010] Step 2: According to the measured channel information, considering that the power distribution between the base station and the user is evenly distributed, model the matching relationship between the user and the base station as a many-to-many graph matching problem, that is, model it as a many-to-many matching model with the maximization of the total capacity as the criterion, and design a many-to-many initialization and exchange matching algorithm.

[0011] Step 3: According to the many-to-many matching scheme between the user and the base station and the received channel information, considering the interference of coherent transmission between users, establish an optimization problem with the sum of user weighted rates as the objective, transform the problem of cooperative scheduling into a non-convex problem with multiple variables using fractional programming, and use the block coordinate descent method to obtain the beamforming matrix of the base station for the user.

[0012] In Step 1, the user local channel estimation uses the minimum mean square error estimation scheme (MMSE), and the upload of channel information considers two schemes: digital upload and analog upload.

[0013] Considering the instability of the many-to-many matching between the user and the base station, in the matching stage, the base station adopts an evenly distributed power allocation scheme for the user, and conjugate beamforming is used for beamforming.

[0014] In Step 2, considering that the connection relationship between the user and the base station affects each other, use the swap-matching scheme to restrict the initialization of the matching scheme between the user and the base station. After initialization, the users will exchange the matched base stations. If the capacity is improved, the exchange is successful. If the system throughput, that is, the capacity, is not improved, the original matching scheme is maintained.

[0015] During the initialization process of the multi - to - multi matching - based solution, it is necessary to first establish a single connection relationship between users and base stations based on the received signal strength of the downlink (RSRP). During the process of establishing multi - to - multi connections, every time a user needs to establish a new connection, it is necessary to determine whether establishing the new connection will bring an increase in the system throughput. If it can bring an increase in throughput, then this solution is maintained; if it cannot bring an increase in throughput, then the number of connections of this user remains unchanged.

[0016] After the edge baseband processing center obtains the channel information of users, it is necessary to perform multi - point cooperative beamforming on users. The fractional programming is used to convex - relax the non - convex objective function, and the method of block - coordinate descent is used to optimize the variables in the problem step by step:

[0017] The optimal solution of the beamforming vector is obtained through a multi - variable cyclic iteration method.

[0018] During the channel estimation process, the pilot sequence needs to be inserted into the downlink frame structure. After the user performs channel estimation based on the received pilot reception strength, the channel needs to be transmitted back to the edge baseband signal processing center.

[0019] The described channel estimation algorithm is different from the existing time - division duplex channel estimation method based on channel reciprocity. The pilot sequence needs to be inserted into the downlink frame structure. After the user performs channel estimation based on the received pilot reception strength, the channel needs to be transmitted back to the edge baseband signal processing center.

[0020] Furthermore, considering the multi - to - multi connection matching relationship between users and base stations, one base station may serve multiple users simultaneously, and one user may be served by multiple base stations simultaneously.

[0021] Furthermore, considering the maximization of spectral efficiency, all users use the same frequency band. Multiple base stations transmit information to multiple users within this frequency band, and when multiple base stations transmit information to a certain user, they are coherent. Considering the scarcity of spectrum resources, it is set that when multiple base stations serve multiple users, they all use the same sub - channel.

[0022] Furthermore, the goal of step 3 is to optimize the weighted value of the throughput of all users in the network. There is mutual interference between base stations and users, which causes both the numerator and denominator of the signal - to - interference - plus - noise ratio in the objective function to contain variables. Therefore, new variables are introduced to express the fraction as a linear quadratic relaxation result.

[0023] Considering the multi - connection relationship between users and base stations, the traditional one - to - one matching and multi - to - multi static matching are no longer applicable. The choices of users interfere with each other. Therefore, a multi - to - multi matching algorithm based on exchange matching is proposed.

[0024] Considering that the uplink transmitting antenna and the receiving antenna are completely independent, and the user's uplink and downlink are connected to different base stations, in order to ensure the continuity of user services, it is considered that there is more than one connection relationship established between the user and the base station.

[0025] There is a high-speed signal path between the base station and the baseband signal processing center, and one baseband signal processing center can be connected to multiple downlink base stations, which is convenient for flexible cooperation and scheduling.

[0026] When the user performs channel estimation locally, there is an error, and there is also an upload error when the estimated channel information is uploaded to the edge baseband signal processing center through the uplink. The digital-analog upload method is considered.

[0027] Furthermore, since the multi-user beamforming scheme of the base station introduces new variables, we use the block coordinate descent scheme to optimize the beamforming.

[0028] Compared with the prior art, the beneficial effects of the present invention are as follows. First, the downlink multi-point cooperation scheduling method for the fully decoupled network can effectively improve the throughput of network users. Second, considering the independent positions of the uplink and downlink, we design a new channel estimation and upload scheme. Third, in the face of user and service guarantee problems, we propose a user-base station multi-connection scheme, and propose a connection scheme based on many-to-many matching. Considering the mutual interference characteristics of user connections, we propose a swap matching algorithm, which can achieve a stable solution of the matching. Finally, in the face of the serious interference problem brought about during the coherent transmission process, we propose a block coordinate descent beamforming scheme, which can significantly improve the throughput of the network. Description of the Drawings

[0029] Figure 1 It is a schematic diagram of downlink multi-point cooperation of the fully decoupled network described in the present invention.

[0030] Figure 2 It is a schematic diagram of multi-user-multi-base station swap matching in an embodiment of the present invention.

[0031] Figure 3 It is a schematic diagram of multi-user beamforming based on block descent in an embodiment of the present invention. Detailed Embodiment

[0032] In order to make the purpose, technical solution and advantages of the present invention clearer, the following describes the embodiments of the present invention in detail with reference to the accompanying drawings: This embodiment is implemented on the premise of the technical solution of the present invention, and gives detailed implementation manners and specific operation processes. It should be understood that the specific examples described here are only used to explain the present invention, but the protection scope of the present invention is not limited to the following embodiments.

[0033] This embodiment considers a scenario of downlink coordinated beamforming, where all access points (APs) can serve all users within their coverage areas. We also consider a fully decoupled network architecture (FD-RAN), in which APs are connected to an edge cloud equipped with a baseband signal processing unit (BBU) pool through high-speed front-ends. Thus, signal processing can be performed on the edge cloud, and signals can be easily exchanged between APs and the cloud. This network setup enables distributed APs to cooperate simultaneously to serve all terminals within the network coverage area. Assume that there are APs in the network, and each AP is equipped with N linearly aligned antennas. We also assume that there are users with single antennas in the network. In the proposed transmission mode, for each user we define a service set denoting the set of base stations that serve this user.

[0034] For each we define the set E m denoting the set of users served by this base station. This set can also be obtained through the expression and as shown in Figure 1 , the coverage areas of APs overlap, a user can connect to multiple base stations for data transmission, and a base station can connect to multiple users. For a single narrowband transmission, such as a single subcarrier in multi-carrier transmission, a terminal may be interfered with by the serving AP and other APs. We also consider a classical block fading model, in which the channel is fixed within a finite-sized time-frequency coherence interval. The received signal of a user can be modeled as

[0035]

[0036] where x k represents the transmitted signal from base station m to user k and the power of the signal is 1; represents the beamforming vector from AP m to user k; represents the channel state information from AP m to user k, and represents the small-scale fading information, ψ m,k and l(d m,k ) represent shadowing and path loss, respectively; represents the additive white Gaussian noise of the user.

[0037] We define the (MN×1)-dimensional complex vector as the set of channels . Similarly, we define w k as the set of beamforming vectors In addition, we also define a block diagonal matrix to represent the connection relationship between users and base stations, where s m,k = 1 represents 0 or 1, that is, whether the user has established a connection with the base station. For the channel generation model, we adopt the 38.901 protocol of the 3GPP standard to generate the channel model between users and base stations.

[0038] Due to the different positions of the transceiver antennas, the channel reciprocity and the channel information acquisition method based on angular reciprocity cannot be realized. Therefore, we consider a typical pilot-based system, where the channel pilot and feedback procedures are described as follows.

[0039] Consider the block fading channel model and orthogonal time-frequency pilot sequences without pilot contamination.

[0040] Channel training part: Base station m sends a pilot sequence of length τ c ≥ 1 on the downlink. And user k estimates the result of this channel as s m,k , where

[0041]

[0042] where Given H m,k and s m,k Minimum mean square error result estimation

[0043]

[0044] where the estimation error is

[0045]

[0046] Channel feedback: Here we consider the analog feedback of the channel. In the analog feedback, each user estimates the channel information when receiving the local signal sent by the base station and then feeds back τ u to represent the channel information, where binary orthogonal amplitude modulation is used. The estimated result and variance are expressed as

[0047]

[0048] where the estimation error is:[[]]

[0049]

[0050] According to the estimated channel information value, we can model the following problem P1, in which our goal is to maximize the throughput of the network, and the constraints include the association constraints between base stations and users and the beamforming constraints of base stations.

[0051]

[0052] P1 is a mixed - integer non - linear optimization problem, where the objective function is also non - convex, indicating that the signal - to - interference - plus - noise ratio in the objective function is non - convex.

[0053] To solve the association problem between base stations and users, we first define the association between base stations and users as a many - to - many matching problem. Each base station and user is selfish, and the evaluation metric for the matching is the user throughput in its own connection relationship. We regard the base stations and users as one side of the matching respectively, and the matching is divided into the following two steps: First, the users stack to initialize their own many - to - many matching results; then the users try to exchange the base station information in their associated base station sets with each other. If the total capacity increases, this exchange is maintained; if the total capacity does not increase, the exchange is not made. The specific algorithm flow of the exchange matching is given in Figure 2 which is given below.

[0054] In power allocation, the complex interference problem among users needs to be considered. First, we perform Lagrangian transformation on the objective function

[0055]

[0056] where γ = γ 1 ,…,γ K represents the SINR vector of users; δ k represents the weighted coefficient of users; represents the beamforming matrix of users; represents the matrix after diagonalization of the user connection relationship; v represents the vector of Lagrangian relaxation. When the beamforming vector and user connection relationship are fixed, the Lagrange multiplier v k can be obtained by differentiating γ k and is expressed as

[0057]

[0058] where, v k represents the Lagrangian relaxation variable of the k - th user. After the transformation, the Lagrangian function is still a non - convex function. Therefore, we use second - order relaxation to relax the objective function to

[0059]

[0060] where the vector β = [β 1 ,…,β K is the introduced relaxation variable. After the transformation, the objective function becomes a convex function. We obtain by differentiating the relaxation variable

[0061]

[0062] To further simplify the problem, we incorporate the power constraint into the Lagrangian function, which is further expressed as

[0063]

[0064] where μ is the introduced Lagrange multiplier for power. Thus, our problem is transformed into an unconstrained optimization problem, and the optimal beamforming vector can be obtained by taking the derivative to get

[0065]

[0066] The optimal solution of the beamforming vector can be obtained through a multi-variable cyclic iteration method. The execution flow chart of the algorithm can be found in Figure 3 as follows.

[0067] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A downlink multi-point cooperative scheduling method for a fully decoupled network, characterized in that: it includes the following steps: Step 1: The user estimates the channel information value based on the pilot sequence in the downlink frame, and then returns the locally estimated channel information to the edge baseband signal processing center of the base station; Step 2: According to the measured channel information, considering that the power allocation between the base station and the user is evenly distributed, the matching relationship between the user and the base station is modeled as a many-to-many graph matching problem, that is, modeled as a many-to-many matching model with maximizing the total capacity as the criterion, and design a many-to-many initialization and swap matching algorithm; Step 3: According to the many-to-many matching scheme between the user and the base station and the received channel information, considering the interference of coherent transmission between users, establish an optimization problem with the sum of user weighted rates as the goal, use fractional programming to transform the cooperative scheduling problem into a non-convex problem with multiple variables, and use the block coordinate descent method to obtain the beamforming matrix of the base station for the user; In Step 1, the user local channel estimation uses the minimum mean square error estimation scheme (MMSE), and the feedback of the channel information considers two schemes: digital feedback and analog feedback; Considering the instability of the many-to-many matching between the user and the base station, in the matching stage, the power allocation of the base station to the user adopts an evenly distributed scheme, and the beamforming uses conjugate beamforming; In Step 2, considering that the connection relationship between the user and the base station affects each other, use the swap-matching scheme to restrict the initialization of the matching scheme between the user and the base station. After initialization, the users will swap the matched base stations. If the capacity is improved, the swap is successful. If the system throughput, that is, the capacity, is not improved, the original matching scheme is maintained; In the initialization process of the many-to-many matching-based scheme, it is necessary to first establish a single connection relationship between the user and the base station based on the received signal strength of the reference signal received power (RSRP). During the establishment of multiple connections, every time a user needs to establish a new connection, it is necessary to judge whether establishing a new connection will bring an increase in the system throughput. If it can bring an increase in throughput, then this scheme is maintained. If it cannot bring an increase in throughput, then the user maintains the original number of connections unchanged; After the edge baseband signal processing center obtains the channel information of the user, it needs to perform multi-point cooperative beamforming on the user. Fractional programming is used to convexify the non-convex objective function, and the block coordinate descent method is used to optimize the variables in the problem step by step: Specifically as follows: In power allocation, it is necessary to consider the complex interference problem between users. First, perform Lagrangian transformation on the objective function where γ = γ 1 , …, γ K represents the SINR vector of the user; δ k represents the weighted coefficient of the user; represents the beamforming matrix of the user; S represents the matrix after diagonalization of the user connection relationship; v represents the vector of Lagrangian relaxation; when the beamforming vector and the user connection relationship are fixed, the Lagrange multiplier v k is obtained by differentiating γ k and is expressed as The Lagrangian function after transformation is still a non-convex function. Therefore, second-order relaxation is used to relax the objective function into where the vector β = [β 1, …, β K is the introduced slack variable. After transformation, the objective function becomes a convex function, and by taking the derivative with respect to the slack variable, we obtain After simplification, the power constraint is penalized into the Lagrangian function, expressed as where μ is the introduced Lagrange multiplier with respect to power; then the problem is transformed into an unconstrained optimization problem, and the optimal beamforming vector is obtained by taking the derivative obtained The optimal solution of the beamforming vector is obtained through a multi-variable cyclic iteration method.

2. A downlink multi-point cooperative scheduling method for a fully decoupled network according to claim 1, characterized in that: During the channel estimation process, the pilot sequence needs to be inserted into the downlink frame structure. After the user performs channel estimation based on the received pilot reception strength, the channel needs to be transmitted back to the edge baseband signal processing center.

3. A downlink multi-point cooperative scheduling method for a fully decoupled network as described in claim 1, characterized in that: The user's uplink and downlink are connected to different base stations.

4. A downlink multi-point cooperative scheduling method for a fully decoupled network as described in claim 1, characterized in that: There is a high-speed signal path between the base station and the baseband signal processing center, and one baseband signal processing center is connected to multiple downlink base stations.

5. A downlink multi-point cooperative scheduling method for a fully decoupled network as described in claim 1, characterized in that: Considering the maximization of spectral efficiency, all users use the same frequency band. Multiple base stations transmit information to multiple users within this frequency band, and when multiple base stations transmit information to a certain user, they are coherent.

Images