Dynamic clustering cellular-removal networking method based on evolution spectral clustering

Abstract

The invention provides a dynamic clustering cellular-removal networking method based on evolution spectral clustering, which comprises the following steps of: firstly, considering a downlink scene of a cellular-removal network and the mobility of a user, and constructing a random road point moving model of the user and a system model of a clustering cellular-removal network; secondly, modeling the problem as a multi-target time-varying smooth clustering cellular-removal networking problem, deducing a system and a rate, and proposing a smoothness index to measure the performance of a clustering cellular-removal networking scheme; related knowledge of the graph theory is introduced again, and the problem is converted into a multi-target time-varying graph segmentation problem; and finally, based on the idea of evolutionary spectral clustering, a time-varying smooth clustering cellular-removal networking algorithm is provided, a networking scheme of smooth change at adjacent moments is obtained, the switching frequency is reduced on the basis of ensuring a certain system and rate, and the cellular-removal networking performance is improved.

Description

Technical Field

[0001] The present invention belongs to the field of mobile communication cellular networking methods, and specifically relates to a decellularized architecture and a dynamic clustering decellularized networking method based on evolutionary spectrum clustering. Background Art

[0002] In recent years, mobile wireless communication technology has made amazing progress, leading the wave of the digital age. From 1G analog communication systems to 5G digital communication systems, mobile communication technology has not only achieved a huge improvement in speed and bandwidth, but also brought unprecedented innovation and convenience. Wireless communication technology has promoted the development of the Internet of Things and provided support for innovations in smart cities, autonomous driving, telemedicine and other fields. At the same time, 5G and 6G technologies will further support emerging technologies and applications such as virtual reality, augmented reality and artificial intelligence, bringing unlimited possibilities for the future.

[0003] The current mobile network adopts a cellular network architecture, which divides the service area into multiple cellular cells. Each cell is covered by a base station to provide a wide coverage range, so that mobile device users can obtain stable communication services in different locations. However, due to the rapid attenuation of the received signal power with the propagation distance, users at the edge of the cell will face a worse signal-to-noise ratio than users at the center of the cell. In order to meet the growing demand for data traffic, the density of base stations will need to be increased in the future to achieve seamless coverage. However, this will result in more users at the edge of the cell, which will be interfered by adjacent cells, thereby reducing the data rate experienced by the users. Therefore, the main goal of future mobile networks is to ensure that users at any location within the coverage area can obtain uniform data rates, which requires the design of a new network architecture to meet the performance requirements of future mobile communication networks.

[0004] In 2010, Marzetta et al. proposed the massive multiple-input multiple-output (MIMO) technology. So far, this technology has become one of the key technologies of 5G and has been tested and deployed in 5G networks. By equipping the base station with a large number of antennas, very high beamforming and spatial multiplexing gains are obtained, and the spectrum and energy efficiency are improved by several orders of magnitude using relatively simple linear processing methods. Despite the configuration of advanced massive MIMO, large data rate differences within the cell cannot be avoided under the cellular architecture. Therefore, Ngo, Ashikhmin, Marzetta et al. proposed the concept of decellularized massive MIMO based on distributed antennas, network MIMO and cloud radio access network (Cloud Radio Access Network, C-RAN). In this network architecture, there are a large number of access points (APs) distributed, each of which is equipped with a large number of antennas, which jointly serve all users in the area on the same time-frequency resources. Each AP is connected to a central processing unit (CPU) through a fronthaul link, and the CPU is responsible for the collaboration between APs. Since the traditional cellular cell division is removed, the surrounding APs can provide services to users at the same time. Therefore, the de-cellularized massive MIMO architecture avoids interference between cells, suppresses problems such as frequent switching and poor edge coverage, and provides users with balanced, stable and reliable services.

[0005] However, in actual scenarios, one CPU manages all APs and processes all data while providing services to all users, which is not scalable for large networks. As the number of users increases, this centralized management method leads to huge signal processing and signaling overhead, which greatly limits the scalability and performance of the network. In order to meet the growing demand and improve the capacity and coverage of the network, researchers have begun to explore new networking mechanisms. User-centric decellularization is a widely studied networking mechanism that takes users as the core, forms virtual cells based on the user's location and needs, and provides services to users by a group of APs with the strongest signals. Compared with traditional cellular networks, decellularization can achieve higher spectrum efficiency, reduce backhaul link overhead, and provide better user experience. In addition, clustered decellularization is another novel networking mechanism that divides the entire network into multiple subnets, divides users with large mutual interference into the same subnet, and uses interference cancellation technology to reduce interference. This networking mechanism can significantly reduce interference and improve the spectrum efficiency of the entire system.

[0006] Although decellularization brings immeasurable potential to improve the performance of next-generation wireless communication systems, it also brings many technical difficulties and challenges. Current research mainly focuses on networking solutions in static scenarios, without considering user mobility and channel time-varying properties. User mobility and changes in channel conditions will lead to dynamic changes in the optimal networking solution, so the networking results need to be dynamically updated. However, frequently and independently solving the optimal networking solution at each moment will consume a lot of time and computing resources, limiting the feasibility of practical applications. In addition, there may be huge differences between different networking solutions, and users frequently switching the best AP group will increase signaling overhead and switching delay, reducing the user's service experience. Summary of the invention

[0007] In view of the problems existing in the prior art, the present invention focuses on more realistic scenarios, considers the mobility of users, and studies the dynamic de-cellular networking mechanism with time-varying smoothness in both spatial and temporal dimensions to solve the existing problems and improve the performance of mobile communication networks. On the basis of clustered de-cellular networking, the user's mobility model is introduced. At the same time, a smoothness constraint index is proposed to measure the change size of the networking at adjacent moments, and the dynamic clustered de-cellular networking problem under the user mobility scenario is modeled based on this. By converting the multi-objective optimization problem into a time-varying graph segmentation problem, a dynamic clustered de-cellular networking algorithm based on evolutionary spectrum clustering is further proposed, so that the networking results at adjacent moments can be smoothly transitioned to balance the total rate and switching cost, improve the performance of the de-cellular system, and reduce the difference and switching frequency of networking schemes at different moments. The present invention has important theoretical significance and practical application value, and can be used to improve the network conditions in hot spots with fast population mobility, and to manage vehicles that need to continuously switch service base stations in intelligent transportation systems, providing them with more stable and reliable data transmission.

[0008] Technical solution of the present invention

[0009] A dynamic clustering and cellular networking method based on evolutionary spectrum clustering comprises the following steps:

[0010] Step 1. Build a decellularized network system model, user mobility model, and smoothness index

[0011] Step 1.1, construct a decellularized network system model;

[0012] Step 1.2: Build a user mobility model;

[0013] Step 1.3: Construct a smoothness index.

[0014] Step 2: Modeling of optimization problems and design of dynamic clustering and decellularization networking scheme based on evolutionary spectrum clustering

[0015] Step 2.1, optimization problem modeling;

[0016] Step 2.2, design a dynamic clustering and decellularization networking solution based on evolutionary spectrum clustering;

[0017] Step 2.3: Solve the optimization problem.

[0018] Beneficial Effects

[0019] The present invention aims at the scenario where users frequently move in the cellular network, and realizes multiple

[0020] Beneficial effects:

[0021] First, reducing the number of switching times helps reduce the risk of communication interruption and connection loss, improves call quality and data transmission stability, and is particularly suitable for scenarios where users move frequently.

[0022] Secondly, the user rate can be guaranteed, and the user will not feel obvious network interruption or quality degradation during the movement, thereby enhancing the continuity and stability of the user experience and improving user satisfaction.

[0023] Finally, the method based on evolutionary spectrum clustering brings a new idea for networking in de-cellular scenarios. Through the relevant knowledge of graph theory and evolutionary clustering, it can effectively solve the cluster networking problem under user mobility, balance the total rate and switching cost, and achieve a smooth transition of networking results at adjacent times.

[0024] These advantages not only improve the performance of cellular networks, but also directly affect the user experience and network quality, bringing positive impacts to the development of the mobile communications field. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a schematic diagram of the processing process of the method of the present invention;

[0026] Figure 2 This is a schematic diagram of a de-cellular downlink scenario when a user moves according to an embodiment of the present invention;

[0027] Figure 3 Snapshot comparison diagram of the networking results of the embodiment of the present invention ((a) networking result at time t-1 (b) networking result of the method proposed by the present invention at time t (c) networking result of the baseline algorithm at time t);

[0028] Figure 4 This is a relationship diagram between different numbers of users and average system and rate in an embodiment of the present invention;

[0029] Figure 5 This is a relationship diagram between different numbers of users and average switching times according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] The technical solution provided by the present application will be further described below in conjunction with specific embodiments and accompanying drawings. The advantages and features of the present application will become more apparent with the following description.

[0031] The present invention selects a downlink scenario based on a decellularized network to perform a smooth transition of the networking solution, establishes a user mobility model and a system model of a clustered decellularized network, proposes a smoothness index to measure the smooth change of the network, and models the time-varying smooth clustered decellularized networking problem. Further citing relevant theories in graph theory, the problem is converted into a multi-objective time-varying graph segmentation problem, and based on the architecture of evolutionary spectrum clustering, a time-varying smooth dynamic clustered decellularized networking algorithm is designed to solve it. The technical solution of the present invention achieves smooth changes in networking at adjacent moments through a dynamic clustered decellularized networking method based on evolutionary spectrum clustering, while reducing the number of switching times and ensuring the sum rate of the entire system.

[0032] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0033] Example

[0034] A dynamic clustering and cellular networking method based on evolutionary spectrum clustering, comprising the following steps (such as Figure 1 ):

[0035] Step 1. Build a decellularized network system model, user mobility model, and smoothness index

[0036] Step 1.1: Build a decellularized network system model:

[0037] Consider the scenario of decellularized downlink under user mobility (e.g. Figure 2 As shown), there are K users and L APs.

[0038] The AP set is User set is

[0039] All APs and users are equipped with single antennas and are randomly distributed throughout the area. Considering the mobility of users, at time t, the entire decellularized network is divided into M non-overlapping subnets, denoted as The entire network is divided into

[0040]

[0041] From AP b l To user u k The channel gain coefficient is h k,k , which can be expressed as

[0042]

[0043] in, represents the large-scale fading coefficient, d k,l Is user u k and AP b l , α′ is the path loss coefficient. Represents the corresponding small-scale fading coefficient.

[0044] According to the above definition, we can deduce that user u k The approximate received signal-to-interference-plus-noise ratio (SINR) is

[0045]

[0046] Where P t represents the transmit power of each AP, σ 2 represents the noise power of additive white Gaussian noise, is with user u k The best AP with the highest channel gain.

[0047] By normalizing the bandwidth, we can get user u k Achievable and rate

[0048]

[0049] At time t, the network is divided into Under this condition, we can get the sum rate of the system as

[0050]

[0051] Step 1.2: Build a user mobility model:

[0052] Users are constantly moving throughout the area, and the random waypoint mobility model is used to model the user's movement.

[0053] Specifically, the random waypoint movement model: at each time t, each user randomly selects a direction and a moving distance that obeys a certain distribution, and then moves to the next waypoint at time t+1, and repeats the above process.

[0054] Step 1.3: Construct smoothness indicator:

[0055] Flexible networking means that the networking results are constantly adjusted as the user moves, which may lead to a large number of handovers. These handovers will cause additional transmission delays and signaling overhead, and ultimately reduce system performance. In order to solve the frequent handover problem, the networking results need to change smoothly with practice. Therefore, the present invention proposes to construct a time smoothness index to measure the results of clustered decellularized networking.

[0056] Specifically, for the dynamic clustering and cellular networking switching overhead after the introduction of the time dimension, the networking solution at time t is adopted at time t-1. Achievable systems and rates To characterize the switching overhead of the networking solution update, denoted as

[0057]

[0058] Based on the basic idea of ​​maximizing the system and rate, The larger the value is, the closer the networking solution at time t is to the networking solution at time t-1, the fewer the number of switching times is, and the smaller the switching overhead is.

[0059] Step 2: Modeling of optimization problems and design of dynamic clustering and decellularization networking scheme based on evolutionary spectrum clustering

[0060] Step 2.1: Optimization problem modeling:

[0061] Based on the above system and rate and smoothness indicators, a mathematical model for the time-varying smooth clustering and cellular networking problem is constructed.

[0062]

[0063] Through the following constraints, it is ensured that all subnets cover all users and APs in the system.

[0064]

[0065] In order to ensure that there is no overlap between each subnet, we define

[0066]

[0067] At the same time, in order to ensure that there is at least one AP in each subnet so that all users in the system can get service, the following constraints can be obtained

[0068]

[0069] The entire network is modeled as a weighted undirected time-varying graph, expressed as

[0070]

[0071] in, is the set of nodes at time t. Each node contains the user and its best AP represents the set of edges, W (t) is the weight matrix;

[0072] Model each edge The weight is

[0073]

[0074] is a weight matrix. Each subnetwork can be represented as a graph Therefore, the node set of the mth subnetwork can be expressed as

[0075] According to the previous research, it can be concluded that maximizing the system sum rate is equivalent to minimizing the packet Next in is the cut function, defined as

[0076]

[0077] Therefore, the above optimization problem can be transformed into a multi-objective time-varying graph segmentation problem, and the problem model is re-established as

[0078]

[0079] Accordingly, formulas (1) and (2) can be rewritten as

[0080]

[0081] The above multi-objectives are converted into a single objective function through the linear weighting method, which can be expressed as

[0082]

[0083] Among them, α∈[0,1] is a weight coefficient used to reflect the importance of the system and rate and smoothness.

[0084] Step 2.2: Design a dynamic clustering and decellularization networking solution based on evolutionary spectrum clustering:

[0085] In many practical applications of clustering, such as community discovery in dynamic social networks and tracking of moving objects, it is necessary to observe the objects to be clustered at multiple time points, with the goal of obtaining clustering results at each time step. One of the simplest methods is to perform clustering using only the most recent data at each time step. However, this method is very sensitive to noise, and the clustering results produced are unstable and inconsistent with the clustering results of adjacent time steps. Evolutionary clustering aims to add a temporal smoothness penalty to the static clustering method. This penalty can prevent the clustering results at any given time from deviating too much from the clustering results of adjacent time steps, thereby being robust to short-term changes. In addition, spectral clustering algorithms have been shown to be able to solve the graph segmentation problem well. Evolutionary spectral clustering combines evolutionary clustering with spectral clustering to solve the clustering problem of dynamic data.

[0086] Evolutionary spectral clustering defines a general cost function to measure the quality of clustering results on evolutionary data points. The cost function contains two costs. The first is the snapshot cost (CS), which is used to measure the clustering quality of the current clustering result relative to the current data features. The higher the snapshot cost, the worse the current clustering quality. The second is the temporal cost (CT), which measures the temporal smoothness based on the fit of the current clustering result to the historical data features or historical clustering results. The higher the temporal cost, the worse the temporal smoothness. Therefore, the total cost function is

[0087] Cost=α0·CS+β0·CT,

[0088] Among them, 0≤α0≤1, β0=1-α0, are user-defined parameters, reflecting the importance of snapshot cost and time cost. The specific implementation of evolutionary spectrum clustering is not limited to one method, and the main difference lies in the definition of time cost. In the present invention, the specific process is as follows:

[0089] Define an indicator matrix The element in row i and column m It is expressed as:

[0090]

[0091] Calculate each node Degree:

[0092]

[0093] Construct the degree matrix at time t:

[0094]

[0095] Computational Graph The Laplace matrix of

[0096] L (t) =D (t) -W (t)

[0097] According to the above definition, the sum cut function in formula (3) can be rewritten as

[0098]

[0099] By substituting formulas (5) and (6) into the function in formula (4), the objective function can be rewritten as

[0100]

[0101] At the same time, the constraints can be rewritten as

[0102]

[0103] Step 2.3: Solve the optimization problem and obtain the best networking result at time t

[0104] Solve for the Laplace matrix αL (t) +(1-α)L (t-1) The eigenvectors corresponding to the first M smallest eigenvalues ​​of are expressed as matrix Y (t) , denoted as

[0105]

[0106] The matrix Y (t) Apply the k-means clustering algorithm to each row of data in the matrix Y (t) Each row of data is assigned to a different cluster center to obtain the graph segmentation result at time t

[0107]

[0108] The best networking result at time t can be obtained according to the following formula

[0109]

[0110] consider Figure 2 In the decellular downlink scenario where the user is constantly moving, the method of the present invention is used to solve the optimal networking result, such as Figure 3 , Figure 4 , Figure 5 As shown in Figure 2, the baseline method used for comparison is based only on maximizing the total rate and independently networking at each moment. It is a special case of the algorithm we proposed, corresponding to α = 1, and provides a reference upper bound for the total rate of the system.

[0111] Figure 3The snapshot diagram of the networking results of the proposed method and the baseline method is shown. In the figure, the circles represent users, the triangles represent APs, and the nodes of the same color represent users and APs in the same subnet. Subgraph (a) is the networking result at time t-1; subgraph (b) is the networking result of the proposed method at time t; subgraph (c) is the networking result of the baseline algorithm at time t. The results show that as time goes by, the proposed method can produce a time-smoothed network partitioning result, greatly reducing the number of switches.

[0112] Figure 4 The relationship between the number of users and the average system sum rate under different weights α is demonstrated. The results show that under different numbers of users, the system sum rate achieved by the algorithm of the present invention is almost close to the sum rate of the benchmark algorithm and still maintains good performance.

[0113] Figure 5 The relationship between the number of users and the average number of switching times under different weights α is depicted. The results show that no matter what the value of α is, the algorithm of the present invention achieves fewer switching times than the benchmark algorithm. Choosing a suitable α value can greatly reduce the number of switching times and achieve a smoother networking result.

[0114] The above description is only a description of the preferred embodiments of the present application, and is not intended to limit the scope of the present application. Any changes or modifications made by any person skilled in the art based on the above disclosed technical contents shall be deemed as equivalent effective embodiments and shall fall within the scope of protection of the technical solution of the present application.

Claims

1. A dynamic clustering and decellularization networking method based on evolutionary spectrum clustering, characterized in that: The following steps are involved: Step 1. Build a decellularized network system model, user mobility model, and smoothness index Step 1.1, construct a decellularized network system model; Step 1.2: Build a user mobility model; Step 1.3: Construct a smoothness index. Step 2: Modeling of optimization problems and design of dynamic clustering and decellularization networking scheme based on evolutionary spectrum clustering; Step 2.1, optimization problem modeling; Step 2.2, design a dynamic clustering and decellularization networking solution based on evolutionary spectrum clustering; Step 2.3: Solve the optimization problem.

2. The method according to claim 1, characterized in that: Step 1.1: Build a decellularized network system model: Consider the decellularized downlink scenario under user mobility, where there are K users and L APs; The AP set is User set is All APs and users are equipped with single antennas and are randomly distributed throughout the area. Considering the mobility of users, at time t, the entire decellularized network is divided into M non-overlapping subnets, denoted as The entire network is divided into From APb l To user u k The channel gain coefficient is h k,l , expressed as in, represents the large-scale fading coefficient, d k,l Is user u k and AP b l The Euclidean distance between ′ is the path loss coefficient, represents the corresponding small-scale fading coefficient; According to the above definition, we can deduce that user u k The approximate received signal-interference-noise ratio SINR is Where P t represents the transmit power of each AP, σ 2 represents the noise power of additive white Gaussian noise, is with user u k The best AP with the highest channel gain; By normalizing the bandwidth, we can get user u k Achievable and rate At time t, the network is divided into The sum rate of the system is 3. The method according to claim 1, characterized in that: Step 1.2: Build a user mobility model: Users are constantly moving throughout the area, and the random waypoint mobility model is used to model the user's movement; Specifically, the random waypoint movement model: at each time t, each user randomly selects a direction and a moving distance that obeys a certain distribution, and then moves to the next waypoint at time t+1, and repeats the above process.

4. The method according to claim 1, characterized in that: Step 1.3: Construct smoothness indicator: Construct a time smoothness index to measure the results of clustering and de-cellular networking. Specifically, for the switching overhead of dynamic clustering and de-cellular networking after the introduction of the time dimension, the networking solution at time t is adopted at time t-1. Achievable systems and rates To characterize the switching overhead of the networking solution update, denoted as Based on the basic idea of ​​maximizing the system and rate, The larger the value is, the closer the networking solution at time t is to the networking solution at time t-1, the fewer the number of switching times is, and the smaller the switching overhead is.

5. The method according to claim 1, characterized in that: Step 2.1: Optimization problem modeling: A mathematical model for the time-varying smooth clustering and cellular networking problem based on system sum rate and smoothness indicators Through the following constraints, it is ensured that all subnets cover all users and APs in the system. In order to ensure that there is no overlap between each subnet, we define At the same time, in order to ensure that there is at least one AP in each subnet so that all users in the system can get service, the following constraints can be obtained The entire network is modeled as a weighted undirected time-varying graph, expressed as in, is the set of nodes at time t; each node contains the user and its best AP represents the set of edges, W (t) is the weight matrix; Model each edge The weight of is a weight matrix. Each subnetwork can be represented as a graph Therefore, the node set of the mth subnetwork can be expressed as Maximizing the system sum rate is equivalent to minimizing the packet The sum of in is the cut function, defined as Therefore, the above optimization problem can be transformed into a multi-objective time-varying graph segmentation problem, and the problem model is re-established as Accordingly, formulas (1) and (2) can be rewritten as The above multi-objectives are converted into a single objective function through the linear weighting method, which can be expressed as Among them, α∈[0,1] is a weight coefficient used to reflect the importance of the system and rate and smoothness.

6. The method according to claim 1, characterized in that: Step 2.2: Design a dynamic clustering and decellularization networking solution based on evolutionary spectrum clustering: Define an indicator matrix The element in row i and column m It is expressed as: Calculate each node Degree: Construct the degree matrix at time t: Computational Graph The Laplace matrix of L (t) =D (t) -W (t) According to the above definition, the sum cut function in formula (3) can be rewritten as By substituting formulas (5) and (6) into the function in formula (4), the objective function can be rewritten as At the same time, the constraints are rewritten as Tr(Z (t)T WITH (t) )=L。 7. The method according to claim 1, characterized in that: Step 2.3: Solve the optimization problem and obtain the best networking result at time t Solve for the Laplace matrix αL (t) +(1-α)L (t-1) The eigenvectors corresponding to the first M smallest eigenvalues ​​of are expressed as matrix Y (t) , denoted as The matrix Y (t) Apply the k-means clustering algorithm to each row of data in the matrix Y (t) Each row of data is assigned to a different cluster center to obtain the graph segmentation result at time t The best networking result at time t can be obtained according to the following formula

Images