Distributed optimization method for spatial traffic-aware wireless network integration relative energy efficiency
Patent Information
- Application Number
- CN202611073398.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-20
- Publication Date
- 2026-10-27
AI Technical Summary
[0003]本发明针对现有技术没有考虑用户的业务流量需求在时空间上的异构性、无法整合流量需求-网络容量不匹配的缺陷以及无法处理不完整用户业务流量需求数据的不足，提出一种面向空间流量感知无线网络集成相对能效的分布式优化方法，以全网可服务流量、容量与流量需求的空间匹配程度以及全网功耗共同构成优化指标，用于指导多基站无线网络中的波束赋形和资源配置，能够保留集成相对能效空间匹配特性的同时，构造可由每个基站本地计算的分布式优化准则并实现低信令开销的波束赋形协同优化
[0019]本发明采用集成相对能效作为无线网络资源配置指标，将全网可服务流量、容量与流量需求的空间匹配程度以及网络功耗统一纳入同一优化框架。针对集成相对能效中JS散度引起的全局耦合问题，通过构造按基站主导子区域可分的JS散度上界，将原本依赖全网容量分布和流量需求分布的空间失配度量转化为每个基站可计算的局部失配项，从而为分布式波束赋形更新提供局部优化依据。通过干扰有向图与标量一致性估计相结合的信息交换机制，每个基站仅与对其容量评估产生显著干扰影响的邻居交换相关波束赋形块，同时通过标量一致性估计获取全网总容量、全网总流量需求和全网总功耗等必要统计量。通过干扰导向增广拉格朗日方法(ID-ALM)，每个基站能够在满足本地功率约束和服务满足度约束的条件下进行分布式波束赋形优化。
Smart Images
Figure CN122994793A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a technology in the field of wireless communication, specifically a distributed optimization method for relative energy efficiency of spatial traffic-aware wireless network integration. Background Technology
[0002] Existing wireless network resource optimization methods struggle to simultaneously achieve spatial matching optimization of capacity supply and traffic demand, as well as low signaling overhead coordination, under conditions of heterogeneous spatial traffic demand, interference coupling among multiple base stations, and limited global information. Summary of the Invention
[0003] This invention addresses the shortcomings of existing technologies, such as failing to consider the spatiotemporal heterogeneity of user service traffic demands, the inability to integrate traffic demand-network capacity mismatch, and the inability to process incomplete user service traffic demand data. It proposes a distributed optimization method for the integrated relative energy efficiency of spatially traffic-aware wireless networks. The method uses the total available network traffic, the spatial matching degree between capacity and traffic demand, and the total network power consumption as optimization indicators to guide beamforming and resource allocation in multi-base station wireless networks. This method retains the spatial matching characteristics of integrated relative energy efficiency while constructing distributed optimization criteria that can be locally calculated by each base station and achieving low-signaling-overhead beamforming collaborative optimization.
[0004] This invention is achieved through the following technical solution:
[0005] This invention relates to a distributed optimization method for relative energy efficiency of spatial traffic-aware wireless network integration, comprising:
[0006] Step 1: Spatial mismatch modeling based on base station service sub-region division: Divide the entire network service area into multiple base station dominant sub-regions, and calculate the total capacity, total traffic demand, capacity weight, demand weight, normalized capacity distribution and normalized traffic demand distribution in each sub-region.
[0007] Step 2: Establish the interference directed graph and the information exchange graph respectively.
[0008] Step 3: Construct the original optimization problem based on the integrated relative energy efficiency index, that is, optimize the beamforming variables of the whole network to maximize the integrated relative energy efficiency of the whole network under the condition that each base station meets the maximum transmit power constraint and the network meets the minimum service satisfaction constraint.
[0009] Step 4: Construct the separable upper bound of JS divergence: Based on the distribution mismatch term within the service sub-region of each base station and the weight mismatch term between sub-regions, construct the separable upper bound of the JS divergence between the capacity distribution and the traffic demand distribution, and form a local mismatch term corresponding to each base station.
[0010] Step 5: Construct a local block update utility function induced by the upper bound of JS divergence, considering the serviceable traffic, local spatial matching factor and local power consumption in the base station's dominant sub-region, so that each base station can locally calculate the optimization criteria for beamforming updates.
[0011] Step 6: Perform beamforming block switching of the base station along the interference directed graph established in Step 2, and perform scalar consensus estimation on the information exchange graph to obtain consensus estimates of the total network capacity, total network traffic demand, and total network power consumption.
[0012] Step 7: Using the Interference-Guided Augmented Lagrangian Optimization (ID-ALM) method, under the conditions of satisfying local power constraints and local service satisfaction constraints, the beamforming variables are updated in a distributed manner, specifically including:
[0013] 7.1 Initialize the beamforming matrix, local power constraint parameters, service satisfaction threshold, augmented Lagrange penalty parameters, dual variables, slack variables, maximum number of iterations, and convergence threshold for each base station.
[0014] 7.2 Each base station calculates capacity weight, demand weight, local mismatch term, local serviceable traffic, local spatial matching factor, local block update utility function, and local service satisfaction based on local sub-region information, neighbor beamforming buffer, and consistency estimation results.
[0015] 7.3 Each base station constructs a local augmented Lagrangian function and approximates the local update subproblem under the local maximum transmit power constraint, updating its own beamforming matrix by projecting gradient.
[0016] 7.4 Each base station recalculates the local service satisfaction and constraint violation based on the updated beamforming results, and updates the slack variables and dual variables sequentially according to the ID-ALM update rules.
[0017] 7.5 Determine whether the change in the beamforming block is less than the preset convergence threshold, or whether the current iteration number has reached the maximum iteration number; if the stopping condition is met, output the updated beamforming matrix of each base station and the corresponding integrated relative energy efficiency result; otherwise, return to step six to continue iterating.
[0018] Technical effect
[0019] This invention employs integrated relative energy efficiency as a wireless network resource allocation indicator, unifying the spatial matching degree between the network's serviceable traffic, capacity, and traffic demand, as well as network power consumption, into a single optimization framework. Addressing the global coupling problem caused by JS divergence in integrated relative energy efficiency, an upper bound for JS divergence, separable by the base station's dominant sub-region, is constructed. This transforms the spatial mismatch metric, originally dependent on the network's capacity and traffic demand distribution, into a local mismatch term calculable by each base station, thus providing a local optimization basis for distributed beamforming updates. Through an information exchange mechanism combining interference directed graphs and scalar consensus estimation, each base station exchanges relevant beamforming blocks only with neighbors that significantly interfere with its capacity assessment. Simultaneously, scalar consensus estimation obtains necessary statistics such as total network capacity, total network traffic demand, and total network power consumption. Using the interference-guided augmented Lagrange method (ID-ALM), each base station can perform distributed beamforming optimization while satisfying local power constraints and service satisfaction constraints. Attached Figure Description
[0020] Figure 1 This is a flowchart of an implementation example;
[0021] Figure 2 This is a schematic diagram of a multi-base station MIMO network system model for an example.
[0022] Figure 3 This is a schematic diagram illustrating the iterative convergence results of the IREE-related indicators in Example 1.
[0023] Figure 4 This is a schematic diagram comparing the performance of different baseline methods under real IREE metrics in the example implementation.
[0024] Figure 5 This is a schematic diagram comparing the signaling overhead of different coordination mechanisms as the base station scale changes, as shown in the example. Detailed Implementation
[0025] like Figure 1 As shown in this embodiment, a distributed optimization method for relative energy efficiency of spatial traffic-aware wireless network integration is provided, comprising:
[0026] Step 1, construct as follows Figure 2 The distributed downlink multi-base station MIMO wireless network model shown includes... Each base station is equipped with multiple transmitting antennas and provides wireless transmission services to receiving entities within its service area through multiple beams, and performs capacity provisioning and local utility calculations within the corresponding sub-area.
[0027] The entire service area of the distributed downlink multi-base station MIMO wireless network ,in: Indicates the entire network service area. This is to divide the entire network service area based on the dominant service range of each base station. Non-overlapping sub-regions; the first From a base station to any location within the service area wireless channels ,in: Indicates the first Base station to location Large-scale channel gain between; Indicates the first Base station to location The normalized array response vector; the first Beamforming matrix of each base station ,in: Indicates the first The first base station The beamforming vector. Each base station is located The capacity provided ,in: Indicates the first Each base station is located The capacity provided at the location; Indicates beam index; Indicates the first The number of beams transmitted by each base station; Indicates system bandwidth; Indicates the first The first base station Each beam at position Signal-to-interference-to-noise ratio at the location; the entire network at the location Total capacity density at the location Total network capacity Total network traffic demand Total power consumption of the entire network ,in: For position Spatial flow demand density at the location; when the first The base station adopts the first When a beam transmits, the power of the useful signal received at that location is determined by the corresponding channel and the beamforming vector. Base station power consumption ,in: Indicates the first The first base station Each beam at position The signal-to-interference-to-noise ratio at the location; Indicates the first Base station to location wireless channels The conjugate transpose of; Indicates the index of interfering base stations; Indicates the interference beam index; Indicates noise power; Indicates the first Power consumption of each base station; Indicates the power consumption factor of the power amplifier; Indicates the first The static circuit power consumption of each base station.
[0028] Step 2: Establish information exchange graphs between base stations for scalar consistency estimation. and the directed interference graph used to define which base stations need to exchange beamforming blocks. Specifically: when the first The beamforming block of the first base station will affect the first... Each base station service sub-area If the capacity assessment causes significant disturbance, then a directed edge is established in the disturbance directed graph. The basis for its boundary construction is the interference intensity. Specifically: when Greater than the preset interference threshold At that time, establish directed edges ,Right now: ;No. Each base station is only included in its neighbor set based on its interference. The relevant beamforming blocks transmitted by the base station are used to calculate the local signal-to-interference-plus-noise ratio, capacity, local mismatch term, and local utility function, without needing to obtain the complete beamforming matrix of the entire network.
[0029] Step 3: Construct the primal optimization problem based on the integrated relative energy efficiency index. Specifically, under the condition that each base station meets the maximum transmit power constraint and the network meets the minimum service satisfaction constraint, optimize the beamforming variable W of the entire network to maximize the integrated relative energy efficiency of the entire network. Its constraints include: , , Among them: integrated relative energy efficiency indicators The entire network beamforming variable set ; The actual volume of traffic that the network can serve; Normalized capacity distribution Normalized flow demand distribution JS divergence and normalized capacity distribution between them Normalized flow demand distribution , The larger the value, the more severe the spatial mismatch between capacity distribution and flow demand distribution; Total power consumption of the entire network; service satisfaction. , For the first The maximum transmit power of each base station This is the preset minimum service satisfaction threshold.
[0030] because Simultaneously, relying on the total network capacity, total network traffic demand, total network power consumption, and normalized spatial distribution, its objective function exhibits significant global coupling; in particular, the JS divergence term causes the beam update of a single base station to affect the normalization result of the network capacity and the spatial mismatch result. Therefore, the original problem is difficult to solve independently by each base station, and a locally computeable objective suitable for distributed optimization needs to be further constructed.
[0031] Step 4: Construct the separable upper bound of the JS divergence. This involves representing the global capacity distribution and flow demand distribution as a mixture of sub-regional conditional distributions and their regional weights. Based on this regionalized representation, and utilizing the mixture decomposition property and log-sum inequality of the Kullback-Leibler divergence (KL divergence), a separable upper bound of the JS divergence is constructed. Specifically, the separable upper bound of the JS divergence can be obtained by summing the local mismatch terms. satisfy: , of which: Local mismatch terms corresponding to each base station , These represent the differences in distribution within the region and sub-regions. Normalized regional capacity distribution Normalized regional flow demand distribution , and The logarithmic ratio term represents the weight difference between regions, for the nth sub-region. Total capacity Total traffic demand The capacity weight of the nth sub-region Traffic demand weight sub-region Local KL divergence .
[0032] The aforementioned upper bound of JS divergence conservatively characterizes the spatial mismatch between capacity distribution and traffic demand distribution from two aspects: regional distribution differences and inter-regional weight differences. This upper bound is not an exact equivalent decomposition of JS divergence, but rather a conservatively separable representation used to construct a distributed optimization objective. Because Local mismatch terms corresponding to each base station By summing the results, each base station can calculate its own local mismatch terms based on the local sub-region capacity, traffic demand, and a small number of global scalars, thus laying the foundation for subsequent distributed beamforming optimization.
[0033] Step 5: Construct an upper-bound induced local block update utility function to guide each base station in performing distributed beamforming updates, i.e., the distributed local update problem, specifically as follows: Its constraints include: , Where: the local block update utility function of the nth base station The local service satisfaction of the nth base station and Locally serviceable traffic of the nth base station , and The first The dominant service sub-area of each base station Total internal capacity and total traffic demand. For local power consumption, For local mismatch terms, the local spatial matching factor of the nth base station .
[0034] The local block update utility function, which considers serviceable traffic, spatial matching degree, and power consumption at the local level, is used to evaluate the first... Current beamforming block of each base station The effect of partial update.
[0035] The local spatial matching factor Used to conservatively portray the first The degree of matching between capacity distribution and flow demand distribution within a sub-region. When there is a local mismatch... When less than 1, Local mismatch items When it is not less than 1, This is to avoid the spatial matching factor taking a negative value.
[0036] Step 6: Perform beamforming block switching along the interference directed graph and perform scalar consistency estimation on the information exchange graph. Specifically, each base station exchanges its own beamforming blocks only with its interference-related neighbors along the edges defined by the interference directed graph. This is used to update the local beamforming buffer and perform local interference assessment. At the same time, each base station exchanges only a small number of scalars such as local capacity, traffic demand, and power consumption on the information exchange graph. Consistent estimates of the total network capacity, total network traffic demand, and total network power consumption are obtained through scalar consistency estimation.
[0037] The exchange of only a few scalars such as local capacity, traffic demand, and power consumption means that during the scalar consensus estimation process, each base station does not send the complete beamforming matrix, complete channel matrix, or complete spatial traffic distribution of the entire network to other base stations. Instead, it uses its local sub-region capacity, local sub-region traffic demand, and local power consumption as the initial quantities for consensus estimation and exchanges the above scalar information with neighboring base stations on the information exchange graph.
[0038] The aforementioned scalar consistency estimation specifically refers to: the total network capacity. Round Consistency Estimate Total network traffic demand Round Consistency Estimate Total power consumption of the entire network Round Consistency Estimate Each iteration is , For consistency weights, the first... Local statistics held by each base station at the initial moment of consensus estimation .
[0039] Step 7: Employ the Interference-Guided Augmented Lagrangian (ID-ALM) method, guiding local beam optimization updates through an upper bound induced local utility function, and handling service satisfaction constraints through a local augmented Lagrangian mechanism. Specifically, this includes:
[0040] 7.1 After initialization, each base station first obtains the necessary neighbor beamforming blocks through the interference directed graph in each outer iteration, and calculates the local sub-region capacity, local power consumption, local mismatch term, local block update utility function, and local service satisfaction based on the scalar consensus estimation results. Specifically, in the t-th iteration, the n-th base station approximates the solution of the local subproblem. Its constraints include Where: the local augmented Lagrangian function of the nth base station , Let the variables be Lagrange dual variables. For positive penalty parameters, Non-negative slack variables and The constraint violation amount of the nth base station and That is, local service satisfaction constraint ;
[0041] 7.2 Update the local beamforming cache using the projection gradient method. ,in: Step size, Indicates projection to the first The feasible set of local power for each base station, the feasible set of local power ;
[0042] 7.3 After completing the local beamforming buffer update, the relaxation variables are updated in closed projection form as follows: The dual variable is updated to .
[0043] Through the above steps, the present invention can achieve distributed beamforming optimization for integrated relative energy efficiency without centrally collecting the complete beamforming matrix of the entire network.
[0044] This embodiment relates to a distributed optimization system for relative energy efficiency of spatial traffic-aware wireless network integration, implementing the above-described method. The system includes: an input configuration unit, a network modeling unit, a graph structure construction unit, a local criterion construction unit, a distributed optimization update unit, and a result output unit. The input configuration unit acquires local channel, traffic demand, initial beam, power constraints, and service thresholds. The network modeling unit constructs a multi-base station MIMO network model and calculates capacity, demand, and power consumption. The graph structure construction unit establishes an interference directed graph and an information exchange graph. The local criterion construction unit constructs the IREE primal optimization problem, the upper bound of JS divergence separability, the local block update utility function, and the local service satisfaction constraint. The distributed optimization update unit exchanges beamforming blocks along the interference directed graph, obtains global statistics through scalar consistency estimation, and performs local optimization updates using ID-ALM. The result output unit outputs the beamforming matrix, IREE results, JS mismatch, and signaling overhead results for each base station after the stopping condition is met.
[0045] Through practical application experiments, on a 64-bit Windows operating system personal computer and in a MATLAB and Python co-simulation platform (hardware and software) environment, the distributed optimization method for relative energy efficiency of spatial traffic-aware wireless network integration proposed in this invention was run according to the following process: "network parameter initialization, interference directed graph construction, beamforming block swapping, scalar consistency estimation, local mismatch term calculation, local utility function calculation, ID-ALM iterative update and convergence output". The experimental data obtained are as follows: the experimental results of the method of this invention in terms of real integration relative energy efficiency, JS spatial mismatch degree, algorithm convergence process, and signaling overhead between base stations, as detailed below. Figures 3 to 5 As shown.
[0046] This embodiment selects as follows: Figure 2The multi-base station MIMO downlink wireless network shown is used as the evaluation scenario in a 5km × 5km evaluation area, with 400 base stations, each 35 meters high, a maximum bandwidth limit of 36GHz, a maximum power limit of 30dBm, a static power consumption of 5W, a power amplifier efficiency of 38%, a path loss of 35+38log10(d), a noise power spectral density of -174dBm / HZ, a total of 2500 users, and a minimum traffic requirement of 0.8.
[0047] The service area was divided into multiple base station-dominated sub-regions, each with different spatial traffic demand densities. Each base station was equipped with multiple transmit antennas and multiple beams, and the receiving entities were distributed throughout the service area. To verify the optimization effect of this invention on spatial capacity-traffic mismatch, simulation experiments were conducted with different degrees of traffic skew to create a significant difference between hotspot areas and low-traffic areas.
[0048] In Experiment 1, the convergence performance of this invention and the effectiveness of the upper bound on the separability of the JS divergence are verified. Under the above parameter configuration, 35 rounds of ID-ALM iterations were performed on the method of this invention, and the changes in the true IREE, conservative IREE, and average local utility over time were recorded. Figure 3 As shown on the left, with increasing runtime, the true IREE increases from approximately 1.5 × 10^9 bit / J to approximately 7.8 × 10^9 bit / J, the conservative IREE increases from approximately 1.1 × 10^9 bit / J to approximately 7.3 × 10^9 bit / J, and the average local utility increases from approximately 2.6 × 10^9 bit / J to approximately 8.1 × 10^9 bit / J. All three metrics stabilize after approximately 30 minutes, indicating that the local block update utility function can effectively guide each base station in performing distributed beamforming updates. Figure 3 As shown on the right, as the number of base stations increases from 5 to 50, the actual JS mismatch decreases from approximately 0.25 to approximately 0.08, and the upper bound of JS divergence separability decreases from approximately 0.32 to approximately 0.09, with the upper bound remaining consistently above the actual JS mismatch. This result demonstrates that the upper bound of JS divergence separability constructed in this invention can conservatively characterize the spatial mismatch between capacity distribution and traffic demand distribution, and can maintain good upper bound tightness as the base station scale increases.
[0049] In Experiment 2, this invention was compared with centralized upper bound schemes, distributed ADMM, distributed fractional programming, traditional flow-aware beamforming, inconsistency-free schemes, and traditional MIMO energy efficiency optimization schemes to verify the real IREE performance of this invention under different spatial flow skew levels. Under the above parameter configuration, the flow skew level was set. The value was increased from 0.2 to 0.9, and the actual IREE results for each method were recorded. For example... Figure 4As shown, when When I = 0.2, the differences in true IREE among the methods are small, and the true IREE of the method of this invention is approximately 7.9 × 10^9 bit / J, close to the approximately 8.1 × 10^9 bit / J of the centralized upper bound scheme; when When the value is increased to 0.9, the heterogeneity of spatial flow demand increases, and the performance differences between various methods widen significantly. The true IREE of the method in this invention is approximately 6.8 × 10^9 bit / J, which is still close to the approximately 7.1 × 10^9 bit / J of the centralized upper bound scheme, and higher than that of distributed ADMM, distributed fractional programming, traditional flow-aware beamforming, inconsistency-free schemes, and traditional MIMO energy efficiency optimization schemes. Compared with distributed ADMM, distributed fractional programming, traditional flow-aware beamforming, inconsistency-free schemes, and traditional MIMO energy efficiency optimization schemes, the method in this invention... The actual IREE is improved by approximately 4.7%, 17.2%, 32.0%, 43.2%, and 88.9% when the value is 0.9, respectively. These results indicate that by constructing an upper bound for JS divergence and a local block update utility function, this invention can maintain a high capacity-flow space matching capability when the heterogeneity of flow demand space is enhanced, thereby improving the actual integration relative energy efficiency.
[0050] In Experiment 3, the signaling overhead of different coordination mechanisms was compared with the base station scale. The number of base stations was increased from 5 to 50 in the experiment, and normalized signaling overhead was used as the evaluation index. The information exchange volume in each round of coordinated update was statistically analyzed for centralized coordination schemes, distributed ADMM, distributed fractional programming, traditional traffic-aware beamforming, traditional gossip-assisted ADMM, and the method of this invention. Figure 5 As shown in Table 1, with the increase in the number of base stations, the normalized signaling overhead of the centralized coordination scheme increases from approximately 45 to approximately 1500 due to the need to collect complete beamforming variables across the entire network. The normalized signaling overhead of distributed ADMM, distributed fractional programming, traditional traffic-aware beamforming, and traditional gossip-assisted ADMM are approximately 125, 95, 70, and 35 respectively when the number of base stations is 50. The normalized signaling overhead of the method of this invention is approximately 20 at this time, significantly lower than the aforementioned comparative schemes. Compared with the centralized coordination scheme, distributed ADMM, distributed fractional programming, traditional traffic-aware beamforming, and traditional gossip-assisted ADMM, the signaling overhead of the method of this invention is reduced by approximately 98.7%, 84.0%, 78.9%, 71.4%, and 42.9% respectively at this time. As shown in Table 1, this invention limits the beamforming block switching range through an interference-directed graph and replaces the propagation of complete global variables with scalar consistency estimation, which can significantly reduce the signaling overhead in the distributed optimization process when the base station scale increases.
[0051] Table 1 Baseline Comparison
[0052] Compared with existing technologies, this invention, by constructing a separable upper bound for JS divergence and a local mismatch term, can further determine whether network capacity is allocated to areas with high actual traffic demand, thereby alleviating the problems of resource waste in areas with excess capacity and insufficient capacity in hotspot areas. By constructing an upper bound-induced local block update utility function, each base station can complete distributed updates based on local sub-region information, interfering neighbor beam information, and a small number of global scalars without directly solving a strongly coupled global optimization objective. Through an information exchange mechanism combining interference directed graphs and scalar consistency estimation, the signaling overhead in the distributed optimization process can be significantly reduced, making it suitable for green resource allocation and distributed beamforming optimization in large-scale multi-base station MIMO networks. Figures 3 to 5 The experimental results show that this method can reduce the JS spatial mismatch between capacity distribution and traffic demand distribution while maintaining high relative energy efficiency of real integration, and reduce the signaling overhead required for inter-base station collaborative updates.
[0053] The above-described specific implementations can be partially adjusted by those skilled in the art in different ways without departing from the principles and purpose of the present invention. The scope of protection of the present invention is defined by the claims and is not limited to the above-described specific implementations. All implementation schemes within the scope of the claims are bound by the present invention.
Claims
1. A distributed optimization method for relative energy efficiency of spatial traffic-aware wireless network integration, characterized in that, include: Step 1: Spatial mismatch modeling based on base station service sub-region division; Step 2: Establish the interference directed graph and the information exchange graph respectively; Step 3: Construct the original optimization problem based on the integrated relative energy efficiency index, that is, optimize the beamforming variables of the whole network to maximize the integrated relative energy efficiency of the whole network under the condition that each base station meets the maximum transmit power constraint and the network meets the minimum service satisfaction constraint. Step 4: Construct the separable upper bound of JS divergence: Based on the distribution mismatch term within the service sub-region of each base station and the weight mismatch term between sub-regions, construct the separable upper bound of the JS divergence between the capacity distribution and the traffic demand distribution, and form a local mismatch term corresponding to each base station. Step 5: Construct a local block update utility function based on the upper bound of JS divergence separability, considering the serviceable traffic, local spatial matching factor and local power consumption in the base station's dominant sub-region, so that each base station can locally calculate the optimization criteria for beamforming update. Step 6: Perform beamforming block switching of the base station along the interference directed graph established in Step 2, and perform scalar consensus estimation on the information exchange graph to obtain consensus estimates of the total network capacity, total network traffic demand, and total network power consumption. Step 7: Using the Interference-Guided Augmented Lagrange Optimization (ID-ALM) method, the beamforming variables are updated in a distributed manner while satisfying local power constraints and local service satisfaction constraints.
2. The distributed optimization method for relative energy efficiency of spatial traffic-aware wireless network integration according to claim 1, wherein step seven specifically includes: 7.1 Initialize the beamforming matrix, local power constraint parameters, service satisfaction threshold, augmented Lagrange penalty parameters, dual variables, slack variables, maximum number of iterations, and convergence threshold for each base station; 7.2 Each base station calculates capacity weight, demand weight, local mismatch term, local serviceable traffic, local spatial matching factor, local block update utility function, and local service satisfaction based on local sub-region information, neighbor beamforming buffer, and consistency estimation results; 7.3 Each base station constructs a local augmented Lagrangian function and approximates the local update subproblem under the local maximum transmit power constraint, updating its own beamforming matrix by projecting gradient. 7.4 Each base station recalculates the local service satisfaction and constraint violation based on the updated beamforming results, and updates the slack variables and dual variables sequentially according to the ID-ALM update rules; 7.5 Determine whether the change in the beamforming block is less than the preset convergence threshold, or whether the current iteration number has reached the maximum iteration number; If the stopping condition is met, output the updated beamforming matrix for each base station and the corresponding integrated relative energy efficiency result; otherwise, return to step six to continue iterating.
3. The distributed optimization method for relative energy efficiency of spatial traffic-aware wireless network integration according to claim 1, characterized in that, The aforementioned spatial mismatch modeling based on base station service sub-region partitioning refers to: constructing a distributed downlink multi-base station MIMO wireless network model, including... Each base station is equipped with multiple transmitting antennas and provides wireless transmission services to receiving entities within its service area through multiple beams, and performs capacity provisioning and local utility calculations within the corresponding sub-area. The entire service area of the distributed downlink multi-base station MIMO wireless network ,in: For the entire network service area, This is to divide the entire network service area based on the dominant service range of each base station. Non-overlapping sub-regions; From a base station to any location within the service area wireless channels ,in: For the first Base station to location Large-scale channel gain between; For the first Base station to location The normalized array response vector; the first Beamforming matrix of each base station ,in: For the first The first base station The beamforming vector, the th beamforming vector Each base station is located The capacity provided ,in: For the first Each base station is located The capacity provided at the location; For beam index; For the first The number of beams transmitted by each base station; For system bandwidth; For the first The first base station Each beam at position Signal-to-interference-to-noise ratio at the location; the entire network at the location Total capacity density at the location Total network capacity Total network traffic demand Total power consumption of the entire network ,in: For position Spatial flow demand density at the location; when the first The base station adopts the first When a beam transmits, the power of the useful signal received at that location is determined by the corresponding channel and the beamforming vector. Base station power consumption ,in: For the first The first base station Each beam at position The signal-to-interference-to-noise ratio at the location; For the first Base station to location wireless channels The conjugate transpose of; For interfering base station index; For interfering beam index; Noise power; For the first Power consumption of each base station; The power consumption factor of the power amplifier; For the first The static circuit power consumption of each base station.
4. The distributed optimization method for relative energy efficiency of spatial traffic-aware wireless network integration according to claim 1, characterized in that, Step 2, namely, establishing information exchange graphs between base stations for scalar consistency estimation. and the directed interference graph used to define which base stations need to exchange beamforming blocks. Specifically: when the first The beamforming block of the first base station will affect the first... Each base station service sub-area If the capacity assessment causes significant disturbance, then a directed edge is established in the disturbance directed graph. The basis for its boundary construction is the interference intensity. Specifically: when Greater than the preset interference threshold At that time, establish directed edges ,Right now: ;No. Each base station is only included in its neighbor set based on its interference. The relevant beamforming blocks transmitted by the base station are used to calculate the local signal-to-interference-plus-noise ratio, capacity, local mismatch term, and local utility function, without needing to obtain the complete beamforming matrix of the entire network.
5. The distributed optimization method for relative energy efficiency of spatial traffic-aware wireless network integration according to any one of claims 1-4, characterized in that, The aforementioned primal optimization problem based on integrated relative energy efficiency indices is specifically as follows: Its constraints include: , , Among them: integrated relative energy efficiency indicators The entire network beamforming variable set ; The actual volume of traffic that the network can serve; Normalized capacity distribution Normalized flow demand distribution JS divergence and normalized capacity distribution between them Normalized flow demand distribution , The larger the value, the more severe the spatial mismatch between capacity distribution and flow demand distribution; Total power consumption of the entire network; service satisfaction , For the first The maximum transmit power of each base station This is the preset minimum service satisfaction threshold.
6. The distributed optimization method for relative energy efficiency of spatial traffic-aware wireless network integration according to claim 5, characterized in that, The construction of the separable upper bound of the JS divergence is specifically as follows: the separable upper bound of the JS divergence can be obtained by summing the local mismatch terms. satisfy: , of which: Local mismatch terms corresponding to each base station , These represent the differences in distribution within the region and sub-regions. Normalized regional capacity distribution Normalized regional flow demand distribution , and The logarithmic ratio term represents the weight difference between regions, for the nth sub-region. Total capacity Total traffic demand The capacity weight of the nth sub-region Traffic demand weight sub-region Local KL divergence .
7. The distributed optimization method for relative energy efficiency of spatial traffic-aware wireless network integration according to claim 6, characterized in that, The construction of the local block update utility function based on the upper bound of JS divergence separability is as follows: Its constraints include: , Where: the local block update utility function of the nth base station The local service satisfaction of the nth base station and Locally serviceable traffic of the nth base station , and The first The dominant service sub-area of each base station Total internal capacity and total traffic demand. For local power consumption, For local mismatch terms, the local spatial matching factor of the nth base station .
8. The distributed optimization method for relative energy efficiency of spatial traffic-aware wireless network integration according to claim 1, characterized in that, Step 6 is specifically as follows: In each distributed iteration, the nth base station sends its own beamforming block to its interfering out-neighbor along the interference directed graph and receives the beamforming block sent by the interfering in-neighbor to update the local beamforming buffer; at the same time, the nth base station uses the total local sub-region capacity, the total local sub-region traffic demand, and the local power consumption as the initial statistics for scalar consistency estimation, exchanges the above scalars with neighboring base stations on the information exchange graph, and after a preset number of consistency iterations, obtains the consistent estimates of the total network capacity, the total network traffic demand, and the total network power consumption; The exchange of only a few scalars such as local capacity, traffic demand, and power consumption means that during the scalar consensus estimation process, each base station does not send the complete beamforming matrix, complete channel matrix, or complete spatial traffic distribution of the entire network to other base stations. Instead, it uses its local sub-region capacity, local sub-region traffic demand, and local power consumption as the initial quantities for consensus estimation and exchanges the above scalar information with neighboring base stations on the information exchange graph.
9. The distributed optimization method for relative energy efficiency of spatial traffic-aware wireless network integration according to claim 5, characterized in that, The aforementioned scalar consistency estimation specifically refers to: the total network capacity. Round Consistency Estimate Total network traffic demand Round Consistency Estimate Total power consumption of the entire network Round Consistency Estimate Each iteration is , For consistency weights, the first... Local statistics held by each base station at the initial moment of consensus estimation .
10. A distributed optimization system for relatively energy-efficient spatial traffic-aware wireless network integration, implementing the method of any one of claims 1-9, characterized in that, include: The system comprises an input configuration unit, a network modeling unit, a graph structure construction unit, a local criterion construction unit, a distributed optimization update unit, and a result output unit. The input configuration unit acquires local channel, traffic demand, initial beam, power constraints, and service thresholds. The network modeling unit constructs a multi-base station MIMO network model and calculates capacity, demand, and power consumption. The graph structure construction unit establishes the interference directed graph and the information exchange graph. The local criterion construction unit constructs the IREE primal optimization problem, the upper bound of JS divergence separability, the local block update utility function, and the local service satisfaction constraints. The distributed optimization update unit exchanges beam blocks along the interference directed graph, obtains global statistics through scalar consistency estimation, and performs local optimization updates using ID-ALM. The result output unit outputs the beamforming matrix, IREE results, JS mismatch, and signaling overhead results for each base station after the stopping condition is met.