Artificial intelligence-based wireless path optimization method and device, and storage medium

CN122891990APending Publication Date: 2026-10-09BEIJING DOUMIAO CLOUD INFORMATION TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610992743.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-06
Publication Date
2026-10-09

AI Technical Summary

Technical Problem

[0004]本发明的目的在于提供一种基于人工智能的无线路径优化方法、装置及存储介质,以解决现有技术存在的问题中传统反应式路由机制在动态遮挡环境下存在探测与收敛延迟,导致5G毫米波多跳回传网络业务中断与用户体验陡降的技术问题

Benefits of technology

[0048]通过融合多源环境感知信息与基带物理层精细参数,构建以IAB节点为顶点、链路状态向量为边特征、节点剩余缓存队列深度为顶点特征的动态时空图,实现了对毫米波链路状态的实时感知与预测。该方案依据动态时空图构建链路可用度矩阵,通过计算回传路径的脆弱性指数主动识别待优化业务流,并基于适应度函数生成最优候选回传路径,采用"先建后断"的软切换策略实现路径的平滑迁移,有效克服了传统反应式路由机制在动态遮挡环境下的探测与收敛延迟问题。在建筑密集或障碍密集的5G毫米波接入回传一体化节点间动态多跳无线回传网络场景中,该方案能够应对大型车辆穿行、临时施工围挡搭建、季节性行道树叶密度变化等动态遮挡因素,确保多跳回传网络在面对突发性链路质量劣化时实现主动、无感的路径切换,维持确定性的容量与延迟性能,显著提升增强移动宽带和超可靠低时延通信混合业务的传输可靠性与用户体验。

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122891990A_ABST
    Figure CN122891990A_ABST
Patent Text Reader

Abstract

The present application relates to the technical field of data signal transmission, and particularly relates to a wireless path optimization method and device based on artificial intelligence and a storage medium, the method comprising: marking an occlusion from point cloud data, extracting a power angle spectrum feature of a link multipath cluster from a radio frequency parameter to generate an occlusion state vector of each link of an IAB node; constructing a dynamic space-time graph with the IAB node as a vertex, the link state vector as an edge feature, and the node residual buffer queue depth as a vertex feature; constructing a link availability matrix, for each path carrying a service flow, calculating a vulnerability index of the backhaul path according to the current hop number and the availability of each link on the backhaul path, and marking the service flow to be optimized; generating a candidate backhaul path of the service flow to be optimized according to the link availability matrix; and issuing the candidate backhaul path to each IAB node to execute path switching. The present application effectively improves the efficiency of path switching in a complex environment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of data signal transmission technology, and in particular to a wireless path optimization method, apparatus and storage medium based on artificial intelligence. Background Technology

[0002] In the deployment of integrated 5G millimeter wave access and backhaul networks, multiple nodes use a self-organizing multi-hop relay method to transmit data from signal dead zones or hotspot areas back to the fiber optic access anchor point. Millimeter wave signals have extremely poor diffraction capabilities and high path loss characteristics, making link quality highly susceptible to dynamic obstructions in the street environment, such as large vehicles passing by, temporary construction barriers erected, and seasonal changes in the density of foliage along street trees.

[0003] Traditional routing protocols rely on hello message probing and fixed threshold-triggered rerouting mechanisms, resulting in significant reactive delays from link quality degradation detection to network-wide path convergence. For backhaul networks carrying a mix of enhanced mobile broadband and ultra-reliable low-latency communication services, this delay is sufficient to cause substantial service interruptions and a sharp decline in user experience, failing to meet the real-time path optimization requirements of dynamic and complex urban environments. Summary of the Invention

[0004] The purpose of this invention is to provide a wireless path optimization method, device, and storage medium based on artificial intelligence, in order to solve the technical problem that the traditional reactive routing mechanism has detection and convergence delays in dynamic occlusion environments, which leads to the interruption of 5G millimeter wave multi-hop backhaul network services and a sharp drop in user experience.

[0005] To achieve the above objectives, according to one aspect of this application, the present invention provides an artificial intelligence-based wireless path optimization method, comprising:

[0006] Obstructions are marked in point cloud data, and the power angle spectrum features of multipath clusters of links are extracted from radio frequency parameters to generate the obstruction state vector of each link of the IAB node.

[0007] Based on the occlusion state vector of each IAB node, a dynamic spatiotemporal graph is constructed with IAB nodes as vertices, link state vectors as edge features, and the remaining cache queue depth of nodes as vertex features.

[0008] Based on the dynamic spatiotemporal graph, a link availability matrix is ​​constructed. For each path carrying a service flow, the vulnerability index of the backhaul path is calculated based on its current hop count and the availability of each link on the backhaul path, and the service flows to be optimized are marked.

[0009] In response to marking the service flows to be optimized, candidate backhaul paths for the service flows to be optimized are generated based on the link availability matrix;

[0010] The candidate backhaul path is distributed to each IAB node to perform path switching.

[0011] As a further improvement of the present invention, the environmental point cloud data collected by the IAB node is clustered and target segmented to identify point cloud clusters and mark each point cloud cluster as an occlusion.

[0012] Power angle spectrum analysis is performed on the link multipath clusters extracted from the channel impulse response to calculate the power ratio of the main path to the strongest reflecting path and the angular spread of the link multipath clusters.

[0013] Calculate the percentage of point clouds marked as occlusions and use the result as the environmental threat coefficient of the link;

[0014] The power ratio, the angular extension of the link multipath cluster, and the environmental threat coefficient are concatenated in the vector dimension to form the link state vector of the link.

[0015] As a further improvement of the present invention, the adjacency table of each IAB node in the target network is obtained, and the active millimeter-wave sector of each IAB node is connected and marked as a directed edge.

[0016] The link state vector reported by each node is used as the feature matrix of the corresponding directed edge;

[0017] The remaining cache queue depth of the node itself is used as the vertex feature vector;

[0018] A network state graph for a given time step is formed by the feature matrices of all directed edges and the feature vectors of all vertices.

[0019] The network state graph is iterated over time to obtain a sequence of network state graphs, which is then used as a dynamic spatiotemporal graph.

[0020] As a further improvement of the present invention, the original availability score of the link is determined based on the directed edges in the dynamic spatiotemporal graph;

[0021] The original availability score is corrected based on the vertex feature vectors in the dynamic spatiotemporal graph to obtain the link availability score;

[0022] The link availability matrix is ​​determined based on the link availability score, and the link availability matrix is ​​a numerical matrix of the availability scores of links between each IAB node.

[0023] As a further improvement of the present invention, for a backhaul path composed of multiple links, the lowest value of the availability score of all links in the backhaul path is extracted according to the link availability matrix, and the number of hops of the backhaul path is counted. Then, the number of hops of the backhaul path is normalized by setting the maximum value of the number of hops.

[0024] The vulnerability index of the backhaul path is set as the lowest value of the availability score of all links in the backhaul path divided by the normalized number of hops in the backhaul path.

[0025] When marking a service flow to be optimized, if the vulnerability index of the backhaul path of any service flow exceeds a preset threshold, the service flow is marked as a service flow to be optimized.

[0026] As a further improvement of the present invention, the source node and target node of the business flow to be optimized are used as constraints. The sum of the availability scores of all reachable paths is calculated based on the link availability matrix. The calculation results are recorded as path availability and sorted in descending order by path availability to obtain a candidate path table.

[0027] Calculate the hop count and routing coupling degree of each path in the candidate path table;

[0028] Construct a fitness function to determine the fitness score of each path in the candidate path table;

[0029] The path with the highest fitness score is selected as the candidate backhaul path.

[0030] As a further improvement of the present invention, the power ratio component, angle spread component and environmental threat coefficient component in the feature matrix of each directed edge in the dynamic spatiotemporal graph are normalized respectively to obtain the power ratio normalization score, angle spread normalization score and environmental threat coefficient normalization score.

[0031] The original availability score is constructed based on the normalized power ratio component, angle spread component, and environmental threat coefficient.

[0032] As a further improvement of the present invention, the routing coupling degree is the number of links shared by each path in the candidate path table and the paths already occupied by other existing service flows;

[0033] The reachable path is a path that satisfies the constraints.

[0034] When constructing the fitness function, the path availability score, hop count and routing coupling degree of each path in the candidate path table are normalized by the maximum and minimum values ​​respectively.

[0035] Setting: Fitness score = Normalized path availability score - Normalized hop count - Normalized route coupling.

[0036] According to another aspect of this application, an artificial intelligence-based wireless path optimization device is provided, comprising:

[0037] The data acquisition unit is used to integrate environmental sensing sensors on the IAB node, and simultaneously collect radio frequency parameters and environmental point cloud data within the millimeter-wave phased array antenna sector of the node, and stamp the two types of data with a unified timestamp.

[0038] The state determination unit is used to mark obstructions from point cloud data and extract the power angle spectrum features of link multipath clusters from radio frequency parameters to generate the obstruction state vector of each link of the IAB node.

[0039] The feature extraction unit is used to construct a dynamic spatiotemporal graph based on the occlusion state vector of each IAB node, with IAB nodes as vertices, link state vectors as edge features, and the remaining cache queue depth of the node as vertex features.

[0040] The service marking unit is used to construct a link availability matrix based on the dynamic spatiotemporal graph. For each path carrying a service flow, it calculates the vulnerability index of the backhaul path based on its current hop count and the availability of each link on the backhaul path, and marks the service flow to be optimized.

[0041] The path optimization unit is used to generate candidate backhaul paths for the marked service flows to be optimized based on the link availability matrix in response to the marked service flows.

[0042] The path switching unit is used to distribute candidate backhaul paths to each IAB node to perform path switching.

[0043] According to another aspect of this application, an electronic device is provided, comprising:

[0044] One or more processors;

[0045] Storage device for storing one or more programs;

[0046] When the one or more programs are executed by the one or more processors, the one or more processors implement the AI-based wireless path optimization method.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0048] By integrating multi-source environmental perception information with fine-grained parameters of the baseband physical layer, a dynamic spatiotemporal graph is constructed with IAB nodes as vertices, link state vectors as edge features, and node remaining buffer queue depth as vertex features, enabling real-time perception and prediction of millimeter-wave link status. Based on the dynamic spatiotemporal graph, this scheme constructs a link availability matrix, proactively identifies service flows to be optimized by calculating the vulnerability index of backhaul paths, and generates optimal candidate backhaul paths based on a fitness function. A "build first, then disconnect" soft handover strategy is adopted to achieve smooth path migration, effectively overcoming the detection and convergence delay problems of traditional reactive routing mechanisms in dynamic obstruction environments. In dynamic multi-hop wireless backhaul network scenarios between 5G millimeter-wave access and backhaul integrated nodes in densely built or obstacle-filled environments, this scheme can cope with dynamic obstruction factors such as large vehicle traffic, temporary construction barriers, and seasonal changes in roadside foliage density. It ensures that the multi-hop backhaul network can achieve proactive and seamless path switching when facing sudden link quality degradation, maintaining deterministic capacity and latency performance, and significantly improving the transmission reliability and user experience of enhanced mobile broadband and ultra-reliable low-latency communication hybrid services. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart illustrating the AI-based wireless path optimization method of this embodiment.

[0051] Figure 2 This is a schematic diagram of the structure of the AI-based wireless path optimization device provided in this embodiment.

[0052] Figure 3 This is a schematic diagram of the structure of the electronic device provided in this embodiment. Detailed Implementation

[0053] To more clearly illustrate the present invention, the following description, in conjunction with preferred embodiments and accompanying drawings, further explains the invention. Similar components in the drawings are indicated by the same reference numerals. Those skilled in the art should understand that the specific description below is illustrative rather than restrictive and should not be construed as limiting the scope of protection of the present invention.

[0054] It should be noted that although the terms first, second, third, etc., may be used in the embodiments of this application for description, these descriptions should not be limited to these terms. These terms are only used to distinguish the descriptions. For example, without departing from the scope of the embodiments of this application, first can also be referred to as second, and similarly, second can also be referred to as first.

[0055] The acquisition, storage, use, and processing of data in this application all comply with the relevant provisions of national laws and regulations.

[0056] This application defines the terms that appear below:

[0057] Link: A point-to-point one-way or two-way wireless connection established between two directly adjacent IAB nodes via a millimeter-wave beam.

[0058] Backhaul path: An ordered sequence of links that connect the source node to the destination node through which data is transmitted.

[0059] Service flow: A sequence of user data packets that are injected into the network from a source node and sent to a destination node.

[0060] Unless otherwise specified in this application, the normalization method is maximum-minimum normalization.

[0061] Example 1:

[0062] It should be noted first that the AI-based wireless routing path optimization method described in this embodiment is applied to dynamic multi-hop wireless backhaul network scenarios between integrated 5G millimeter-wave access and backhaul nodes in densely built-up or obstacle-heavy environments. In this application scenario, multiple IAB nodes use self-organized multi-hop relay to backhaul data traffic from signal blind spots or hotspot areas to the nearest fiber optic access anchor point. Due to the extremely poor diffraction capability and high path loss of millimeter-wave signals, their links are highly susceptible to dynamic obstructions in the street environment, such as the passage of large vehicles, the erection of temporary construction barriers, and even seasonal changes in the density of foliage on roadside trees, leading to sudden quality degradation or even interruption of one or more established backhaul paths. Traditional mechanisms based on routing protocol hello message detection and fixed threshold-triggered rerouting have reactive delays from detecting link quality degradation to completing network-wide path convergence. For backhaul networks carrying a mix of enhanced mobile broadband and ultra-reliable low-latency communication services, this is sufficient to cause significant service interruptions and a sharp decline in user experience.

[0063] In response, this embodiment provides a method that can pre-calculate the optimal path in complex urban street environments by fusing multi-source environmental perception information with fine parameters of the baseband physical layer. This enables proactive, smooth, and seamless switching of the backhaul path between IAB nodes, ensuring that the multi-hop backhaul network maintains deterministic capacity and latency performance when facing dynamic occlusion.

[0064] To apply the above-mentioned application scenarios, this application provides a wireless routing path optimization method based on artificial intelligence, specifically as follows: Figure 1 As shown, it includes:

[0065] Step S1: Integrate environmental sensing sensors on the IAB node to synchronously collect radio frequency parameters and environmental point cloud data within the millimeter-wave phased array antenna sector of the node, and assign a unified timestamp to the two types of data.

[0066] Step S2: Mark the obstructions from the point cloud data and extract the power angle spectrum features of the link multipath cluster from the radio frequency parameters to generate the obstruction state vector of each link of the IAB node.

[0067] Step S3: Based on the occlusion state vector of each IAB node, construct a dynamic spatiotemporal graph with IAB nodes as vertices, link state vectors as edge features, and the remaining cache queue depth of nodes as vertex features.

[0068] Step S4: Construct a link availability matrix based on the dynamic spatiotemporal graph. For each path carrying a service flow, calculate the vulnerability index of the backhaul path based on its current hop count and the availability of each link on the backhaul path, and mark the service flows to be optimized.

[0069] Step S5: In response to marking the service flow to be optimized, generate candidate backhaul paths for the service flow to be optimized based on the link availability matrix;

[0070] Step S6: Distribute the candidate backhaul paths to each IAB node to perform path switching.

[0071] It should be noted that this embodiment does not limit the type of environmental perception sensor integrated on the IAB node. It can be a low-cost automotive-grade multi-line LiDAR or a binocular stereo vision camera module. As long as it can perceive the dynamic obstacle distribution in the block space centered on the IAB node and covering its adjacent nodes with centimeter-level accuracy in real time, those skilled in the art can flexibly select the type according to cost and performance requirements.

[0072] The radio frequency parameters described in this embodiment include at least: the channel impulse response in each probe beam direction, the number of multipath clusters extracted based on the channel impulse response, the power of each multipath cluster, the angle of arrival, the angle of departure, and the Doppler shift.

[0073] The process of marking occlusions from point cloud data includes:

[0074] Clustering and target segmentation are performed on the environmental point cloud data collected by the IAB nodes to identify point cloud clusters and mark each point cloud cluster as an occlusion.

[0075] The process of generating the occlusion-aware link state vector is as follows:

[0076] Power angle spectrum analysis is performed on the link multipath clusters extracted from the channel impulse response to calculate the power ratio of the main path to the strongest reflecting path and the angular spread of the link multipath clusters.

[0077] Calculate the percentage of point clouds marked as occlusions and use the result as the environmental threat coefficient of the link;

[0078] The power ratio, the angular extension of the link multipath cluster, and the environmental threat coefficient are concatenated in the vector dimension to form the link state vector of the link.

[0079] The strongest reflection path refers to the multipath cluster with the highest power other than the main path; the main path refers to the multipath cluster with the highest power in the power angle spectrum.

[0080] It should be noted that the power angular spectrum analysis of the extracted multipath clusters includes the following steps:

[0081] The angular detection range of the receiving antenna array is divided into several angular units, each corresponding to a small angular interval. All extracted multipath clusters are traversed, and each multipath cluster is classified into the corresponding angular unit according to its angle of arrival. If there are multiple multipath clusters in the same angular unit, the power of the unit is the sum of the powers of these multipath clusters. If there are no multipath clusters in an angular unit, the power of the unit is zero. With the angular unit as the horizontal axis and the normalized or absolute power value of each unit as the vertical axis, the power angular spectrum of the link at the current moment is obtained.

[0082] Specifically, the process of constructing the dynamic spatiotemporal graph is as follows:

[0083] Obtain the adjacency table of each IAB node in the target network, and mark the active millimeter-wave sectors of each IAB node as a directed edge;

[0084] The link state vector reported by each node is used as the feature matrix of the corresponding directed edge;

[0085] The remaining cache queue depth of the node itself is used as the vertex feature vector;

[0086] A network state graph for a given time step is formed by the feature matrices of all directed edges and the feature vectors of all vertices.

[0087] The network state graph is iterated over time to obtain a sequence of network state graphs, which is then used as a dynamic spatiotemporal graph.

[0088] It should be noted that the active millimeter-wave sector connection described in this embodiment can be determined according to the following steps:

[0089] After initial access or beam reconfiguration, the IAB node reports "IAB node establishes connection" to the centralized intelligent controller via control plane signaling. After confirmation, the controller adds the link to the adjacency table and issues measurement configuration for the sector to the IAB node. The IAB node continuously performs reference signal transmission and reception on the sector according to the configuration and periodically reports measurement results and extracted channel impulse response data. When measurement results are continuously reported and the link does not receive a release command, the controller determines that the sector is in an active state, that is, the link between the IAB node and the sector is an active millimeter-wave sector connection.

[0090] In an exemplary embodiment, the time step is the time discretization unit used when constructing the dynamic spatiotemporal graph sequence. In this embodiment, it is set to 100ms to synchronize with the common lidar frame rate.

[0091] The process of constructing the link availability matrix based on the dynamic spatiotemporal graph is as follows:

[0092] Based on the directed edges in the dynamic spatiotemporal graph, determine the original availability score of the link;

[0093] The original availability score is corrected based on the vertex feature vectors in the dynamic spatiotemporal graph to obtain the link availability score;

[0094] The link availability matrix is ​​determined based on the link availability score, and the link availability matrix is ​​a numerical matrix of the availability scores of links between each IAB node.

[0095] Specifically, the process for determining the initial availability score of a link is as follows:

[0096] The power ratio component, angle spread component, and environmental threat coefficient component in the feature matrix of each directed edge in the dynamic spatiotemporal graph are normalized respectively to obtain the power ratio normalization score, angle spread normalization score, and environmental threat coefficient normalization score.

[0097] Based on the normalized power ratio component, angle spread component, and environmental threat coefficient, the original availability score is constructed, and the original availability score is set as α × power ratio normalized score + β × angle spread normalized score + γ × environmental threat coefficient normalized score.

[0098] Where α, β, and γ are weighting coefficients, satisfying α+β+γ=1.

[0099] It should be noted that the weighting coefficients are predetermined based on the predicted contribution of each feature component in historical link interruption events. In an exemplary embodiment, α is set to 0.4, β to 0.2, and γ to 0.4, so that features directly related to environmental perception have a higher weight in the decision-making process.

[0100] The process of correcting the original availability score is as follows:

[0101] Read the vertex features of the link target node j from the network state graph, i.e., the remaining cache queue depth of node j;

[0102] Calculate the cache queue depth penalty factor: When the remaining cache queue depth of node j is greater than or equal to the safe depth threshold, the penalty factor is 1; when the remaining cache queue depth is less than the safe depth threshold, the penalty factor is the ratio of the current remaining cache queue depth to the safe depth threshold.

[0103] Availability score = Original availability score × Cache queue depth penalty factor.

[0104] The security depth threshold is an empirical value preset based on the maximum buffer capacity of the IAB node. In an exemplary embodiment, the security depth threshold is set to 30% of the maximum buffer capacity. Based on this security depth threshold, even if the wireless quality of a link itself is excellent, if the target node's buffer is about to overflow, data packets will be forced to be discarded after reception, rendering the link unusable from an end-to-end service perspective.

[0105] Traverse all directed edges in the network state graph and calculate the final availability score for each edge. Construct an N×N link availability score matrix using all N IAB nodes in the network as row and column indices. The element in the i-th row and j-th column of the matrix represents the availability score of the link from node i to node j within a future preset time window. If there is no direct directed edge connection between node i and node j, the corresponding matrix element is set to 0.

[0106] The process of calculating the vulnerability index of the backhaul path is as follows:

[0107] For a backhaul path consisting of multiple links, the minimum availability score of all links in the backhaul path is extracted based on the link availability matrix, and the number of hops in the backhaul path is counted. Then, the number of hops in the backhaul path is normalized by setting a maximum value for the number of hops.

[0108] The vulnerability index of the backhaul path is set as the lowest value of the availability score of all links within the backhaul path divided by the normalized number of hops in the backhaul path.

[0109] When marking a service flow to be optimized, if the vulnerability index of the backhaul path of any service flow exceeds a preset threshold, the service flow is marked as a service flow to be optimized.

[0110] In an exemplary embodiment, the preset threshold can be set through the following process: select a stable operating period of the backhaul network in history that is determined to be fault-free and without link interruption, calculate the vulnerability index of all active backhaul paths in the entire network at each sampling time during the period, and form a normal behavior sample set.

[0111] Calculate the statistical characteristics of the sample set: calculate the mean μ and standard deviation σ of all vulnerability indices in the sample set; set the preset threshold as μ + k × σ, where k is an empirical coefficient with a value range of [2,3]. For example, k is set to 2.5. The statistical meaning is: when the network is in a normal state, the probability of the vulnerability index falling below this threshold is extremely high; once the vulnerability index exceeds this threshold, it can be determined with high confidence that the vulnerability of the current path has significantly deviated from the normal fluctuation range, belonging to an abnormal state, and optimization needs to be triggered.

[0112] Specifically, the process of generating candidate backhaul paths for the service flow to be optimized based on the link availability matrix is ​​as follows:

[0113] Using the source and target nodes of the business flow to be optimized as constraints, the sum of availability scores of all reachable paths is calculated based on the link availability matrix. The calculation results are recorded as path availability, and the paths are sorted in descending order of availability to obtain a candidate path table.

[0114] Calculate the hop count and routing coupling degree of each path in the candidate path table;

[0115] Construct a fitness function to determine the fitness score of each path in the candidate path table;

[0116] The path with the highest fitness score is selected as the candidate backhaul path.

[0117] The routing coupling degree is the number of links shared by each path in the candidate path table with existing paths already occupied by other service flows;

[0118] The reachable path is a path that satisfies the constraints.

[0119] When constructing the fitness function, the path availability score, hop count and routing coupling degree of each path in the candidate path table are normalized by the maximum and minimum values ​​respectively.

[0120] Setting: Fitness score = Normalized path availability score - Normalized hop count - Normalized route coupling.

[0121] Specifically, the process of distributing candidate backhaul paths to each IAB node to perform path switching is as follows:

[0122] The topological differences between the new candidate path and the current path are compared to identify the links that need to be added and the links that need to be removed. For the links that need to be added, the controller pre-configures the millimeter-wave beamforming parameters, time and frequency resources and quality of service policies of the target node through the management and control interface of the IAB node, so that it enters the waiting state.

[0123] Once all new links are ready, the controller adopts a "build first, disconnect later" soft switching strategy, issuing traffic scheduling instructions to the source IAB node to seamlessly and sequentially migrate the data packets of the specified service flow from the old path to the new path; after there are no more data packets in transit on the old path, instructions are issued to dismantle the old link.

[0124] Example 2:

[0125] Please see Figure 2 The diagram shown is a structural schematic of the artificial intelligence-based wireless path optimization device provided in this application, including:

[0126] The data acquisition unit is used to integrate environmental sensing sensors on the IAB node, and simultaneously collect radio frequency parameters and environmental point cloud data within the millimeter-wave phased array antenna sector of the node, and stamp the two types of data with a unified timestamp.

[0127] The state determination unit is used to mark obstructions from point cloud data and extract the power angle spectrum features of link multipath clusters from radio frequency parameters to generate the obstruction state vector of each link of the IAB node.

[0128] The feature extraction unit is used to construct a dynamic spatiotemporal graph based on the occlusion state vector of each IAB node, with IAB nodes as vertices, link state vectors as edge features, and the remaining cache queue depth of the node as vertex features.

[0129] The service marking unit is used to construct a link availability matrix based on the dynamic spatiotemporal graph. For each path carrying a service flow, it calculates the vulnerability index of the backhaul path based on its current hop count and the availability of each link on the backhaul path, and marks the service flow to be optimized.

[0130] The path optimization unit is used to generate candidate backhaul paths for the marked service flows to be optimized based on the link availability matrix in response to the marked service flows.

[0131] The path switching unit is used to distribute candidate backhaul paths to each IAB node to perform path switching.

[0132] The AI-based wireless path optimization device provided in this application can execute the AI-based wireless path optimization method provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects of executing the method.

[0133] Example 3:

[0134] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0135] If the functions implemented by the method are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art or the current technical solution, can be embodied in the form of a software product. This current computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0136] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0137] Example 4:

[0138] From a hardware perspective, to realize the functionality of the AI-based wireless path optimization method in a computer, this application also provides an electronic device, which can be a server, and its internal structure diagram can be as follows. Figure 3 As shown:

[0139] The computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores server data. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements an artificial intelligence-based wireless path optimization method.

[0140] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0141] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A wireless path optimization method based on artificial intelligence, characterized in that, include: Obstructions are marked in point cloud data, and the power angle spectrum features of multipath clusters of links are extracted from radio frequency parameters to generate the obstruction state vector of each link of the IAB node. Based on the occlusion state vector of each IAB node, a dynamic spatiotemporal graph is constructed with IAB nodes as vertices, link state vectors as edge features, and the remaining cache queue depth of nodes as vertex features. Based on the dynamic spatiotemporal graph, a link availability matrix is ​​constructed. For each path carrying a service flow, the vulnerability index of the backhaul path is calculated based on its current hop count and the availability of each link on the backhaul path, and the service flows to be optimized are marked. In response to marking the service flows to be optimized, candidate backhaul paths for the service flows to be optimized are generated based on the link availability matrix; The candidate backhaul path is distributed to each IAB node to perform path switching.

2. The wireless path optimization method based on artificial intelligence according to claim 1, characterized in that, Clustering and target segmentation are performed on the environmental point cloud data collected by the IAB nodes to identify point cloud clusters and mark each point cloud cluster as an occlusion. Power angle spectrum analysis is performed on the link multipath clusters extracted from the channel impulse response to calculate the power ratio of the main path to the strongest reflecting path and the angular spread of the link multipath clusters. Calculate the percentage of point clouds marked as occlusions and use the result as the environmental threat coefficient of the link; The power ratio, the angular extension of the link multipath cluster, and the environmental threat coefficient are concatenated in the vector dimension to form the link state vector of the link.

3. The wireless path optimization method based on artificial intelligence according to claim 2, characterized in that, Obtain the adjacency table of each IAB node in the target network, and mark the active millimeter-wave sectors of each IAB node as a directed edge; The link state vector reported by each node is used as the feature matrix of the corresponding directed edge; The remaining cache queue depth of the node itself is used as the vertex feature vector; A network state graph for a given time step is formed by the feature matrices of all directed edges and the feature vectors of all vertices. The network state graph is iterated over time to obtain a sequence of network state graphs, which is then used as a dynamic spatiotemporal graph.

4. The wireless path optimization method based on artificial intelligence according to claim 3, characterized in that, Based on the directed edges in the dynamic spatiotemporal graph, determine the original availability score of the link; The original availability score is corrected based on the vertex feature vectors in the dynamic spatiotemporal graph to obtain the link availability score; The link availability matrix is ​​determined based on the link availability score, and the link availability matrix is ​​a numerical matrix of the availability scores of links between each IAB node.

5. The wireless path optimization method based on artificial intelligence according to claim 4, characterized in that, For a backhaul path consisting of multiple links, the minimum availability score of all links in the backhaul path is extracted based on the link availability matrix, and the number of hops in the backhaul path is counted. Then, the number of hops in the backhaul path is normalized by setting a maximum value for the number of hops. The vulnerability index of the backhaul path is set as the lowest value of the availability score of all links in the backhaul path divided by the normalized number of hops in the backhaul path. When marking a service flow to be optimized, if the vulnerability index of the backhaul path of any service flow exceeds a preset threshold, the service flow is marked as a service flow to be optimized.

6. The wireless path optimization method based on artificial intelligence according to claim 5, characterized in that, Using the source and target nodes of the business flow to be optimized as constraints, the sum of availability scores of all reachable paths is calculated based on the link availability matrix. The calculation results are recorded as path availability, and the paths are sorted in descending order of availability to obtain a candidate path table. Calculate the hop count and routing coupling degree of each path in the candidate path table; Construct a fitness function to determine the fitness score of each path in the candidate path table; The path with the highest fitness score is selected as the candidate backhaul path.

7. The wireless path optimization method based on artificial intelligence according to claim 6, characterized in that, The power ratio component, angle spread component, and environmental threat coefficient component in the feature matrix of each directed edge in the dynamic spatiotemporal graph are normalized respectively to obtain the power ratio normalization score, angle spread normalization score, and environmental threat coefficient normalization score. The original availability score is constructed based on the normalized power ratio component, angle spread component, and environmental threat coefficient.

8. The wireless path optimization method based on artificial intelligence according to claim 7, characterized in that, The routing coupling degree is the number of links shared by each path in the candidate path table with existing paths already occupied by other service flows; The reachable path is a path that satisfies the constraints. When constructing the fitness function, the path availability score, hop count and routing coupling degree of each path in the candidate path table are normalized by the maximum and minimum values ​​respectively. Setting: Fitness score = Normalized path availability score - Normalized hop count - Normalized route coupling.

9. An artificial intelligence-based wireless path optimization device, applied to the artificial intelligence-based wireless path optimization method as described in any one of claims 1-8, characterized in that, include: The data acquisition unit is used to integrate environmental sensing sensors on the IAB node, and simultaneously collect radio frequency parameters and environmental point cloud data within the millimeter-wave phased array antenna sector of the node, and stamp the two types of data with a unified timestamp. The state determination unit is used to mark obstructions from point cloud data and extract the power angle spectrum features of link multipath clusters from radio frequency parameters to generate the obstruction state vector of each link of the IAB node. The feature extraction unit is used to construct a dynamic spatiotemporal graph based on the occlusion state vector of each IAB node, with IAB nodes as vertices, link state vectors as edge features, and the remaining cache queue depth of the node as vertex features. The service marking unit is used to construct a link availability matrix based on the dynamic spatiotemporal graph. For each path carrying a service flow, it calculates the vulnerability index of the backhaul path based on its current hop count and the availability of each link on the backhaul path, and marks the service flow to be optimized. The path optimization unit is used to generate candidate backhaul paths for the marked service flows to be optimized based on the link availability matrix in response to the marked service flows. The path switching unit is used to distribute candidate backhaul paths to each IAB node to perform path switching.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program is used to control the electronic device on which the computer-readable storage medium resides to perform the artificial intelligence-based wireless path optimization method according to any one of claims 1-8 during runtime.