A Channel Resource Scheduling Method for Asynchronous Cooperative Positioning Networks

By constructing a hypergraph neural network to optimize resource block scheduling, the problem of high computational overhead in asynchronous cooperative positioning networks is solved, achieving faster scheduling and more efficient resource utilization, thus improving positioning and communication performance.

CN123124102APending Publication Date: 2026-11-17BEIJING INST OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611367316.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-04
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In asynchronous cooperative positioning networks, existing resource scheduling methods have high computational overhead, making it difficult to meet real-time scheduling requirements and failing to balance communication and positioning performance.

Method used

By constructing a hypergraph neural network, information is aggregated using the hypergraph neural network to generate link selection indicators, optimize resource block scheduling, and reduce redundant searches and iterative calculations.

Benefits of technology

It improves the positioning and communication performance of the network, reduces computation time, is suitable for network scenarios where the number of nodes and resource blocks changes frequently, and improves the efficiency of channel resource utilization.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123124102A_ABST
    Figure CN123124102A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of mobile communication and navigation positioning technology, specifically relating to a channel resource scheduling method for asynchronous cooperative positioning networks. The method includes: based on the current network state and resource block information, using directed cooperative ranging links on resource blocks as hypergraph nodes, constructing hyperedges according to the resource sharing relationship of links within the same resource block, the half-duplex conflict relationship of links formed by the same node, the cumulative positioning information relationship of the same receiving node, and the geometric direction relationship between transmitting and receiving nodes, thus forming a hypergraph structure oriented towards resource scheduling; inputting the initial features of the hypergraph nodes and a hypergraph association matrix constructed from multiple types of hyperedge sets into a hypergraph neural network, and using the hypergraph neural network to aggregate information between hypergraph nodes and hyperedges to obtain the scheduling score of the directed links in each resource block; and generating resource block scheduling results based on the scheduling scores of the directed links in each resource block and the link positioning benefits.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of mobile communication and navigation positioning technology, and specifically relates to a channel resource scheduling method for asynchronous cooperative positioning networks. Background Technology

[0002] High-precision position awareness is a fundamental capability in mobile ad hoc networks such as spacecraft networks, drone networks, and vehicle-to-everything (V2X) networks. In these networks, typically only the positions of some nodes are known and can serve as anchor nodes to provide positioning references; the remaining nodes need to estimate their own positions based on wireless measurement information. When the number of anchor nodes is limited, their geometric distribution is not ideal, or link conditions are constrained, relying solely on anchor node information often fails to yield stable positioning results. To improve positioning reliability, nodes with unknown positions can obtain inter-node measurement information through cooperative ranging, which, along with anchor node reference information, is used to calculate the positions of multiple nodes. Since it is usually difficult for nodes in a network to maintain strict global clock synchronization, the aforementioned inter-node cooperative ranging needs to be completed under asynchronous conditions; this type of positioning method is commonly referred to as asynchronous cooperative positioning.

[0003] Asynchronous cooperative positioning requires acquiring cooperative measurement information through inter-node ranging links, which consume limited wireless channel resources. For a single ranging link, bandwidth, transmit power, and propagation distance affect the quality of ranging information; for the entire positioning network, the direction of different links and the distribution of receiving nodes affect the positioning geometry. In practical networks, available resource blocks, total bandwidth, and transmit power are all limited, and ordinary nodes within the same resource block must also satisfy half-duplex constraints, preventing them from arbitrarily participating in multiple ranging links simultaneously. Therefore, resource scheduling is necessary to determine the selection of inter-node ranging links, resource block occupancy, and corresponding resource allocation methods. Different scheduling results will change the amount of ranging information and the positioning geometry distribution in the network, thus affecting communication performance and overall positioning accuracy.

[0004] Existing resource scheduling methods typically employ exhaustive search, greedy algorithms, alternating optimization, convex approximation, or genetic algorithms to solve link selection and resource allocation schemes. While these methods can improve cooperative localization performance to some extent, they usually require searching for a large number of candidate link combinations or iterative calculations. As the number of ordinary nodes and resource blocks increases, or as the network topology changes frequently, the number of candidate link combinations grows rapidly, leading to significant computational overhead and making it difficult to meet the real-time scheduling requirements of asynchronous cooperative localization networks. Therefore, it is necessary to research a resource scheduling method for integrated communication and localization networks that is independent of initial values, has a fast convergence speed, and can simultaneously balance communication and localization performance.

[0005] This invention is a technical invention based on the above-mentioned problems. As for the content of this invention, there are no reports at home and abroad. Summary of the Invention

[0006] In view of this, the present invention provides a channel resource scheduling method for asynchronous cooperative positioning networks. This method can select more suitable resource blocks for ranging links between ordinary nodes when channel resources are limited, making ranging communication between nodes more reasonable, thereby improving the positioning and communication effects of the network.

[0007] The technical solution for implementing the present invention is as follows: A channel resource scheduling method for asynchronous cooperative positioning networks includes: Based on the current network status and resource block information, directed cooperative ranging links on resource blocks are used as hypergraph nodes. Hyperedges are constructed according to the resource sharing relationship of links within the same resource block, the half-duplex conflict relationship of links formed by the same node, the cumulative relationship of the location information of the same receiving node, and the geometric direction relationship between the sending and receiving nodes, forming a hypergraph structure oriented towards resource scheduling. The initial features of hypergraph nodes and the hypergraph association matrix constructed from the sets of various types of hyperedges are input into the hypergraph neural network. The hypergraph neural network aggregates information between hypergraph nodes and hyperedges to obtain the scheduling score of directed links in each resource block. The scheduling score and link positioning benefit of the directed links in each resource block are fused to obtain the link selection index. Based on the link selection index, feasible links are selected to be added to the corresponding resource block. Under the constraint that nodes in the same resource block do not occupy the same resource block repeatedly, the resource block scheduling result is generated. The scheduling result is used to reasonably arrange the cooperative ranging links between ordinary nodes under the condition of limited channel resources, so as to improve network communication performance and positioning performance.

[0008] Optionally, the initial features of the hypergraph nodes in this invention are as follows: For ordinary nodes in the network, all directed cooperative ranging links are enumerated to obtain the geometric features of the directed links; for resource blocks in the network, a normalized resource block index is constructed to form resource block features, and these features are mapped to obtain resource block embeddings used to distinguish different resource blocks; each resource block embedding and the geometric features of each directed link are concatenated to obtain the directed cooperative ranging link representation on the resource block. As a hypergraph node; directed link representation The features are mapped to serve as the initial features of the hypergraph nodes.

[0009] Optionally, the specific process for obtaining the geometric features of directed links in this invention is as follows: for ordinary nodes in the network, enumerate all directed cooperative ranging links; for any directed link... Obtain the embedded data of the sending node respectively. and receiving node embedding and will , The difference between the two Element-wise product of the two and normalized link distance By splicing the links together, the geometric features of the directed link are obtained.

[0010] Optionally, the specific process of resource block embedding acquisition according to the present invention is as follows: for the first For each resource block, construct a normalized resource block index.

[0011] in, The total number of resource blocks; and the characteristics of the resource blocks are formed. The resource block features are input into the embedding mapping network to obtain resource block embeddings used to distinguish different resource blocks.

[0012] Optionally, the present invention describes information aggregation between hypergraph nodes and hyperedges using a hypergraph neural network, specifically as follows: After performing a linear transformation on the hypergraph node features, the node features are aggregated into hyperedges and normalized according to the hyperedge degree matrix. The hyperedge messages are then back to the hypergraph nodes according to the hyperedge weights and normalized according to the node degree matrix. The hypergraph node features are then normalized, nonlinearly activated, and randomly deactivated, and residual connections are made with the original hypergraph node features to update the hypergraph node features.

[0013] Optionally, the present invention inputs the hypergraph node features after information aggregation into a scoring network to obtain the corresponding scalar scores, and rearranges all the scalar scores of the hypergraph nodes according to the resource block index and the directed link index to obtain a scoring matrix. Each row of the scoring matrix corresponds to a resource block, and each column corresponds to a non-self-loop directed link. The scoring matrix serves as the output of the hypergraph neural network.

[0014] Optionally, this invention fuses the scheduling score and link location benefit of the directed links in each resource block to obtain a link selection index. The specific process is as follows: Calculate candidate link gain: For candidate links By comparing the changes in SPEB before and after adding the link, the link positioning benefit can be obtained. The scoring matrix output by the hypergraph neural network is organized into a directed link scoring matrix corresponding to each resource block. By fusing the hypergraph neural network scoring matrix with the link localization benefit, a link selection index is obtained. : .

[0015] Optionally, during training, the hypergraph neural network model of the present invention constructs a supervised loss function, which includes link selection loss. Link quantity constraint loss Ranking loss and half-duplex conflict soft constraint loss The total loss is expressed as:

[0016] in, This is used to constrain the model to determine whether a candidate directed link on each resource block is selected. Used to constrain the number of links selected in each resource block. Used to improve the score of selected links relative to unselected links. Used to prevent the same ordinary node from participating in multiple links simultaneously within the same resource block. , , These are the weighting coefficients; Based on total losses parameters of the hypergraph neural network model Perform backpropagation and iterative updates:

[0017] in, For learning rate, This represents the gradient of the loss function with respect to the model parameters.

[0018] Beneficial effects Compared to existing technologies, this invention can select more suitable resource blocks for ranging links between ordinary nodes under limited channel resources, making ranging communication between nodes more rational and thus improving the positioning and communication performance of the network. Compared to traditional methods such as exhaustive search, alternating optimization, and genetic algorithms, this invention uses a trained hypergraph neural network to directly provide link scores and then generates scheduling results based on the scores, reducing a large amount of repetitive search and iterative calculation, thus resulting in faster calculation speed and better real-time performance. At the same time, the scheduling results obtained by this invention can better meet constraints such as power, bandwidth, and half-duplex, reduce the occupation of resources by invalid links, improve the utilization efficiency of limited channel resources, and are applicable to network scenarios where the number of nodes and the number of resource blocks change frequently. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the 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.

[0020] Figure 1 This is a schematic diagram of resource block scheduling constraints provided by the present invention; Figure 2The flowchart illustrates the channel resource scheduling method for asynchronous cooperative positioning networks provided by this invention. Detailed Implementation

[0021] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0022] It should be noted that, in the absence of conflict, the following embodiments and features can be combined with each other; and, based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0023] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0024] The asynchronous cooperative positioning network involved in this invention is described as follows: Assume the network includes K ordinary nodes with unknown locations and A anchor nodes with known locations. The system's available channel resources are divided into R resource blocks. Anchor nodes provide positioning reference information and do not participate in resource scheduling among ordinary nodes; ordinary nodes obtain inter-node measurement information through directed cooperative ranging links. Multiple directed cooperative ranging links are allowed to be scheduled within the same resource block, but the same ordinary node can participate in at most one link to satisfy the half-duplex constraint.

[0025] This invention discloses a channel resource scheduling method for asynchronous cooperative localization (ALO) networks, comprising a model training phase and a resource scheduling inference phase. The model training phase obtains hypergraph neural network parameters, enabling it to output directed link scores on each resource block based on the state of the ALO network. The resource scheduling inference phase utilizes the trained hypergraph neural network to generate resource block scheduling results for the current scenario.

[0026] The model training phase includes the following steps: Step 1: Import the dataset and construct training samples Import an existing labeled dataset, which includes system parameters, network status information, and directed link labels for resource blocks. The system parameters include the number of ordinary nodes, the number of anchor nodes, the number of resource blocks, power constraints, and bandwidth constraints. The network status information includes the distance matrix between ordinary nodes, the direction vector between ordinary nodes, the distance matrix from ordinary nodes to anchor nodes, and the direction vector from ordinary nodes to anchor nodes. The directed link labels for resource blocks are used to indicate the selected directed cooperative ranging links on each resource block.

[0027] Step 2: Construct ordinary node geometric encoding, i.e., node embedding representation; Step 201: Read the network state data from the training samples.

[0028] Step 202: Construct ordinary node features. Based on the distance matrix and direction vector input in step 201, construct a node feature vector for each ordinary node. The node feature vector includes distance statistical features, direction statistical features, and node index features.

[0029] Step 203: Generate ordinary node embedding representations. Input the ordinary node features obtained in step 202 into the node encoder, and perform weighted message passing by combining the distance between ordinary nodes to obtain the node embedding representation of each ordinary node.

[0030] Step 3: Construct initial features for hypergraph nodes Step 301: Enumerate non-self-loop directed links. For Given a set of ordinary nodes, enumerate all nodes that satisfy the condition. Directed cooperative ranging link This forms a set of directed links. The set of directed links contains a total of 10 candidate links.

[0031] Step 302: Construct the geometric features of the directed link. For any directed link... Obtain the embedded data of the sending node respectively. and receiving node embedding and will , The difference between the two Element-wise product of the two and normalized link distance By splicing the links together, the geometric features of the directed link are obtained.

[0032] Step 303: Construct resource block encoding. For the first... For each resource block, construct a normalized resource block index.

[0033] And form resource block characteristics The resource block features are input into a trainable embedding mapping network to obtain resource block embeddings that distinguish different resource blocks. This mapping mechanism is set up mainly to allow the network to remember the location of each resource block.

[0034] Step 304: Construct initial features for hypergraph nodes. For each resource block... and every directed link The directed link geometric features obtained in step 302 are concatenated with the resource block embeddings obtained in step 303 to obtain the directed link representation on the resource block. Subsequently, the directed link is represented. The input consists of a feature mapping layer composed of linear transformation, normalization, and nonlinear activation, which yields the initial features of the hypergraph nodes.

[0035] Step 4: Construct hyperedges, incidence matrices, and weight matrices Based on the initial features of the hypergraph nodes obtained in step 3, multiple types of hyperedges are constructed according to resource block sharing relationships, half-duplex conflict relationships, receiver information accumulation relationships, sender role relationships, and geometric direction relationships. The hypergraph association matrix is ​​then generated from these hyperedges. and hyperedge weight matrix .

[0036] Step 401: Define hypergraph nodes. For each resource block... directed links on express As a hypergraph node, based on the above link structure, the total number of hypergraph nodes is:

[0037] Step 402: Construct a shared hyperedge for resource blocks. For each resource block... Construct a resource block sharing hyperedge .

[0038] Step 403: Construct half-duplex conflicting hyperedges. For each resource block and each ordinary node Constructing a half-duplex conflict hyperborder .

[0039] Step 404: Construct the receiver information accumulation superedge. For each receiver node... Construct a super-edge for accumulating information at the receiving end. .

[0040] Step 405: Construct the sender role hyperedge. For each sender node... Constructing a hyperedge for the sending role .

[0041] Step 406: Construct geometrically oriented hyperedges. Calculate candidate links. The geometric orientation angle, and according to the receiving node and direction sector Candidate scheduling actions are categorized into geometric direction hyperedges. This is used to characterize the geometric correlation between links with similar orientations.

[0042] Step 407: Construct the association matrix. Construct the hypergraph association matrix based on the multi-type hyperedge set. The rows of the association matrix correspond to candidate directed cooperative link nodes, and the columns correspond to hyperedges. When a candidate directed cooperative link node belongs to a hyperedge, the corresponding element in the association matrix is ​​set to 1; otherwise, it is set to 0.

[0043]

[0044] in, Indicates a candidate directed cooperative link node. Indicates the super-edge.

[0045] Step 408: Construct the hyperedge weight matrix. Assign corresponding hyperedge weights to various types of hyperedges, and construct the hyperedge weight matrix based on these weights. Each element in the hyperedge weight vector corresponds to a hyperedge, which is used to characterize the degree of influence of the hyperedge in the hypergraph information propagation process.

[0046] Step 5: Perform Hypergraph message passing The hypergraph incidence matrix obtained in step 4 The initial features of the hypergraph nodes obtained in step 3 are input into the hypergraph neural network, and the hypergraph node representations are updated through the propagation method from nodes to hyperedges and from hyperedges to nodes.

[0047] Step 501: Perform a linear transformation on the hypergraph node features:

[0048] in, For the hypergraph neural network, the first The trainable parameters of the layer, This represents the characteristics of a hypergraph node.

[0049] Step 502: Aggregate node information to hyperedges and sort them according to the hyperedge degree matrix. Normalize:

[0050] in, The hypergraph incidence matrix; Step 503: Return the hyperedge messages to the hypergraph nodes according to the hyperedge weights, and then return them according to the node degree matrix. Normalize:

[0051] in, This is the hyperedge weight matrix; Step 504: Normalize, nonlinearly activate, and randomly deactivate the node messages, and perform residual connections with the original node features:

[0052] in, Representation layer normalization, Represents a non-linear activation function. This represents the random deactivation operation during the training phase.

[0053] Therefore, the hypergraph message passing layer transmits the resource sharing relationship, half-duplex conflict relationship, receiver information accumulation relationship, sender role relationship and geometric direction relationship within the same hyperedge to the relevant hypergraph nodes, and obtains the updated hypergraph node representation, that is, the updated hypergraph node features.

[0054] Step 6: Output directed link score Input the hypergraph node representation obtained in step 5 into the scoring network to obtain the scalar score corresponding to each hypergraph node.

[0055] Step 601: Calculate the hypergraph node score. Input each hypergraph node representation into a scoring network consisting of linear transformation, nonlinear activation, random deactivation, and linear transformation to obtain the corresponding scalar score.

[0056] Step 602: Form a resource block rating matrix. Rearrange the ratings of all hypergraph nodes according to their resource block indices and directed link indices, resulting in a matrix of shape [formula missing]. rating matrix Each row corresponds to a resource block, and each column corresponds to a non-self-loop directed link. The scoring matrix serves as the output of the hypergraph neural network.

[0057] Step 7: Train the hypergraph neural network model and save the model parameters Step 701: Construct the supervised loss function. Based on the resource block directed link scoring matrix obtained in Step 6. And obtain the supervision label according to step 1. The model training loss is constructed. The training loss includes link selection loss, link quantity constraint loss, ranking loss, and half-duplex conflict soft constraint loss. The total loss is expressed as:

[0058] in, This is used to constrain the model to determine whether a candidate directed link on each resource block is selected. Used to constrain the number of links selected in each resource block. Used to improve the score of selected links relative to unselected links. Used to prevent the same ordinary node from participating in multiple links simultaneously within the same resource block. , , These are the weighting coefficients. During training, they are determined based on the total loss. Update the model parameters and save the model weights that meet the validation conditions.

[0059] Step 702: Update the hypergraph neural network parameters and save the model weights. Based on the total loss... parameters of the hypergraph neural network model Perform backpropagation and iterative updates:

[0060] in, For learning rate, This represents the gradient of the loss function with respect to the model parameters. After training, the model weights that meet the preset validation conditions are saved for resource block scheduling in the actual scheduling phase.

[0061] After the above training, the hypergraph neural network can output a scheduling score for the directed links of each resource block based on the current network state. Based on this score, resource block scheduling decoding is further performed to obtain the actual scheduling scheme. The specific process of the inference phase is as follows: Step 8: Perform model inference. Load the trained hypergraph neural network model and input the current network state and resource block information into the model to obtain a shape of... The directed link scoring matrix. Specifically: Based on the current network status and resource block information, directed cooperative ranging links on resource blocks are used as hypergraph nodes. Hyperedges are constructed according to the resource sharing relationship of links within the same resource block, the half-duplex conflict relationship of links formed by the same node, the cumulative relationship of the location information of the same receiving node, and the geometric direction relationship between the sending and receiving nodes, forming a hypergraph structure oriented towards resource scheduling. The initial features of hypergraph nodes and the hypergraph association matrix constructed from the sets of various types of hyperedges are input into the hypergraph neural network. The hypergraph neural network aggregates information between hypergraph nodes and hyperedges to obtain the scheduling score of directed links in each resource block. Step 9: Link scheduling decoding. Based on the directed link scoring matrix obtained in Step 8, and combined with the positioning performance evaluation and resource constraints, the final resource block scheduling result is generated.

[0062] Step 901: Organize the candidate link scores output by the hypergraph neural network in Step 8 into a directed link score matrix corresponding to each resource block:

[0063] in, Indicates the resource block number. Indicates from node Pointing to node Candidate directed links, Indicates the first Nodes on each resource block To the node Priority rating.

[0064] Step 902: Calculate the location information gain. Based on the distance, direction, and anchor point information in the current environment, construct the initial equivalent Fisher information matrix for each node:

[0065] For candidate links Calculate the incremental positioning information for receiving node j per unit power, and use SPEB as the positioning performance evaluation index:

[0066] Step 903: Calculate the candidate link gain. For the candidate links... By comparing the changes in SPEB before and after adding this link, the link positioning benefits can be obtained:

[0067] like This indicates that the link can improve positioning performance.

[0068] Step 904: Generate the final scheduling result. The hypergraph neural network score and localization reward are fused to obtain the link selection index:

[0069] in, This represents the weighting factor obtained from the hypergraph neural network scoring. The feasible link with the highest score is selected. The resource is added to the corresponding resource block, while ensuring that no node within the same resource block occupies the same resource block repeatedly. The final resource block scheduling result is then obtained. :

[0070] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A channel resource scheduling method for asynchronous cooperative positioning networks, characterized in that, include: Based on the current network status and resource block information, directed cooperative ranging links on resource blocks are used as hypergraph nodes. Hyperedges are constructed according to the resource sharing relationship of links within the same resource block, the half-duplex conflict relationship of links formed by the same node, the cumulative relationship of the location information of the same receiving node, and the geometric direction relationship between the sending and receiving nodes, forming a hypergraph structure oriented towards resource scheduling. The initial features of hypergraph nodes and the hypergraph association matrix constructed from the sets of various types of hyperedges are input into the hypergraph neural network. The hypergraph neural network aggregates information between hypergraph nodes and hyperedges to obtain the scheduling score of directed links in each resource block. The scheduling score and link location benefit of the directed links in each resource block are fused to obtain the link selection index. Based on the link selection index, feasible links are selected to be added to the corresponding resource block. Under the constraint that nodes in the same resource block do not occupy the same space repeatedly, the resource block scheduling result is generated.

2. The channel resource scheduling method for asynchronous cooperative positioning networks according to claim 1, characterized in that, The initial features of the hypergraph nodes are as follows: For ordinary nodes in the network, all directed cooperative ranging links are enumerated to obtain the geometric features of the directed links; for resource blocks in the network, a normalized resource block index is constructed to form resource block features, and these features are mapped to obtain resource block embeddings used to distinguish different resource blocks; each resource block embedding and the geometric features of each directed link are concatenated to obtain the directed cooperative ranging link representation on the resource block. As a hypergraph node; directed link representation The features are mapped to serve as the initial features of the hypergraph nodes.

3. The channel resource scheduling method for asynchronous cooperative positioning networks according to claim 2, characterized in that, The specific process for obtaining the geometric features of a directed link is as follows: For a regular node in the network, enumerate all directed cooperative ranging links; for any directed link... Obtain the embedded data of the sending node respectively. and receiving node embedding and will , The difference between the two Element-wise product of the two and normalized link distance By splicing the links together, the geometric features of the directed link can be obtained.

4. The channel resource scheduling method for asynchronous cooperative positioning networks according to claim 2, characterized in that, The specific process for obtaining the resource block embedding is as follows: For the first... For each resource block, construct a normalized resource block index. in, The total number of resource blocks; and the characteristics of the resource blocks are formed. The resource block features are input into the embedding mapping network to obtain resource block embeddings used to distinguish different resource blocks.

5. The channel resource scheduling method for asynchronous cooperative positioning networks according to claim 2, characterized in that, The process of aggregating information between hypergraph nodes and hyperedges using a hypergraph neural network specifically involves: After performing a linear transformation on the hypergraph node features, the node features are aggregated into hyperedges and normalized according to the hyperedge degree matrix. The hyperedge messages are then back to the hypergraph nodes according to the hyperedge weights and normalized according to the node degree matrix. The hypergraph node features are then normalized, nonlinearly activated, and randomly deactivated, and residual connections are made with the original hypergraph node features to update the hypergraph node features.

6. The channel resource scheduling method for asynchronous cooperative positioning networks according to claim 5, characterized in that, The aggregated hypergraph node features are input into the scoring network to obtain the corresponding scalar scores. The scalar scores of all hypergraph nodes are rearranged according to the resource block index and the directed link index to obtain the scoring matrix. Each row of the scoring matrix corresponds to a resource block, and each column corresponds to a non-self-loop directed link. The scoring matrix serves as the output of the hypergraph neural network.

7. The channel resource scheduling method for asynchronous cooperative positioning networks according to claim 6, characterized in that, The link selection index is obtained by fusing the scheduling score and link location benefit of the directed links in each resource block. The specific process is as follows: Calculate candidate link gain: For candidate links By comparing the changes in SPEB before and after adding the link, the link positioning benefit can be obtained. The scoring matrix output by the hypergraph neural network is organized into a directed link scoring matrix corresponding to each resource block. By fusing the hypergraph neural network scoring matrix with the link localization benefit, a link selection index is obtained. : 。 8. The channel resource scheduling method for asynchronous cooperative positioning networks according to claim 5, characterized in that, During training, the hypergraph neural network model constructs a supervised loss function, which includes link selection loss. Link quantity constraint loss Ranking loss and half-duplex conflict soft constraint loss The total loss is expressed as: in, This is used to constrain the model to determine whether a candidate directed link on each resource block is selected. Used to constrain the number of links selected in each resource block. Used to improve the score of selected links relative to unselected links. Used to prevent the same ordinary node from participating in multiple links simultaneously within the same resource block. , , These are the weighting coefficients; Based on total losses parameters of the hypergraph neural network model Perform backpropagation and iterative updates: in, For learning rate, This represents the gradient of the loss function with respect to the model parameters.