A low-orbit satellite network user terminal batch switching method

Abstract

This application provides a method for batch handover of user terminals in a low-Earth orbit satellite network, comprising: a source satellite-borne base station sending measurement indicators of multiple user terminals entering its edge radii to multiple neighboring satellites to a satellite-borne decision center; the satellite-borne decision center inputting the measurement indicators into a packet handover decision model, outputting packet handover decision results, and returning the packet handover decision results to the source satellite-borne base station; the source satellite-borne base station determining the target neighboring satellites to which the multiple user terminals should handover based on the packet handover decision results and initiating batch handover request messages; and each target neighboring satellite, based on the batch handover request messages, cooperating with the source satellite-borne base station and each user terminal to complete the service handover. The dedicated signaling process for batch handover and the accompanying signaling compression method designed above can significantly reduce the transmission pressure of inter-satellite links. Furthermore, by using a deep learning model to consider the complex states and constraints of multiple users and multiple satellites for joint decision-making, it can more accurately select the satellites for handover.

Description

Technical Field

[0001] This invention relates to the field of satellite communication technology, and more specifically, to a method for batch switching of user terminals in a low-Earth orbit satellite network. Background Technology

[0002] Low Earth Orbit (LEO) communication satellites are increasingly becoming an important component of mobile communications due to their wide coverage, low latency, and high throughput. However, because LEO satellites move at high speeds relative to the ground (up to 7.6 kilometers per second), meaning each satellite provides service to user equipment (UE), a large number of satellites must be deployed to continuously provide service to UEs on the ground in order to ensure service continuity. This process inevitably generates a large amount of handover signaling. Furthermore, since LEO satellites serve UEs in units of wavebands (approximately 15km), during ephemeris updates, all user terminals at a given waveband may simultaneously initiate handover. Therefore, it is essential to consider how to solve the handover problem from an architectural perspective.

[0003] Due to the limited scale of current satellite network construction and the lack of uniformity in the size and topology of satellite constellations among various operators, 3GPP's current description of 5G non-terrestrial networks (NTN) does not systematically discuss solutions for the large number of handovers caused by the rapid movement of LEO satellites relative to the ground.

[0004] The current mainstream solution adopts conditional handover, which is a standardized preventive handover procedure defined in 3GPP. The network reserves resources for the UE in neighboring stations in advance, and the handover is automatically performed when the UE measurement report meets the conditions. However, this solution requires reserving resources for multiple UEs on multiple target satellites in advance, resulting in a huge waste of satellite resources. Summary of the Invention

[0005] In view of this, the present invention provides a batch handover method for user terminals in low-Earth orbit satellite networks. By designing a dedicated signaling process for batch handover and a corresponding signaling compression method, it can significantly reduce the transmission pressure of inter-satellite links. At the same time, by using a deep learning model to consider the complex states and constraints of multiple users and multiple satellites for joint decision-making, it can more accurately select the satellites for handover.

[0006] A method for batch handover of satellite network user terminals according to an embodiment of the present invention includes: The source satellite base station identifies multiple user terminals that have entered its edge bands and sends the measurement indicators of these multiple user terminals for multiple neighboring satellites to the satellite decision center. After receiving the measurement indicators, the onboard decision center inputs the measurement indicators into a pre-trained packet handover decision model, which outputs a packet handover decision result. The onboard decision center then returns the packet handover decision result to the source onboard base station. The packet handover decision result includes recommendations for one or more neighboring satellites to which the multiple user terminals should handover. The source satellite base station determines one or more target neighboring satellites to which the multiple user terminals need to hand over based on the packet handover decision result, and initiates a batch handover request message to each of the one or more target neighboring satellites. The batch handover request message includes a targetCellGlobalID field for representing the global identifier of the target cell and a UEContextInfoHORequest field for representing the context information of the user terminal. The targetCellGlobalID field is the global identifier of the target cell of the multiple target cells of the target neighboring satellites that receive the batch handover request message, and the UEContextInfoHORequest field uses the index of the global identifier of the target cell to identify the target cell. The one or more target neighboring satellites, based on the batch handover request message, collaborate with the source satellite base station and the multiple user terminals to complete the service handover.

[0007] Furthermore, the one or more target neighboring satellites, based on the batch handover request message, collaborate with the source satellite-borne base station and the multiple user terminals to complete the service handover, including: The one or more target neighboring satellites reserve resources for the multiple user terminals according to the batch handover request message, and reply to the source satellite base station with a batch handover request confirmation message, carrying resource reservation completion information; After receiving the batch handover request confirmation message, the source satellite base station sends the Packet Data Convergence Protocol Sequence Number (PDCP) SN Status Mapping Table of multiple user terminals that are granted handover to the one or more target neighboring satellites, and issues a handover command to each user terminal that is granted handover. After receiving the handover command, the multiple user terminals that are granted handover permission initiate a random access procedure to their corresponding target neighboring satellite. During the handover process of the target neighboring satellite, the data forwarding process is completed with the core network to confirm the batch handover user plane path and send a batch user terminal release message to the source satellite base station. The source satellite-borne base station releases resources based on the batch user terminal release message.

[0008] Furthermore, the method also includes: the source satellite base station determining multiple user terminals that have entered its secondary edge position, and sending a Radio Resource Control (RRC) connection reconfiguration message to the multiple user terminals that have entered the secondary edge position, instructing the multiple user terminals that have entered the secondary edge position to shorten the measurement reporting cycle and include their location information in their measurement reports.

[0009] Furthermore, the batch handover request message also includes a GUAMI field, which includes multiple globally unique core network identifiers, and the core network is identified by the index of the core network identifier in the UEContextInfoHORequest field.

[0010] Furthermore, the UEContextInfoHORequest field includes a UESecurityCapabilities field, which includes a 4-bit binary index indicating the security capabilities of the user terminal. This 4-bit binary index indicates the 16 most frequently occurring combinations selected from the user terminal capability reporting messages {nea1-128, nea2-128, nea3-128} and {nia1-128, nia2-128, nia3-128} as integrity protection encryption algorithms.

[0011] Furthermore, the batch handover request message also includes a PDUsessionResourcesToBeSetup-List field indicating the PDU session resource establishment list. The PDUsessionResourcesToBeSetup-List field includes a Radio Bearer Configuration (RB) field. The RB field is used to indicate the template identifier and differential parameters used by the Data Radio Bearer (DRB). Each template identifier corresponds to a template, and each template includes a set of radioBearerConfig parameters. The radioBearerConfig parameter set includes Radio Link Control (RLC) mode, Packet Data Convergence Protocol (PDCP) configuration, and Logical Channel Configuration. The differential parameters include a dedicated drb-Identity and a logicalChannelIdentity.

[0012] Furthermore, the measurement indicators include the elevation angle of the multiple neighboring satellites observed by the multiple user terminals, the duration for which the multiple neighboring satellites can serve the multiple user terminals, the signal strength of the multiple neighboring satellites acquired by the multiple user terminals, the QoS level of the multiple user terminals, and the load status of the multiple neighboring satellites.

[0013] Furthermore, the group handover decision model adopts the standard Transformer encoder structure, and in the self-attention layer, the (UE_i, Sat_j) term interacts with all other terms in the sequence to compare the link quality of the same user terminal to different neighboring satellites, compare the competitive relationship of different user terminals to the same neighboring satellite, and capture complex global dependencies.

[0014] Furthermore, hard constraints are set through the attention mask of the attention layer. The hard constraints include: the neighboring satellite elevation angle is less than 30°, the RSRP is less than -90dB, the remaining service time of the target neighboring satellite is less than 10s, and the load of the neighboring satellite reaches 90%. Before calculating the attention weight, the attention logits corresponding to the (UE_i, Sat_j) tokens that violate the above hard constraints are set to a maximum negative value.

[0015] Furthermore, after completing the service handover, the multiple user terminals will, before or during the next handover, feed back the service quality and high-priority service satisfaction obtained on one or more target neighboring satellites to the onboard decision center in order to optimize the parameters of the group handover decision model.

[0016] According to the scheme of the embodiments of this application, by designing a dedicated signaling process for batch switching and a corresponding signaling compression method, the transmission pressure of inter-satellite links can be significantly reduced. At the same time, by using a deep learning model to consider the complex states and constraints of multiple users and multiple satellites for joint decision-making, the selection of switching satellites can be made more accurately. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of wave position division according to an embodiment of the present invention.

[0018] Figure 2 This is a flowchart of a method for batch switching of user terminals in a low-Earth orbit satellite network according to an embodiment of the present invention.

[0019] Figure 3 This is a signaling flowchart for batch handover of user terminals in a low-Earth orbit satellite network according to an embodiment of this application.

[0020] Figure 4 This is a schematic diagram of a batch switching request message according to an embodiment of this application.

[0021] Figure 5 This is a schematic diagram of the input data of a group handover decision model according to an embodiment of this application.

[0022] Figure 6 This is a schematic diagram of the output data of a group handover decision model according to an embodiment of this application. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the spirit of the content disclosed in this application will be clearly explained below with reference to the accompanying drawings and detailed description. After understanding the embodiments of this application, any person skilled in the art can make changes and modifications based on the technology taught in this application without departing from the spirit and scope of this application.

[0024] The illustrative embodiments and descriptions provided in this application are for explaining the application, but are not intended to limit the application. Furthermore, elements / components using the same or similar reference numerals in the drawings and embodiments are used to represent the same or similar parts.

[0025] The terms “first,” “second,” etc., used in this document are not intended to specifically refer to order or sequence, nor are they used to limit this application; they are merely used to distinguish elements or operations described using the same technical terms.

[0026] The terms “include,” “including,” “have,” “contain,” etc., used in this article are all open-ended terms, meaning that they include but are not limited to.

[0027] The term "and / or" as used herein includes any or all of the things mentioned.

[0028] The term "multiple" in this article includes "two" and "more than two"; the term "multiple groups" in this article includes "two groups" and "more than two groups".

[0029] Certain terms used to describe this application will be discussed below or elsewhere in this specification to provide additional guidance to those skilled in the art in describing the application.

[0030] This application provides a method for batch handover of user terminals in a low-Earth orbit satellite network. By designing a dedicated signaling process for batch handover and a corresponding signaling compression method, it can significantly reduce the transmission pressure on the inter-satellite interface. At the same time, by using a deep learning model to consider the complex states and constraints of multiple users and multiple satellites for joint decision-making, it can more accurately select the satellites for handover.

[0031] In one embodiment of this application, based on the relative positional relationship between the satellite beam position and the UE, the upcoming batch handover requirements are predicted and data collection configuration is performed in advance.

[0032] Figure 1 A schematic diagram of wave position division according to an embodiment of this application is shown. Figure 1As shown, the coverage area of the satellite is divided into multiple wave positions, identified as A to I in the figure. These wave positions are arranged in rows. For example, ABC, DEF, and GHI are each in a row, and the arrangement direction of the rows is approximately perpendicular to the satellite's moving direction. As the satellite moves, the positions of the wave positions relative to the satellite coverage edge change dynamically. The positions of the wave positions can be defined in three states: the central wave position, the sub-edge wave position, and the edge wave position. Specifically, the central wave position is located in the central area of the satellite coverage, and the UEs therein perform regular mobility management. The sub-edge wave position is the wave position adjacent to the edge wave position that is about to enter the coverage edge, and it is the prediction and preparation area for batch handovers. For the edge wave position, the UEs in the edge wave position of the current satellite coverage need to immediately perform handovers.

[0033] In one embodiment, assume that the satellite coverage area is circular with a radius of R; each wave position is approximately a circular or hexagonal shape of equal size, and its equivalent radius is r (or take half of the width of the wave position in the satellite moving direction as r). Taking the satellite's sub-satellite point (i.e., the center of the coverage area) as the reference point and the satellite's moving direction as the positive direction. Define the projection distance of the line connecting the center of the wave position to the center of the satellite in the opposite direction of the satellite's movement (i.e., the distance by which the center of the wave position lags behind the center of the satellite) as D. When the center of the wave position is exactly below the satellite, D = 0, and when it is behind the moving direction, D > 0. Based on the relationship between D, the coverage radius R, and the wave position radius r, the following judgments are made: When R - r < D < R, it is determined that the wave position is partially or entirely outside the coverage edge, and it is marked as the edge wave position.

[0034] When R - 2r < D <= R - r, it is determined that the wave position is adjacent to the edge, and it is marked as the sub-edge wave position.

[0035] When D < R - 2r, it is determined that the wave position is located in the central area of the coverage, and it is marked as the central wave position.

[0036] As Figure 1 shown, at time T1, GHI is the edge wave position; DEF is the sub-edge wave position; ABC is the relative central wave position. At this time, wave position C is the central wave position, and the UEs in this wave position perform regular mobility management. Specifically, the UEs in this wave position are all configured for regular location-based measurements, that is, the measurement event is the D2 event. In this stage, if a UE detects the D2 event, that is, the neighboring satellite is closer, it triggers the regular handover process.

[0037] At time T2, the service wave position information of the satellite changes: GHI moves out of the service range, DEF becomes the new edge wave position, and ABC becomes the new sub-edge wave position.

[0038] When wave position C becomes a sub-edge wave position, the handover preparation process of this application is triggered, and the base station begins to perform handover prediction and resource preparation for UEs within that wave position. Specifically, the network side sends Radio Resource Control (RRC) connection reconfiguration messages to all UEs within that wave position one by one. In the reconfiguration message: the UE's measurement reporting period is shortened from the normal period to a more frequent period, for example, adjusted from 1280 milliseconds to 200 milliseconds, so as to obtain channel state information more frequently; and the UE is instructed to include its location information in its measurement reports.

[0039] When the target waveform (such as waveform C) enters the edge waveform state, the network immediately triggers the batch packet handover process.

[0040] Specifically, such as Figure 2 As shown, in one embodiment of this application, the method for batch handover of user terminals in a low-Earth orbit satellite network includes: S21, the source satellite base station identifies multiple user terminals that have entered its edge position and sends the measurement indicators of multiple user terminals that have entered its edge position to multiple neighboring satellites to the satellite decision center. S22, after receiving the measurement indicators, the onboard decision center inputs the measurement indicators into the packet handover decision model, the packet handover decision model outputs the packet handover decision result, and returns the packet handover decision result to the source onboard base station. The packet handover decision result includes the recommendation result of one or more neighboring satellites to which multiple user terminals want to hand over. S23, the source satellite base station determines one or more target neighboring satellites to which multiple user terminals need to be handed over based on the packet handover decision results, and sends batch handover request messages to the one or more target neighboring satellites respectively; S24, the one or more target neighboring satellites, based on the batch handover request message, work together with the source satellite base station and the user terminal to complete the service handover.

[0041] Furthermore, one or more target neighboring satellites, in accordance with the batch handover request message, coordinate with the source satellite-borne base station and user terminal to complete the service handover, including: The one or more target neighboring satellites reserve resources for multiple user terminals according to the batch handover request message, and reply to the source satellite base station with a batch handover request confirmation message, carrying resource reservation completion information; After receiving the batch handover request confirmation message, the source satellite base station sends the Packet Data Convergence Protocol Sequence Number (PDCP) SN status mapping table of multiple user terminals that are granted handover to one or more target neighboring satellites, and issues a handover command to each user terminal that is granted handover. After receiving the handover command, multiple user terminals that have been granted handover permission initiate a random access procedure to their corresponding target neighboring satellite. During the handover process initiated by the target neighboring satellite, data forwarding is performed, and the access and mobility management function (AMF) confirms the user plane path for batch handover and sends batch user terminal release messages to the source satellite base station. The source satellite base station releases resources based on the batch user terminal release message.

[0042] The switching process will be described below using a specific example. Figure 3 As shown, the switching process includes: S1: The source satellite base station sends the measurement index information of each UE to each neighboring satellite through a handover decision request message to the satellite decision center located in medium Earth (MEO) orbit.

[0043] S2: After receiving the handover decision request message, the onboard decision center calculates the optimal packet handover decision result based on its deployed packet handover decision model, and returns the packet handover decision result to the source onboard base station through the handover decision response message.

[0044] S3: After obtaining the packet handover decision result, the source satellite base station sends batch handover request messages to each target neighboring satellite one by one.

[0045] S4: After receiving the batch handover request message, each target satellite-borne centralized unit (CU) sends a batch bearer context establishment request (BATCH BEARER CONTEXT SETUP REQUEST) message to its user plane (UP), carrying the data radio bearer (DRB) context and security context of multiple UEs, so as to reserve UP side resources at one time.

[0046] S5: Each target satellite UP creates bearers for multiple UEs and replies to its CU with a BATCH BEARER CONTEXT SETUP RESPONSE message, carrying information such as the tunnel endpoint identifier (TEID) of the newly created DRB.

[0047] S6: Each target satellite-borne CU simultaneously sends a BATCH UECONTEXT SETUP REQUEST message to its distribution unit (DU), carrying the DRB context and security context, in order to reserve DU-side resources in one go.

[0048] S7: Each target satellite DU creates contexts for multiple UEs and assigns cell radio network temporary identifiers (C-RNTI) to these UEs, and then replies to its CU with a BATCH UE CONTEXT SETUP RESPONSE message.

[0049] S8: Each target satellite's onboard CU, based on the response returned by the DU (obtaining the DU-side TEID of the newly created DRB), assembles a BATCH BEARER MODIFICATION REQUEST message and notifies its UP of the DU-side data plane tunnel information.

[0050] S9: Each target satellite UP, based on the previously set downlink tunnel endpoint, replies to its CU with a BATCH BEARER MODIFICATION RESPONSE message.

[0051] S10: The target onboard CU replies to the source CU via the Xn link with a BATCH HANDOVERREQUEST ACKNOWLEDGE message, carrying information on resource reservation completion.

[0052] S11: After receiving the response, the source onboard CU decides to issue a handover command based on the resource reservation results of each UE. For UEs that are granted handover, the source CU sends a BATCH BEARER CONTEXTMODIFICATION REQUEST message to its UP to obtain source-side tunnel information for data forwarding.

[0053] S12: While sending a request to the UP, the source CU sends an RRC reconfiguration message (i.e., a handover command) to each UE that is granted handover permission.

[0054] S13: The source UP replies with a BATCH BEARER CONTEXTMODIFICATION RESPONSE message, informing the source CU of information such as the Packet Data Convergence Protocol (PDCP) sequence number (SN) cached within it.

[0055] S14: After receiving the handover command, the UE initiates a random access procedure to the target satellite.

[0056] S15: The source satellite CU sends a BATCH SN STATUS TRANSFER message to the target satellite CU, carrying the PDCP SN status mapping table of the multiple UEs to be switched.

[0057] S16: After receiving the PDCP SN list, the target satellite CU initiates data forwarding during the handover period.

[0058] S17: After the data forwarding is completed, the target satellite CU sends a BATCH PATHSWITCH REQUEST message to the core network to request a batch switching of user plane paths.

[0059] S18: AMF replies with a BATCH PATH SWITCH REQUEST ACK message, confirming the path switch.

[0060] S19: The target satellite CU sends a BATCH UE RELEASE message to the source CU to release the source satellite resources.

[0061] S20: The source satellite CU receives the batch UE release message and sends batch context (bearer) release requests to the DU and CU in sequence to release local resources. At this point, the handover is complete.

[0062] Those skilled in the art will understand that some steps in the above process can be combined or eliminated under certain conditions, such as co-location of onboard CU / DU or the use of a tighter interface. For example, resource reservation requests to UP and DU may be sent in a combined manner.

[0063] As mentioned above, due to the large amount of UE context carried by a single signaling message during batch UE handover, problems such as excessive data volume, complex encoding and decoding, and decreased transmission reliability can occur. In one embodiment of this application, a method combining classification merging and multiple compression is used in the Xn interface signaling to solve the above problems.

[0064] The biggest challenge in batch handover is that transmitting a large number of UE contexts simultaneously can lead to large data volumes in a single signaling message, high encoding / decoding overhead, and increased risk of data transmission packet loss. Therefore, in this embodiment, the UE contexts are categorized and merged in the Xn signaling for subsequent batch handovers. Several data compression methods are employed to minimize the data volume of a single batch handover signaling message, such as a "dictionary + index" method or bit compression. This embodiment uses the batch handover request message, which has the largest number of contexts, as an example to illustrate the batch handover signaling data compression method. Of course, other messages transmitted via Xn can also be compressed using a similar method to the batch handover request message.

[0065] In one embodiment of this application, each batch Xn handover signaling process can be constrained to contain a maximum of 16 UE contexts.

[0066] In one embodiment of this application, the batch handover request message includes a targetCellGlobalID field for representing the global identifier of the target cell and a UEContextInfoHORequest field for representing the user terminal context information. The targetCellGlobalID field includes the global identifier of the target cell of a plurality of target neighboring satellites that receive the batch handover request message, and the UEContextInfoHORequest field uses the index of the global identifier of the target cell to identify the target cell.

[0067] Specifically, assuming each base station has a maximum of eight cells, and the global identifier of each cell is represented by 1 byte, the target cell set can be represented by 8 bytes at the very beginning of the message body. Then, in the subsequent list section, such as in the UEContextInfoHORequest field representing user terminal context information, only the index is used to identify which cell the UE belongs to.

[0068] In one embodiment, the index can be the location index of each target cell's global identifier in the targetCellGlobalID field. If each index occupies 3 bits, 16 UEs will occupy a total of 4.5 bits. Thus, using this "dictionary + index" approach can significantly reduce the data volume of batch handover request messages.

[0069] Batch handover request messages can also include a GUAMI field. The GUAMI field includes multiple globally unique core network identifiers. The UEContextInfoHORequest field uses the index of the core network identifier to identify the corresponding core network. Specifically, similar to the approach used for targetCellGlobalID above, assuming each base station has a maximum of eight cells, and each cell's AMF identifier is represented by 1 byte, the AMF identifier set can be represented in the message body's front end using 8 bytes. Then, in the subsequent list section, such as in the UEContextInfoHORequest field representing user terminal context information, only the index (also 4.5 bytes) is needed to identify which AMF the UE belongs to. In one embodiment, this index can be the position index of each core network identifier in the GUAMI field. Thus, using this "dictionary + index" approach can significantly reduce the data volume of the batch handover request message.

[0070] The UEContextInfoHORequest information element may also include a UESecurityCapabilities field. The UESecurityCapabilities field includes a 4-bit binary index indicating the user terminal's security capabilities. This 4-bit binary index indicates the 16 most frequent combinations selected from the user terminal capability reporting messages {nea1-128, nea2-128, nea3-128} and {nia1-128, nia2-128, nia3-128} as integrity protection encryption algorithms. Specifically, a "integrity protection encryption algorithm combination dictionary" needs to be maintained, and subsequent indices are used to represent the security capabilities of each UE. For example, {nea1-128, nea2-128, nea3-128} represents three encryption algorithms with eight possibilities, and {nia1-128, nia2-128, nia3-128} represents three integrity protection algorithms with eight possibilities. Therefore, the theoretical total number of combinations is 8×8=64. Based on the actual UE capability reporting in the network, the 16 most frequent combinations are selected as the "complete encryption algorithm dictionary". To uniquely identify these 16 combinations, a 4-bit binary index is needed. In this way, the data volume of batch handover request messages is further reduced.

[0071] Batch switching request messages may also include the PDUessionResourcesToBeSetup-List field, which indicates the list of PDU session resources to be established.

[0072] The PDUSessionResourcesToBeSetup-List field includes a Radio Bearer Configuration (RB) field. The RB field indicates the template identifier and differential parameters used by the Data Radio Bearer (DRB). Each template identifier corresponds to a template, and each template includes a set of radioBearerConfig parameters, which may include Radio Link Control (RLC) mode, Packet Data Convergence Protocol (PDCP) configuration, and logical channel configuration. Differential parameters may include dedicated drb-Identity and logicalChannelIdentity. Specifically, the RB context is compressed using a "template-based + differential encoding" approach. First, a predefined configuration template library is established, and an RB-Config-Template library is pre-stored on the network side (source and target base stations). Each template is identified by a 1-byte templateId and corresponds to a complete set of radioBearerConfig parameters, including RLC mode, PDCP configuration, logical channel configuration, etc. A simplified list is transmitted for each UE, indicating which template is used for each DRB and the necessary differential parameters. For example, dedicated drb-Identity and logicalChannelIdentity. Under this scheme, the RB configuration for 16 UEs only requires 200B. This further reduces the data volume of batch handover request messages.

[0073] The following is combined Figure 4 This article will use a specific example to explain in detail the format of batch switching request messages.

[0074] In this example, a batch handover request message is constructed by analyzing and reconstructing the required fields of the standard handover request message in the 3GPP 38.423 protocol, as follows: The batch switching request message described above can be encoded into two messages and sent: The first piece of information (main context) consists of: 8B (cell target set) + 8B (AMF target set) + 32B (SourceNG-RANnodeUEXnAPID List) + 4.5B (UE Cell Id List) + 4.5B (UE AMF Index) + 32B (AMF UE NGAP ID set for 16 UEs) + 16B (Ue Sec Info) + 640B (PDU Info List) + 200B (RB Info) ≈ 945B. That is, the size of the first piece of information is nearly 1KB.

[0075] Second item (key information): approximately 4K.

[0076] Thus, by designing a set of dedicated signaling procedures for batch handover and a corresponding signaling compression method—dictionary indexing and templated encoding—the size of a single batch signaling message can be reduced by an order of magnitude, significantly alleviating the transmission pressure on the inter-satellite interface and avoiding signaling congestion during intensive handovers.

[0077] Furthermore, those skilled in the art will understand that the aforementioned signaling compression methods, such as dictionary indexing, templates, field discarding, and key separation, can be used individually or in combination in any logically feasible manner. The contents of the dictionary and templates can be pre-configured and static, or dynamically generated and updated based on actual network conditions. The specific bit field division for bit-level compression can be customized according to the protocol version or operator policy. The aforementioned signaling compression function can be centralized in the source CU, or it can be completed collaboratively by the source and target CUs.

[0078] The following section details the group handover decision model described above. This model is a deep learning model. It can be an independent network element or integrated into a spaceborne CU or a core network-specific function, such as NWDAF.

[0079] Figure 5 A schematic diagram of the input data for a group handover decision model according to an embodiment of this application is shown. Figure 5As shown, the correlation feature matrix formed by each UE and each neighboring satellite can be used as input to the group handover decision model. This matrix has dimensions M×N×F, where M and N represent the number of UEs and the number of neighboring satellites, respectively, both of which can change dynamically; F is the feature dimension, representing the measurement indicators observed by each user terminal on each neighboring satellite. In one embodiment, the measurement indicators may include the elevation angle of each neighboring satellite observed by each user terminal, the duration for which each neighboring satellite can serve each user terminal, the signal strength of each neighboring satellite obtained by each user terminal, the QoS level of each user terminal, and the load status of each neighboring satellite, etc. Furthermore, the measurement indicators may further include the moving speed and direction of each UE relative to each neighboring satellite, the interference level between neighboring satellites, etc. Through embedding encoding, this variable-dimensional matrix is ​​converted into a fixed-dimensional tensor representation.

[0080] Figure 6 A schematic diagram of the output data of a group handover decision model according to an embodiment of this application is shown. Figure 6 As shown, the output of the group handover decision model is the joint scheduling decision for all UEs in the batch, in the form of an M×N decision matrix. In this matrix, each column corresponds to a UE, and its elements identify the recommended allocation of the UE to each neighboring satellite, such as a probability distribution or binary identifier. Each row has exactly one valid output, indicating the target neighboring satellite to which the UE should handover. Figure 6 The matrix shown indicates that a 1 indicates the UE corresponding to the vertical axis of that cell should switch to the satellite corresponding to that cell, while a 0 indicates that switching is not recommended. There can be at most one 1 in the same column, meaning that for a given UE, only one target satellite is recommended for switching. Those skilled in the art will understand that the model's output is not limited to a deterministic 0 / 1 scheduling matrix. In another embodiment, the model's output can be a probability distribution matrix, for example, indicating the probability of each UE switching to each satellite for further policy decisions; or it can be grouping labels, for example, dividing UEs into several groups and recommending a target satellite for each group.

[0081] The aforementioned group switching decision model needs to be able to receive and process three-dimensional feature tensors; adapt to dynamic changes in the length and width dimensions of the input feature matrix; support retraining based on new samples and achieve effective parameter adjustment; and be able to incorporate expert experience into the model in the form of rules (such as constraints) to guide and constrain its decision-making behavior.

[0082] The Transformer is a deep learning model based on an attention mechanism. Its core idea is to replace the recursive structure in traditional sequence models (such as RNNs and LSTMs) with an attention mechanism, overcoming the limitations of traditional sequence models in long-range dependency modeling and achieving more efficient parallel computation and longer-range dependency capture capabilities. The Transformer consists of an encoder and a decoder, each composed of multiple identical layers stacked together. The encoder maps the input sequence to a high-dimensional feature representation, and the decoder generates the final output based on the encoder's output and the target sequence. Each decoder layer includes three sub-layers: a masked multi-head self-attention mechanism, an encoder-decoder attention mechanism, and a feedforward neural network.

[0083] Based on the above characteristics of Transformer, it can meet some of the model requirements of the packet handover decision model. However, when applied to the packet handover decision task of the low-Earth orbit satellite network in this application, since it is an "encode-decision" task and does not require a decoder, in one embodiment of this application, a standard Transformer encoder structure is used to construct the packet handover decision model. Furthermore, to better adapt to packet handover decisions, this embodiment further improves the Transformer encoder structure as follows: In the self-attention layer, the (UE_i, Sat_j) token interacts with all other tokens in the sequence, enabling the model to compare the link quality of the same UE to different satellites, compare the competitive relationships of different UEs to the same satellite, and capture complex global dependencies, such as: "Assigning UE1 to satellite A will cause an increase in the load on satellite A, which may affect the subsequent allocation decision of UE2." In the above model, (UE_i, Sat_j) represents the association pair formed by the i-th user terminal and the j-th neighboring satellite, i.e., a "UE-satellite" combination to be evaluated. In the packet handover decision model, each such "UE-satellite" combination corresponds to a token in the feature matrix.

[0084] In addition, hard constraints can be set through attention masks, which means that the following situations are considered to violate the constraints and should not be set as the target neighbor of the UE: - When the elevation angle of the neighboring star is less than 30°; - When RSRP is below -90dB; - When the remaining service time of the target neighboring satellite is less than 10 seconds; - When the neighboring star's load reaches 90%; Then, before calculating the attention weights, the attention logits corresponding to those (UE_i, Sat_j) tokens that violate the hard constraints are set to a very large negative value (such as -1e9). In this way, when the model aggregates information, it will automatically ignore these invalid options, fundamentally avoiding making illegal decisions.

[0085] After selecting a model, it needs to be trained before use. The training process for the Transformer-based low-Earth orbit satellite network packet handover decision model is as follows: Step 1: Create a simulation environment that can simulate UE movement, channel changes, service generation, and satellite dynamics.

[0086] Step 2: Based on the simulation environment, generate 10,000 samples (the feature space is relatively small and there is a need for dynamic training later, so the number of samples should not be too large). The sample description is as follows: Input characteristics: List<UeInfo,Satlite> The decision matrix consists of the UE and its neighboring satellites. (See attached image.) Figure 5 The matrix is ​​divided into horizontal UE replacement and vertical satellite replacement. Each cell in the grid stores various indicators observed by a certain UE to a certain satellite. For example, gray cells represent the elevation angle of satellite 2 observed by UE3, the duration that satellite 2 can serve UE3, the signal strength of satellite 2 obtained by UE3, the UE's QoS level, and the load status of satellite 2.

[0087] Sample label: Simulated overall benefits after switching: R_t = a × total system throughput + β × high priority service satisfaction - γ × switching overhead - δ × load imbalance, where α, β, γ and δ are the weights of each indicator, and are all set to 0.25 in the initial training phase, i.e., equal weight.

[0088] Step 3: Train a Transformer neural network with attention mask using simulated samples.

[0089] Step 4: Real-world environment prediction. After the satellite-based base station goes online, if waveband C becomes an edge waveband, its 100 UEs need to be simultaneously switched to the target satellite. (List)<UeInfo,Satlite> The input model of the indicator is used to obtain the output scheduling matrix, which indicates which satellite each UE should switch to. For example... Figure 6 The matrix shown indicates that a 1 indicates the UE corresponding to the vertical axis of that cell should switch to the satellite corresponding to that cell, while a 0 indicates that switching is not recommended. There can be at most one 1 in the same column, meaning that for a given UE, only one target satellite is recommended for switching. For example... Figure 6 As shown, UE1 and UE2 will switch to satellite 1.

[0090] Step 5: Once these 100 UEs have completed service on the target satellite (marked by the next handover), they all report their actual service quality to the onboard decision center, including their throughput, satisfaction with high-priority services, and their previous handover overhead. This process is completed sequentially, meaning that these 100 UEs report their actual service quality sequentially.

[0091] Step 6: After receiving the service reports from all UEs in this batch (if a UE loses contact or is switched prematurely, the service CU should inform the decision center and no longer wait), the decision center will combine the reports from each UE and its own recorded load imbalance to assemble and calculate the actual handover benefit feedback: R_t. Step 7: As the satellite moves, the decision center continuously receives such samples. When the number of samples reaches 10% of the total number of pre-trained samples, 10% is removed from the original sample library and real samples are added to form a new sample library.

[0092] Step 8: Retrain an offline model using the new sample library. Once the training is successful, deploy the model online to replace the original model.

[0093] Through the aforementioned model definition, structural improvements, and application process, this invention constructs a complete closed-loop system encompassing simulation pre-training, online intelligent decision-making, and continuous feedback optimization. This scheme not only achieves global optimization decision-making in multi-UE / multi-satellite scenarios through an improved Transformer model, but also ensures the engineering feasibility of the decision-making through a hard-constraint attention mask mechanism. Furthermore, it leverages the online feedback closed loop to achieve continuous self-evolution of the model, thereby significantly improving the overall efficiency, reliability, and long-term adaptability of low-Earth orbit satellite network handover at the system level.

[0094] Furthermore, in the embodiments of this application, during the actual use of the trained group handover decision, when the service obtained by the UE after handover is about to terminate on the target satellite, that is, during the next handover, the UE can also inform the onboard decision center of the service quality obtained on the satellite, i.e., the total throughput and high-priority service satisfaction, through the handover decision request message of the next round of handover, so as to optimize the decision model based on the actual handover benefits.

[0095] In this way, by using deep learning models to simultaneously consider the complex states and constraints of multiple users and multiple satellites to make joint decisions, multiple objectives such as total system throughput, high-priority service guarantee and load balancing are comprehensively optimized, overcoming the limitations of traditional decision-making based on local or fixed rules.

[0096] As shown above, this application provides a method for batch packet handover signaling interaction and compression for low-Earth orbit satellite networks: First, a complete inter-satellite batch signaling interaction process and compression mechanism are proposed. The batch handover signaling process is adopted to reduce the signaling storm on the Xn link caused by a large number of UE handovers in scenarios of rapid satellite movement; and a semantic-based hybrid compression method is proposed for batch handover signaling, which significantly reduces the signaling load from the source by using public information dictionary, radio bearer configuration template, and security key separation transmission.

[0097] Second, a global optimization decision-making method for satellite handover based on hard-constrained attention masks is proposed. Addressing the local optima problem caused by existing methods relying on local indicators or sequential decisions, this invention proposes a novel decision model. This model performs full-matrix joint calculations on multiple UEs and multiple satellites through an attention mechanism, evaluating all matching possibilities at once, thereby achieving system-level global optimization.

[0098] Third, a dynamic optimization system for satellite handover decision-making models based on online feedback is constructed. This application embodiment constructs a model continuous optimization system containing an online feedback closed loop. The system automatically collects real network performance data after user handover as feedback and uses this feedback data to periodically perform incremental training or retraining of the decision-making model, thereby achieving dynamic updating and adaptive optimization of model parameters during operation.

[0099] Specifically, the embodiments of this application involve: Intelligent Grouping Decision-Making Method: This method deploys a deep learning model based on the Transformer architecture at the onboard decision center, using the feature matrix of multiple users (UEs) against multiple candidate neighboring satellites collected in the early stage based on wave position state transitions as input. The model uses its attention mechanism to globally model and optimize complex UE-satellite associations, directly outputting the system-level optimal UE-satellite matching scheme. In addition, this method constructs an online learning closed loop, which can continuously train and update the model based on real network performance data fed back after the handover, thereby achieving adaptive optimization of the decision-making strategy.

[0100] Batch Signaling Process and Compressed Transmission Method: This method defines and executes a complete inter-satellite signaling interaction sequence specifically designed for batch handover operations. Based on the output of the aforementioned intelligent decision-making, the source satellite and target satellite complete context transfer and resource reservation for multiple UEs through optimized batch request and response messages. During this process, various targeted compression techniques, such as dictionary indexing, field simplification, configuration templates, and key separation transmission, are comprehensively utilized to efficiently compress the batch signaling payload, thereby significantly reducing transmission overhead and ensuring the feasibility of the process under limited inter-satellite bandwidth.

[0101] This invention achieves a complete closed loop from optimized decision-making to reliable execution through the synergy of the aforementioned intelligent decision-making and efficient signaling execution, effectively improving switching efficiency, resource utilization, and overall system performance.

[0102] The above description is a further detailed explanation of the present invention in conjunction with specific preferred embodiments. It should not be considered that the specific embodiments of the present invention are limited to this. For those skilled in the art, several simple deductions or substitutions can be made without departing from the concept of the present invention, and all such deductions or substitutions should be considered to fall within the scope of protection of the present invention as defined by the submitted claims.

Claims

1. A method for batch handover of user terminals in a low-Earth orbit satellite network, characterized in that, include: The source satellite base station identifies multiple user terminals that have entered its edge bands and sends the measurement indicators of these multiple user terminals for multiple neighboring satellites to the satellite decision center. After receiving the measurement indicators, the onboard decision center inputs the measurement indicators into a pre-trained packet handover decision model. The packet handover decision model outputs a packet handover decision result and returns the packet handover decision result to the source onboard base station. The packet handover decision result includes a recommendation result for one or more neighboring satellites to which the multiple user terminals want to handover. The source satellite base station determines one or more target neighboring satellites to which the multiple user terminals need to hand over based on the packet handover decision result, and initiates batch handover request messages to the one or more target neighboring satellites respectively; the batch handover request message includes a targetCellGlobalID field for representing the global identifier of the target cell and a UEContextInfoHORequest field for representing the user terminal context information, wherein the targetCellGlobalID field is the global identifier of the target cell of the multiple target neighboring satellites receiving the batch handover request message, and the UEContextInfoHORequest field uses the index of the global identifier of the target cell to identify the target cell; and The one or more target neighboring satellites, based on the batch handover request message, collaborate with the source satellite base station and the multiple user terminals to complete the service handover.

2. The method according to claim 1, characterized in that, The one or more target neighboring satellites, based on the batch handover request message, collaborate with the source satellite base station and the multiple user terminals to complete the service handover, including: The one or more target neighboring satellites reserve resources for the multiple user terminals according to the batch handover request message, and reply to the source satellite base station with a batch handover request confirmation message, while carrying resource reservation completion information; After receiving the batch handover request confirmation message, the source satellite base station sends the Packet Data Convergence Protocol Sequence Number (PDCP) SN Status Mapping Table of multiple user terminals that are granted handover to the one or more target neighboring satellites, and issues a handover command to each user terminal that is granted handover. After receiving the handover command, the multiple user terminals that are granted handover permission initiate a random access procedure to their corresponding target neighboring satellite. During the handover process of the target neighboring satellite, the data forwarding process is completed with the core network to confirm the batch handover user plane path and send a batch user terminal release message to the source satellite base station. The source satellite-borne base station releases resources based on the batch user terminal release message.

3. The method according to claim 1, characterized in that, The method further includes: The source satellite base station identifies multiple user terminals that have entered its secondary edge position and sends Radio Resource Control (RRC) connection reconfiguration messages to these terminals, instructing them to shorten their measurement reporting cycle and include their location information in their measurement reports.

4. The method according to claim 1, characterized in that, The batch handover request message also includes a GUAMI field, which includes multiple globally unique core network identifiers. The core network is identified by the index of the core network identifier in the UEContextInfoHORequest field.

5. The method according to claim 1, characterized in that, The UEContextInfoHORequest field includes a UESecurityCapabilities field, which includes a 4-bit binary index indicating the security capabilities of the user terminal. This 4-bit binary index indicates the 16 most frequent combinations selected from the user terminal capability reporting messages {nea1-128, nea2-128, nea3-128} and {nia1-128, nia2-128, nia3-128} as the integrity protection encryption algorithm.

6. The method according to claim 1, characterized in that, The batch handover request message also includes a PDUsessionResourcesToBeSetup-List field indicating the PDU session resource establishment list. The PDUsessionResourcesToBeSetup-List field includes a Radio Bearer Configuration (RB) field, which indicates the template identifier and differential parameters used by the Data Radio Bearer (DRB). Each template identifier corresponds to a template, and each template includes a set of radioBearerConfig parameters. The radioBearerConfig parameter set includes Radio Link Control (RLC) mode, Packet Data Convergence Protocol (PDCP) configuration, and Logical Channel Configuration. The differential parameters include dedicated drb-Identity and logicalChannelIdentity.

7. The method according to claim 1, characterized in that, The measurement indicators include the elevation angle of the multiple neighboring satellites observed by the multiple user terminals, the duration for which the multiple neighboring satellites can serve the multiple user terminals, the signal strength of the multiple neighboring satellites acquired by the multiple user terminals, the QoS level of the multiple user terminals, and the load status of the multiple neighboring satellites.

8. The method according to claim 7, characterized in that, The packet handover decision model adopts the standard Transformer encoder structure. In the self-attention layer, the (UE_i, Sat_j) term interacts with all other terms in the sequence to compare the link quality of the same user terminal to different neighboring satellites, compare the competitive relationship of different user terminals to the same neighboring satellite, and capture complex global dependencies.

9. The method according to claim 8, characterized in that, Hard constraints are set through the attention mask of the attention layer. The hard constraints include: the elevation angle of the neighboring satellite is less than 30°, the RSRP is less than -90dB, the remaining service time of the target neighboring satellite is less than 10s, and the load of the neighboring satellite reaches 90%. Before calculating the attention weight, the attention logits corresponding to the (UE_i, Sat_j) tokens that violate the above hard constraints are set to a maximum negative value.

10. The method according to claim 1, characterized in that, After completing the service handover, the multiple user terminals will, before or during the next handover, feed back the service quality and high-priority service satisfaction obtained on one or more target neighboring satellites to the onboard decision center in order to optimize the parameters of the group handover decision model.

Images