An eSIM global operator dynamic OTA switching system based on AI strategy

Abstract

The present invention relates to the field of wireless communication network technology, and specifically to an AI-based strategy-based eSIM global operator dynamic OTA switching system, comprising an environmental data acquisition module that collects bandwidth, packet loss rate, network stability, and switching success rate on the device side to generate local network status data; a behavior and tariff modeling module that collects communication behavior and tariff information to generate user preference data; an AI strategy generation module that outputs OTA control instructions based on a graph neural network and attention mechanism; an OTA strategy synchronization module that issues and parses instructions to generate execution parameter packets; and a dual-link switching execution module that completes switching and fallback configuration. The present invention achieves intelligent, low-latency, and high-reliability dynamic switching of eSIM operators by integrating graph neural network strategy modeling, terminal AI and OTA collaborative mechanisms, and a dual-link redundant switching process.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication network technologies, and in particular to an AI-based eSIM global operator dynamic OTA switching system. Background Art

[0002] With the widespread application of eSIM technology, users can quickly switch between multiple operators without having to replace the physical SIM card, which has become an important development direction in the field of mobile communications. Scenarios such as global mobile office, smart terminal overseas expansion, and cross-border logistics have put forward higher requirements for network quality, tariff optimization, and connection stability, which has promoted the continuous increase in demand for eSIM in dynamic switching among multiple operators. In order to achieve flexible network switching and ensure communication quality, it is urgent to build an intelligent switching system that takes into account user preferences, network status, and policy execution efficiency.

[0003] Existing eSIM operator switching solutions are mostly based on static threshold judgment and single-link interruption switching, lacking in-depth modeling of user behavior patterns and tariff sensitivity. In addition, most switching strategies are difficult to adapt to real-time changes in dynamic network environments. During the switching process, if the target link is unstable or policy generation is delayed, it is easy to cause data interruption, switching failure or service unavailability. At the same time, traditional solutions lack redundancy mechanisms and policy fallback capabilities, and cannot effectively guarantee communication continuity in high-reliability scenarios. Summary of the Invention

[0004] The present invention provides an AI-based eSIM global operator dynamic OTA switching system, which implements policy delivery and local adaptation in seconds, adopts a dual-link redundancy mechanism to ensure uninterrupted switching, and uses fallback protection to improve overall system stability, thereby achieving smarter, more stable, and more economical operator switching capabilities worldwide.

[0005] An AI-based eSIM global operator dynamic OTA switching system includes an environmental data acquisition module, a behavior and tariff modeling module, an AI policy generation module, an OTA policy synchronization module, and a dual-link switching execution module, wherein;

[0006] The environmental data collection module deploys an AI unit on the user device side to collect local network quality data in real time, including current bandwidth, packet loss rate, network stability score and handover success rate, and generates local network status data;

[0007] The behavior and tariff modeling module collects the user's communication behavior data in different time periods and the tariff information of the connected operator, builds a user behavior profile and operator tariff model, and generates user operation preference data;

[0008] The AI ​​strategy generation module uploads local network status data and user operation preference data to the cloud strategy engine. The cloud strategy engine generates the optimal operator switching decision based on the trained global machine learning model graph neural network (GNN) and attention mechanism fusion model, and outputs the OTA switching control instructions including the target operator identification, switching strategy weight, and redundancy mechanism settings.

[0009] The OTA policy synchronization module sends the OTA handover control instruction to the user equipment through the OTA mechanism, triggering the local AI unit to analyze and perform secondary confirmation based on the current local network status to form a local execution handover parameter package;

[0010] The dual-link switching execution module executes the switching parameter package locally, and the device starts the dual-link redundant switching process, maintains the current connection while establishing the target operator link, releases the original link after confirming that the target link is stable, and completes the fallback configuration according to the policy settings.

[0011] Optionally, the environmental data acquisition module includes:

[0012] Data sampling: every preset time interval Collect the original data packet of network status once, which includes the downlink rate of the device access link , Uplink rate , signal reception strength , RTT delay , Packet Loss And the total number of packets sent ;

[0013] Multi-dimensional network indicator calculation: Normalizes and extracts indicators from network status raw data packets to generate multi-dimensional local network quality data, including current bandwidth, packet loss rate, and network stability score;

[0014] Historical switching success rate estimation: The AI ​​unit maintains the status records of the last N operator switchings and uses a sliding window average to calculate the switching success rate. ;

[0015] Local network status data generation: encapsulate the current bandwidth, packet loss rate, network stability score, and switching success rate into a structured data vector and output it as local network status data .

[0016] Optionally, the behavior and tariff modeling module includes:

[0017] Communication behavior data collection: Periodically record the user's communication behavior information in each time period, including communication type (voice / data), communication duration , data usage , access network type (such as 4G, 5G, WiFi) and the connected operator number , forming a communication behavior record table;

[0018] User communication behavior collection: Periodically record communication behavior logs on the user's device, including timestamp, communication type (voice / data), and communication duration. , data usage , Current access operator and communication link quality labels , build a behavior record set ,in, The total number of communication behavior events recorded by the user's device during the preset behavior collection period;

[0019] Behavioral data clustering and feature statistics: Communication behavior records are divided into daily time periods Aggregate and count each time period - operator combination Unit communication cost , average communication quality , Time period activity distribution ;

[0020] User operation preference data estimation: based on statistical unit communication cost , average communication quality , Time period activity distribution , calculate the user operation preference data, including the user's sensitivity to tariffs, communication quality and time period, expressed as:

[0021] ;

[0022] ;

[0023] ;

[0024] in, The user's sensitivity to tariffs, is the user's sensitivity to communication quality, is the user's sensitivity to the availability of the usage time period, is the upper limit of the unit communication cost, Score for maximum communication quality.

[0025] Optionally, the AI ​​strategy generation module includes:

[0026] Feature processing: Normalize local network status data and user operation preference data to form a structured feature vector;

[0027] Strategy Generation Model: This model is based on the fusion of graph neural networks (GNNs) and attention mechanisms. It inputs a structured feature vector and outputs the optimal switching target and the corresponding strategy score.

[0028] Instruction encapsulation and distribution: Based on the results of the policy generation model, the target operator identification, handover priority, and redundancy configuration are generated, encapsulated as OTA control instructions, and securely distributed to the device to drive local handover execution.

[0029] Optionally, the feature processing includes:

[0030] Normalization processing: Apply minimum-maximum normalization processing to local network status data and user operation preference data;

[0031] Structured feature concatenation: concatenates the normalized local network state data and user preference data into a structured feature vector .

[0032] Optionally, the strategy generation model includes:

[0033] Graph structure construction and node feature injection: building heterogeneous graph structures , where the node set Includes a user node With multiple operator nodes , edge set Represents the switchable relationship between users and operators, and assigns a structured feature vector to each node, including user node features , Operator Node Characteristics ;

[0034] Graph neural network propagation and attention weight calculation: Applying graph attention neural network (GAT) on the graph structure, for each edge Calculating attention weights Indicates the preference fit between users and operators;

[0035] Strategy score generation and optimal target output: Calculate the final strategy score for each operator based on the attention weight and operator feature aggregation results .

[0036] Optionally, the instruction encapsulation and issuance includes:

[0037] Policy result analysis: Receives the output from the policy generation model, including the target operator identifier, corresponding policy score, and candidate operator ranking information, and analyzes the optimal switching target and its priority.

[0038] Control parameter generation: Generates a control parameter structure based on the parsing results, including the target operator identifier, switching priority level, dual-link redundancy enable flag, fallback mechanism configuration flag, and policy effective timestamp;

[0039] OTA command encapsulation: encapsulates control parameters into a data format that complies with the OTA protocol, uses an encrypted signature mechanism to verify the integrity and source of the command data, and generates an OTA handover control command data packet with a version identifier and checksum field.

[0040] Delivery and response processing: The control command data packet is securely delivered to the user device through the OTA channel, and the ACK response and execution status on the device side are monitored. If the device confirms that the data packet is received correctly, the local AI module is triggered to start the operator switching process.

[0041] Optionally, the OTA policy synchronization module includes:

[0042] Local state adaptation and policy confirmation: After the local AI unit receives the switching control instruction issued by OTA, it extracts the network state when the policy is generated in the local cache in the instruction , and combined with the current local network state vector Calculating consistency scores ,when When the policy is still applicable, execution is allowed;

[0043] Generate a local execution handover parameter package: When the policy is available, a local execution handover parameter package is generated, including the target operator number, handover priority, redundancy configuration, fallback mechanism and execution timestamp. When the policy is not available, the current policy is not executed and the current round of handover process is terminated.

[0044] Optionally, the dual-link switching execution module includes:

[0045] Target link stability determination: within the set observation window Continuously collect the network state vector of the target link , and calculate the average network stability score of the target link within the window ,like , then the link is determined to be stable and the switching process continues, where is the real-time bandwidth, is the real-time packet loss rate, Score the real-time stability. is the stability scoring threshold;

[0046] Link switching and resource release: When the target link passes the stability assessment, the device releases the original carrier link, switches to the target link, and sets the link as the default primary channel.

[0047] Beneficial effects of the present invention:

[0048] This invention, by constructing a strategy generation model based on the fusion of graph neural networks and attention mechanisms, can effectively integrate local network status data and user operation preference information, realize dynamic scoring and switching decision generation for multiple operators, break away from the limitations of traditional threshold judgment, and improve the intelligence, adaptability and personalization of switching strategies.

[0049] The present invention establishes a two-layer structure that coordinates global cloud modeling and local terminal decision-making by deploying a lightweight AI unit on the user device side and cooperating with the OTA policy delivery mechanism. It ensures the optimality of the strategy while taking into account the execution efficiency. The device side can perform secondary confirmation of the delivered strategy based on the current real-time network status, effectively reducing the risk of switching misjudgment or failure due to state deviation.

[0050] The present invention designs a dual-link redundant switching process and a fallback protection mechanism. During the switching process, the system maintains a stable connection of the old link while establishing a new link. The main channel switching and the original link release operations are performed only after the stability of the target link meets a preset threshold. This ensures the continuity and stability of the eSIM communication service during the switching process, and improves the overall fault tolerance and user experience of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 Schematic diagram of system function modules according to an embodiment of the present invention;

[0053] Figure 2 Schematic diagram of a behavior and tariff modeling module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. Those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0055] like Figure 1-Figure 2As shown, an AI-based eSIM global operator dynamic OTA switching system includes an environmental data acquisition module, a behavior and tariff modeling module, an AI policy generation module, an OTA policy synchronization module, and a dual-link switching execution module, wherein;

[0056] The environmental data collection module deploys an AI unit on the user device side to collect local network quality data in real time, including current bandwidth, packet loss rate, network stability score, and handover success rate, and generates local network status data;

[0057] The behavior and tariff modeling module collects users' communication behavior data in different time periods and the tariff information of the connected operators, builds user behavior profiles and operator tariff models, and generates user operation preference data;

[0058] The AI ​​policy generation module uploads local network status data and user operation preference data to the cloud policy engine. The cloud policy engine generates the optimal operator switching decision based on a trained global machine learning model, a graph neural network (GNN), and an attention mechanism fusion model. It also outputs OTA switching control instructions, including the target operator identifier, switching policy weight, and redundancy mechanism settings.

[0059] The OTA policy synchronization module sends the OTA handover control command to the user device through the OTA mechanism, triggering the local AI unit to analyze and perform secondary confirmation based on the current local network status to form a local execution handover parameter package;

[0060] The dual-link switching execution module executes the switching parameter package locally. The device initiates the dual-link redundant switching process, maintains the current connection while establishing the target operator link. After confirming that the target link is stable, the original link is released and the fallback configuration is completed according to the policy settings to ensure that the service is not interrupted during the switching process.

[0061] The environmental data acquisition module includes:

[0062] Data sampling: every preset time interval Collect the original data packet of network status once, which includes the downlink rate of the device access link , Uplink rate , signal reception strength , RTT delay , Packet Loss And the total number of packets sent ;

[0063] Multi-dimensional network indicator calculation: Normalize and extract indicators from the original network status data packets to generate multi-dimensional local network quality data, including current bandwidth, packet loss rate, and network stability score, expressed as:

[0064] Current bandwidth: ;

[0065] Packet loss rate: ;

[0066] Network stability score: ;

[0067] Historical switching success rate estimation: The AI ​​unit maintains the status records of the last N operator switchings and uses a sliding window average to calculate the switching success rate. , expressed as:

[0068] ;

[0069] in, is the number of successful switches in the past N times, The number of failed switching in the last N times;

[0070] Local network status data generation: encapsulate the current bandwidth, packet loss rate, network stability score, and switching success rate into a structured data vector and output it as local network status data .

[0071] The behavior and tariff modeling module includes:

[0072] Communication behavior data collection: Periodically record the user's communication behavior information in each time period, including communication type (voice / data), communication duration , data usage , access network type (such as 4G, 5G, WiFi) and the connected operator number , forming a communication behavior record table;

[0073] User communication behavior collection: Periodically record communication behavior logs on the user's device, including timestamp, communication type (voice / data), and communication duration. , data usage , Current access operator and communication link quality labels , build a behavior record set ,in, The total number of communication behavior events recorded by the user's device during the preset behavior collection period;

[0074] Behavioral data clustering and feature statistics: Communication behavior records are divided into daily time periods Aggregate and count each time period - operator combination Unit communication cost , average communication quality , Time period activity distribution , expressed as:

[0075] ;

[0076] in, 、 The unit prices of data and voice are: The number of communication records with the operator during this time period;

[0077] ;

[0078] ;

[0079] in, For time period The total number of communication records in It is the sum of communication records in all time periods;

[0080] User operation preference data estimation: based on statistical unit communication cost , average communication quality , Activity distribution over time periods , calculate the user operation preference data, including the user's sensitivity to tariffs, communication quality and time period, expressed as:

[0081] ;

[0082] ;

[0083] ;

[0084] in, The user's sensitivity to tariffs, is the user's sensitivity to communication quality, is the user's sensitivity to the availability of the usage time period, is the upper limit of the unit communication cost, Score for maximum communication quality.

[0085] The AI ​​strategy generation module includes:

[0086] Feature processing: Normalize local network status data and user operation preference data to form a structured feature vector;

[0087] Strategy Generation Model: This model is based on the fusion of graph neural networks (GNNs) and attention mechanisms. It inputs a structured feature vector and outputs the optimal switching target and the corresponding strategy score.

[0088] Instruction encapsulation and distribution: Based on the results of the policy generation model, the target operator identification, handover priority, and redundancy configuration are generated, encapsulated as OTA control instructions, and securely distributed to the device to drive local handover execution.

[0089] Feature processing includes:

[0090] Normalization: Minimum-maximum normalization is applied to local network status data and user operation preference data, expressed as:

[0091] ;

[0092] in, is the normalized result, is the original eigenvalue, including 、 、 、 , 、 are the minimum and maximum values ​​of the feature in the historical data respectively;

[0093] Structured feature concatenation: concatenates the normalized local network state data and user preference data into a structured feature vector , expressed as:

[0094] ;

[0095] in, 、 、 、 are the corresponding normalized values ​​respectively.

[0096] The policy generation model includes:

[0097] Graph structure construction and node feature injection: building heterogeneous graph structures , where the node set Includes a user node With multiple operator nodes , edge set Represents the switchable relationship between users and operators, and assigns a structured feature vector to each node, including user node features , Operator Node Characteristics , expressed as:

[0098] ;

[0099] ;

[0100] in, is the average communication cost, is the historical average communication quality, is the regional activity (normalized);

[0101] Graph neural network propagation and attention weight calculation: Applying graph attention neural network (GAT) on the graph structure, for each edge Calculating attention weights It represents the preference fit between users and operators, expressed as:

[0102] ;

[0103] ;

[0104] in, is the shared linear transformation matrix, is the learnable attention parameter vector, 、 is the edge attention score, is the normalized attention weight;

[0105] Strategy score generation and optimal target output: Calculate the final strategy score for each operator based on the attention weight and operator feature aggregation results , expressed as:

[0106] ;

[0107] in, is the number of attention heads, For the Attention heads for operators The weight of For operators The feature vector after being encoded by the graph neural network, is the output weight vector, is the ReLU activation function;

[0108] ;

[0109] in, Operator Node The set of neighbor nodes, including the user node , For the Neighbor nodes under the attention head and The attention weights between is the original input feature vector of the neighbor node (such as user), For the The weight matrix of the attention head.

[0110] Instruction packaging and issuance include:

[0111] Policy result parsing: Receives the output from the policy generation model, including the target operator ID, the corresponding policy score, and the ranking information of the candidate operators, and parses the optimal switching target and its priority, which is expressed as:

[0112] ;

[0113] in, The set of operator scores output by the policy generation model, is the number of optional operators;

[0114] ;

[0115] in, Number the target operator;

[0116] ;

[0117] in, A list of descending indices of scores for all operators, Indicates that the elements of the collection are sorted from large to small (descending order);

[0118] Control parameter generation: Generate a control parameter structure based on the parsing results, including the target operator identifier, switching priority level, dual-link redundancy enable flag, fallback mechanism configuration identifier, and policy effective timestamp, expressed as:

[0119] ;

[0120] in, is the identifier of the target operator, The priority of the target operator in the scoring order, Enables the flag for redundant switching, indicating whether to use dual-link switching protection. Enable the flag for the fallback mechanism, indicating whether to automatically fallback when the switch fails. The effective timestamp of the policy instruction. The target operator number, To obtain the target operator The index of the position in the sorted list, Score the target operator's strategy, To enable redundancy, The local historical switching success rate, The switch success rate threshold for enabling the fallback mechanism. is the current system timestamp, The delay time for the policy to take effect;

[0121] OTA command encapsulation: encapsulates control parameters into a data format that complies with the OTA protocol, uses an encrypted signature mechanism to verify the integrity and source of the command data, and generates an OTA handover control command data packet with a version identifier and checksum field.

[0122] Delivery and response processing: The control command data packet is securely delivered to the user device through the OTA channel, and the ACK response and execution status on the device side are monitored. If the device confirms that the data packet is received correctly, the local AI module is triggered to start the operator switching process.

[0123] The OTA policy synchronization module includes:

[0124] Local state adaptation and policy confirmation: After the local AI unit receives the switching control instruction issued by OTA, it extracts the network state when the policy is generated in the local cache in the instruction , and combined with the current local network state vector Calculating consistency scores ,when , the policy is considered to still apply and execution is allowed. The consistency score is expressed as:

[0125] ;

[0126] in, is the current state value, The network status when the policy is generated in the local cache. The preset acceptable error range ( Set as 10%, Set as 1.5%, Set as 0.1, Set as 0.05);

[0127] Generate a local execution handover parameter package: When the policy is available, a local execution handover parameter package is generated, including the target operator number, handover priority, redundancy configuration, fallback mechanism and execution timestamp. When the policy is not available, the current policy is not executed and the current round of handover process is terminated.

[0128] The dual-link switching execution module includes:

[0129] Target link stability determination: within the set observation window Continuously collect the network state vector of the target link , and calculate the average network stability score of the target link within the window ,like , then the link is determined to be stable and the switching process continues, where is the real-time bandwidth, is the real-time packet loss rate, Score the real-time stability. is the stability score threshold, expressed as:

[0130] ;

[0131] in, Score real-time stability;

[0132] Link switching and resource release: When the target link passes the stability assessment, the device releases the original carrier link, switches to the target link, and sets the link as the default primary channel.

[0133] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.

[0134] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. An AI-based eSIM global operator dynamic OTA switching system, characterized by: It includes an environmental data acquisition module, a behavior and tariff modeling module, an AI strategy generation module, an OTA strategy synchronization module, and a dual-link switching execution module, among which; The environmental data collection module deploys an AI unit on the user device side to collect local network quality data in real time, including current bandwidth, packet loss rate, network stability score and handover success rate, and generates local network status data; The behavior and tariff modeling module collects the user's communication behavior data in different time periods and the tariff information of the connected operator, builds a user behavior profile and operator tariff model, and generates user operation preference data; The AI ​​strategy generation module uploads local network status data and user operation preference data to the cloud strategy engine. The cloud strategy engine generates the optimal operator switching decision based on the trained global machine learning model graph neural network and attention mechanism fusion model, and outputs the OTA switching control instructions including the target operator identification, switching strategy weight, and redundancy mechanism settings. The OTA policy synchronization module sends the OTA handover control instruction to the user equipment through the OTA mechanism, triggering the local AI unit to analyze and perform secondary confirmation based on the current local network status to form a local execution handover parameter package; The dual-link switching execution module executes the switching parameter package locally, and the device starts the dual-link redundant switching process, maintains the current connection while establishing the target operator link, releases the original link after confirming that the target link is stable, and completes the fallback configuration according to the policy settings.

2. The AI-based eSIM global operator dynamic OTA switching system according to claim 1, characterized in that: The environmental data acquisition module includes: Data sampling: every preset time interval Collect the original data packet of network status once, which includes the downlink rate of the device access link , Uplink rate , signal reception strength , RTT delay , Packet Loss And the total number of packets sent ; Multi-dimensional network indicator calculation: Normalizes and extracts indicators from network status raw data packets to generate multi-dimensional local network quality data, including current bandwidth, packet loss rate, and network stability score; Historical switching success rate estimation: The AI ​​unit maintains the status records of the last N operator switchings and uses a sliding window average to calculate the switching success rate. ; Local network status data generation: encapsulate the current bandwidth, packet loss rate, network stability score, and switching success rate into a structured data vector and output it as local network status data .

3. The AI-based eSIM global operator dynamic OTA switching system according to claim 2, characterized in that: The behavior and tariff modeling module includes: Communication behavior data collection: Periodically record the user's communication behavior information in each time period, including communication type and communication duration , data usage , access network type and connected operator number , forming a communication behavior record table; User communication behavior collection: Periodically record communication behavior logs on the user's device, including timestamp, communication type, and communication duration , data usage , Current access operator and communication link quality labels , build a behavior record set ,in, The total number of communication behavior events recorded by the user's device during the preset behavior collection period; Behavioral data clustering and feature statistics: Communication behavior records are divided into daily time periods Aggregate and count each time period - operator combination Unit communication cost , average communication quality , Time period activity distribution ; User operation preference data estimation: based on statistical unit communication cost , average communication quality , Time period activity distribution , calculate the user operation preference data, including the user's sensitivity to tariffs, communication quality and time period, expressed as: ； ； ； in, The user's sensitivity to tariffs, is the user's sensitivity to communication quality, is the user's sensitivity to the availability of the usage time period, is the upper limit of the unit communication cost, Score for maximum communication quality.

4. The AI-based eSIM global operator dynamic OTA switching system according to claim 3, characterized in that: The AI ​​strategy generation module includes: Feature processing: Normalize local network status data and user operation preference data to form a structured feature vector; Strategy Generation Model: This model is built based on the fusion of graph neural networks and attention mechanisms. It inputs a structured feature vector and outputs the optimal switching target and the corresponding strategy score. Instruction encapsulation and distribution: Based on the results of the policy generation model, the target operator identification, handover priority, and redundancy configuration are generated, encapsulated as OTA control instructions, and securely distributed to the device to drive local handover execution.

5. The AI-based eSIM global operator dynamic OTA switching system according to claim 4, characterized in that: The feature processing includes: Normalization processing: Apply minimum-maximum normalization processing to local network status data and user operation preference data; Structured feature concatenation: concatenates the normalized local network state data and user preference data into a structured feature vector .

6. The AI-based eSIM global operator dynamic OTA switching system according to claim 5, characterized in that: The strategy generation model includes: Graph structure construction and node feature injection: building heterogeneous graph structures , where the node set Includes a user node With multiple operator nodes , edge set Represents the switchable relationship between users and operators, and assigns a structured feature vector to each node, including user node features , Operator Node Characteristics ; Graph neural network propagation and attention weight calculation: Apply graph attention neural network on the graph structure, for each edge Calculating attention weights Indicates the preference fit between users and operators; Strategy score generation and optimal target output: Calculate the final strategy score for each operator based on the attention weight and operator feature aggregation results .

7. The AI-based eSIM global operator dynamic OTA switching system according to claim 6, characterized in that: The instruction packaging and issuing include: Policy result analysis: Receives the output from the policy generation model, including the target operator identifier, corresponding policy score, and candidate operator ranking information, and analyzes the optimal switching target and its priority. Control parameter generation: Generates a control parameter structure based on the parsing results, including the target operator identifier, switching priority level, dual-link redundancy enable flag, fallback mechanism configuration flag, and policy effective timestamp; OTA command encapsulation: encapsulates control parameters into a data format that complies with the OTA protocol, uses an encrypted signature mechanism to verify the integrity and source of the command data, and generates an OTA handover control command data packet with a version identifier and checksum field. Delivery and response processing: The control command data packet is securely delivered to the user device through the OTA channel, and the ACK response and execution status on the device side are monitored. If the device confirms that the data packet is received correctly, the local AI module is triggered to start the operator switching process.

8. The AI-based eSIM global operator dynamic OTA switching system according to claim 7, characterized in that: The OTA policy synchronization module includes: Local state adaptation and policy confirmation: After the local AI unit receives the switching control instruction issued by OTA, it extracts the network state when the policy is generated in the local cache in the instruction , and combined with the current local network state vector Calculating consistency scores ,when When the policy is still applicable, execution is allowed; Generate a local execution handover parameter package: When the policy is available, a local execution handover parameter package is generated, including the target operator number, handover priority, redundancy configuration, fallback mechanism and execution timestamp. When the policy is not available, the current policy is not executed and the current round of handover process is terminated.

9. The AI-based eSIM global operator dynamic OTA switching system according to claim 8, characterized in that: The dual-link switching execution module includes: Target link stability determination: within the set observation window Continuously collect the network state vector of the target link , and calculate the average network stability score of the target link within the window ,like , then the link is determined to be stable and the switching process continues, where is the real-time bandwidth, is the real-time packet loss rate, Score the real-time stability. is the stability scoring threshold; Link switching and resource release: When the target link passes the stability assessment, the device releases the original carrier link, switches to the target link, and sets the link as the default primary channel.

Images