A call forwarding method, network device, medium and computer program product

Abstract

This disclosure provides a call forwarding method, network device, medium, and computer program product. It constructs a call forwarding path prediction model and uses this model to predict multiple candidate forwarding paths in real time. When a call is received, the optimal primary forwarding path is selected from the multiple candidate paths, and forwarding resources are negotiated with the operator through a preset communication protocol, thereby achieving a smooth forwarding. This avoids reliance on preset operator lines and enables dynamic adjustment of the forwarding strategy based on actual network conditions.

Description

Technical Field

[0001] This disclosure relates to the field of communication technology, and in particular to a call transfer method, network device, medium, and computer program product. Background Technology

[0002] Most existing systems rely on preset operator line priorities for call routing planning, or simply select base stations based on signal strength. For example, some systems attempt different operator lines sequentially in a pre-defined order, without considering real-time network congestion or line quality. This approach struggles to achieve efficient call routing in complex and dynamic network environments, such as scenarios with frequent switching between 5G, 4G, and Wi-Fi networks, easily leading to increased call latency and decreased call quality. Summary of the Invention

[0003] This disclosure provides a call transfer method, network device, medium, and computer program product. By using a transfer path prediction model to predict the main transfer path in real time, session quality is effectively improved.

[0004] In one aspect, this embodiment provides a call forwarding method, which includes: constructing a forwarding path prediction model based on multiple communication data; predicting at least one candidate forwarding path based on the forwarding path prediction model; selecting a primary forwarding path from multiple candidate forwarding paths when a terminal device initiates a call access; and negotiating forwarding resources with the operator using a preset communication protocol based on the primary forwarding path to connect the call access.

[0005] In embodiments of this disclosure, the method further includes: acquiring various communication data based on network quality information, wherein the communication data includes at least one or more of physical layer data, network layer data, and service layer data; assigning at least a first dynamic weight to the physical layer data, assigning a second dynamic weight to the network layer data, and assigning a third dynamic weight to the service layer data; and generating transfer path prediction data based at least on the physical layer data and the first dynamic weight, the network layer data and the second dynamic weight, and the service layer data and the third dynamic weight, wherein the transfer path prediction data is used at least to generate a transfer path prediction model.

[0006] In the embodiments of this disclosure, the switching path prediction model includes at least a hybrid neural network module and a reinforcement learning module. The hybrid neural network module consists of at least a front-end feature extraction submodule, a medium-to-long-term feature learning submodule, and an attention mechanism submodule.

[0007] In embodiments of this disclosure, predicting at least one candidate transfer path based on a transfer path prediction model includes: inputting transfer path prediction data into the transfer path prediction model, and having the transfer path prediction model output a quality prediction value for at least one candidate transfer path; generating a two-layer transfer path routing table based on multiple constraints and the quality prediction value of at least one candidate transfer path. The two-layer transfer path routing table includes a first-layer path matrix and a second-layer path matrix. The first-layer path matrix is ​​used to store multiple candidate primary transfer paths, and the second-layer path matrix is ​​used to store multiple candidate backup transfer paths.

[0008] In embodiments of this disclosure, selecting a primary transfer path from multiple candidate transfer paths includes: selecting a candidate primary transfer path that meets a first preset condition from multiple candidate primary transfer paths, and determining the candidate primary transfer path as the primary transfer path when the candidate primary transfer path is available; or selecting a candidate backup transfer path that meets a second preset condition from multiple candidate backup transfer paths when the candidate primary transfer path is unavailable.

[0009] In embodiments of this disclosure, the preset communication protocol includes at least a transport layer that supports fast reconnection and a session layer that includes a transfer tag.

[0010] In embodiments of this disclosure, the method includes: after a call is connected, sliding a preset time window and sequentially calculating the network information indicators of the current session within the preset time window; determining whether each network information indicator exceeds a network information indicator threshold; and when a series of consecutive network information indicators exceed the network information indicator threshold, and the number of consecutive indicators is greater than or equal to a preset number, switching from the primary transfer path to a candidate backup transfer path.

[0011] On the other hand, embodiments of this disclosure provide a network device, including: a communication interface configured to perform wireless communication with a terminal device; a memory for storing computer-readable instructions; and a processor for executing the computer-readable instructions, causing the network device to perform the call transfer method.

[0012] In another aspect, embodiments of this disclosure provide a non-transitory computer-readable storage medium for storing computer-readable instructions that, when executed by a processor, cause the processor to perform the aforementioned call transfer method.

[0013] In another aspect, embodiments of this disclosure provide a computer program product, including a computer program that, when executed by a processor, implements the above-described call transfer method. Attached Figure Description

[0014] The above and other objects, features, and advantages of this disclosure will become more apparent from the more detailed description of the embodiments thereof in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this disclosure and form part of the specification. They are used together with the embodiments of this disclosure to explain the disclosure and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0015] Figure 1 A flowchart illustrating a call transfer method according to an embodiment of the present disclosure is shown.

[0016] Figure 2 A flowchart illustrating another call transfer method according to an embodiment of the present disclosure is shown.

[0017] Figure 3 A flowchart illustrating another call transfer method according to an embodiment of the present disclosure is shown.

[0018] Figure 4 A schematic diagram of a switching path prediction model according to an embodiment of the present disclosure is shown.

[0019] Figure 5 A flowchart illustrating the prediction of at least one candidate transfer path according to an embodiment of the present disclosure is shown.

[0020] Figure 6 The flowchart illustrating the selection of a primary transfer path from a plurality of candidate transfer paths according to an embodiment of the present disclosure is shown in the illustration.

[0021] Figure 7 A schematic diagram illustrating the selection of the primary routing path according to an embodiment of the present disclosure is provided.

[0022] Figure 8 A schematic diagram illustrating a preset communication protocol according to an embodiment of the present disclosure is provided.

[0023] Figure 9 A flowchart illustrating another call transfer method according to an embodiment of the present disclosure is shown.

[0024] Figure 10 A block diagram of a network device according to an embodiment of the present disclosure is shown schematically.

[0025] Figure 11 A block diagram illustrating a non-transitory computer-readable storage medium according to an embodiment of the present disclosure is shown.

[0026] Figure 12 A block diagram illustrating a computer program product according to an embodiment of the present disclosure is shown schematically. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this disclosure more apparent, exemplary embodiments according to this disclosure will now be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this disclosure, and not all embodiments of this disclosure. It should be understood that this disclosure is not limited to the exemplary embodiments described herein.

[0028] Most existing systems rely on preset operator line priorities for call routing planning, or simply select base stations based on signal strength. For example, some systems attempt different operator lines sequentially in a pre-defined order, without considering real-time network congestion or line quality. This approach struggles to achieve efficient call routing in complex and dynamic network environments, such as scenarios with frequent switching between 5G, 4G, and Wi-Fi networks, easily leading to increased call latency and decreased call quality.

[0029] Furthermore, when interacting with operator systems, common communication protocols such as SIP (Session Initiation Protocol) or H.323 are often used. These protocols cannot achieve real-time, efficient data exchange with operator systems, nor can they dynamically adjust call transfer strategies based on real-time network conditions and user needs. For example, when network congestion occurs, it is impossible to promptly negotiate with operators to adjust call transfer paths or priorities.

[0030] It is evident that the above path switching method has the following shortcomings: 1. Low call transfer efficiency: Due to the limitations of fixed call transfer path planning, it is impossible to quickly find the optimal call transfer path based on real-time network conditions in complex network environments, resulting in high call transfer delays and affecting users' real-time communication experience.

[0031] 2. Poor stability: In cases of signal fluctuations or base station switching, call interruptions are likely to occur, making it impossible to provide users with stable and reliable communication services.

[0032] 3. Lack of intelligent interaction: The interaction with the operator's system lacks intelligence and cannot dynamically adjust the switching strategy based on factors such as real-time network conditions, user behavior habits, and business needs, thus failing to meet the needs of modern users for efficient and intelligent communication.

[0033] Based on this, embodiments of this disclosure provide a call transfer method that constructs a transfer path prediction model using various communication data, and uses the transfer path prediction model to predict candidate transfer paths, thereby determining the primary transfer path. This allows for dynamic determination of the transfer path based on real-time network conditions, ensuring smooth communication.

[0034] Figure 1A flowchart illustrating a call transfer method according to an embodiment of the present disclosure is shown.

[0035] like Figure 1 As shown, the call transfer method of this embodiment includes S101, S102 and S103: S101. Construct a transfer path prediction model based on various communication data.

[0036] In this embodiment of the disclosure, the communication data can be data used to reflect real-time network information. The communication data can be multi-dimensional network data, reflecting multi-dimensional real-time network information. The communication data includes, but is not limited to, data such as network signal strength, base station load, user historical call forwarding preferences, and real-time network congestion status.

[0037] In the above, network signal strength can be obtained through real-time monitoring of signals in different frequency bands, and it directly affects the smoothness of call transfers and subsequent communications. Base station load refers to the real-time operating load of the base stations currently connected to / accessible by the terminal device. This can be obtained from the base station status data interface provided by the operator, avoiding the selection of overloaded base stations as transfer nodes and preventing situations where only base station signal strength is considered during path transfer, ignoring base station congestion. User historical transfer preferences refer to personalized data such as call transfer records, path selection behavior, and frequently used transfer methods over a past period. This can be obtained by analyzing users' transfer records and selection behavior over a historical period. Real-time network congestion status refers to the traffic load, data transmission latency, and network congestion trends in the network area where the terminal device is located. This can be predicted using network traffic monitoring data and machine learning algorithms. Utilizing real-time network congestion status allows for advance prediction of network conditions and avoidance of congested network lines.

[0038] In this embodiment of the disclosure, the transfer path prediction model can calculate and predict the optimal transfer path in real time based on communication data such as current network conditions and user needs. For example, deep neural network algorithms (such as LSTM, Long-Short Term Memory) can be used to perform in-depth analysis and learning on various communication data. By training the model with a large amount of historical and real-time communication data, a transfer path prediction model can be constructed.

[0039] S102. Based on the transfer path prediction model, predict at least one candidate transfer path, and when the terminal device initiates a call access, select one main transfer path from multiple candidate transfer paths.

[0040] In this embodiment of the disclosure, real-time communication data can be analyzed and predicted using a constructed transfer path prediction model. The transfer path prediction model can output quality prediction values ​​for multiple transfer paths, and at least one candidate transfer path can be determined based on the quality prediction values ​​of the transfer paths. For example, multiple transfer paths can be filtered using a quality prediction value threshold, and transfer paths with quality prediction values ​​greater than the quality prediction value threshold can be selected as candidate transfer paths.

[0041] In this embodiment, a terminal device can establish a session connection with an incoming call device through candidate transfer paths. Multiple candidate transfer paths can be included, such as a primary candidate transfer path and a backup candidate transfer path. The primary candidate transfer path indicates the path through which the terminal device establishes a session connection with the incoming call device, resulting in the best call quality. When selecting a primary transfer path from multiple candidate transfer paths, real-time network information and user preferences can be used to evaluate or filter the multiple candidate transfer paths, selecting the optimal candidate transfer path as the current primary transfer path from multiple evaluation results. Simultaneously, a corresponding backup candidate transfer path can be selected from multiple candidate transfer paths based on multiple evaluation results. When the primary transfer path fluctuates, the connection can be switched to a backup candidate transfer path, thereby ensuring the smooth continuation of the session.

[0042] S103. Based on the main transfer path, negotiate transfer resources with the operator using a preset communication protocol to connect the call.

[0043] In this embodiment, the aforementioned switching path prediction model can be deployed within a computing node, such as an edge computing node. An edge computing node refers to a computing node located near the user / base station. Edge computing nodes have short data transmission distances with terminal devices, meeting millisecond-level response requirements. The operator can refer to a system providing mobile communication services. The operator can provide the computing node with communication data such as base station load and network topology, and can also allocate dedicated bandwidth and base station resources for switching to establish session connections or perform resource release operations to terminate sessions.

[0044] In this embodiment, when the computing node selects a primary transfer path using a transfer path prediction model, it can negotiate transfer resources with the operator through a preset communication protocol. This includes features such as dynamic and rapid bandwidth matching and digital signature verification of identity compliance. The computing node then obtains a dedicated QoS tunnel and transfer resource allocation receipt confirmed by the operator. Furthermore, the computing node can send a primary transfer path execution command to the operator, who can then complete the line switch according to the negotiation results, establishing a session link, thereby successfully establishing a session between the calling and called parties.

[0045] According to embodiments of this disclosure, a transfer path prediction model is constructed, and multiple candidate transfer paths are predicted in real time using this model. When a call is received, an optimal primary transfer path is selected from the multiple candidate transfer paths, and transfer resources are negotiated with the operator through a preset communication protocol, thereby achieving a smooth transfer. This avoids reliance on preset operator lines and enables dynamic adjustment of the transfer strategy based on actual network conditions.

[0046] Figure 2 A flowchart illustrating another call transfer method according to an embodiment of the present disclosure is shown.

[0047] like Figure 2 As shown, the call transfer method of this disclosure includes S201, S202 and S203: S201. Based on network quality information, acquire various communication data, including at least one or more of physical layer data, network layer data, and service layer data. In this embodiment, a lightweight, resource-efficient small data acquisition tool can be installed on the user's mobile phone (terminal device) and the operator's edge server (edge ​​node). Communication data is collected synchronously through the terminal device and edge node, acquiring data from the nearest available location to reduce latency. Network quality information can be used to reflect the current network connectivity level, such as whether the network connection is stable or fluctuating. When the network connection is stable, communication data can be acquired at a first acquisition frequency, for example, once every 100 milliseconds, balancing accuracy with minimal strain on the terminal device's performance. When the network connection fluctuates, communication data can be acquired at a higher frequency than the first acquisition frequency. For example, when the network connection fluctuates, such as when the terminal device moves to a location with poor signal, like an elevator or underground parking garage, the acquisition frequency automatically increases to once every 50 milliseconds.

[0048] In some embodiments, communication data can be categorized into physical layer data, network layer data, and service layer data. Physical layer data is used to determine user movement information. It can be collected via GPS+BeiDou dual-mode positioning and a mobile phone's three-axis accelerometer. Physical layer data includes, for example, precise latitude and longitude, and user movement speed. By analyzing physical layer data, the user's network environment can be predicted; for example, signals may switch rapidly on high-speed trains, allowing for the preparation of alternative routes. Network layer data reflects network signal conditions and can be collected using a dedicated software development kit (SDK). Network layer data may include parameters of all network signals that the mobile phone can detect, such as LTE / Wi-Fi 6 / 5G NSA / SA multi-mode signal parameters. By acquiring and analyzing network layer data, the quality of all available networks can be assessed, not just a single network. Service layer data reflects operator resource conditions and can be obtained by directly connecting to the operator's official open API. Service layer data may include, for example, precise cell-level base station load rates and VoLTE HD call success rates. By acquiring and analyzing business layer data, congested base stations can be avoided, reducing call interruption / transfer delays at the source.

[0049] S202, at least assign a first dynamic weight to physical layer data, a second dynamic weight to network layer data, and a third dynamic weight to service layer data.

[0050] In this embodiment of the disclosure, physical layer data, network layer data, and service layer data can be integrated together to serve as training data for constructing a transit path prediction model. Before integration, processing can be performed such as synchronizing communication data from different sources according to collection time and removing erroneous data to construct a unified time series.

[0051] Furthermore, the computing node can determine the normal data range for the current moment based on the actual data from the physical layer, network layer, and service layer. For example, it can predict the normal data range for the current moment based on the actual data from the previous moment (such as the previous signal strength and location coordinates) and the physical characteristics of the sensors. Then, it can compare the communication data collected in real time with the normal data range for the current moment. When the communication data is within the normal data range, it can be determined that the communication data is normal; when the communication data is outside the normal data range, it can be determined that the communication data is abnormal.

[0052] When the communication data is normal, the weight can be increased; when the communication data is abnormal, the weight can be decreased. This allows for the allocation of a first dynamic weight to physical layer data, a second dynamic weight to network layer data, and a third dynamic weight to service layer data. When network connectivity fluctuates, such as in elevators or tunnels where signals become inconsistent, sensors collect a lot of noise data. By assigning dynamic weights to the communication data, the interference from this noise data can be effectively reduced.

[0053] S203. Based at least on physical layer data and a first dynamic weight, network layer data and a second dynamic weight, and service layer data and a third dynamic weight, generate transfer path prediction data, which is used at least to generate a transfer path prediction model.

[0054] In this embodiment, all layered communication data can be weighted and summed according to their respective dynamic weights to obtain comprehensive transfer path prediction data, such as comprehensive signal quality score, comprehensive network stability score, and other indicators. The transfer path prediction data can be used to train a model, thereby generating a transfer path prediction model.

[0055] Figure 3 A flowchart illustrating another call transfer method according to an embodiment of the present disclosure is shown.

[0056] like Figure 3 As shown in Figure S301, distributed probes are deployed to acquire multi-source data. Multi-source data can include, for example... Figure 3 The data shown includes physical layer data, network layer data, and service layer data.

[0057] S302. Dynamic weight allocation and fusion of multi-source data. After obtaining multiple communication data, the multiple communication data can be fused using the methods shown in S202 and S203 above to generate transfer path prediction data.

[0058] In this embodiment of the disclosure, the switching path prediction model includes at least a hybrid neural network module and a reinforcement learning module. The hybrid neural network module consists of at least a front-end feature extraction submodule, a medium-to-long-term feature learning submodule, and an attention mechanism submodule.

[0059] The transition path prediction data can be input into the front-end feature extraction submodule, which can be used as the first layer of the hybrid neural network module. The front-end feature extraction submodule can use a three-layer 1D-CNN to process the transition path prediction data and extract local key features from the transition path prediction data.

[0060] The mid-to-long-term feature learning submodule serves as the second layer of the hybrid neural network module. It captures user behavior patterns by inputting local key features extracted from the transition path prediction data. This submodule includes bidirectional LSTM layers. The forward LSTM layer captures past network trends, such as signal fluctuations over the past 10 seconds. The backward LSTM layer captures potential future network trends, such as whether congestion will continue or whether the signal will worsen. The output features of the bidirectional LSTM layers are concatenated to obtain complete long-term temporal context information, enabling the model to understand the changing patterns of the network state.

[0061] The attention mechanism submodule can serve as the third layer in a hybrid neural network. As the attention mechanism layer, it dynamically assigns network quality index weights to the output features of the bidirectional LSTM layer. By assigning higher weights to features that have a greater impact on path decisions, the influence of irrelevant noise can be effectively reduced.

[0062] The reinforcement learning module can be an online learning framework based on DQN. The features weighted by the attention mechanism submodule are fed into the DQN-based online learning framework. At the same time, the reward function R=α×(1-latency rate)+β×signal stability+γ×cost coefficient is combined to guide the optimization of the switching path prediction model: the lower the latency, the higher the signal stability, and the lower the cost, the higher the reward obtained by the model. The path selection of "high latency, easy interruption, and high cost" is penalized, so that the switching path prediction model learns the optimal decision of "low latency + high stability + low cost".

[0063] In some embodiments, the transit path prediction model can be deployed on edge computing nodes. This means that edge computing nodes can establish local transit path prediction models, bringing them closer to the user and avoiding cloud transmission delays, thus achieving low-latency inference. Simultaneously, differential privacy technology can be used to encrypt model parameter updates, enabling collaborative training while protecting user privacy. This involves adding controllable noise during model parameter updates, preventing attackers from retrieving individual users' original data from the update results. This privacy-preserving collaborative training improves model accuracy without leaking sensitive user information.

[0064] Figure 4 A schematic diagram of a switching path prediction model according to an embodiment of the present disclosure is shown.

[0065] like Figure 4 As shown, the transition path prediction model 400 includes a hybrid neural network module 401 and a reinforcement learning module 402. The hybrid neural network module 401 includes a front-end feature extraction submodule 4011, a mid-to-long-term feature learning submodule 4012, and an attention mechanism submodule 4013.

[0066] The generated transition path prediction data can be input into the front-end feature extraction submodule 4011 to extract local key features. For example, the transition path prediction data sequentially passes through the input layer, Conv1D 64@5, Conv1D 128@3, and Conv1D 256@3, thus outputting the extracted local key features. The extracted local key features are input into the mid-to-long-term feature learning submodule 4012, where the back LSTM layer and forward LSTM layer are used to concatenate the local key features. The concatenated local key features are then input into the attention mechanism submodule 4013, where a weighted feature is output after passing through the weight allocation matrix.

[0067] The reinforcement learning module 402, based on the DQN framework, optimizes the output of the hybrid neural network module 401. It uses a reward function to align the prediction results of the entire hybrid neural network towards the goals of "low latency, high stability, and low cost." The output of the hybrid neural network module 401, after optimization using the DQN framework and reward function, outputs the final result.

[0068] Figure 5 A flowchart illustrating the prediction of at least one candidate transfer path according to an embodiment of the present disclosure is shown.

[0069] like Figure 5 As shown, in S102 above, based on the transfer path prediction model, at least one candidate transfer path is predicted, including S501 and S502: S501. Input the transfer path prediction data into the transfer path prediction model, and the transfer path prediction model outputs the quality prediction value of at least one candidate transfer path.

[0070] In this embodiment of the disclosure, the transfer path prediction data is input into the transfer path prediction model, and the transfer path prediction model can output the quality prediction value of each candidate transfer path, including information such as latency, stability, cost, and power consumption.

[0071] S502. Based on multiple constraints and the quality prediction value of at least one candidate transfer path, generate a two-layer transfer path routing table. The two-layer transfer path routing table includes a first-layer path matrix and a second-layer path matrix. The first-layer path matrix is ​​used to store multiple candidate primary transfer paths, and the second-layer path matrix is ​​used to store multiple candidate backup transfer paths.

[0072] In this embodiment of the disclosure, the computing node can also obtain multiple constraints, such as inputting multiple constraints into the switching path selection engine included in the computing node. The constraints include latency requirements (e.g., latency <150ms to ensure low call latency), tariff models (which help control call costs), and energy consumption constraints (which help limit the power consumption of terminals / edge nodes).

[0073] In this embodiment, the solution can be obtained using the predicted quality value of each candidate transfer path and multiple constraints, seeking a balance among conflicting objectives (such as the often contradictory goals of "lower latency" and "lower cost"). A set of Pareto front solutions is obtained, which are solutions that cannot be "completely surpassed" by other solutions on multiple objectives, representing a batch of optimal candidate transfer paths under the current conditions. Further, it is determined whether a valid Pareto solution exists. If so, this batch of optimal candidate paths is organized into a two-layer transfer path matrix to generate a two-layer transfer path routing table. If not, the process returns to the input, readjusting constraints or algorithm parameters until a usable path is generated.

[0074] In this embodiment, the two-layer transit path routing table may include a first-layer path matrix and a second-layer path matrix. The first-layer path matrix includes multiple high-quality, complete candidate transit main paths (satisfying all constraints and having the best overall performance) that have been filtered. The second-layer path matrix includes candidate transit backup paths that correspond one-to-one with the candidate transit main paths or are alternatives within the same region. The candidate transit backup paths are a set of suboptimal paths composed of paths that minimize overlap with the network-dependent nodes (base stations, edge nodes) of the candidate transit main paths. The candidate transit backup paths can be used to ensure that when the candidate transit main paths are abnormal, they can be quickly activated and are less susceptible to the impact of the same network failure.

[0075] In some embodiments, the path matrix can be arranged in a certain structure, such as a structured arrangement of "node → base station → link," enabling rapid retrieval and retrieval of paths when needed. The two-layer transfer path routing table can be updated in real time, thus providing the latest path library for the next call or handover.

[0076] Figure 6 The flowchart illustrating the selection of a primary transfer path from a plurality of candidate transfer paths according to an embodiment of the present disclosure is shown in the illustration.

[0077] like Figure 6 As shown, in S102 above, a primary transfer path is selected from multiple candidate transfer paths, including S601a or S601b: S601a. ​​Select a candidate transit main path that meets the first preset condition from multiple candidate transit main paths, and determine the candidate transit main path as the transit main path when the candidate transit main path is available.

[0078] In this embodiment of the disclosure, among multiple candidate transfer paths, a computing node, such as a transfer path selection engine included in the computing node, can select a candidate transfer path that meets a first preset condition based on preset service priorities or weights. For example, preset weights can be assigned to each optimization objective, and then the comprehensive score of each candidate transfer path can be calculated, selecting the candidate transfer path with the highest comprehensive score. Alternatively, the path can be filtered sequentially according to service priorities, such as a first priority of latency <150ms, a second priority of highest stability (ensuring uninterrupted calls), etc., thereby selecting the candidate transfer path that best meets the conditions from multiple candidate transfer paths.

[0079] After selecting a candidate primary routing path that meets the first preset condition, the availability of the candidate primary routing path can be determined. If the candidate primary routing path is determined to be available, it can be designated as the primary routing path, routing can be executed, and a session link can be established.

[0080] S601b: When the candidate primary transfer path is unavailable, select the candidate backup path that meets the second preset condition from multiple candidate backup transfer paths as the primary transfer path.

[0081] If the candidate primary transfer path is determined to be unavailable (e.g., due to base station congestion or link failure), the candidate backup path that meets the second preset condition can be selected as the primary transfer path from multiple candidate backup paths. For example, a preset weight can be assigned to each optimization objective, and then the comprehensive score of each candidate backup path can be calculated, selecting the candidate backup path with the highest comprehensive score. Alternatively, the path can be filtered according to service priority, such as first priority being latency <150ms, second priority being highest stability (ensuring uninterrupted calls), etc., thus selecting the candidate backup path that best meets the conditions from multiple candidate backup paths.

[0082] Figure 7 A schematic diagram illustrating the selection of the primary routing path according to an embodiment of the present disclosure is provided.

[0083] like Figure 7 As shown, S701 is a multi-constraint input. The computation node can acquire multiple constraints and input them into the transition path selection engine. The transition path selection engine can solve for a set of Pareto front solutions based on the quality prediction value and constraints of each candidate transition path output by the transition path prediction model.

[0084] S702, Two-layer transit path routing table. The transit path selection algorithm can determine whether a valid Pareto solution exists. If it does, a two-layer transit path matrix is ​​generated to produce a two-layer transit path routing table. If it does not exist, the algorithm loops back to the input, readjusting constraints or algorithm parameters until a usable path is generated.

[0085] S703. Determine candidate primary transfer paths. Based on the two-layer transfer path routing table, select candidate primary transfer paths that meet the first preset condition through preset service priorities or weights.

[0086] S704. Determine the availability of the candidate transit main path. After selecting the candidate transit main path, its availability can be determined. If the candidate transit main path is determined to be available, proceed to S705a; if the candidate transit main path is determined to be unavailable, proceed to S705b.

[0087] S705a, Execute candidate transfer primary path routing.

[0088] S705b, Enable candidate transfer backup path. If the candidate transfer primary path is unavailable, a candidate transfer backup path that meets the second preset condition can be selected as the transfer primary path from multiple candidate transfer backup paths.

[0089] S706, Disaster Recovery Monitoring. Real-time monitoring of the communication quality of the current session can be performed by sliding a preset time window.

[0090] In this embodiment of the disclosure, the preset communication protocol includes at least a transport layer that supports fast reconnection and a session layer that includes a transfer tag.

[0091] The preset communication protocol can be a dedicated communication protocol set for the call transfer method provided in the embodiments of this disclosure. The preset communication protocol abandons the cumbersome process of traditional general protocols, optimizes from the transport layer, session layer and application layer, and, with dedicated encapsulation and QoS guarantee, achieves millisecond-level interaction with operators and high-definition call transmission.

[0092] In some embodiments, the protocol stack architecture of the preset communication protocol includes a transport layer, a session layer, and an application layer. The transport layer is based on the QUIC protocol and supports 0-RTT fast reconnection. 0-RTT fast reconnection allows for immediate reconnection after network outages, such as in elevators or tunnels where calls can be instantly reconnected after signal recovery. The session layer is a customized lightweight SIP extension protocol. By adding the X-CallTransfer extension header field, a transfer label is added, allowing operators to directly obtain important information such as call transfer and destination path. The application layer is a JSON-RPC 2.0 service instruction set, defining a simple and standardized instruction format, thereby ensuring unified instructions, extremely fast parsing, zero misunderstanding, and efficient execution of interactions between computing nodes and operators.

[0093] In some embodiments, the preset communication protocol also includes a specific data encapsulation format, including signaling frames using a TLV structure with a format of Type-Length-Value, where the type field contains an 8-bit protocol version and a 24-bit instruction encoding. The media stream supports adaptive selection of AV1 (High Definition Video) / EVS (Ultra-High Definition Audio) codecs, allowing for automatic switching of video / audio quality over the network, and the encapsulation format is SRTP over UDP.

[0094] In some embodiments, the preset communication protocol also includes a QoS guarantee mechanism to ensure smooth sessions with low latency. The QoS guarantee mechanism obtains the operator's entire network topology map in real time through the BGP-LS protocol, thereby knowing the congestion status of all base stations and lines, and establishing an end-to-end QoS tunnel by implementing a dynamic bandwidth reservation strategy.

[0095] According to embodiments of this disclosure, a preset communication protocol enables efficient and real-time data interaction between computing nodes and operator systems. This preset communication protocol allows for real-time synchronization of key information such as user call status (e.g., call suggestions, call in progress, call ended) and number status (e.g., number validity, whether marked as abnormal). During call transfer, the computing node can promptly negotiate and adjust transfer paths, priorities, and other parameters with the operator system based on real-time network conditions and user needs, ensuring a smooth transfer process. Simultaneously, digital signatures and certificate verification technologies are employed to guarantee the integrity and authenticity of the interactive data.

[0096] In embodiments of this disclosure, the computing node may further include an intelligent negotiation system, which may be deployed together with the transfer path selection engine within the computing node to collaboratively complete the path transfer process.

[0097] The intelligent negotiation system can negotiate with operators through a strategy reasoning engine. This engine can incorporate expert experience rules, a domain knowledge graph including operator policies, pricing plans, and network topology, and use the Takagi-Sugeno model to transform fuzzy network states into precise quantitative scores. In the pre-negotiation phase, operator resources are pre-locked through blockchain intelligent contracts. During real-time negotiation, a two-sided auction mechanism based on game theory is designed to complete resource bidding and matching within 50ms. Post-negotiation verification utilizes zero-knowledge proof technology to verify the compliance of the negotiation process.

[0098] Figure 8 A schematic diagram illustrating a preset communication protocol according to an embodiment of the present disclosure is provided.

[0099] like Figure 8As shown, the default communication protocol stack architecture includes an application layer 801, a session layer 802, and a transport layer 803. The application layer 801 includes the JSON-RPC 2.0 service instruction set, the session layer 802 is a custom SIP protocol including the X-CallTransfer extension header, and the transport layer 803 is the QUIC protocol, including 0-RTT fast reconnection. The network bearer includes signaling frames 804 and media streams 805. Signaling frames 804 use TLV encapsulation and BGP-LS QoS tunneling, while media streams 805 use SRTP over UDP encapsulation and support AV1 / EVS encoding / decoding.

[0100] Figure 9 A flowchart illustrating another call transfer method according to an embodiment of the present disclosure is shown.

[0101] like Figure 9 As shown, the call transfer method of this embodiment includes S901, S902 and S903: S901. After the call is connected, slide the preset time window and calculate the network information indicators of the current session within the preset time window in sequence.

[0102] In the embodiments of this disclosure, after the terminal device (calling end) and the called device establish a session connection, the session quality can be monitored in real time. The call quality can be monitored in real time by sliding a preset time window. The preset time window can be 10 seconds, 15 seconds, etc., and the network information indicators of the current session, such as QoS indicators (latency, packet loss rate, signal stability, etc.), are calculated within the time length corresponding to the preset time window.

[0103] S902. Determine whether each network information indicator exceeds the network information indicator threshold.

[0104] Furthermore, the network information indicators for the current session within each preset time window are compared with the network information indicator threshold. When the network information indicator exceeds the threshold, the corresponding time period can be marked as an abnormal time.

[0105] S903. When a number of consecutive network information indicators exceed the network information indicator threshold, and the number of consecutive indicators is greater than or equal to the preset number, switch from the primary transfer path to the candidate backup transfer path.

[0106] In the embodiments of this disclosure, a path switch is triggered when a series of consecutive time periods are marked as abnormal times, and the number of consecutive periods exceeds a preset number. For example, if three consecutive time periods are marked as abnormal times, a switch from the primary transfer path to a candidate backup transfer path can be triggered. Triggering the switch only when the number of consecutive periods exceeds the preset number effectively reduces the impact of occasional fluctuations on the switching process. When a session connection based on the primary transfer path experiences persistent abnormalities, a path switch is triggered, thereby facilitating seamless switching of the session link, ensuring uninterrupted calls and a seamless user experience.

[0107] Figure 10 A block diagram of a network device according to an embodiment of the present disclosure is schematically illustrated; like Figure 10 As shown, the network device 1000 of this embodiment includes a communication interface 1001, a memory 1002, and a processor 1003.

[0108] Communication interface 1001 is configured to perform wireless communication with a terminal device. Memory 1002 is used to store computer-readable instructions. Processor 1003 is used to execute the aforementioned computer-readable instructions, causing the network device to perform the aforementioned call transfer method.

[0109] Figure 11 A block diagram illustrating a non-transitory computer-readable storage medium according to an embodiment of the present disclosure is shown schematically. like Figure 11 As shown, a non-transitory computer-readable storage medium 1100 of this disclosure embodiment is used to store computer-readable instructions 1101, which, when executed by a processor, cause the processor to perform the call transfer method as described above.

[0110] Figure 12 A block diagram illustrating a computer program product according to an embodiment of the present disclosure is shown schematically. like Figure 12 As shown, a computer program product 1200 according to an embodiment of the present disclosure includes a computer program 1201, which, when executed by a processor, implements the call transfer method as described above.

[0111] The above description, with reference to the accompanying drawings, illustrates a call forwarding method, network device, medium, and computer program product according to embodiments of the present disclosure. Through the collection of multi-dimensional communication data and the construction of a forwarding path prediction model, the optimal primary forwarding path can be predicted in real time based on network conditions, user conditions, etc., thereby effectively improving overall communication efficiency. Simultaneously, a backup path can be determined at the same time as the primary forwarding path. When the primary forwarding path becomes unavailable or a degradation in session quality based on the primary forwarding path is detected through a disaster recovery mechanism, the system can promptly switch to the backup path, effectively addressing signal fluctuations and base station switching, reducing the probability of call interruption, and ensuring communication stability. During negotiations with operators, the use of preset communication protocols can optimize interaction with the operator's system, enabling dynamic adjustment of the forwarding strategy based on real-time conditions, thereby improving the intelligence level of communication.

[0112] The basic principles of this disclosure have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0113] The block diagrams of devices, apparatuses, devices, and systems disclosed herein are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0114] Additionally, as used herein, the "or" used in a list of items beginning with "at least one" indicates a separate list, such that a list of, for example, "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Furthermore, the word "exemplary" does not imply that the described example is preferred or better than other examples.

[0115] It should also be noted that in the systems and methods of this disclosure, the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered as equivalent solutions to this disclosure.

[0116] Various changes, substitutions, and modifications can be made to the technology described herein without departing from the teachings defined by the appended claims. Furthermore, the scope of the claims of this disclosure is not limited to the specific aspects of the processes, machines, manufactures, events, means, methods, and actions described above. Currently existing or later-developed processes, machines, manufactures, events, means, methods, or actions that perform substantially the same function or achieve substantially the same result as the corresponding aspects described above can be utilized. Therefore, the appended claims include such processes, machines, manufactures, events, means, methods, or actions within their scope.

[0117] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of this disclosure. Therefore, this disclosure is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0118] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.

Claims

1. A call forwarding method, characterized by, The method includes: A transfer path prediction model is constructed based on various communication data. Based on the aforementioned transfer path prediction model, at least one candidate transfer path is predicted. When a terminal device initiates a call access, a primary transfer path is selected from the multiple candidate transfer paths. Based on the main transfer path, transfer resources are negotiated with the operator using a preset communication protocol to connect the call.

2. The call forwarding method of claim 1, wherein, The method further includes: Based on network quality information, the various communication data are obtained, wherein the communication data includes at least one or more of physical layer data, network layer data, and service layer data. At least a first dynamic weight is assigned to the physical layer data, a second dynamic weight is assigned to the network layer data, and a third dynamic weight is assigned to the service layer data; and Based at least on the physical layer data and the first dynamic weight, the network layer data and the second dynamic weight, and the service layer data and the third dynamic weight, transfer path prediction data is generated, and the transfer path prediction data is used at least to generate the transfer path prediction model.

3. The call forwarding method according to claim 2, characterized by, The switching path prediction model includes at least a hybrid neural network module and a reinforcement learning module. The hybrid neural network module consists of at least a front-end feature extraction submodule, a medium-to-long-term feature learning submodule, and an attention mechanism submodule.

4. The call forwarding method according to claim 3, characterized by, The prediction of at least one candidate transfer path based on the transfer path prediction model includes: The transfer path prediction data is input into the transfer path prediction model, and the transfer path prediction model outputs the quality prediction value of at least one candidate transfer path. Based on multiple constraints and the quality prediction value of at least one candidate transfer path, a two-layer transfer path routing table is generated. The two-layer transfer path routing table includes a first-layer path matrix and a second-layer path matrix. The first-layer path matrix is ​​used to store multiple candidate primary transfer paths, and the second-layer path matrix is ​​used to store multiple candidate backup transfer paths.

5. The call forwarding method according to claim 4, characterized by, The step of selecting a primary transfer path from the multiple candidate transfer paths includes: Select the candidate transit main path that meets the first preset condition from multiple candidate transit main paths, and determine the candidate transit main path as the transit main path when the candidate transit main path is available; or When the candidate primary routing path is unavailable, select the candidate backup routing path that meets the second preset condition from among the multiple candidate backup routing paths as the primary routing path.

6. The call forwarding method of claim 1, wherein, The preset communication protocol includes at least a transport layer that supports fast reconnection and a session layer that includes a transfer tag.

7. The call forwarding method of claim 5, wherein, The method includes: After the call is connected, a preset time window is slid, and the network information indicators of the current session within the preset time window are calculated sequentially. Determine whether each of the network information indicators exceeds a network information indicator threshold; and When a number of consecutive network information indicators exceed the network information indicator threshold, and the number of consecutive indicators is greater than or equal to a preset number, the system switches from the primary transfer path to the candidate backup transfer path.

8. A network device, comprising: include: The communication interface is configured to perform wireless communication with the terminal device. Memory, used to store computer-readable instructions; as well as A processor for executing the computer-readable instructions, causing the network device to perform the call transfer method as described in any one of claims 1 to 7.

9. A non-transitory computer-readable storage medium storing computer-readable instructions, the computer-readable instructions comprising: When computer-readable instructions are executed by the processor, the processor performs the call transfer method as described in any one of claims 1 to 7.

10. A computer program product comprising a computer program, characterized in that, When a computer program is executed by a processor, it implements the call transfer method as described in any one of claims 1 to 7.

Images