Seamless cloud connection method, system, terminal device, gateway device and medium for hybrid multipath transmission in satellite communication scenario

Abstract

This application provides a seamless cloud connection method, system, terminal device, gateway device, and medium for hybrid multipath transmission in satellite communication scenarios. The method involves the terminal device and the cloud access gateway negotiating multipath transmission enable to determine the set of available physical interfaces; sending path challenge frames through each physical interface to obtain round-trip time and congestion window status, and performing delay prediction processing to determine the predicted delay for the next moment; establishing the target transmission path based on the service type of the data packets to be sent, the predicted delay, and the congestion window status; calculating out-of-order compensation delay based on the delay difference and injecting it into the low-latency path interface, ensuring that data packets arrive synchronously at the gateway. This application alleviates the problems of line congestion caused by heterogeneous delays in satellite and ground links and connection interruptions caused by IP changes, achieving bandwidth aggregation and millisecond-level seamless switching, and improving transmission stability in hybrid networking environments.

Description

Technical Field

[0001] This application relates to the fields of transport layer protocols and hybrid cloud connectivity technology. Specifically, it relates to a seamless cloud connectivity method, system, terminal device, gateway device, and medium for hybrid multipath transmission in satellite communication scenarios. Background Technology

[0002] In modern enterprise communication architectures, to improve office needs in remote areas or mobile scenarios, a hybrid networking mode consisting of satellite links and terrestrial fiber optic leased lines is typically adopted. This mode aims to provide reliable connectivity support for accessing cloud virtual private networks (VPCs) through redundant backup and traffic sharing of multiple physical links.

[0003] Existing hybrid path transmission schemes are mostly based on the traditional five-tuple binding mechanism for data routing. This scheme first establishes a fixed transmission path between the terminal and the gateway; then, by monitoring the link connectivity, it triggers a switchover action when the primary link fails; finally, it redirects traffic to the backup link.

[0004] However, the aforementioned existing technologies have serious drawbacks in specific satellite communication scenarios. Due to the high physical dynamism of satellite links, link switching is often accompanied by changes in Internet Protocol (IP) addresses, causing traditional five-tuple-based connections to be directly disconnected, resulting in low service continuity. At the same time, there is a significant latency heterogeneity between satellite links and terrestrial fiber optic links (usually satellite latency is much greater than fiber optic latency). Direct multipath parallel transmission will cause severe out-of-order congestion at the receiving end. Data packets must wait in the buffer for a long time for the arrival of "slow path" packets, causing serious line-end congestion problems and severely affecting the interactive experience of hybrid cloud connections. Summary of the Invention

[0005] This application provides a seamless cloud connection method, system, terminal device, gateway device, and medium for hybrid multipath transmission in satellite communication scenarios, so as to at least alleviate the above-mentioned technical problems.

[0006] A seamless cloud connectivity method for hybrid multipath transmission in satellite communication scenarios includes: Step 1: The terminal device and the cloud access gateway perform multipath transmission enable negotiation to determine the set of available physical interfaces consisting of satellite links and terrestrial leased lines; Step 2: The terminal device sends path challenge frames through each physical interface in the set of available physical interfaces to obtain the round-trip time and congestion window status of each path, and performs prediction processing on the round-trip time of each path to determine the predicted delay at the next moment. Step 3: The terminal device performs scheduling decisions based on the service type of the data packets to be sent, the predicted delay of the next time step for each path, and the congestion window status, in order to establish the target transmission path for each data packet to be sent. Step 4: The terminal device calculates the out-of-order compensation delay based on the delay difference in the set of available physical interfaces, and injects it into the interface of the target transmission path that belongs to the low-latency path, so that the data packet to be sent arrives synchronously at the cloud access gateway.

[0007] Optionally, in step 1, the terminal device and the cloud access gateway perform multipath transmission enable negotiation to determine the set of available physical interfaces, including: the terminal device and the cloud access gateway negotiate multipath transmission extension parameters through handshake messages and allocate globally unique connection identifiers to establish a multipath transmission protocol connection; the terminal device, in response to the connection identifier, maintains the logical continuity of the multipath transmission protocol connection when the physical Internet Protocol address changes; in the multipath transmission protocol connection, the terminal device, in response to the connection identifier, obtains the physical Internet Protocol addresses of the satellite link and the terrestrial leased line link and performs an interface registration action to establish the set of available physical interfaces composed of the satellite link and the terrestrial leased line link after the interface registration action is performed.

[0008] Optionally, in step 2, sending path challenge frames through each physical interface in the set of available physical interfaces to obtain the round-trip time of each corresponding path includes: the terminal device periodically sending path challenge frames carrying transmission timestamps on the satellite link and the terrestrial leased line link included in the set of available physical interfaces, and obtaining path response frames fed back by the cloud access gateway; the terminal device captures the arrival time of the path response frame, and performs a difference operation based on the arrival time and the transmission timestamp to calculate the round-trip time of each corresponding physical interface.

[0009] Optionally, in step 2, performing prediction processing on the round-trip delay of each path to determine the predicted delay at the next moment includes: the terminal device acquiring the current round-trip delay observation value of each path, and performing a difference operation between the current round-trip delay observation value and the state prediction value at the previous moment to extract the prediction residual; the terminal device acquiring the Kalman gain corresponding to each path, and using the Kalman gain to perform weighted correction processing on the prediction residual to generate an updated delay state estimate value and determine it as the predicted delay at the next moment.

[0010] Optionally, in step 3, a scheduling decision is made based on the service type of the data packet to be sent to establish a target transmission path, including: if the service type of the data packet to be sent is identified as a low-latency interactive service, then the physical interface with the smallest predicted latency at the next moment is selected as the target transmission path from the set of available physical interfaces; if the service type is identified as a high-bandwidth throughput service, then the physical interface with the largest remaining capacity in the congestion window state is selected as the target transmission path from the set of available physical interfaces.

[0011] Optionally, executing scheduling decisions and establishing a target transmission path further includes: when the congestion window state reaches a preset full load threshold, the terminal device triggers an overflow scheduling mechanism and uses the overflow scheduling mechanism to redirect subsequently arriving data packets to physical interfaces belonging to high-latency paths in the set of available physical interfaces, so as to update the target transmission path.

[0012] Optionally, in step 4, calculating the out-of-order compensation delay based on the delay difference in the set of available physical interfaces includes: the terminal device obtaining the first round-trip delay corresponding to the satellite link and the second round-trip delay corresponding to the ground leased link in the set of available physical interfaces, calculating the algebraic difference between the first round-trip delay and the second round-trip delay, and establishing the algebraic difference as the out-of-order compensation delay.

[0013] Optionally, the terminal device runs a multi-path management layer, which maintains a path list consisting of the identity identifiers of each path, and monitors the congestion window status and round-trip delay corresponding to each path in real time based on the path list, so as to generate path quality assessment parameters and feed them back to the scheduling decision process in step 3.

[0014] Optionally, the cloud access gateway is further configured with an out-of-order compensation buffer. The cloud access gateway receives the data packets to be sent transmitted through the set of available physical interfaces, and uses the out-of-order compensation buffer to perform reordering processing on the data packets to be sent, so as to restore the ordered sub-service flow and deliver it to the cloud virtual private cloud.

[0015] Optionally, before the data packet to be sent is physically sent through the target transmission path, the method further includes: the terminal device assigning consecutive packet sequence numbers to the data packet to be sent within the multipath transmission protocol stack, and using the multipath transmission protocol stack to establish an association mapping relationship between the packet sequence numbers and each physical interface in the target transmission path, so that the cloud access gateway performs data integrity verification based on the association mapping relationship.

[0016] Optionally, in step 2, obtaining the congestion window state for each path includes: the terminal device capturing service confirmation packets fed back by the cloud access gateway through the set of available physical interfaces in real time, and extracting the round-trip byte weights and packet loss awareness features for each path from the service confirmation packets; the terminal device using the round-trip byte weights and packet loss awareness features to drive a preset congestion control logic to perform iterative calculation processing to generate dynamic values ​​reflecting the real-time path throughput capacity and establish them as the congestion window state for each path.

[0017] A seamless cloud connectivity method for hybrid multipath transmission in a satellite communication scenario includes: Step 1: A cloud access gateway and a terminal device perform multipath transmission enable negotiation to determine a set of available physical interfaces consisting of satellite links and terrestrial leased lines; Step 2: The cloud access gateway receives path challenge frames sent by the terminal device through each physical interface in the set of available physical interfaces and feeds back path response frames, so that the terminal device can obtain the round-trip delay and congestion window status of each path and perform prediction processing on the round-trip delay of each path to determine the predicted delay at the next moment; Step 3: The cloud access gateway receives data packets to be transmitted by the terminal device through the established target transmission path after the terminal device performs scheduling decisions based on the service type of the data packets to be transmitted and the predicted delay at the next moment and the congestion window status of each path; Step 4: The cloud access gateway receives multiple data packets to be transmitted synchronously after the terminal device calculates based on the delay difference and injects out-of-order compensation delay, reorders the multiple data packets based on the packet sequence number of the data packets to be transmitted using an out-of-order compensation buffer, and performs data integrity verification.

[0018] A seamless cloud connectivity system for hybrid multipath transmission in a satellite communication scenario includes a terminal device and a cloud access gateway. The terminal device and the cloud access gateway perform multipath transmission enable negotiation to determine a set of available physical interfaces consisting of satellite links and terrestrial leased lines. The terminal device sends path challenge frames through each physical interface in the set of available physical interfaces to obtain the round-trip time and congestion window status of each path, and performs prediction processing on the round-trip time of each path to determine the predicted delay for the next moment. The terminal device performs scheduling decisions based on the service type of the data packets to be sent, the predicted delay for the next moment of each path, and the congestion window status to establish the target transmission path for each data packet to be sent. The terminal device calculates out-of-order compensation delay based on the delay difference in the set of available physical interfaces and injects it into the interfaces of the target transmission path that belong to the low-latency path, so that the data packets to be sent arrive synchronously at the cloud access gateway.

[0019] A terminal device supporting hybrid multipath transmission includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement a seamless cloud connection method for hybrid multipath transmission in a satellite communication scenario as described in any one of the claims of this application.

[0020] A cloud access gateway device includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement a seamless cloud connection method for hybrid multipath transmission in a satellite communication scenario as described in any one of the claims of this application.

[0021] A computer storage medium having computer-executable instructions stored thereon, the computer-executable instructions being executed to perform the method described in any of the claims of this application.

[0022] A computer storage medium having computer-executable instructions stored thereon, the computer-executable instructions being executed to perform the method described in any of the claims of this application.

[0023] The technical advantages of the technical solution provided in this application are: This application provides a seamless cloud connectivity method for hybrid multipath transmission in satellite communication scenarios. Addressing the connection interruptions and line-end blockages inherent in traditional solutions in hybrid satellite-ground networks, this method significantly improves transmission performance by introducing multipath management and heterogeneous latency prediction mechanisms. Compared to traditional five-tuple-based binding techniques, this application establishes multipath protocol connections based on global connection identifiers, decoupling the transport layer from the underlying Internet Protocol (IP) addresses. This ensures that service connections maintain a high degree of logical continuity even with address drift caused by frequent satellite handovers, mitigating session interruptions caused by physical link jumps.

[0024] To address the heterogeneity of time delay between satellite and ground, this application utilizes the Kalman filter algorithm to predict the round-trip time delay for each path. Compared to simple instantaneous sampling, the predicted delay state values ​​exhibit higher smoothness and jitter resistance. Combined with service type-aware scheduling decisions, it can accurately guide delay-sensitive services to fiber optic paths while mapping high-bandwidth services to satellite links.

[0025] The core innovation lies in the out-of-order compensation delay injection mechanism proposed in step 4. Traditional multipath protocols, unable to coordinate the arrival times of different paths, cause severe congestion in the receiver's buffer due to waiting for "slow satellite packets." This application proactively introduces a tiny artificial delay into the "fast ground path" by calculating the algebraic delay difference between the satellite and ground links word-to-word, enabling data packets from different paths to arrive at the gateway synchronously with extremely high consistency. This "delay-for-synchronization" strategy eliminates receiver-end congestion at the source, significantly reduces reordering overhead, and improves seamless switching and efficient bandwidth aggregation in hybrid cloud connections across heterogeneous cross-domain environments. Attached Figure Description

[0026] Figure 1 This is a flowchart of a seamless cloud connection method for hybrid multipath transmission in a satellite communication scenario, as described in this application. Detailed Implementation

[0027] like Figure 1 As shown in the figure, a seamless cloud connection method for hybrid multipath transmission in a satellite communication scenario is provided in an embodiment of this application, comprising: Step 1: The terminal device and the cloud access gateway perform multipath transmission enable negotiation to determine the set of available physical interfaces consisting of satellite links and terrestrial leased lines; Step 2: The terminal device sends path challenge frames through each physical interface in the set of available physical interfaces to obtain the round-trip time and congestion window status of each path, and performs prediction processing on the round-trip time of each path to determine the predicted delay at the next moment. Step 3: The terminal device performs scheduling decisions based on the service type of the data packets to be sent, the predicted delay of the next time step for each path, and the congestion window status, in order to establish the target transmission path for each data packet to be sent. Step 4: The terminal device calculates the out-of-order compensation delay based on the delay difference in the set of available physical interfaces, and injects it into the interface of the target transmission path that belongs to the low-latency path, so that the data packet to be sent arrives synchronously at the cloud access gateway.

[0028] In this application, the multipath transport protocol may include, but is not limited to, QUIC or MP-QUIC.

[0029] Optionally, in step 1, the terminal device and the cloud access gateway perform multipath transmission enable negotiation to determine the set of available physical interfaces, including: the terminal device and the cloud access gateway negotiate multipath transmission extension parameters through handshake messages and allocate globally unique connection identifiers to establish a multipath transmission protocol connection; the terminal device, in response to the connection identifier, maintains the logical continuity of the multipath transmission protocol connection when the physical Internet Protocol address changes; in the multipath transmission protocol connection, the terminal device, in response to the connection identifier, obtains the physical Internet Protocol addresses of the satellite link and the terrestrial leased line link and performs an interface registration action to establish the set of available physical interfaces composed of the satellite link and the terrestrial leased line link after the interface registration action is performed.

[0030] Preferably, the multi-path management layer performs a low-level scan of the physical communication units carried by the terminal through its configured link determination interface to determine the currently available transmission path resources. Within the determined implementation, it extracts the low-level link characteristics associated with the high-throughput satellite communication antenna and the electrical signal physical state associated with the fiber optic access unit. This essentially provides the transport layer protocol stack with a low-level resource view possessing physical attributes, enabling real-time capture of the connectivity status of satellite links and terrestrial leased lines. Determining these raw hardware states provides a deterministic hardware foundation for subsequent construction of logical connections across heterogeneous media, establishing the boundaries of physical layer resources, thereby supporting subsequent processing actions to perform efficient traffic scheduling across multiple physical paths.

[0031] Preferably, based on the physical carrier identified above, the terminal device drives its internal protocol control logic to generate multipath enable configuration parameters that are discretely distributed along the time axis according to a preset multipath transmission mechanism. In the specific technical implementation of this generation, the multipath enable request flag, protocol negotiation sequence, and preset congestion control initial strategy execution logic are encapsulated. Since the coverage intensity generated by the satellite during its orbital operation dynamically evolves over time, corresponding connectivity parameters are generated for each discrete time slice according to a preset time step. This achieves a transformation from physical interface awareness to protocol logic definition. The generated set of parameters with temporal characteristics enables the terminal and gateway to reach a consensus on multipath extension characteristics, providing logical guidance for establishing a parallel transmission environment on heterogeneous links.

[0032] Preferably, the multipath management layer performs identity establishment processing for logical sessions while generating the aforementioned protocol parameters, thereby generating a globally unique connection identifier. Unlike the traditional five-tuple binding method using the Internet Protocol (IP) source address, destination address, and port number, the generated connection identifier, as a logical label that does not change with the underlying network topology, is directly locked within the lifetime of the transport layer session. This effectively achieves complete decoupling of the transport connection from the underlying physical network addressing characteristics. This discretely generated identifier can cope with coverage beam switching caused by high-speed satellite movement and IP address drift. Responding to transient changes in the underlying address, the logical continuity of the transport connection can still be maintained, alleviating the technical bottleneck of connection interruption in hybrid satellite-ground networks.

[0033] Preferably, the terminal device drives its communication control module to perform continuous stream output in response to the generated negotiation signaling containing the connection identifier. During execution, the negotiation parameter set is encapsulated into a transport layer handshake message, and the User Datagram Protocol (UDP) is used as the physical bearer channel to continuously pump the negotiation data stream to the cloud access gateway. This action realizes the physical migration of control commands from the terminal edge to the cloud gateway. By executing this continuous signaling stream output, the gateway can obtain and confirm the terminal's concurrent path requests in real time, thereby establishing a logical framework at the transport layer capable of accommodating multiple physical sub-streams. This streaming output method avoids the handshake frequency limitation problem caused by the large satellite-to-ground round-trip delay, providing immediate signaling stream support for subsequent bandwidth aggregation and sub-stream registration.

[0034] Preferably, after confirming the completion of the above negotiation, the terminal device utilizes the established logical channel to continuously stream the interface registration signal for the captured physical access address. In this step, the physical Internet Protocol (IP) addresses assigned to satellite links and terrestrial leased line links are read respectively, and these address information is mapped to the subordinate entries of the connection identifier. For this purpose, the underlying interface identity characteristics are sent to the path management matrix via the signaling interface. By performing this streaming output and logical registration of address information, a dynamic conversion from raw physical hardware addressing to multipath protocol stack logical paths is achieved. This action enables the gateway to accurately identify the source of different paths based on the registration information in subsequent forwarding stages, improving the correct reassembly and delivery of heterogeneous data streams at the receiving end.

[0035] Preferably, the multi-path management layer aggregates the multi-dimensional path resources after the aforementioned execution interface registration actions, thereby ultimately forming a set of available physical interfaces characterizing the terminal's concurrent transmission capabilities. In the specific technical implementation, a digital resource index master is established, containing physical attributes of satellite links, physical attributes of terrestrial leased links, and a unified associated connection identifier. This formed set is no longer an isolated hardware list, but a logical resource library with multi-path concurrent forwarding semantics, realizing the integration of satellite and terrestrial fiber optic resources within the protocol stack. By forming this set of available physical interfaces, a unified operational benchmark is provided for subsequent execution of Kalman filter-based latency prediction and service type-based scheduling decisions, improving the connection stability of hybrid multi-path transmission in highly dynamic, large-latency-difference scenarios.

[0036] In summary, this application demonstrates that the determination of physical interfaces provides the initial resource view for the entire process, directly determining the validity of subsequent configuration parameter generation. The generated discrete parameters and connection identifiers provide the core content for stream output, establishing the logical foundation of the transport layer protocol stack. Next, continuous signaling stream output actions complete protocol negotiation and address registration from the terminal to the gateway. Finally, the orderly organization of this information forms a set of available physical interfaces, enabling the transformation of heterogeneous physical links into a unified transport logical space. This series of collaborative actions leverages the stability of transport layer identifiers to compensate for the volatility of underlying physical addresses, achieving transparent support for cloud connectivity from the satellite network.

[0037] Optionally, in step 2, sending path challenge frames through each physical interface in the set of available physical interfaces to obtain the round-trip time of each corresponding path includes: the terminal device periodically sending path challenge frames carrying transmission timestamps on the satellite link and the terrestrial leased line link included in the set of available physical interfaces, and obtaining path response frames fed back by the cloud access gateway; the terminal device captures the arrival time of the path response frame, and performs a difference operation based on the arrival time and the transmission timestamp to calculate the round-trip time of each corresponding physical interface.

[0038] Preferably, the multi-path management layer, through its configured link determination interface, performs a real-time scan of the underlying electrical characteristics of the satellite communication unit and the ground access unit carried by the terminal to determine the initial physical connectivity of each member link in the set of available physical interfaces. In the determined specific implementation, the signal-to-noise ratio distribution and synchronization state characteristics of the satellite link interface and the ground leased network card interface at the physical layer are extracted. This essentially establishes a deterministic physical resource entry point for the transmission protocol stack, enabling real-time capture of underlying state changes caused by high-speed satellite motion or ground fiber optic jitter. Determining these original hardware indicators provides a physical basis for subsequent injection of probe payloads, enabling subsequent processing actions to be precisely anchored to a transmission channel with carrying capacity.

[0039] Preferably, within its transmission protocol stack, the terminal device generates a set of probe commands discretely distributed along the time axis for the determined link physical resources using preset path verification logic. In the specific technical implementation, each probe command unit is injected with an original transmission timestamp generated by a local high-precision timer. Since the coverage relationship between the satellite and the terminal exhibits dynamic characteristics as the orbit evolves, these generated probe commands are arranged into a series of path challenge frames discretely distributed along the time axis. This step realizes the transformation from a static physical interface view to logical probe commands with time-domain characteristics. These time-stamped frame data provide a digital measurement benchmark for subsequent quantitative measurement of the round-trip delay of the hybrid link.

[0040] Preferably, the terminal device drives its internal signaling scheduling engine to perform continuous stream output processing on the generated set of probe commands containing discrete timestamps. During the execution of continuous stream output, the path challenge frames are continuously pumped into the space network via satellite links and ground-based dedicated lines at a preset sampling frequency. This action achieves the physical diffusion of the probe signal from the terminal edge to the cloud access gateway, forming an observation stream characterizing the real-time features of the path. By executing this continuous stream output, even in transient situations with fluctuating link quality, the gateway can receive a sufficient density of sampled data, providing immediate sampling stream support for subsequently establishing a high-reliability time-delay evolution model.

[0041] Preferably, the protocol control logic of the terminal device continuously monitors the downlink channels of each physical interface to determine the feedback signaling transmitted back by the cloud access gateway in response to the aforementioned probe stream. In this step, a protocol feature recognition model is used to extract the corresponding path response frame from the downlink bitstream. In the determined specific implementation, the echo timestamp encapsulated in the feedback frame is extracted synchronously, and the instantaneous local time at which the frame reaches the terminal hardware interface is determined. This determination action captures the physical boundary points where the signaling completes round-trip interaction on the heterogeneous satellite-ground path. By capturing these feedback samples, the physical feedback closed loop of the probe logic is established, providing raw observational data for subsequent calculation of the dynamic operating parameters of each path.

[0042] Preferably, the path quality calculation module built into the terminal device aggregates the determined feedback time and transmission timestamp, generates the real-time round-trip delay for each physical interface by performing differential comparison, and simultaneously extracts the congestion window state for each path. During the generation process, the algebraic difference between the transmission and reception timestamps of each pair is calculated, and combined with the number of acknowledgment bytes from the gateway feedback, a dynamic value reflecting real-time throughput capability is generated. This processing step transforms discrete message interactions into quantitative path quality characteristics, forming link profile data characterizing the transmission capability of each sub-stream. This data result not only reflects the current path load but also serves as a core input in the subsequent time-domain prediction process.

[0043] Preferably, the latency prediction engine of the terminal device acquires the generated link profile data and performs recursive correction processing on the round-trip latency using a preset state estimation model, thereby ultimately forming the next-moment predicted latency characterizing the path evolution trend. In the specific technical implementation, the difference between the current round-trip latency observation and the previous moment's state prediction value is calculated to extract the prediction residual, and a weighted correction is performed on the predicted state using the generated gain factor. This is equivalent to using the latency stability generated by orbit determinism to smooth out random disturbances caused by the wireless environment. By forming the next-moment predicted latency, a high-precision forward-looking parameter reference is provided for subsequent scheduling decisions with service-aware characteristics, alleviating the data synchronization problem in environments with large latency differences between satellite and fiber optic heterogeneous paths.

[0044] In summary, this application provides the hardware foundation for the entire detection chain by defining the physical interface; the generated discrete detection commands provide the payload for the streaming output. By executing continuous streaming output and corresponding feedback, a closed-loop observation network spanning the heterogeneous space between space and Earth is constructed. Finally, real-time digital twin characterization of the hybrid path quality is achieved through the generated round-trip delay characteristics and the formed predicted delay. Therefore, this application enables the scheduling decision-making process to be based on path quality parameters with predictable attributes, improving the transmission stability and low-latency characteristics of hybrid cloud connections in complex inter-satellite topologies.

[0045] Optionally, in step 2, performing prediction processing on the round-trip delay of each path to determine the predicted delay at the next moment includes: the terminal device acquiring the current round-trip delay observation value of each path, and performing a difference operation between the current round-trip delay observation value and the state prediction value at the previous moment to extract the prediction residual; the terminal device acquiring the Kalman gain corresponding to each path, and using the Kalman gain to perform weighted correction processing on the prediction residual to generate an updated delay state estimate value and determine it as the predicted delay at the next moment.

[0046] Preferably, in the specific implementation process, the terminal device uses its built-in path monitoring module to perform real-time capture of the protocol response frames fed back by the cloud access gateway to determine the link status. In the specific technical implementation of the determination, the path monitoring module uses a high-speed signaling listening interface to determine the path response frames containing echo timestamps transmitted back through the set of available physical interfaces. This is equivalent to capturing the physical boundaries of the signaling traveling back and forth in the heterogeneous satellite-ground path, thereby obtaining the original time-domain pulse signal for subsequent quantitative measurement of link quality. By determining these feedback payloads, the on / off status and response rhythm of the current physical link can be identified in real time, providing data source support for the subsequent generation of delay characteristics with definite physical semantics.

[0047] Preferably, the terminal device acquires the determined feedback payload and drives its internal protocol solution logic to perform differential operations to generate the round-trip delay observations discretely distributed along the time axis. During the generation process, the transmission time record in the response frame is retrieved, and the instantaneous local time at which the frame reaches the terminal hardware interface is determined. An algebraic subtraction of the local time and the transmission time is performed to generate a numerical sample reflecting the transient quality of the link in each detection cycle. Since the spatiotemporal displacement generated by the satellite during its orbital operation is continuous but exhibits discrete sampling characteristics at the detection level, these samples form the current round-trip delay observations discretely distributed along the time axis. This step realizes the conversion from the raw bitstream signal to quantitative path characteristics, enabling the capture of micro-delay jumps caused by deep fading in the satellite wireless channel or network congestion.

[0048] Preferably, the latency prediction engine built into the terminal device acquires the generated current round-trip latency observation value and performs iterative comparison processing between this observation value and the pre-stored reference state to execute continuous stream output. In the specific process of executing continuous stream output, the latency prediction engine retrieves the previous time-stamped state prediction value recorded in memory and performs a difference operation with the current round-trip latency observation value to extract the prediction residual reflecting the model prediction bias. This action achieves precise separation of the difference between real-time observed data and the expected state. This uninterrupted residual calculation constitutes a prediction data stream with temporal closed-loop characteristics, providing a quantitative input reflecting the intensity of environmental fluctuations for subsequent weight allocation via Kalman gain.

[0049] Preferably, the delay prediction engine synchronously drives its internal Kalman filter operator to perform gain-weighted correction processing on the aforementioned prediction residuals, thereby further executing continuous stream output. In this stage, adaptive calculation is performed based on the current prediction error covariance and observation noise variance to obtain the Kalman gain reflecting the reliability weights of each path. Subsequently, the delay prediction engine uses the Kalman gain to perform weighted correction processing on the prediction residuals, generating a set of dynamic correction streams reflecting the path evolution trend. This is equivalent to using the delay stability generated by orbital operation to smooth out random disturbances caused by the wireless environment, thereby identifying the true performance trend in complex mixed path environments and alleviating the problem of single sampling being susceptible to sudden interference in traditional schemes.

[0050] Preferably, the latency prediction engine aggregates the previous time-stamped state prediction value and the weighted corrected value, and performs state update synthesis processing to form the next-time predicted latency, which characterizes the dynamic profile of hybrid link latency. In the specific technical implementation, the prediction benchmark and correction component are logically fused to generate an updated latency state estimate reflecting the endpoint of the new time slot evolution. This value not only encompasses the smoothing characteristics of historical paths but also proactively includes the expected direction of the next sampling period, thus ultimately forming the next-time predicted latency characterizing the dynamic characteristics of the heterogeneous links between satellite and ground-based dedicated lines. This processing step marks the final closed loop from discrete observation samples to a spatiotemporally predictive path profile, providing a digital foundation for achieving accurate traffic sharing across hybrid multipaths.

[0051] In summary, this application firstly provides a hardware-level trigger source for the entire prediction chain by determining the feedback signaling action; secondly, the generated discrete observations transform physical phenomena into digital features, directly determining the original input quality of subsequent stream output stages. Next, noise filtering and trend capture are achieved through continuous Kalman state iteration stream output; finally, the resulting next-moment prediction delay acts inversely on the scheduling decision module. This cross-level closed-loop support enables the pre-establishment of optimal transmission paths for different service types, alleviating the line-end congestion problem in environments with large space-to-ground latency differences, and achieving seamless hybrid cloud connectivity in scenarios with drastic topology changes.

[0052] Optionally, in step 3, a scheduling decision is made based on the service type of the data packet to be sent to establish a target transmission path, including: if the service type of the data packet to be sent is identified as a low-latency interactive service, then the physical interface with the smallest predicted latency at the next moment is selected as the target transmission path from the set of available physical interfaces; if the service type is identified as a high-bandwidth throughput service, then the physical interface with the largest remaining capacity in the congestion window state is selected as the target transmission path from the set of available physical interfaces.

[0053] Preferably, the packet scheduler uses its built-in message monitoring interface to perform feature sampling on the raw bitstream flowing from the user application layer to determine the service attribution characteristics of the currently pending message. In the determined implementation, deep message detection logic is invoked to determine the protocol feature bits in the data packet to be sent and matched against a pre-set service profile library to identify the service type associated with the message (e.g., jitter-sensitive real-time audio / video streams or large file transfer streams with high bandwidth requirements). This essentially transforms disordered application traffic into service vectors with scheduling semantics, enabling the pre-establishment of forwarding priority weights before physical layer transmission. Determining the service type establishes the logical trigger source for differentiated scheduling decisions, providing definite service feature support for subsequent processing actions to accurately guide to the matching physical link.

[0054] Preferably, upon determining the aforementioned business requirements, the multi-path management layer drives its configured path quality observation operator to perform high-frequency probing on the real-time load and performance evolution of each link in the available physical interface set, generating path performance evaluation vectors discretely distributed along the time axis. In the specific technical implementation, the predicted delay for the next moment calculated for each physical interface and the feedback congestion window state are retrieved and mapped into a set of discrete state codes describing the link's carrying capacity according to a preset time step. Since the channel quality of the satellite-to-ground link exhibits dynamic fluctuations with the satellite's orbital position, these generated discrete vectors achieve a time-domain digital representation of the dynamics of the hybrid link. This step realizes the conversion from physical channel observations to transport layer scheduling instructions, providing a time-predictable quantitative basis for establishing the final target transmission path.

[0055] Preferably, in response to the determined service type being a low-latency interactive service, the packet scheduler performs a filtering and mapping process for the optimal low-latency path. In this step, the generated path performance evaluation vector is retrieved, and discrete state values ​​reflecting latency characteristics are compared between the satellite link and the terrestrial leased line link. The packet scheduler selects the physical interface with the lower predicted latency value for the next moment and locks it as the target transmission path for the current data packet to be sent. This effectively utilizes the high time-efficiency physical characteristics of terrestrial optical fiber to perform targeted acceleration of low-latency messages, thereby effectively shielding the inherent long transmission latency of the satellite link. Establishing this target path improves the ability of interactive control signaling or voice streams to reach the cloud access gateway with better timing consistency.

[0056] Preferably, in response to the determined service type being a high-bandwidth throughput service, the packet scheduler performs path load balancing processing to maximize throughput. In this step, the congestion window status of each path in the set of available physical interfaces is determined, and by calculating the remaining transmission quota for each path, the physical interface with the highest remaining capacity is selected as the target transmission path. If the congestion window of the ground fast path is found to have reached the full load threshold, the packet scheduler automatically triggers sub-flow sharing logic, redirecting subsequent high-bandwidth service flows to satellite links with high-throughput characteristics. This achieves full utilization of redundant bandwidth in hybrid networking, avoids queuing congestion caused by single-path bandwidth limitations, and improves the overall throughput stability of cloud connectivity services in a heterogeneous space-ground environment.

[0057] Preferably, the packet scheduler drives its built-in protocol distribution engine to perform continuous stream output to the physical interface buffer for the message sequence with the established target mapping relationship. During the continuous stream output process, the data packets to be sent are continuously pumped to the associated satellite communication component or terrestrial communication component according to the transmission frequency corresponding to each physical interface. This action realizes the real-time migration of logical forwarding instructions to the physical signal transmission queue, forming a service code stream representing the multi-path cooperative state. By performing this uninterrupted stream output processing, each frame of message can be injected into the corresponding transmission channel in real time according to the scheduling decision, alleviating the service interruption problem caused by a sudden drop in link quality in traditional single-path transmission, and providing a definite data input for the receiving end to perform the final reassembly.

[0058] Preferably, the multi-path management layer aggregates the output status and mapping feedback of the aforementioned interfaces to ultimately form benchmark traffic mapping data characterizing the logical topology of multi-path transmission. In the specific technical implementation, a traffic tracing master based on connection identifiers is established, creating a chained index between each data packet assigned a sequence number and the selected physical path. This is equivalent to constructing a "forwarding profile" reflecting the dynamic evolution of hybrid cloud connectivity, enabling real-time perception of the physical distribution of service flows in space and terrestrial networks. By forming the benchmark traffic mapping data, not only is the allocation model for the current period established, but a feedback mechanism also provides state-closed-loop support for updating the path evaluation vector for the next period. This synergy of these steps alleviates the problems of connection interruptions and heterogeneous latency fluctuations caused by path switching in satellite communications, achieving highly reliable and energy-efficient cloud connectivity in a hybrid multi-path environment.

[0059] In summary, this application first establishes the application-layer semantics for the entire scheduling logic by determining the service type, serving as the logical starting point for the decision-making process. Second, the generated discrete evaluation vector transforms dynamic path characteristics into digital computational inputs, directly determining the predictability of the scheduling decision. Next, the physical migration of data packets between the satellite and ground-based heterogeneous media is completed by executing continuous stream output actions. Finally, the generated baseline splitting mapping data closes the entire transmission management loop in reverse. This series of coordinated actions leverages the flexibility of multipath management to compensate for the instability of satellite links, effectively shielding the complex underlying topology.

[0060] Optionally, executing scheduling decisions and establishing a target transmission path further includes: when the congestion window state reaches a preset full load threshold, the terminal device triggers an overflow scheduling mechanism and uses the overflow scheduling mechanism to redirect subsequently arriving data packets to physical interfaces belonging to high-latency paths in the set of available physical interfaces, so as to update the target transmission path.

[0061] Preferably, in the specific implementation process, the packet scheduler determines the original message flow flowing from the user application layer through its built-in service identification interface to determine the attribution characteristics of the currently pending message. In the specific technical implementation, the packet scheduler uses deep message detection logic to determine the service priority flag bit in the data packet to be sent and confirms it as the corresponding service type (e.g., identified as a real-time instruction service sensitive to interaction latency). Simultaneously, the packet scheduler determines the next-moment predicted latency of each path fed back by the multi-path management layer through the internal communication bus. This is equivalent to establishing an application-layer semantic reference and an underlying physical performance benchmark for scheduling decisions, ensuring that subsequent actions can perform resource allocation based on real-time satellite-to-ground link performance. By capturing these discrete service and link indicators, it provides definite service characteristic inputs for constructing differentiated traffic splitting strategies.

[0062] Preferably, the packet scheduler obtains the determined service requirements and predicted latency, and drives its built-in decision model to perform path mapping calculations to generate initial path allocation control codes that are discretely distributed along the time axis. In the specific technical implementation of the generation, based on the time-domain matrix composed of predicted latency, for each discrete sampling period, packets of the service type being low-latency interactive are mapped to the member path with the lowest predicted latency value. Due to the dynamic displacement of satellite orbits causing fluctuations in the latency of each link, these generated allocation control codes realize the digital discrete representation of the scheduling logic on the time axis. Therefore, it is equivalent to transforming the macroscopic path selection strategy into microscopic packet-level forwarding instructions, realizing precise coupling between distribution timing and link performance evolution, and providing a logical template for finally establishing the target transmission path corresponding to each packet.

[0063] Preferably, the multi-path management layer, through its configured protocol stack status monitoring module, performs feature sampling on the buffer occupancy of each link in the available physical interface set in real time to determine the congestion window status of each corresponding path. In the determined implementation, the weight of bytes sent but not acknowledged on the current path is extracted and compared with a preset full-load threshold in a pre-stored configuration table. When the congestion window value of a terrestrial leased line link, originally the preferred path, reaches or exceeds the preset full-load threshold, the physical characteristic of that path entering forwarding saturation is captured. This determination action achieves instantaneous quantitative perception of network congestion intensity, providing a deterministic physical basis for triggering subsequent overflow sharing logic.

[0064] Preferably, in response to the determined preferred path saturation signal, the packet scheduler drives its internal routing redirection operator to perform overflow scheduling for subsequent arriving packet flows. In the specific implementation, the packet scheduler uses an overflow scheduling mechanism to redirect the data packets to be sent, which should originally be injected into the saturated interface, mapping them to physical interfaces (e.g., satellite link interfaces) belonging to high-latency paths within the set of available physical interfaces. This effectively utilizes redundant satellite bandwidth resources to carry overflowing service loads, mitigating the buffer overflow risk caused by single-path congestion through dynamic rebalancing across heterogeneous paths. This overflow scheduling generates a set of redirection instruction flows with load-sharing characteristics, improving the throughput continuity of critical services in hybrid networking environments.

[0065] Preferably, the terminal device drives its underlying communication control engine to continuously stream output the message sequence after overflow processing to the physical transmission queue. During this continuous stream output, the redirected data packets are continuously pumped to the associated satellite communication components according to the modulation frequency and transmission window corresponding to each physical interface. This action achieves real-time migration from logical path decisions to physical stream transmission, constituting a physical service flow characterizing the multi-path collaborative state between satellite and ground. By performing this uninterrupted physical transmission, even in the transient state of full terrestrial path load, subsequent messages can still utilize the satellite link to achieve millisecond-level overflow carrying, alleviating the forwarding jitter problem caused by traditional protocols when the link is saturated, and providing immediate physical payload support for the receiving end to obtain the complete data stream.

[0066] Preferably, the multi-path management layer aggregates the output status, congestion feedback, and redirection records of the aforementioned interfaces to ultimately form the target transmission path characterizing the cross-media distribution trajectory of the service flow. In the specific technical implementation, a traffic tracing master based on connection identifiers is established, creating a chained index between the message sequence in an overflow state and the currently accessed physical interface. This is equivalent to constructing a "forwarding profile" reflecting the dynamic resource utilization of the hybrid network, enabling each discrete forwarding action to be summarized into a definite routing trajectory. By forming the updated target transmission path, not only is the allocation model for the current period established, but a feedback mechanism also provides closed-loop support for generating a more accurate allocation code for the next time slot. The synergy of this series of steps enables high-efficiency multi-path transport in a satellite network environment.

[0067] Optionally, in step 4, calculating the out-of-order compensation delay based on the delay difference in the set of available physical interfaces includes: the terminal device obtaining the first round-trip delay corresponding to the satellite link and the second round-trip delay corresponding to the ground leased link in the set of available physical interfaces, calculating the algebraic difference between the first round-trip delay and the second round-trip delay, and establishing the algebraic difference as the out-of-order compensation delay.

[0068] Preferably, the multi-path management layer uses its configured latency detection interface to determine the transmission characteristics of each member link in the set of available physical interfaces in real time, thereby determining the physical latency index of each path. In the specific technical implementation, the feedback timestamp obtained through satellite link interaction is captured, and the dwell time of the feedback payload in the onboard communication module is determined, thus calculating the first round-trip latency reflecting the long latency characteristics of the satellite. Simultaneously, the second round-trip latency reflecting the low latency characteristics of the optical fiber is determined through interaction via the terrestrial leased line link. The technical essence of this lies in capturing the inherent time deviation generated at the physical level by the heterogeneous satellite-ground links. By determining these multi-dimensional latency parameters, a definite time reference coordinate is established for subsequent refined out-of-order compensation, enabling subsequent processing actions to be precisely anchored to the transmission rhythm of different media.

[0069] Preferably, the packet scheduler acquires the determined first round-trip delay and second round-trip delay, and drives its built-in compensation operator to perform algebraic difference operations to generate a discretely distributed out-of-order compensated delay sequence along the time axis. During the generation process, the real-time difference between the larger round-trip delay corresponding to the satellite link and the smaller round-trip delay corresponding to the terrestrial dedicated line link is calculated, and this difference is mapped to a set of discrete status codes describing the time offset. Because the spatial distance changes caused by satellite orbital motion are dynamic, these generated hysteresis parameters achieve a time-domain digital representation of the asymmetric delay of the hybrid link. This step realizes the transformation from raw physical observation to transport layer control logic, providing a time-deterministic control template for ultimately forming a target transmission path with phase alignment characteristics.

[0070] Preferably, based on the generated compensation parameters, the packet scheduler drives the internal hysteresis execution logic to perform targeted injection processing on the transmission queue of the low-latency path. In the specific processing action, the low-latency physical interface corresponding to the second round-trip delay in the target transmission path is identified, and the associated out-of-order compensation delay sequence is retrieved. A controllable hysteresis clock is established within the driver layer of the terrestrial leased link, and the calculated compensation value is injected into the trigger step of this clock. This is equivalent to using the generated "hysteresis" to compensate for the long transmission delay caused by the physical distance to the satellite. By performing this targeted injection, the synchronization rhythm of the fast and slow paths at the logical level is established, providing a physical execution basis for subsequent cross-medium synchronous stream output.

[0071] Preferably, the terminal device drives its configured protocol stack scheduling engine to perform continuous stream output processing on the packets to be sent after the compensation parameters have been injected. During the specific process of continuous stream output, according to the transmission frequency corresponding to each physical interface, the high-priority sub-streams in the data packets to be sent are time-shifted backward. Because the packet transmission time on the terrestrial leased link is subject to an offset matching the delay difference, the originally sent terrestrial packets are shifted in the time domain to the interval synchronized with the satellite packets. This achieves real-time migration from logical-level time difference compensation to physical-level code stream output, constituting a physical service flow representing multi-path phase alignment between satellite and ground, enabling data packets from different paths to be injected into the complex hybrid network space with a preset synchronization phase.

[0072] Preferably, by executing the aforementioned continuous stream output actions, a synchronous arrival and forwarding configuration representing heterogeneous link collaboration is ultimately formed at the cloud access gateway side. In the specific technical implementation, since the path transmission time difference is pre-offset at the terminal side, the data packets to be sent, carried by the satellite link and the terrestrial leased line respectively, can reach the inbound buffer of the cloud access gateway with extremely high consistency after traversing heterogeneous network media. This is equivalent to realizing the distribution of transmission payloads in the heterogeneous space dimension between satellite and ground and their convergence in the time dimension of the cloud. This synchronous arrival state alleviates the problem of out-of-order delivery at the receiving end caused by severe fluctuations in link latency, providing a data foundation with ordered physical characteristics for the gateway side to perform lightweight protocol reassembly.

[0073] In summary, this application firstly provides accurate physical quantity input for the entire compensation chain by determining the round-trip delay of heterogeneous paths; secondly, the generated discrete compensation sequence transforms dynamic performance differences into programmable control parameters. Next, the physical mapping from "time offset" to "synchronous transmission" is completed by executing continuous stream output actions. Finally, the formation of the synchronous arrival and forwarding configuration completely eliminates the head-end congestion problem caused by the heterogeneity of satellite and fiber optic path delays. This series of coordinated actions leverages the flexibility of the multipath transmission protocol stack to compensate for the transmission shortcomings of the underlying satellite link, achieving high-reliability transmission of hybrid cloud connections in complex environments and improving the long-term connection stability and service continuity of enterprise-level private cloud services during satellite-to-ground handover.

[0074] Optionally, the terminal device runs a multi-path management layer, which maintains a path list consisting of the identity identifiers of each path, and monitors the congestion window status and round-trip delay corresponding to each path in real time based on the path list, so as to generate path quality assessment parameters and feed them back to the scheduling decision process in step 3.

[0075] Preferably, in the specific implementation process, the multi-path management layer performs low-level state polling on the satellite channels and terrestrial fiber optic channels currently carrying transmission services through its configured link feature acquisition interface to determine the physical existence of each member path. In the specific technical implementation, the multi-path management layer retrieves the hardware driver identifier of its physical network interface card, identifies the first interface belonging to the satellite communication module and the second interface belonging to the terrestrial leased line module, and determines the registration status of these interfaces in the kernel protocol stack. This is equivalent to establishing a resource view with multiple physical attributes for the transport layer protocol, enabling real-time capture of connection topology changes caused by link cold starts or dynamic hot-plugging. Continuously determining the identity characteristics of these physical media provides a deterministic hardware indexing foundation for constructing a dynamically updated path list.

[0076] Preferably, the multi-path management layer obtains the identified link identity information and drives its internal path registration engine to perform orchestration processing for logical connections, thereby forming a path list composed of the identity identifiers of each member path. During this formation process, satellite link identifiers, terrestrial leased line link identifiers, and sub-flow indexes associated with connection identifiers are bound together using triples to generate a set of memory indexes describing the multi-path topology. The resulting path list physically covers all available forwarding paths in the hybrid cloud connection, enabling subsequent processing actions to perform concurrent operations on heterogeneous resources based on this list. This processing step realizes the transformation from fragmented physical interfaces to structured logical resources, alleviating the technical challenge of the difficulty in solidifying and binding physical paths and transmission logic due to the high dynamism of satellites.

[0077] Preferably, the multi-path management layer, based on the path list formed above, drives its built-in state detection operator to perform discretized monitoring processing for path quality, generating path performance observations discretely distributed over time. In the specific implementation of this generation, based on the round-trip delay calculated from the path challenge frame and the congestion window state output by the congestion control algorithm, transient numerical samples reflecting link health are calculated in each sampling period. Since the physical distance changes caused by the satellite's orbital operation are continuous, but exhibit on-demand sampling characteristics at the protocol detection level, these samples form a link state description discretely distributed over time. Therefore, this is equivalent to quantifying continuous physical channel performance changes into a time-deterministic logical evaluation vector, achieving a time-domain digital representation of the dynamic nature of hybrid links.

[0078] Preferably, the multi-path management layer drives its configured signaling synchronization pipeline to perform continuous streaming output to the scheduling plane for the generated discretely distributed state samples. During the continuous streaming output process, the sampled data, including the round-trip time and the congestion window state, is encapsulated into an internal control message stream and continuously pumped to the scheduling decision-making process. To improve the continuity and low jitter characteristics of the output, a weighted average smoothing process is used in the output stage to ensure that even in transients with sudden changes in channel quality, the scheduling side can receive a quality feedback stream with consistent characteristics. This streaming output method achieves efficient data migration between the path management plane and the scheduling decision-making plane, alleviating the evaluation parameter lag problem caused by the traditional single-query mechanism in highly dynamic environments.

[0079] Preferably, the multi-path management layer aggregates the continuously output status data and performs a comprehensive evaluation of the link carrying capacity using a preset quality scoring model, thereby ultimately forming the path quality evaluation parameters that characterize the dynamic performance profile of the hybrid link. In the specific technical implementation, the predicted latency trend, window capacity margin, and historical packet loss rate are subjected to multi-dimensional logical weighting processing to generate a graded score reflecting the applicability of the target path. The generated path quality evaluation parameters describe the "performance trajectory" of each member path at the current spatiotemporal coordinates at the physical level, providing a direct logical basis for achieving accurate traffic distribution for different service types.

[0080] In summary, this application firstly provides the hardware source for the formation of the entire path list by determining the physical interface; secondly, the formed path list provides logical anchors for generating discrete performance observations, establishing the execution scope of the probing actions. Next, the physical transmission of performance data from the acquisition end to the application end is completed by executing continuous streaming output. Finally, the formation of path quality assessment parameters supports the scheduling decision-making process. This series of coordinated actions utilizes the fine-grained feedback of multipath management to compensate for the long latency characteristics of satellite links, achieving high-reliability transport of hybrid cloud connections in complex environments and improving the low-order and high-throughput characteristics of enterprise private cloud services during satellite-to-ground handover.

[0081] Optionally, the cloud access gateway is further configured with an out-of-order compensation buffer. The cloud access gateway receives the data packets to be sent transmitted through the set of available physical interfaces, and uses the out-of-order compensation buffer to perform reordering processing on the data packets to be sent, so as to restore the ordered sub-service flow and deliver it to the cloud virtual private cloud.

[0082] Preferably, the cloud access gateway, through its built-in traffic capture interface, monitors in real time the inbound data streams converged from terminal devices via a set of available physical interfaces (including satellite links and terrestrial leased lines) to determine the arrival characteristics of each packet. In the specific technical implementation, the cloud access gateway identifies the physical path identifier in the header of each data packet to be sent and simultaneously determines the instantaneous local reception time of the packet reaching the gateway hardware. The essence of this technique lies in establishing the real-time distribution of data payloads in the heterogeneous satellite-terrestrial network, thereby obtaining the original time-domain sampled signal for subsequent quantitative measurement of out-of-order deviations. By determining these microscopic packet arrival sequences, the transient arrival delay fluctuations of the satellite link relative to the terrestrial leased line can be identified in real time, providing data source support for subsequent processing actions to be accurately anchored to the correct buffering logic.

[0083] Preferably, the cloud access gateway obtains the determined arrival characteristics and drives its internal out-of-order analysis operator to perform mapping calculations on the logical attribution relationships in the message sequence to generate the message reordering guidance parameters discretely distributed along the time axis. During the generation process, the packet sequence number carried in the message is retrieved and differentially compared with the expected received sequence pre-stored by the gateway. Due to the significant difference in round-trip delay between the satellite and ground links, a set of discrete status codes describing the current memory hole location and time difference offset is generated for each discrete sampling step. This essentially transforms the macroscopic link delay deviation into a microscopic protocol stack logical index, achieving a temporal digital representation of the message out-of-order degree. These generated parameter sets with temporal characteristics provide a quantitative logical reference for subsequent precise placeholder reordering within the out-of-order compensation buffer.

[0084] Preferably, in response to the reordering guidance parameters generated above, the out-of-order compensation buffer drives the internal memory scheduling engine to perform continuous stream output processing on the non-ordered payloads residing within it. In the specific implementation of continuous stream output, based on the continuity judgment result of packet sequence numbers, packets that originally arrived intermittently on the timeline are logically aligned in ascending numerical order. Even if some packets from the satellite link experience long delays due to physical distance, the buffer will still maintain the logically reserved bit at that position and continuously pump the aligned packet payloads to the upper-layer protocol parsing unit in the form of a bit stream. This action achieves real-time migration from discrete, out-of-order physical payloads to logically ordered service bitstreams, alleviating the forwarding performance bottleneck caused by severe path performance jitter in satellite hybrid multipath transmission scenarios.

[0085] Preferably, the cloud access gateway aggregates the reordered message sequence and performs integrity restoration processing for business semantics using a preset flow aggregation operator, thereby ultimately forming the baseline reassembled business data stream that represents the temporal consistency of the business flow. In the specific technical implementation, the sorted payload undergoes decapsulation, and the physical splicing boundaries between data frames are established based on the packet sequence number, restoring the fragments originally scattered across the satellite and ground interfaces into semantically continuous sub-business flows. This is equivalent to completing a closed loop mapping from heterogeneous media offloading payloads to a single tenant business logic. The formed baseline reassembled business data stream eliminates the negative interference caused by link differences at the physical level, enabling the data subsequently delivered to the cloud virtual private cloud to possess highly reliable ordered characteristics.

[0086] In summary, this application firstly provides a hardware-level trigger cycle for the entire reassembly chain by determining the characteristics of inbound packets; secondly, the generated discrete reordering guidance parameters transform complex latency heterogeneity into simple logical addressing indices, directly determining the accuracy of buffering operations. Next, by executing continuous stream output actions, the ordered arrangement of data packets from physical distribution to logical alignment is completed in memory space. Finally, the deep organization of these ordered payloads forms the final baseline reassembly service data stream, achieving transport layer support transparent to tenant services. This series of coordinated actions leverages the buffering flexibility of the gateway side to compensate for the latency shortcomings of the underlying satellite link.

[0087] In summary, this application establishes an elastic compensation mechanism based on packet sequence numbers and guidance parameters on the cloud access gateway side, enabling efficient packet sorting even in scenarios involving frequent physical link switching or retransmissions due to satellite link errors. This reassembly process based on out-of-order compensation buffers allows user-side raw data requests to enter the cloud virtual private cloud for business processing in an orderly manner after traversing complex inter-satellite topologies. This processing logic essentially utilizes the synergistic effect of spatial computing and logical buffering, mitigating the impact of heterogeneous link head congestion on cloud connection stability and improving high-throughput delivery of hybrid multipath transmission protocols in complex space networking environments.

[0088] Optionally, before the data packet to be sent is physically sent through the target transmission path, the method further includes: the terminal device assigning consecutive packet sequence numbers to the data packet to be sent within the multipath transmission protocol stack, and using the multipath transmission protocol stack to establish an association mapping relationship between the packet sequence numbers and each physical interface in the target transmission path, so that the cloud access gateway performs data integrity verification based on the association mapping relationship.

[0089] Preferably, in the specific implementation process, the multipath transmission protocol stack performs feature sampling on the raw bit stream sent from the application layer through its built-in service capture interface to determine the data packets to be sent in the transmission queue and their associated traffic attributes. Within the determined technical implementation, the message length, priority label, and the target transmission path identifier determined at the scheduling level are identified. The technical essence of this lies in establishing the initial association between logical service flows and physical transmission resources, enabling real-time awareness of the scheduling destination of each message frame. By determining these micro-level service characteristics, a data source foundation is provided for subsequent globally unified sequence number distribution, ensuring that subsequent processing actions are accurately anchored within the correct service context.

[0090] Preferably, the protocol orchestration unit built into the terminal device acquires the determined service flow and drives the internal counting logic to perform identification processing for data integrity, thereby generating the continuous packet sequence numbers discretely distributed along the time axis. In the specific technical implementation of generation, each packet belonging to the same connection session is assigned a sequentially increasing logical tag. These tags serve as discrete scales representing service evolution in the time domain, acting similarly to observing the logical position of the data flow. Since data packets are generated in the form of discrete pulses on the time axis, these generated continuous packet sequence numbers constitute a set of digital indexes describing service progress, realizing a quantitative representation of the logical topology of the service code stream. Therefore, it is equivalent to using discretely generated digital tags as observation points to record the logical position of packets within the protocol stack, providing a globally unique timing reference for subsequent reassembly and verification on heterogeneous paths.

[0091] Preferably, the packet scheduler obtains the generated consecutive packet sequence numbers and associated service payloads, drives its configured distribution operator to perform dynamic mapping processing for the physical layer interfaces, and accordingly performs continuous stream output for the set of available physical interfaces consisting of satellite links and terrestrial leased lines. During the specific process of executing continuous stream output, according to the scheduling decision instructions, the data packets to be sent, injected with sequence number identifiers, are continuously pumped to the associated physical network interface card buffer. This action realizes the real-time migration from logical layer packet distribution to physical layer code stream conversion, enabling packets carrying different logical locations to be orderly injected into the transmission medium according to a preset load balancing strategy, even in scenarios with significant differences in round-trip latency between satellite and ground links. This provides immediate physical support for improving the continuous flow of data payloads in heterogeneous spaces.

[0092] Preferably, the multipath management layer of the terminal device aggregates the output status and scheduling records of the aforementioned interfaces, and uses its internal log organization engine to perform chained storage processing based on path affiliation, thereby ultimately forming the baseline association mapping data characterizing the dynamic distribution of terminal service flows across paths. In the specific technical implementation, the generated consecutive packet sequence numbers are bound to the actually selected physical paths using triplet binding, constructing a dynamic mapping model describing the physical migration patterns of service payloads in heterogeneous spatial networks. This model logically and semantically fully characterizes the "transmission track" of each packet in a multipath environment. The formed baseline association mapping data achieves deep coupling between logical sequences and physical interface locations, providing a deterministic digital foundation for the receiving end to identify the source path and original timing relationship of each packet.

[0093] In summary, this application firstly provides the original triggering source and business context for the entire transmission chain by determining the service traffic; secondly, the generated discrete sequence number provides the core identity identifier for the stream output, establishing the logical ranking of the message in multipath transmission. Next, by executing continuous stream output actions, the substantial transfer of data packets from the logical plane to the physical heterogeneous medium is completed. Finally, the formed reference association mapping data reverses and closes the entire protocol encapsulation and path association monitoring loop. This series of coordinated actions utilizes the flexibility of multipath protocol orchestration to compensate for the topology dynamics caused by the high-speed movement of satellite links, achieving transparent support for private cloud connections in complex network environments.

[0094] Preferably, the automated execution of this series of technical processing actions alleviates the receiver reordering problem caused by the heterogeneity of satellite and ground link performance in satellite communication scenarios. Because a baseline association mapping data between consecutive packet sequence numbers and the target transmission path is pre-established before physical transmission, the cloud access gateway, upon receiving interleaved packet payloads, can perform rapid data integrity verification and logical reordering based on this mapping relationship. This end-to-end sequence alignment mechanism avoids the risk of out-of-order packet loss due to frequent path switching or latency fluctuations, improving the high-throughput delivery of hybrid multipath transmission protocols in complex topology environments.

[0095] Optionally, in step 2, obtaining the congestion window state for each path includes: the terminal device capturing service confirmation packets fed back by the cloud access gateway through the set of available physical interfaces in real time, and extracting the round-trip byte weights and packet loss awareness features for each path from the service confirmation packets; the terminal device using the round-trip byte weights and packet loss awareness features to drive a preset congestion control logic to perform iterative calculation processing to generate dynamic values ​​reflecting the real-time path throughput capacity and establish them as the congestion window state for each path.

[0096] Preferably, in the specific implementation process, the service acknowledgment message capture interface configured inside the terminal device monitors in real time the downlink bitstream transmitted back by the cloud access gateway through the set of available physical interfaces to determine the feedback activity of each member path. In the specific technical implementation, protocol layer parsing is performed on the received user datagram protocol payload to identify the service acknowledgment packet encapsulated within, containing the acknowledgment sequence number and window update information. The technical essence of this lies in establishing the physical interaction feedback boundary between the terminal and the gateway at the transmission layer, enabling the capture of the confirmation signal after each frame of data successfully arrives on the satellite link or terrestrial leased line. By determining these feedback payloads, the terminal device obtains the original signal source for quantitatively evaluating the real-time load strength of the path, providing deterministic data payload input for subsequently generating feature samples reflecting the channel throughput capability.

[0097] Preferably, the path feature extraction module built into the terminal device acquires the determined feedback payload and drives the internal calculation operator to perform differential mapping on the message header and fields to generate the round-trip byte weights and packet loss awareness features discretely distributed along the time axis. During the generation process, the cumulative byte acknowledgment identifier in the message is retrieved, and the incremental amount of data successfully delivered within the current observation period is calculated to generate the round-trip byte weights reflecting the transient throughput capability of the link. The logical continuity in the message stream sequence is analyzed synchronously to determine the packet loss locations caused by satellite link signal fading, thereby generating packet loss awareness features describing path reliability. Since the feedback signaling is generated in the form of discrete pulses along the time axis, these generated features achieve a time-domain digital discrete representation of the physical channel quality, enabling the capture of micro-performance jumps caused by inter-satellite topology variations.

[0098] Preferably, the congestion control processing unit built into the terminal device acquires the performance samples with discrete distribution characteristics generated above, drives its internal parameter scheduling logic, and performs continuous streaming output processing on the extracted data sequence. In the specific technical implementation of executing continuous streaming output, the generated round-trip byte weights and packet loss awareness features are encapsulated into a high-frequency updated internal configuration instruction stream according to a preset sampling clock step, and continuously pumped to the congestion control logic to execute parameters. This is equivalent to providing a time-series continuous service load for the iterative calculation process, alleviating the problem of limited sampling accuracy caused by the single-request mechanism in traditional solutions. Through this uninterrupted streaming output, a time-domain digital twin of the dynamic performance of hybrid links is realized, providing an instantaneous computational stream input for achieving dynamic tracking of transmission rates.

[0099] Preferably, the congestion control operation unit aggregates the continuously output performance characteristic streams and drives the internally configured congestion control logic to perform iterative calculations on the transmission window quota, thereby ultimately forming the congestion window state characterizing the throughput capacity of each path. In the specific technical implementation, the current observation weight is used as a feedback gain term, and pre-stored historical state parameters are retrieved to perform incremental recursive calculations, thereby calculating a dynamic value reflecting the real-time carrying capacity boundary of the physical path. This formed congestion window state is no longer a fixed hard-coded value, but a logical control variable that deeply follows the actual physical performance of the satellite and ground links. This processing step marks the final closed loop from discrete observation samples to deterministic scheduling criteria, providing a direct quantitative indicator for subsequent precise path resource matching based on the service requirements of the data packets to be transmitted.

[0100] In summary, this application demonstrates that, firstly, the determination of the business confirmation packet provides real-time feedback pulses for the entire state deduction process, serving as the logical starting point of the determination chain. Secondly, the generated discrete round-trip byte weights and packet loss awareness features quantify complex physical interactions into digital inputs, directly determining the raw input quality of subsequent stream output stages. Next, the execution of continuous stream output actions completes the physical signal migration from the raw data capture end to the computational decision end. Finally, the generated congestion window state is injected into the scheduling decision module, achieving deep linkage between path state awareness and traffic allocation execution.

[0101] 10. A seamless cloud connectivity system for hybrid multipath transmission in a satellite communication scenario, characterized in that it includes a terminal device and a cloud access gateway, wherein the terminal device and the cloud access gateway perform multipath transmission enable negotiation to determine a set of available physical interfaces consisting of satellite links and terrestrial leased lines; the terminal device sends path challenge frames through each physical interface in the set of available physical interfaces, and the cloud access gateway receives the path challenge frames and feeds back path response frames, so that the terminal device can obtain the round-trip time and congestion window status of each path, and perform prediction processing on the round-trip time of each path to determine the predicted delay at the next moment; the terminal device, according to the service type of the data packet to be sent, negotiates with each of the available physical interfaces to determine the path response frame. The next-moment prediction delay and the congestion window state of the path are used to make scheduling decisions to establish the target transmission path for each of the data packets to be sent, and the data packets to be sent are transmitted to the cloud access gateway through the target transmission path; the terminal device calculates the out-of-order compensation delay based on the delay difference in the set of available physical interfaces, and injects the out-of-order compensation delay into the interfaces of the target transmission path that belong to the low-latency path, so that the data packets to be sent arrive synchronously at the cloud access gateway; the cloud access gateway receives the synchronously arriving multiple data packets to be sent, uses the out-of-order compensation buffer to reorder the multiple data based on the packet sequence number of the data packets to be sent, and performs data integrity verification.

[0102] This application embodiment also provides a seamless cloud connection method for hybrid multipath transmission in a satellite communication scenario, comprising: a cloud access gateway and a terminal device performing multipath transmission enable negotiation to determine a set of available physical interfaces consisting of satellite links and terrestrial leased lines; the cloud access gateway receiving path challenge frames sent by the terminal device through each physical interface in the set of available physical interfaces, and feeding back path response frames, so that the terminal device can obtain the round-trip delay and congestion window status of each path, and perform prediction processing on the round-trip delay of each path to determine the predicted delay at the next moment; the cloud access gateway receiving data packets to be transmitted by the terminal device through the established target transmission path after the terminal device performs scheduling decisions based on the service type of the data packets to be transmitted and the predicted delay at the next moment and the congestion window status of each path; the cloud access gateway receiving multiple data packets to be transmitted synchronously after the terminal device calculates based on the delay difference and injects out-of-order compensation delay, reordering the multiple data packets based on the packet sequence number of the data packets to be transmitted using an out-of-order compensation buffer, and performing data integrity verification.

[0103] This application also provides a terminal device that supports hybrid multipath transmission, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement a seamless cloud connection method for hybrid multipath transmission in a satellite communication scenario as described in any one of this application.

[0104] This application also provides a cloud access gateway device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement a seamless cloud connection method for hybrid multipath transmission in a satellite communication scenario as described in any one of this application.

[0105] This application also provides a computer storage medium storing computer-executable instructions thereon, which, when executed, perform the methods described in any one of this application.

[0106] In summary, this application improves the high-throughput stability of enterprise-level private cloud services in hybrid multipath environments. Because this application establishes a proactive determination mechanism based on real-time feedback and utilizes a congestion control operation unit to perform high-frequency iterative processing on round-trip byte weights and packet loss awareness features, it establishes a congestion window state that can track the actual capacity evolution of the physical link with high sensitivity. This dynamic feedback-based state formation process effectively suppresses transmission performance drops caused by high latency or random bit errors in satellite links. This processing logic essentially utilizes digital feature extraction to compensate for the non-stationary characteristics of the physical channel, achieving high-throughput delivery of hybrid multipath transmission in complex space communication scenarios.

[0107] The above description is merely an exemplary embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A seamless cloud connection method for hybrid multipath transmission in a satellite communication scenario, characterized in that, include: Step 1: The terminal device and the cloud access gateway perform multipath transmission enable negotiation to determine the set of available physical interfaces consisting of satellite links and terrestrial leased lines; Step 2: The terminal device sends path challenge frames through each physical interface in the set of available physical interfaces to obtain the round-trip time and congestion window status of each path, and performs prediction processing on the round-trip time of each path to determine the predicted delay at the next moment; Step 3: The terminal device performs scheduling decisions based on the service type of the data packet to be sent and the predicted delay at the next moment and the congestion window status of each path to establish the target transmission path for each data packet to be sent; Step 4: The terminal device calculates the out-of-order compensation delay based on the delay difference in the set of available physical interfaces, and injects it into the interface of the target transmission path that belongs to the low-latency path, so that the data packet to be sent arrives synchronously at the cloud access gateway.

2. The method of claim 1, wherein, In step 1, the terminal device and the cloud access gateway perform multipath transmission enable negotiation to determine the set of available physical interfaces, including: the terminal device and the cloud access gateway negotiate multipath transmission extension parameters through handshake messages and allocate globally unique connection identifiers to establish a multipath transmission protocol connection; in response to the connection identifier, the terminal device maintains the logical continuity of the multipath transmission protocol connection when the physical Internet Protocol address changes; during the connection process based on the multipath transmission protocol, in response to the connection identifier, the terminal device obtains the physical Internet Protocol addresses of the satellite link and the terrestrial leased line link and performs an interface registration action to determine the set of available physical interfaces composed of the satellite link and the terrestrial leased line link after the interface registration action is performed. 3.The seamless cloud connection method of hybrid multipath transmission in a satellite communication scenario of claim 1, wherein, In step 2, sending path challenge frames through each physical interface in the set of available physical interfaces to obtain the round-trip time of each corresponding path includes: the terminal device periodically sending path challenge frames carrying transmission timestamps on the satellite link and the terrestrial leased line link included in the set of available physical interfaces, and obtaining path response frames fed back by the cloud access gateway; the terminal device captures the arrival time of the path response frame, and performs a difference operation based on the arrival time and the transmission timestamp to calculate the round-trip time of each corresponding physical interface.

4. The method of claim 1, wherein, In step 2, a prediction process is performed on the round-trip delay of each path to determine the predicted delay for the next moment. This includes: the terminal device acquiring the current round-trip delay observation value for each path and performing a difference operation between the current round-trip delay observation value and the state prediction value at the previous moment to extract the prediction residual; the terminal device acquiring the Kalman gain corresponding to each path and performing a weighted correction process on the prediction residual using the Kalman gain to generate an updated delay state estimate value and determining it as the predicted delay for the next moment.

5. The method of claim 1, wherein, In step 3, a scheduling decision is made based on the service type of the data packet to be sent to establish the target transmission path, including: if the service type of the data packet to be sent is identified as a low-latency interactive service, then the physical interface with the smallest predicted latency at the next moment is selected from the set of available physical interfaces as the target transmission path; if the service type is identified as a high-bandwidth throughput service, then the physical interface with the largest remaining capacity in the congestion window state is selected from the set of available physical interfaces as the target transmission path.

6. The method of claim 5, wherein, Executing scheduling decisions and establishing a target transmission path also includes: when the congestion window state reaches a preset full load threshold, the terminal device triggers an overflow scheduling mechanism and uses the overflow scheduling mechanism to redirect subsequently arriving data packets to physical interfaces belonging to high-latency paths in the set of available physical interfaces, so as to update the target transmission path.

7. The method of claim 1, wherein the satellite communication scenario is a hybrid multi-path transmission seamless cloud connection method. In step 4, calculating the out-of-order compensation delay based on the delay difference in the set of available physical interfaces includes: the terminal device obtains the first round-trip delay corresponding to the satellite link and the second round-trip delay corresponding to the ground leased link in the set of available physical interfaces, calculates the algebraic difference between the first round-trip delay and the second round-trip delay, and establishes the algebraic difference as the out-of-order compensation delay.

8. The method of claim 1, wherein the method is used in a satellite communication scenario. The terminal device runs a multi-path management layer, which maintains a path list consisting of the identity identifiers of each path, and monitors the congestion window status and round-trip time of each path in real time based on the path list to generate path quality assessment parameters and feed them back to the scheduling decision process in step 3. 9.A seamless cloud connection method of hybrid multipath transmission in a satellite communication scenario, characterized in that, include: Step 1: The cloud access gateway and the terminal device perform multipath transmission enable negotiation to determine the set of available physical interfaces consisting of satellite links and terrestrial leased lines; Step 2: The cloud access gateway receives path challenge frames sent by the terminal device through each physical interface in the set of available physical interfaces and feeds back path response frames, so that the terminal device can obtain the round-trip time and congestion window status of each path, and perform prediction processing on the round-trip time of each path to determine the predicted time delay at the next moment; Step 3: The cloud access gateway receives the data packets to be sent by the terminal device after performing scheduling decisions based on the service type of the data packets to be sent and the predicted time delay at the next moment and the congestion window status of each path, and transmits them through the established target transmission path; Step 4: The cloud access gateway receives multiple data packets to be sent that arrive synchronously after being calculated by the terminal device based on the delay difference and injected with out-of-order compensation delay. It then uses the out-of-order compensation buffer to reorder the multiple data packets based on the packet sequence number of the data packets to be sent and performs data integrity verification.

10. A seamless cloud connectivity system for hybrid multipath transmission in a satellite communication scenario, characterized in that, The system includes a terminal device and a cloud access gateway. The terminal device and the cloud access gateway perform multipath transmission enable negotiation to determine a set of available physical interfaces consisting of satellite links and terrestrial leased lines. The terminal device sends path challenge frames through each physical interface in the set of available physical interfaces. The cloud access gateway receives the path challenge frames and sends back path response frames, allowing the terminal device to obtain the round-trip time and congestion window status for each path. The terminal device then performs prediction processing on the round-trip time of each path to determine the predicted delay for the next moment. The terminal device determines the predicted delay for the next moment based on the service type of the data packet to be sent and the predicted delay for each path. The congestion window state executes scheduling decisions to establish target transmission paths for each of the data packets to be sent, and transmits the data packets to be sent to the cloud access gateway through the target transmission paths; the terminal device calculates out-of-order compensation delay based on the latency difference in the set of available physical interfaces, and injects the out-of-order compensation delay into the interfaces belonging to the low-latency path in the target transmission path, so that the data packets to be sent arrive synchronously at the cloud access gateway; the cloud access gateway receives the synchronously arriving multiple data packets to be sent, reorders the multiple data packets based on the packet sequence number of the data packets to be sent using the out-of-order compensation buffer, and performs data integrity verification.

11. A terminal device supporting hybrid multi-path transmission, characterized by, It includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a seamless cloud connection method for hybrid multipath transmission in a satellite communication scenario as described in any one of claims 1-8.

12. A cloud access gateway device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the seamless cloud connection method for hybrid multipath transmission in a satellite communication scenario as described in claim 9.

13. A computer storage medium, characterized in that, It stores computer-executable instructions thereon, which, when executed, perform the method according to any one of claims 1-9.

Images