An intelligent collaborative transmission method for space-ground-ground three-dimensional heterogeneous networks

Abstract

The present invention provides an intelligent and collaborative transmission method for a three-dimensional heterogeneous network of air, space, and ground. It adopts a decision condition constructed based on the Lyapunov function to constrain the selection range of node reconfiguration updates, thereby ensuring that only network nodes that do not meet service performance and stability requirements are selected for update. This solves the problem of excessively high costs caused by traditional algorithms relying on global node traversal algorithms for node resource scheduling, and significantly improves the adaptability to changes in the network environment as well as the efficiency and stability of data stream transmission.

Description

Technical Field

[0001] The present invention belongs to the field of network communication and data processing technology, and in particular relates to an intelligent integrated collaborative transmission method for an air-space-ground-three-dimensional heterogeneous network. Background Art

[0002] With the rapid development of communication technologies, a three-dimensional, heterogeneous network architecture combining space, air, and ground has become a key infrastructure for the widespread interconnection of devices and intelligent services in the future. This three-dimensional, heterogeneous network architecture primarily consists of a multi-orbit satellite network, a near-Earth drone network, and a traditional terrestrial network. Future 6G wireless communication networks will integrate ground-based and space-based communication nodes to achieve full coverage. While the integration of space, air, and ground networks offers many advantages and potential, the heterogeneity of multiple communication formats, the dynamic changes in network topology, the instability of link performance, and the long number of routing hops all impact the performance of the Transmission Control Protocol (TCP), designed for the Internet, resulting in poor performance in this complex environment.

[0003] The existing Internet transmission control method is centered on the "end-to-end" design. The sender relies on the ACK feedback obtained from the receiver after a round-trip delay to adjust the sending rate, but this interaction method is sluggish and lagging in the aerospace three-dimensional heterogeneous network spanning a vast space, and cannot respond to the rapid changes in the network environment in a timely manner. Since the nodes in the aerospace network (such as low-orbit satellites and drones) move at high speed, the rapid change of relative position will cause TCP connections to be frequently disconnected and rebuilt in the aerospace three-dimensional heterogeneous network, which may cause the TCP probe data packet to fail to be delivered on time, resulting in abnormal session interruption. In addition, the congestion detection mechanism in the current transmission control method usually regards the disorder caused by packet loss as network congestion. In the high bit error rate environment of the aerospace link, link packet loss can easily lead to misjudgment, frequently reducing the congestion control window, and hindering users from fully utilizing network bandwidth resources.

[0004] To meet the demands of three-dimensional, space-ground, and heterogeneous networks, current research trends have introduced centralized resource management technologies to address the complex challenges posed by heterogeneous network convergence. Smart Integration Identifier Networking (SINET-I) is a novel network architecture that effectively guarantees the transmission performance of heterogeneous networks through intelligent control server cluster scheduling at the resource adaptation layer, providing a research foundation for achieving efficient and reliable three-dimensional, heterogeneous network transmission. This architecture enables network scheduling to not only deploy and enforce node behavior rules in real time but also efficiently manage resources through a centralized control server. Other network optimization efforts include core network functions such as routing, address translation, and performance enhancement agents, encompassing satellite and UAV nodes. Furthermore, some work at the physical and link layers has considered characteristics of satellite and UAV communication systems, such as Doppler shift and transmission delay jitter, to adaptively improve routing and location selection algorithms. However, existing research often overlooks the characteristics of intermediate nodes, and few studies have considered these in the design of transmission control methods for three-dimensional, space-ground, and heterogeneous networks. This results in most research on this scenario lacking collaboration with the transport layer. Similarly, some research at the network layer or lower levels has focused solely on the impact of local algorithm improvements on overall network system performance, with little attention paid to analyzing the interactions and collaborations between algorithms at different layers. In short, most existing research fails to take a holistic perspective on heterogeneous networks across space, space, and ground, lacking in-depth analysis of the interactions and collaborations between algorithms at different layers.

[0005] The core challenge of the three-dimensional heterogeneous network of air, space and ground is the networking coordination of aerial nodes. Figure 1 A typical scenario is shown, where the data stream is initially transmitted to the receiver along a path consisting of nodes such as low-orbit satellites and high-orbit satellites (shown by black arrows); during the transmission process, the low-orbit satellite causes a link interruption due to high dynamic changes, and the network node needs to react quickly and try to re-establish the link through other nodes, such as Figure 1 The new path indicated by the red arrow continues to support the transmission of the original data flow. Recent research has achieved traffic reconfiguration by using a centralized control server to allocate node resources on the control side and coordinate network nodes at a large spatial scale. However, these works only consider reachability at the routing level and ignore the interaction between flow reconfiguration behavior and transmission control algorithms.

[0006] Unlike terrestrial systems, space-ground heterogeneous networks face challenges such as dynamic topology, long latency, and high bit error rates. These factors, particularly at the transport layer, require ACK (Acknowledgement) confirmations. These factors make traffic management sluggish and unreliable. However, existing research has limited in-depth analysis of the characteristics of space-ground heterogeneous networks from a transport layer perspective. For example, some work has introduced varying packet loss rates as a parameter in the model, superimposing errors caused by mobility for dynamic analysis. While the proposed algorithm demonstrates stability in high-packet-loss environments, it overlooks the issue of unreachable paths within the network caused by dynamic topology, leading to burst buffer accumulation at intermediate nodes. Furthermore, existing end-to-end transmission control methods fail to capture important influencing factors at the network level. According to statistics from professional organizations, every 1,000 km increase in inter-satellite distance increases single-hop latency by approximately 20 milliseconds; additional satellite hops also significantly increase latency. This demonstrates that the transmission performance of space-ground heterogeneous networks relies heavily on dynamic network-level evolution, making it difficult to accurately perceive through upper-layer predictions or receiver feedback. Therefore, there is an obvious need to develop new transmission control methods that can adapt to these dynamic network characteristics to improve the efficiency and reliability of the entire network system.

[0007] Another notable characteristic of the heterogeneous, space-air-ground network is its high degree of heterogeneity. Its constituent subnetworks are supported by a variety of communication protocols. This heterogeneity encompasses a large number of node devices connected by diverse configuration interfaces, increasing the complexity of network integration and management. Therefore, new technical approaches are needed to address complex issues such as routing management and coordinated transmission control.

[0008] Although the TCP / IP protocol has been widely used in terrestrial fixed and mobile networks, classic transmission control schemes such as Cubic and latency-friendly BBR perform well in terrestrial networks but perform poorly in air-ground collaborative network environments. With the development of infrastructure such as low-orbit satellites and drones, a number of transmission control methods specifically designed for these applications have begun to emerge. Emerging research, such as CTCP, introduces intermediate proxy nodes into transmission control to rapidly adapt to network environment changes caused by aerial node mobility. While these new methods additionally consider the impact of the wireless environment, they do not delve into the specific causes of these impacts at the network level. As a result, their perception of factors such as packet loss and congestion often relies on theoretically pre-defined random variables, making them less effective in real-world network environments.

[0009] With the development of technology, the concept of identity-based networking has been widely applied in heterogeneous network management, providing a variety of innovative solutions for optimizing traffic management and network transmission control. Leveraging the "dynamic resource adaptation" mechanism within the network architecture of intelligent identity-based networks, network status is dynamically perceived and customized network encoding and decoding capabilities are built. This logically isolated design based on intelligent identity-based networks facilitates the efficient integration and management of network resources. Furthermore, recent research has begun exploring cross-layer design to improve the performance of heterogeneous networks spanning air, space, and ground. These efforts attempt to establish closer collaboration between the different layers of the TCP / IP model to achieve more efficient data transmission and management. Despite progress in cross-layer design, significant gaps remain at the transmission control level, particularly in bottom-up research that addresses both network and transport layer issues.

[0010] In summary, existing transmission control methods for heterogeneous air-space-ground networks face three major problems: 1) They cannot quickly adjust the sending rate in networks with large spatial scales; 2) They require a central control server to schedule global node resources, but do not consider the increase in scheduling overhead brought about by the increase in network scale and the possible resource waste caused by incorporating all nodes into the set of devices serving a single flow; 3) They do not support collaboration between the transport layer and the network layer and lack specialized cross-layer design, which may lead to problems such as network layer routing reconfiguration causing transport layer session interruption. Summary of the Invention

[0011] To solve the above problems, the present invention provides an intelligent and collaborative transmission method for a three-dimensional heterogeneous network of air, space, and ground. It adopts a decision condition based on the Lyapunov function to constrain the selection range of node reconfiguration and update, which significantly improves the adaptability to changes in the network environment and the efficiency and stability of data stream transmission.

[0012] A method for intelligent fusion collaborative transmission in a three-dimensional air-ground-heterogeneous network, comprising the following steps:

[0013] S1: When a new data stream needs to be transmitted in the network, a breadth-first routing search algorithm is used to obtain the initial path for the new data stream to be transmitted from the source node to the target node in the network;

[0014] S2: Filter out existing data flows from the network that have path nodes that overlap with the new data flow as overlapping data flows;

[0015] S3: Obtain the influence of the new data stream on each overlapping data stream, and set the influence greater than the set threshold R min The overlapping data flow is recorded as the reconfiguration data flow;

[0016] S4: Using a decision condition constructed based on the Lyapunov function, determine in sequence whether each intermediate node in the transmission path of each reconfigured data flow is retained, and obtain a retained path corresponding to each reconfigured data flow, where the retained path is a discontinuous path that has undergone node hopping relative to the transmission path;

[0017] S5: For each jump that occurs in each reserved path, the node before the jump is recorded as the jump start node, and the node after the jump is recorded as the jump end node. Then, a breadth-first routing search algorithm is used to obtain the reconfiguration sub-path from the jump start node to the jump end node of each jump. Each reserved path is combined with its corresponding reconfiguration sub-path to obtain a continuous reconfiguration path for each reconfigured data stream.

[0018] Furthermore, in step S3, the method for obtaining the influence of the new data stream on any overlapping data stream is as follows:

[0019]

[0020] Among them, R(f a ,f b ) is the influence of the new data stream a on the overlapping data stream b, f a is the size of the new data stream a, f b is the size of the overlapping data stream b, △f a is the change in the size of the new data stream relative to the overlapping data stream b, r a The remaining transmission potential after removing the throughput of data flow a from the overlapped portion of the path nodes between the new data flow a and the overlapped data flow b, B(p a ) is the bottleneck congestion degree of the new data flow a, B(p b ) is the bottleneck congestion degree of the overlapping data flow b.

[0021] Furthermore, the bottleneck congestion level B(p) of any data flow p is calculated as follows:

[0022]

[0023] Where v is the node on the transmission path corresponding to data stream p, γ is the sum of the congestion window sizes of the nodes on the transmission path corresponding to data stream p, D is the size of the data packet in data stream p, C v is the throughput upper limit of node v, and x is the actual throughput of node v.

[0024] Furthermore, the decision condition constructed based on the Lyapunov function in step S4 is:

[0025]

[0026] Among them, Evo(t-1) is the evolution penalty term of time slot t-1, δ * (t) is a variable whose range is (0,1) for time slot t, is the global optimality gap of the space-ground-space heterogeneous network in time slot t-1, λ v To represent the percentage of node service capacity occupied, Q v (t) is the local updated virtual queue associated with the Lyapunov function in time slot t, and w is the set weight;

[0027] The calculation method of Evo(t-1) is:

[0028]

[0029] in, To update the flag bits of each node in the transmission path in a global manner within time slot t-1, To update the flag of each node in the transmission path in a local manner within time slot t-1, is the local optimality gap of the space-ground-space heterogeneous network in time slot t-1, Evo(t-2) is the penalty term in time slot t-2, is the global optimality gap of the space-ground-space heterogeneous network within time slot t-2; δ * (t-1) is the (0,1) variable at time slot t-1;

[0030] The variable δ is in the range (0,1) * The calculation method of (t) is:

[0031]

[0032] Among them, t0 is the time slot when the data stream of the current node joins the space-ground-space three-dimensional heterogeneous network, and λ is a set constant;

[0033] Global Optimality Gap The calculation method is:

[0034]

[0035] Where J is a hopping constant used to quantify the impact of a new data stream on the network, and a(t-1) is a random variable describing whether a new data stream arrives at the network in time slot t-1. If it arrives, a(t-1) = 1; if it does not arrive, a(t-1) = 0.

[0036] Locally update the virtual queue Q v The calculation method of (t) is:

[0037]

[0038] Among them, the coefficient λv =f v / C v , and C v is the throughput upper limit of node v, f v is the number of data flows of node v.

[0039] Furthermore, if all the intermediate nodes of any transmission path of the reconfigured data flow do not meet the judgment condition, all the nodes of the transmission path are updated in a global manner, and the flag is set to Flag

[0040] If any intermediate node in the transmission path of any reconfigured data flow meets the judgment condition, the nodes in the transmission path that do not meet the judgment condition are updated in a local manner, and the flag bit is set to Flag

[0041] Furthermore, the evolution penalty term Evo(t-1) is a bounded function, and the upper bound of the drift penalty function P(t-1) corresponding to the evolution penalty term is:

[0042]

[0043] Among them, f * is the frequency of global update, f is the frequency of local update, is the global updated virtual queue related to the Lyapunov function in time slot t-1, Indicates expectation, Q v (t-1) is the local update virtual queue related to the Lyapunov function in time slot t-1, and:

[0044]

[0045] in, is the global updated virtual queue related to the Lyapunov function in time slot t-2, The flag bits of each node in the transmission path are updated globally in time slot t-2.

[0046] Furthermore, in step S5, the method for obtaining the reconfiguration sub-path from the jump start node to the jump end node of each jump using the breadth-first routing search algorithm is specifically as follows:

[0047] S51: Add the jump start node to the queue as the starting point of the search;

[0048] S52: Count all unvisited neighbor nodes of the jump starting node, evaluate the priority of each neighbor node according to the heuristic function, and use the neighbor node with the highest priority as the next hop of the jump starting node;

[0049] S53: Filter out all unvisited neighbor nodes of the next-hop node selected in step S52, and evaluate the priority of each neighbor node according to the heuristic function, and use the neighbor node with the highest priority as the next-hop node; and so on, until the jump end node is encountered, and reconstruct the reconfiguration sub-path from the jump start node to the jump end node by backtracking the path.

[0050] Furthermore, the heuristic function is as follows:

[0051]

[0052] Among them, x v is the actual throughput of node v, N(v) is the set of neighbor nodes of node v, u is a node in the set of neighbor nodes, α1 is the set weight related to the throughput, and β1 is the set weight related to the penalty term.

[0053] Furthermore, after obtaining the continuous reconfiguration path of each reconfigured data flow, the intermediate nodes on each reconfiguration path independently confirm the reception status of the data packets in the data flow. After each intermediate node receives the data packet, it returns the ACK message to the previous hop node. The ACK message will continue to be fed back to the sending end of the data flow, and the ACK message carries the congestion level information of the intermediate node.

[0054] Furthermore, after obtaining the continuous reconfiguration paths of each reconfigured data stream, the method for determining the data transmission volume in each time slot between any two adjacent intermediate nodes on each reconfiguration path is as follows:

[0055]

[0056] Among them, condition 1 is that the congestion level of all nodes in the current transmission path does not exceed the set threshold, condition 2 is that there are nodes in the current transmission path whose congestion level exceeds the set threshold, condition 3 is that the node in front of the adjacent nodes receives two duplicate ACK messages or the RTO times out, cwnd(t) is the congestion window size in time slot t, RTT is the round-trip delay of the entire transmission process, cwnd(t+RTT) is the congestion window size in time slot t+RTT, α is the growth coefficient, ε is the protection factor, and β is the multiplicative reduction coefficient.

[0057] Beneficial effects:

[0058] 1. The present invention provides an intelligent and collaborative transmission method for a three-dimensional heterogeneous network of air, space, and ground. It adopts a decision condition based on the Lyapunov function to constrain the selection range of node reconfiguration and update, thereby ensuring that only network nodes that do not meet service performance and stability requirements are selected for update. This solves the problem of high cost caused by traditional algorithms relying on global node traversal algorithms for node resource scheduling, and significantly improves the adaptability to changes in the network environment as well as the efficiency and stability of data stream transmission.

[0059] 2. The present invention provides an intelligent collaborative transmission method for a three-dimensional heterogeneous network of air, space, and ground. The heuristic evaluation function adopted by the breadth-first routing search algorithm measures the remaining transmission capacity of each node in the network and the real-time workload of the node respectively, so that the breadth-first routing search algorithm of the present invention can intelligently give priority to nodes with strong transmission capabilities and light current loads during the search process, significantly improving the coordination of data flow transmission in the network.

[0060] 3. The present invention provides an intelligent collaborative transmission method for a three-dimensional heterogeneous network of air, space, and ground. By integrating an intelligent heuristic breadth-first routing search algorithm within the SINET-I framework, transport layer information is taken into account in the routing update process, and a new transmission performance standard - "network loop-free and unobstructed" is defined. This ensures that there are no routing loops and network black holes in the three-dimensional heterogeneous network of air, space, and ground, while avoiding local link congestion, effectively overcoming the challenges faced by the collaborative network layer and the transport layer.

[0061] 4. The present invention provides an intelligent and collaborative transmission method for three-dimensional heterogeneous networks of air, space, and ground, replacing the traditional TCP's "end-to-end" solution with a finer-grained "hop-to-hop" design model. It uses a mixed method of linear increase, logarithmic increase, and multiplicative decrease to adjust the congestion control window, significantly improving the adaptability to changes in the network environment.

[0062] 5. The present invention provides an intelligent collaborative transmission method for a three-dimensional heterogeneous network of air, space, and ground. Unlike the traditional end-to-end confirmation mechanism, the hop-to-hop confirmation mechanism of the present invention allows each intermediate node to independently confirm the reception status of the data packet. This means that if an error occurs during the transmission process, it only needs to be agilely retransmitted from the previous hop node without waiting for the SACK containing error information or RTO timeout, which greatly reduces the retransmission overhead caused by the accumulation of multi-hop transmission errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 This is a schematic diagram of three-dimensional heterogeneous network transmission in space, air, and ground;

[0064] Figure 2 A schematic diagram of the network topology and node control flow provided by the present invention;

[0065] Figure 3 The first network state of the network topology and node control flow provided by the present invention;

[0066] Figure 4 The second network state of the network topology and node control flow provided by the present invention;

[0067] Figure 5 The third network state of the network topology and node control flow provided by the present invention;

[0068] Figure 6 A flowchart of a method for intelligent collaborative transmission in an air-ground-space heterogeneous network provided by the present invention;

[0069] Figure 7 This is a schematic diagram of the data transmission workflow under the hop-to-hop asynchronous confirmation mechanism provided by the present invention. DETAILED DESCRIPTION

[0070] In order to enable people skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0071] The present invention first establishes a mathematical model for the transmission control problem existing in the space-ground-air three-dimensional heterogeneous network, and then defines the concept of network looplessness and unobstructedness in the present invention, as follows:

[0072] The space-ground three-dimensional heterogeneous network is defined in this invention as an undirected graph G = (V, E), where V is the set of network nodes and the set of available bidirectional links. The capacity of each link is determined by the minimum network port rate C of the nodes on both sides of the communication. v Decision,Before introducing the transmission control problem, we will provide an example of common network state changes in,air-space-ground networks from the perspective of the resource adaptation layer in,SINET-I networks.

[0073] Figure 2 This figure shows a simple network topology consisting of ground, drone, and satellite nodes. It assumes that the network cards of all nodes support a maximum throughput of 20 Mbps. The black dashed lines represent available bidirectional links. From the perspective of the abstract "dynamic resource adaptation layer," continuous update events arrive randomly. Each update event is triggered by the data sender or the network node itself and includes one or a group of flow routing update operations. Specifically, these update operations act on heterogeneous nodes in the network and can be roughly divided into three categories: adding, modifying, or deleting network rules. The execution of these update operations reconfigures the data flows in the network in real time to optimize the network's load balancing and average transmission performance.

[0074] Figure 3The figure depicts an initial network state. The solid blue arrow represents data flow 2, transmitted across a three-dimensional, space-air, and ground-ground heterogeneous network. Its real-time throughput is 10 Mbps, and the receiving end is a ground user. Simultaneously, data flow 1, represented by the dashed orange arrow, appears. The dashed lines indicate that this is a set of network update configurations that have not yet been deployed. Before it is actually executed, it is necessary to determine whether it conflicts with existing flows in the network. For example, if the throughput of a node with multiple flows coexists exceeds the throughput limit of the network card,

[0075] exist Figure 4 In the network state 2 shown, data streams 1 and 2 are already in the transmitting state; at this time, a transmission request (data stream 3) from the satellite node to the ground user appears. The present invention uses a green dotted arrow to indicate this network update event that has not yet been executed. However, if this update event is executed, the three data streams will pass through a drone node at the same time, and the sum of their throughput requirements exceeds the upper limit of the network card processing capacity of a single node. Therefore, in Figure 5 The three data streams of network status shown are updated in a coordinated manner, as shown in Figure 5 The scheme shown changes the way limited resources are utilized. The new network update scheme avoids the transmission performance degradation that may be caused by traffic competition by reconfiguring data flows.

[0076] In this example of network state change, the present invention envisions a special scenario to describe the problems that are frequently encountered in the actual use of the three-dimensional heterogeneous network of air, space and ground. Figure 5 This paper presents an ingenious solution. Some existing research works attempt to use a centralized control server to solve the problem of data flow reconfiguration, and aim to design a solution with shorter path hops and no routing loops. This invention innovatively proposes to combine the congestion window information of the transport layer with the update operation of the network layer, such as Figure 2 As shown in the red box, this invention implements hop-by-hop congestion control at each network node and maintains an interface for interacting with a centralized control server. The goal of the proposed HWCTC (HopWise Consistent Transmission Control) is to achieve a "loop-free and unobstructed" network for heterogeneous, space-air, and ground-based networks. The following provides clear definitions of some of the technical concepts in this invention.

[0077] Definition 1. Network state and update event. The present invention uses a set of update events to describe the network state N. state ={O e , O ne}, where O e Represents the set of executed update events, O neRepresents a set of update events that have not yet been executed. Assuming there are n update events in the network, each update event contains several update operations. The two update event sets can be represented as follows:

[0078]

[0079]

[0080] Among them O event n represents the update operation set of the nth update event. In this invention, the addition of a new data stream is an update event. An update operation means that the addition of a new data stream causes a node to be reconfigured. The number of nodes that need to be reconfigured is recorded as the number of update operations. Indicates the set of update operations that have not been executed in the nth update event. When the update event arrives at the centralized control server and has not been executed, there is As time passes, the update operations are executed sequentially until If the set of no update events executed in the current network state at a certain moment is an empty set, the present invention defines it as the network initial state.

[0081] Definition 2. Loop-free and unobstructed transmission control. This collaborative decision-making process combines information from the transport and network layers to achieve loop-free and unobstructed transmission control in heterogeneous air-ground, space-ground networks. "Loop-free and unobstructed" refers to a scheduling decision-making scheme that ensures that network routing is loop-free, black-hole-free, and congestion-free.

[0082] Definition 3. Bounded Update. This refers to the fact that the update scope and resource usage of transmission strategies in heterogeneous air-space-ground networks are bounded. Strategies are formulated based on the use of limited resources, rather than searching and scheduling node resources from a global perspective.

[0083] The above definition lays the foundation for the transmission control method proposed later in this invention. It is worth noting that HWCTC is a cross-layer collaborative transmission control method, and its performance metrics are not entirely defined at the transport layer. Furthermore, the design philosophy behind HWCTC distinguishes it from other transmission control methods by incorporating transport-layer traffic information into the routing algorithm influenced by the centralized control server and by incorporating routing information from network nodes into transport-layer congestion control.

[0084] As mentioned above, the present invention mentions an example of multi-flow coexistence to illustrate the importance of throughput information to node coordination. However, there are still two problems in its actual application in air-space-ground three-dimensional transmission control. 1) First, this new flow configuration process may update all flows in the network system. Although the present invention Figure 2Only three data streams are mentioned in the article, but in real scenarios the number of data streams depends on the number of demands at that moment. If the resource boundaries of the update strategy are not restricted, then the addition of new data streams at any time may affect all existing data streams in the network. This update strategy that requires global information is not practical. 2) User demands in the three-dimensional heterogeneous network of space, air, ground and space are huge and dynamic. In general, new data stream demands will arrive before the optimal solution for the next network state flow configuration is solved. Therefore, the present invention attempts to propose a smart and collaborative transmission method for the three-dimensional heterogeneous network of space, air, ground and space, so as to realize traffic management based on the mutual influence of data streams.

[0085] It should be noted that in the transmission process of the three-dimensional heterogeneous network of air, space and ground, two key problems need to be solved: 1) How to design a data flow reconfiguration algorithm to determine which flows need to be reconfigured within a bounded resource range and ensure that the new path is loop-free; 2) How to design a hop-to-hop congestion control algorithm to quickly respond to the network environment to avoid transmission congestion. At the same time, the intelligent fusion of the present invention is derived from the intelligent identification network proposed by Academician Zhang Hongke of Beijing Jiaotong University, which is similar to the concept of SDN; fusion can be understood as the fusion of transport layer and routing layer algorithms (routing algorithm and congestion control algorithm), and collaboration can be understood as the collaboration of nodes in the network or on the network path.

[0086] Based on this, Figure 6 As shown, a method for intelligent fusion collaborative transmission for a three-dimensional air-ground-heterogeneous network includes the following steps:

[0087] S1: When a new data stream needs to be transmitted in the network, a breadth-first routing search algorithm is used to obtain the initial path for the new data stream to be transmitted from the source node to the target node in the network.

[0088] S2: Filter out existing data flows from the network that have path nodes that overlap with the new data flow as overlapping data flows.

[0089] S3: Obtain the influence of the new data stream on each overlapping data stream, and set the influence greater than the set threshold R min The overlapping data stream is recorded as the reconfiguration data stream.

[0090] It should be noted that the bottleneck congestion level of a data flow's path is determined by the most congested node along the path. A centralized control server maintains bottleneck congestion information for each data flow. Furthermore, when data flows in the network change and the paths of two or more data flows partially overlap, this invention defines a method for calculating their mutual influence. This invention treats these data flows as several nodes and calculates the influence of the new data flow on any overlapping data flow using the following method:

[0091]

[0092] Among them, R(f a ,f b ) is the influence of the new data stream a on the overlapping data stream b, f a is the size of the new data stream a, f b is the size of the overlapping data stream b, △f a is the change in the size of the new data stream relative to the overlapping data stream b, r a The remaining transmission potential after removing the throughput of data flow a from the overlapped portion of the path nodes between the new data flow a and the overlapped data flow b, B(p a ) is the bottleneck congestion degree of the new data flow a, B(p b ) is the bottleneck congestion degree of the overlapping data flow b. This calculation method not only takes into account the change in the throughput of the data flow, but also takes into account the load baseline of the flow in a specific bottleneck transmission environment and the carrying capacity of the network node. In addition, the present invention introduces a parameter R min This influence threshold is used to determine whether a configuration update is required for a specific flow. If the calculated influence value exceeds this threshold, the present invention updates the configuration of the data flow during the next network reconfiguration; otherwise, the current status is maintained. This enables HWCTC to dynamically adjust to the real-time needs of the network and achieve a balance between performance and cost.

[0093] It should be noted that in order to improve network performance and maintain efficient global flow configuration, the controller in the three-dimensional heterogeneous network of air, space and ground needs to solve the traffic distribution problem under strict time constraints. The present invention proposes a data flow reconfiguration algorithm under the HWCTC framework. Unlike existing work that relies on global information, HWCTC balances the impact of global and local network information on decision-making. The present invention introduces the idea of ​​studying shared link bottlenecks at the transport layer into the data flow reconfiguration at the network layer. Before the flow reconfiguration is performed, the data flow p is transmitted along the initial path in the network. Based on this, the present invention proposes the following mathematical model to measure the bottleneck congestion degree B(p) of any data flow p:

[0094]

[0095] Where v is the node on the transmission path corresponding to data stream p, γ is the sum of the congestion window sizes of the nodes on the transmission path corresponding to data stream p, D is the size of the data packet in data stream p, C v is the throughput upper limit of node v, and x is the actual throughput of node v.

[0096] It should be noted that after identifying the data flows that require configuration updates, the next step is to determine how these data flows should be updated. These update methods are categorized as either global node updates or local node updates. While global updates can deliver relatively superior performance, they also come with higher control costs. In contrast, local node flow configuration updates, while offering only moderate performance, are less expensive. See step S4 for details.

[0097] S4: Using the decision conditions constructed based on the Lyapunov function, it is determined in turn whether each intermediate node of the transmission path of each reconfigured data flow is retained, and the retained path corresponding to each reconfigured data flow is obtained, wherein the retained path is a non-continuous path in which a node jump has occurred relative to the transmission path.

[0098] It should be noted that if all the intermediate nodes of any transmission path of the reconfigured data flow do not meet the judgment condition, all the nodes of the transmission path are updated in a global manner, and the flag bit is set to Flag

[0099] If any intermediate node in the transmission path of any reconfigured data flow meets the judgment condition, the nodes in the transmission path that do not meet the judgment condition are updated in a local manner, and the flag bit is set to Flag

[0100] That is to say, the present invention sets the update operation of the time slot t on the data stream as two binary variables when When , it means that the network processes the update event in a global way at time slot t; when When , it means that the update event is processed locally in time slot t. When both are 0, it means that the system does not need to update the configuration of any node in this time slot.

[0101] Furthermore, the decision condition constructed based on the Lyapunov function is:

[0102]

[0103] Among them, Evo(t-1) is the evolution penalty term of time slot t-1, δ * (t) is a variable whose range is (0,1) for time slot t, is the global optimality gap of the space-ground-space heterogeneous network in time slot t-1, λ v To represent the percentage of node service capacity occupied, Q v (t) is the local updated virtual queue associated with the Lyapunov function in time slot t, and w is the set weight;

[0104] It can be seen that the present invention is based on the idea of ​​Lyapunov control theory and introduces two variables: and They represent the global and local optimality gaps of the space-ground-space heterogeneous network in time slot t, respectively, and are defined as:

[0105]

[0106] Where J is a jump constant used to quantify the impact of new data flows on the network, a(t) is a random variable describing whether a new data flow arrives at the network in time slot t, where a(t) = 1 if it arrives and a(t) = 0 if it does not arrive; k and d represent the number of local flow configuration updates and the impact weight since the last global update, respectively; δ(t), δ * (t) is a variable in the range (0,1) for time slot t and is calculated as follows:

[0107]

[0108]

[0109] Here, t0 is the time slot when the data stream of the current node joins the space-space-ground heterogeneous network, and λ is a set constant. It should be emphasized that the optimal data stream reconfiguration is global node-oriented. This paper defines the gap between global and local optimality so that after the update strategy performs local updates a reasonable number of times, the algorithm can adaptively perform a global node-oriented reconfiguration of the data stream. According to Definition 1, this paper represents the evolution of the network state as changes in the set of update events.

[0110] Update policies are related to update events, which in turn are generated by specific data flow requirements. Therefore, HWCTC defines two virtual queues for the network nodes that data flows through:

[0111]

[0112] When the centralized control server updates the global and data flows containing node v, the virtual queues will increase respectively, where λ v =f v / C v Represents the service capability of the node, which is between 0 and 1, and C v is the throughput upper limit of node v, f v is the number of data flows of node v. Inspired by the latest research on defining local Lyapunov functions, the present invention defines the Lyapunov function L of a node based on the square form of a virtual queue. v (t):

[0113]

[0114] In order to ensure the stability of the update strategy of local nodes and optimize the long-term transmission performance, the present invention adopts the method of minimizing the system drift, that is, minimizing the expected change of the Lyapunov function over time, so as to promote the system to evolve towards a stable state. v (t) can be defined as:

[0115]

[0116] in The operator representing the statistical expected value defines the Lyapunov drift of a node system, which can quantify the stability of the system state over time. Based on this, the present invention introduces a penalty term that indicates the system's evolutionary deviation from the optimal data flow configuration. The evolution penalty term Evo(t) is defined as follows:

[0117]

[0118] in, To update the flag of each node in the transmission path in a global manner within time slot t, To update the flag of each node in the transmission path in a local manner within time slot t, is the local optimality gap of the space-ground-space heterogeneous network in time slot t, Evo(t-1) is the penalty term of time slot t-1, is the global optimality gap of the space-ground-space heterogeneous network in time slot t-1; δ * (t) is the (0,1) variable at time slot t;

[0119] The design of the penalty term follows the idea of ​​balancing global and local update strategies. On this basis, the drift plus penalty function is further defined:

[0120]

[0121] The above formula comprehensively considers the system's immediate stability changes and long-term performance expectations. Here, w is an importance weight that balances the system's stability and performance requirements.

[0122] By observing the above four formulas, we can infer the upper bound of the drift plus penalty function:

[0123]

[0124] Among them, f * is the frequency of global update, f is the frequency of local update, is the global updated virtual queue associated with the Lyapunov function in time slot t, Indicates expectation, Q v(t) is the local updated virtual queue associated with the Lyapunov function in time slot t.

[0125] Since update events in the space-ground-air three-dimensional heterogeneous network are random, it is difficult to predict the exact value of the drift plus penalty function.

[0126] Therefore, the present invention attempts a simple threshold strategy to minimize the upper bound of the drift plus penalty function. If the conditions are met:

[0127]

[0128] In this data flow update, the node is excluded from the set of optional nodes, that is, the node needs to be reconfigured and is not a node in the reserved path; if all nodes on this path need to be reconfigured, then On the contrary If no node meets the conditions, the bottleneck node of the old flow is eliminated. This paper has thus far described in detail how the controller in the HWCTC framework determines which data flows require updating, and within these updated data flows, which nodes require reconsideration or remain unchanged. The method proposed in this paper enables bounded updates in heterogeneous networks with three-dimensional space-ground and space-ground architectures. By introducing the concept of balance from Lyapunov control theory before reconfiguring paths, HWCTC effectively limits the range of resources that need to be considered during the update process to a reasonable range through two-stage screening.

[0129] S5: For each jump that occurs in each reserved path, the node before the jump is recorded as the jump start node, and the node after the jump is recorded as the jump end node. Then, a breadth-first routing search algorithm is used to obtain the reconfiguration sub-path from the jump start node to the jump end node of each jump. Each reserved path is combined with its corresponding reconfiguration sub-path to obtain a continuous reconfiguration path for each reconfigured data stream.

[0130] That is, after determining the update range of network nodes, the next task of the centralized control server is to rebuild or repair the network routing to achieve the path update of the data flow. The present invention proposes a breadth-first routing search algorithm that performs heuristic routing node search by evaluating the evolution status of network nodes.

[0131] The challenge faced by the three-dimensional heterogeneous network of air, space and ground in this link is how to search for a series of intermediate nodes in a highly dynamic network environment. Due to the high mobility of nodes such as drones and satellites, traditional static routing algorithms cannot adapt to the frequent changes in node locations and network topology. In addition, the complexity of the actual data flow source may also cause the node to become congested due to multiple flow loads. After screening the data flow to be updated and the nodes that need to be updated on its path, the present invention starts from the source node or the source node or the last intermediate node connected to the source node, and searches outward layer by layer until the target node is found or all reachable nodes are covered.

[0132] For example, assuming that the transmission path of a reconfigured data flow is from node 1 to node 10 in sequence, and nodes 3, 4, 5, and 8 do not meet the judgment conditions constructed based on the Lyapunov function, then there are two jumps in the reserved path, namely the jump between node 2 and node 6, and the jump between node 7 and node 9. Nodes 2 and 7 are the jump start nodes, and nodes 6 and 9 are the jump end nodes. At this time, it is necessary to execute the breadth-first routing search algorithm twice to re-search nodes in the space-space-ground three-dimensional heterogeneous network to establish the reconfiguration sub-path between node 2 and node 6 and the reconfiguration sub-path between node 7 and node 9. Finally, the two obtained reconfiguration sub-paths are combined with the reserved path to obtain a continuous reconfiguration path from source node 1 to target node 10 of the current reconfigured data flow. It should be noted that the number of nodes in the reconfiguration path may be less than 10, or may be equal to 10 or greater than 10.

[0133] It should be noted that the breadth-first routing search algorithm describes the process of finding new routing paths in an air-space-ground-air network using breadth-first search techniques. The algorithm's input is the network topology G = (V, E), and its output is a routing path p from a source node s to a destination node o. The algorithm first initializes a queue and marks the starting node as visited. Then, based on the network's connectivity structure, it searches layer by layer within the set of selectable network nodes selected in the previous section until the destination node is found or all reachable nodes are exhausted. This algorithm is applicable to dynamically changing, heterogeneous air-space-ground networks, ensuring real-time and accurate routing.

[0134] Specifically, the method for obtaining the reconfiguration sub-path from the jump start node to the jump end node of each jump using the breadth-first routing search algorithm is as follows:

[0135] S51: Add the jump start node to the queue as the starting point of the search;

[0136] S52: Count all unvisited neighbor nodes of the jump starting node, evaluate the priority of each neighbor node according to the heuristic function, and use the neighbor node with the highest priority as the next hop of the jump starting node;

[0137] S53: Filter out all unvisited neighbor nodes of the next-hop node selected in step S52, and evaluate the priority of each neighbor node according to the heuristic function, and use the neighbor node with the highest priority as the next-hop node; and so on, until the jump end node is encountered, and reconstruct the reconfiguration sub-path from the jump start node to the jump end node by backtracking the path.

[0138] The heuristic function is as follows:

[0139]

[0140] Among them, x v is the actual throughput of node v, N(v) is the set of neighbor nodes of node v, u is the node in the neighbor node set, α1 is the set weight related to throughput, β1 is the set weight related to the penalty term, and α1 and β1 are the importance weights for balancing the remaining capacity of reachable nodes and evolutionary cheapness. The evolutionary penalty term Evo(t) represents the load of the node in time slot t. It is worth noting that since the routing search algorithm adopts the design concept of incremental construction, routing loops can be avoided under any circumstances. This layer-by-layer expansion feature also gives the algorithm a natural advantage in reducing the number of routing hops. The routing search algorithm of the present invention introduces heuristic evaluation functions under the framework of breadth-first search, which respectively measure the remaining transmission capacity of each node in the network and the real-time workload of the node, so that the algorithm can intelligently give priority to those nodes with strong transmission capabilities and light current loads during the search process.

[0141] It should be noted that in the three-dimensional heterogeneous network of air, space and ground, due to the complex network topology and diverse communication environment, the data transmission packet loss rate between intermediate nodes is high, and the probability of such errors will increase exponentially with the increase in the number of intermediate nodes. In order to meet this challenge, the hop-to-hop asynchronous confirmation mechanism of the HWCTC of the present invention provides a unique solution. Unlike the traditional end-to-end confirmation mechanism, the hop-to-hop confirmation mechanism allows each intermediate node to independently confirm the reception status of the data packet. This means that if an error occurs during the transmission process, it only needs to be retransmitted agilely from the previous hop node without waiting for the SACK containing the error information or the RTO timeout, which greatly reduces the retransmission overhead caused by the accumulation of multi-hop transmission errors.

[0142] Figure 7 This paper shows an example of a hop-to-hop asynchronous confirmation mechanism. The present invention describes the process of data packet transmission between multiple nodes in the form of a swim lane diagram. The data transmission and single-hop confirmation occur between the sending node and the next node. As the transmission distance increases, the round-trip delay (RTT) of a single hop increases. hopThe proportion of the round-trip time (RTT) of the entire transmission process will be smaller. The red arrow in the figure shows that when data is lost at a node, if the sending node can still receive ACKs for the same data flow (indicating that the nodes are still connected), hop-to-hop retransmission will be performed after receiving two duplicate ACKs. If the drone node is unreachable as shown in the figure, retransmission traffic will begin to accumulate at the sending end until the flow reconfiguration threshold is reached. Figure 7 The two blue dotted lines in the figure represent two newly added intermediate nodes. The newly added nodes will fill the previous routing gaps without the need to trace the network error back to the sender and re-establish the session like the traditional TCP protocol. Figure 7 The black arrows in the figure represent the transmission behavior if no flow reconfiguration occurs. It can be intuitively observed that under the hop-to-hop asynchronous confirmation mechanism, the network recovery process is shortened to a smaller range. Figure 7 The green arrow in the figure represents the ACK feedback flow. It is worth mentioning that after the confirmation information is returned to the previous hop node through the ACK packet, it will continue to be fed back to the sender of the data flow; it carries the congestion level information of the intermediate nodes. This helps the sender obtain the bottleneck load information of the link and helps it adaptively adjust the congestion control window.

[0143] Therefore, after obtaining the continuous reconfiguration path of each reconfigured data flow, the intermediate nodes on each reconfiguration path independently confirm the reception status of the data packets in the data flow. After each intermediate node receives the data packet, it returns the ACK message to the previous hop node. The ACK message will continue to be fed back to the sending end of the data flow, and the ACK message carries the congestion level information of the intermediate node.

[0144] At the same time, the method for determining the data transmission volume in each time slot between any two adjacent intermediate nodes on each reconfiguration path is as follows:

[0145]

[0146] Among them, condition 1 is that the congestion level of all nodes in the current transmission path does not exceed the set threshold, condition 2 is that there are nodes in the current transmission path whose congestion level exceeds the set threshold, condition 3 is that the node in front of the adjacent nodes receives two duplicate ACK messages or the RTO times out, cwnd(t) is the congestion window size in time slot t, RTT is the round-trip delay of the entire transmission process, cwnd(t+RTT) is the congestion window size in time slot t+RTT, α is the growth coefficient, ε is the protection factor, and β is the multiplicative reduction coefficient.

[0147] It can be seen that in the HWCTC framework provided by the present invention, the congestion window adjustment strategy combines three methods: linear growth, logarithmic growth, and multiplicative reduction. The three update methods provided by the present invention, when the congestion level of a node on the observation path p exceeds a preset threshold, in order to maintain network stability, the congestion window cwnd will be adjusted by logarithmic growth, and the logarithmic growth has a relatively mild window growth rate. On the contrary, if the node congestion level is lower than the threshold, HWCTC encourages the occupation of more network resources, so a more radical linear growth strategy is adopted, in which the growth coefficient α is 2, which is larger than the default value of traditional TCP (AIMD algorithm), and the protection factor ε is set to 1. This design is to provide a faster bandwidth utilization improvement in a low congestion state, and can more effectively utilize idle bandwidth compared to the traditional TCP solution. The present invention observes that under different network conditions (such as low delay and high bandwidth and high delay and low bandwidth), setting α to 2 can significantly improve network throughput without significantly increasing the risk of congestion. In addition, considering the sensitivity to network status feedback, the present invention refers to the mature Cubic algorithm in the multiplicative reduction stage and sets the multiplicative reduction coefficient β to 0.7. Through comparative experiments, the present invention finds that 0.7 achieves a good balance between reducing network congestion recovery time and improving overall transmission efficiency.

[0148] It's worth emphasizing that this invention recognizes that due to the diversity and unpredictability of real-world network environments, it's difficult to establish a universally applicable set of optimal parameter configurations. Therefore, while this invention demonstrates that the aforementioned parameter selections achieve good performance in a simulation environment, this doesn't necessarily mean they are optimal for all scenarios. It's important to note that the core contribution of this invention isn't detailed parameter tuning, but rather demonstrating the design principles and effectiveness of HWCTC.

[0149] It can be seen that although there are some transmission control methods that take into account the impact of the wireless environment, they are limited to the transmission layer and do not go deep into the network layer to analyze the specific reasons for these impacts. For example, the perception of factors such as packet loss and congestion in these methods often relies on theoretically preset random variables, which makes it difficult for them to achieve the expected results in the actual network environment. The HWCTC method proposed in the present invention incorporates the routing search algorithm of the network layer into the transmission algorithm architecture for the first time, replacing the traditional TCP's "end-to-end" solution with a finer-grained "hop-to-hop" design model, and uses a mixed method of linear increase, logarithmic increase and multiplicative decrease to adjust the congestion control window. The hop-to-hop asynchronous confirmation mechanism is used to cope with the challenges brought by the harsh wireless transmission environment of the three-dimensional air-space-ground network, significantly improving the adaptability to changes in the network environment.

[0150] At the same time, previous methods in this field, such as the CT-RMU method, have demonstrated how to effectively apply identification controllers to these fields. Although this method is theoretically advanced, it requires the controller to collect extensive global network information, resulting in high control overhead and synchronization costs that limit its application value in practical environments. The HWCTC method of the present invention, combined with the threshold method designed based on Lyapunov control theory, can limit the scheduled node resources to a smaller range, thereby avoiding the resource consumption and feasibility issues that may be caused by traditional traversal algorithms at the network scale of practical applications.

[0151] In summary, the present invention proposes a smart and collaborative transmission method for heterogeneous air-space-ground networks. This method is a transmission control method based on a bounded update mechanism. It provides more efficient and robust transmission services through intelligent control servers in the smart identification network, providing technical support for data interaction in the Internet of Things environment. Specific advantages are as follows:

[0152] The present invention collaborates with the network layer across layers. This method introduces the concept of "loop-free and unobstructed network" for the first time, ensuring that there are no forwarding loops or network black holes in the network, and avoiding link congestion. By screening a bounded set of resources through control theory, it achieves efficient point-to-point collaboration that balances global and local influences, ultimately significantly improving the overall network transmission performance.

[0153] In the three-dimensional space-air-ground network, node configuration updates are performed within a constrained range, and a data flow reconfiguration algorithm is designed. This algorithm measures the service performance and stability provided by network nodes based on the Lyapunov drift plus penalty method, thereby ensuring that only a subset of network devices that meet specific conditions are selected for update. This solves the problem of traditional algorithms relying on global node traversal algorithms for node resource scheduling.

[0154] HWCTC effectively overcomes the challenges faced by the collaborative network and transport layers. By integrating an intelligent, heuristic breadth-first routing search algorithm within the SINET-I framework, it incorporates transport layer information into the routing update process and defines a new transmission performance standard: "network loop-free and unobstructed." This ensures that there are no routing loops or network black holes in the three-dimensional heterogeneous network of air, space, and ground, while avoiding local link congestion.

[0155] In other words, HWCTC not only effectively addresses the challenges that may be encountered in three-dimensional heterogeneous networks in the air, space, and ground, but also improves transmission performance while ensuring loop-free and unobstructed network operation.

[0156] Of course, the present invention may have many other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art may of course make various corresponding changes and modifications based on the present invention, but these corresponding changes and modifications should all fall within the scope of protection of the claims attached to the present invention.

Claims

1. A method for intelligent fusion and collaborative transmission in a three-dimensional space-ground heterogeneous network, characterized in that: The following steps are involved: S1: When a new data stream needs to be transmitted in the network, a breadth-first routing search algorithm is used to obtain the initial path for the new data stream to be transmitted from the source node to the target node in the network; S2: Filter out existing data flows from the network that have path nodes that overlap with the new data flow as overlapping data flows; S3: Obtain the influence of the new data stream on each overlapping data stream, and set the influence greater than the set threshold R min The overlapping data flow is recorded as the reconfiguration data flow; S4: Using a decision condition constructed based on the Lyapunov function, determine in sequence whether each intermediate node in the transmission path of each reconfigured data flow is retained, and obtain a retained path corresponding to each reconfigured data flow, where the retained path is a discontinuous path that has undergone node hopping relative to the transmission path; S5: For each jump that occurs in each reserved path, the node before the jump is recorded as the jump start node, and the node after the jump is recorded as the jump end node. Then, a breadth-first routing search algorithm is used to obtain the reconfiguration sub-path from the jump start node to the jump end node of each jump. Each reserved path is combined with its corresponding reconfiguration sub-path to obtain a continuous reconfiguration path for each reconfigured data stream.

2. The intelligent fusion collaborative transmission method for a three-dimensional air-ground-heterogeneous network according to claim 1, characterized in that: The method for obtaining the influence of the new data stream on any overlapping data stream in step S3 is as follows: Among them, R(f a ,f b ) is the influence of the new data stream a on the overlapping data stream b, f a is the size of the new data stream a, f b is the size of the overlapping data stream b, △f a is the change in the size of the new data stream relative to the overlapping data stream b, r a The remaining transmission potential after removing the throughput of data flow a from the overlapped portion of the path nodes between the new data flow a and the overlapped data flow b, B(p a ) is the bottleneck congestion degree of the new data flow a, B(p b ) is the bottleneck congestion degree of the overlapping data flow b.

3. The intelligent fusion collaborative transmission method for a three-dimensional air-ground-heterogeneous network according to claim 2, characterized in that: The calculation method of the bottleneck congestion degree B(p) of any data flow p is: Where v is the node on the transmission path corresponding to data stream p, γ is the sum of the congestion window sizes of the nodes on the transmission path corresponding to data stream p, D is the size of the data packet in data stream p, C v is the throughput upper limit of node v, and x is the actual throughput of node v.

4. The intelligent fusion collaborative transmission method for a three-dimensional air-ground-heterogeneous network according to claim 1, characterized in that: The decision condition based on the Lyapunov function described in step S4 is: Among them, Evo(t-1) is the evolution penalty term of time slot t-1, δ * (t) is a variable whose range is (0,1) for time slot t, is the global optimality gap of the space-ground-space heterogeneous network in time slot t-1, λ v To represent the percentage of node service capacity occupied, Q v (t) is the local updated virtual queue associated with the Lyapunov function in time slot t, and w is the set weight; The calculation method of Evo(t-1) is: in, To update the flag bits of each node in the transmission path in a global manner within time slot t-1, To update the flag of each node in the transmission path in a local manner within time slot t-1, is the local optimality gap of the space-ground-space heterogeneous network in time slot t-1, Evo(t-2) is the penalty term in time slot t-2, is the global optimality gap of the space-ground-space heterogeneous network within time slot t-2; δ * (t-1) is the (0,1) variable at time slot t-1; The variable δ is in the range (0,1) * The calculation method of (t) is: Among them, t0 is the time slot when the data stream of the current node joins the space-ground-space three-dimensional heterogeneous network, and λ is a set constant; Global Optimality Gap The calculation method is: Where J is a hopping constant used to quantify the impact of a new data stream on the network, and a(t-1) is a random variable describing whether a new data stream arrives at the network in time slot t-1. If it arrives, a(t-1) = 1; if it does not arrive, a(t-1) = 0. Locally update the virtual queue Q v The calculation method of (t) is: Among them, the coefficient λ v =f v / C v , and C v is the throughput upper limit of node v, f v is the number of data flows of node v.

5. The intelligent integrated collaborative transmission method for a three-dimensional space-ground-heterogeneous network according to claim 4, characterized in that: If all the intermediate nodes of any transmission path of the reconfigured data flow do not meet the judgment condition, all the nodes of the transmission path are updated in a global manner, and the flag bit is set to Flag If any intermediate node in the transmission path of any reconfigured data flow meets the judgment condition, the nodes in the transmission path that do not meet the judgment condition are updated in a local manner, and the flag bit is set to Flag 6. The intelligent integrated collaborative transmission method for a three-dimensional space-ground-heterogeneous network according to claim 4, characterized in that: The evolution penalty term Evo(t-1) is a bounded function, and the upper bound of the drift penalty function P(t-1) corresponding to the evolution penalty term is: Among them, f * is the frequency of global update, f is the frequency of local update, is the global updated virtual queue related to the Lyapunov function in time slot t-1, Indicates expectation, Q v (t-1) is the local update virtual queue related to the Lyapunov function in time slot t-1, and: in, is the global updated virtual queue related to the Lyapunov function in time slot t-2, The flag bits of each node in the transmission path are updated globally in time slot t-2.

7. The intelligent integrated collaborative transmission method for a three-dimensional space-ground-heterogeneous network according to claim 1, characterized in that: The method of using the breadth-first routing search algorithm in step S5 to obtain the reconfiguration sub-path from the jump start node to the jump end node of each jump is specifically as follows: S51: Add the jump start node to the queue as the starting point of the search; S52: Count all unvisited neighbor nodes of the jump starting node, evaluate the priority of each neighbor node according to the heuristic function, and use the neighbor node with the highest priority as the next hop of the jump starting node; S53: Filter out all unvisited neighbor nodes of the next-hop node selected in step S52, and evaluate the priority of each neighbor node according to the heuristic function, and use the neighbor node with the highest priority as the next-hop node; and so on, until the jump end node is encountered, and reconstruct the reconfiguration sub-path from the jump start node to the jump end node by backtracking the path.

8. The intelligent integrated collaborative transmission method for a three-dimensional space-ground-heterogeneous network according to claim 7, characterized in that: The heuristic function is as follows: Among them, x v is the actual throughput of node v, N(v) is the set of neighbor nodes of node v, u is a node in the set of neighbor nodes, α1 is the set weight related to the throughput, and β1 is the set weight related to the penalty term.

9. The intelligent integrated collaborative transmission method for a three-dimensional space-ground heterogeneous network according to claim 1, characterized in that: After obtaining the continuous reconfiguration path of each reconfigured data flow, the intermediate nodes on each reconfiguration path independently confirm the reception status of the data packets in the data flow. After each intermediate node receives the data packet, it returns an ACK message to the previous hop node. The ACK message will continue to be fed back to the sending end of the data flow, and the ACK message carries the congestion level information of the intermediate node.

10. The intelligent fusion collaborative transmission method for a three-dimensional air-ground-heterogeneous network according to claim 9, characterized in that: After obtaining the continuous reconfiguration paths of each reconfigured data stream, the method for determining the data transmission volume in each time slot between any two adjacent intermediate nodes on each reconfiguration path is as follows: Among them, condition 1 is that the congestion level of all nodes in the current transmission path does not exceed the set threshold, condition 2 is that there are nodes in the current transmission path whose congestion level exceeds the set threshold, condition 3 is that the node in front of the adjacent nodes receives two duplicate ACK messages or the RTO times out, cwnd(t) is the congestion window size in time slot t, RTT is the round-trip delay of the entire transmission process, cwnd(t+RTT) is the congestion window size in time slot t+RTT, α is the growth coefficient, ε is the protection factor, and β is the multiplicative reduction coefficient.

Images