Deterministic transmission scheduling method and system for differentiated service flow of satellite-ground convergence network

Abstract

The invention provides a deterministic transmission scheduling method and a deterministic transmission scheduling system for differentiated service flows of a satellite-ground convergence network, which belong to the technical field of communication networks, and are characterized in that complementary satellite-ground cross-domain peak cache resources are calculated and reserved by utilizing an elastic transmission algorithm based on genetic cache according to the total demand of the service flows, and adaptive optimization of the cache resources is carried out in a domain; obtaining a domain node resource allocation result; and according to a node resource allocation result, transient routing based on deep reinforcement learning and a variable queue algorithm, carrying out routing-queue two-dimensional scheduling and deterministic transmission scheduling on the service flow so as to meet different cache capacity requirements and time requirements of the service flow. According to the method, the defects that a traditional STIN network architecture lacks a differentiated deterministic service concept and adaptation between protocol stack levels is loose are overcome, the deterministic communication guarantee capability of the network in different scenes is remarkably improved, and powerful support is provided for efficient operation and reliable transmission of the STIN.

Description

Technical Field

[0001] The present invention relates to the technical field of communication networks, and particularly to a method and system for deterministic transmission scheduling of differentiated service flows in a satellite-terrestrial integrated network. Background Art

[0002] Driven by the large-scale access of terminals, 5G networks have developed rapidly and been deployed on a large scale. Under this background, the demand for wide geographical coverage in communication networks has become increasingly urgent. At the same time, the focus of network functions has shifted from simply "reliable transmission" to providing "differentiated and customizable services". New network applications have put forward more stringent service quality standards for network latency and jitter, prompting the evolution of communication networks from the traditional best-effort service mode to a deterministic service mode.

[0003] However, there are currently many challenges. Firstly, existing terrestrial mobile communication systems are limited by their limited coverage and it is difficult to achieve the goal of global seamless communication. Although the satellite-terrestrial integrated network (STIN) can provide universal, consistent and scalable services by integrating satellite networks and terrestrial networks, and effectively alleviates the coverage blind spot problem of terrestrial networks to a certain extent, STIN itself does not have the corresponding support capabilities for customized and differentiated services. Secondly, the traditional Internet can only reduce the end-to-end average delay to the order of dozens of milliseconds, and cannot achieve strict upper limit guarantees for latency. Although existing deterministic technologies have made certain progress in specific scenarios, there is no mature system in the transmission scheduling scheme that is applicable to the huge STIN system. Thirdly, the adaptation between protocol stack levels is loose, lacking global control and local optimization means, resulting in a lack of coordination between upper-layer network services and lower-layer network resources, and between large-scale transmission and small-scale scheduling, making it difficult to ensure the stability and continuity of transmission scheduling.

[0004] In view of the above defects, it is urgent to explore and construct a more complete satellite-terrestrial integrated network architecture to meet the differentiated needs of massive services for deterministic services. In addition, due to the characteristics of communication networks as cross-disciplinary complex systems, it is necessary to start from application requirements and conduct top-down system design to make the results widely applicable. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for deterministic transmission scheduling of differentiated service flows in a satellite-terrestrial integrated network, so as to solve at least one technical problem such as the traditional STIN network architecture being unable to achieve deterministic communication guarantee as described in the above background art.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a method for deterministic transmission scheduling of differentiated service flows in a satellite-ground integrated network, including: based on the total demand of the service flow, using a genetic-based elastic transmission algorithm for cache utilization to calculate and reserve complementary satellite-ground cross-domain peak cache resources, and perform adaptive optimization of cache resources within the domain to obtain the domain's node resource allocation result; according to the node resource allocation result, using a transient routing and variable queue algorithm based on deep reinforcement learning to perform two-dimensional routing-queue scheduling on the service flow for deterministic transmission scheduling to meet the differentiated cache capacity requirements and time requirements of the service flow; the transient routing and variable queue algorithm based on deep reinforcement learning includes: on the basis of the deep Q network, introducing the idea of Double DQN to form a dual-network structure in which action selection and target Q-value calculation are separated. The structures of these two networks are the same but the parameters are different. Action selection is completed by the online network, while the calculation of the target Q-value is completed by the target network.

[0008] In a second aspect, the present invention provides a system for deterministic transmission scheduling of differentiated service flows in a satellite-ground integrated network, including: a resource adaptation module for calculating and reserving complementary satellite-ground cross-domain peak cache resources based on the total demand of the service flow using a genetic-based elastic transmission algorithm for cache utilization, and performing adaptive optimization of cache resources within the domain to obtain the domain's node resource allocation result; a differentiated scheduling module for performing two-dimensional routing-queue scheduling on the service flow based on the node resource allocation result using a transient routing and variable queue algorithm based on deep reinforcement learning for deterministic transmission scheduling to meet the differentiated cache capacity requirements and time requirements of the service flow; wherein, the transient routing and variable queue algorithm based on deep reinforcement learning includes: on the basis of the deep Q network, introducing the idea of Double DQN to form a dual-network structure in which action selection and target Q-value calculation are separated. The structures of these two networks are the same but the parameters are different. Action selection is completed by the online network, while the calculation of the target Q-value is completed by the target network.

[0009] In a third aspect, the present invention provides a non-transitory computer-readable storage medium for storing computer instructions, which when executed by a processor, implement the method for deterministic transmission scheduling of differentiated service flows in a satellite-ground integrated network as described in the first aspect.

[0010] In a fourth aspect, the present invention provides a computer device including a memory and a processor, where the processor and the memory communicate with each other, the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the method for deterministic transmission scheduling of differentiated service flows in a satellite-ground integrated network as described in the first aspect.

[0011] Fifth aspect, the present invention provides an electronic device, comprising: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device runs, the processor executes the computer program stored in the memory, so that the electronic device executes the instructions for implementing the satellite-ground integrated network differentiated service flow deterministic transmission scheduling method described in the first aspect.

[0012] Advantages of the present invention: Starting from the actual business requirements, through two modules of resource adaptation and differentiated scheduling, the differentiated cache capacity requirements and time requirements of transmission scheduling are respectively met, and end-to-end low-latency and high-reliability transmission of differentiated service flows is achieved. This transmission scheduling method differentially and concurrently supports applications and services with different deterministic requirements. While maintaining low network configuration overhead, it improves the effect and efficiency of service flow transmission scheduling, enhances the operability and resource utilization rate of the network, and makes up for the deficiency of the strict time scheduling mechanism in terms of scalability.

[0013] The advantages of the additional aspects of the present invention will be more clearly given in the following description part, or can be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0015] Figure 1 It is a schematic diagram of the deterministic satellite-ground integrated network architecture described in the embodiments of the present invention.

[0016] Figure 2 It is a flowchart of the deterministic technology described in the embodiments of the present invention.

[0017] Figure 3 It is a flowchart of the resource adaptation module described in the embodiments of the present invention.

[0018] Figure 4 It is a flowchart of the differentiated scheduling module described in the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] To facilitate the understanding of the present invention, the following will further explain the present invention with specific embodiments in conjunction with the drawings, and the specific embodiments do not constitute a limitation to the embodiments of the present invention.

[0020] Those skilled in the art should understand that the drawings are only schematic diagrams of the embodiments, and the components in the drawings are not necessarily essential for implementing the present invention.

[0021] The Deterministic Satellite - terrestrial Network (DetSTIN) architecture and service flow transmission scheduling method provided by the present invention. The DetSTIN architecture includes: a deterministic technology framework and a Satellite - terrestrial Network (STIN) entity. The deterministic technology framework realizes the smooth interconnection and integration of heterogeneous networks by providing hierarchical deterministic services. The STIN entity network is composed of fixed - mobile terminals, terrestrial networks, and satellite networks, and realizes end - to - end low - latency and high - reliability transmission of a large number of differentiated services under the support of the deterministic technology framework. The service flow transmission scheduling method includes: 1) Resource adaptation: Dynamically allocate the cache resources of nodes by inter - domain peak cache resource reservation and intra - domain adaptive cache optimization, and minimize the resource allocation overhead while maintaining network performance; 2) Differentiated scheduling: Optimize the transmission scheduling benefit while ensuring the transmission effect of the service flow by dynamically determining the transmission path and queue selection of the service flow. The DetSTIN architecture and transmission scheduling method provided by the present invention make up for the defects of the traditional STIN network architecture lacking the concept of differentiated deterministic services and loose adaptation between protocol stack levels, significantly improve the deterministic communication guarantee ability of the network in different scenarios, and provide strong support for the efficient operation and reliable transmission of STIN.

[0022] Example 1

[0023] In this Example 1, first, a deterministic transmission scheduling system for differentiated service flows in a satellite - terrestrial network is provided, including: a resource adaptation module (i.e., the Centralized Intelligent Control Center (CICC)), which is used to calculate and reserve complementary satellite - terrestrial cross - domain peak cache resources based on the genetic - based cache utilization elastic transmission algorithm according to the total demand of the service flow, and perform adaptive optimization of the cache resources within the domain to obtain the domain - wide node resource allocation result; among them, the genetic - based cache utilization elastic transmission algorithm considers the cache resource configuration overhead The throughput rate τ and the resource utilization rate η, that is:

[0024]

[0025] s.t. τ≥τ req , η≥η req

[0026] Among them, Λ, Γ, Π are the weight coefficients of each index, with Λ>Γ>Π, Λ + Γ+Π = 1; represents the normalized cache resource configuration overhead.

[0027] The differential scheduling module (i.e., the Domain Controller, DC) is used to perform two-dimensional routing-queue scheduling on service flows based on the transient routing and variable queue algorithm of deep reinforcement learning according to the node resource allocation result, and perform deterministic transmission scheduling to meet the differentiated cache capacity requirements and time requirements of service flows; among them, the transient routing and variable queue algorithm based on deep reinforcement learning includes: on the basis of the deep Q network, introducing the idea of Double DQN to form a dual-network structure with separate action selection and target Q value calculation. The structures of these two networks are the same but the parameters are different. Action selection is completed by the online network, while the calculation of the target Q value is completed by the target network.

[0028] In this Embodiment 1, using the above system, a method for differential service flow transmission scheduling based on a satellite-ground integrated network is implemented, including: using the resource adaptation module to calculate the reserved complementary satellite-ground cross-domain peak cache resources based on the genetic cache utilization elastic transmission algorithm according to the total requirements of the service flow, and performing adaptive optimization of the cache resources within the domain to obtain the node resource allocation result within the domain; among them, the genetic cache utilization elastic transmission algorithm considers the cache resource configuration overhead The throughput τ and the resource utilization rate η, that is:

[0029]

[0030] s.t. τ≥τ req , η≥η req

[0031] Among them, Λ, Γ, and Π are the weight coefficients of each index, with Λ>Γ>Π and Λ + Γ + Π = 1; represents the normalized cache resource configuration overhead. Use the differential scheduling module to perform two-dimensional routing-queue scheduling on service flows based on the transient routing and variable queue algorithm of deep reinforcement learning according to the node resource allocation result, and perform deterministic transmission scheduling to meet the differentiated cache capacity requirements and time requirements of service flows; among them, the transient routing and variable queue algorithm based on deep reinforcement learning includes: on the basis of the deep Q network, introducing the idea of Double DQN to form a dual-network structure with separate action selection and target Q value calculation. The structures of these two networks are the same but the parameters are different. Action selection is completed by the online network, while the calculation of the target Q value is completed by the target network.

[0032] The genetic-based cache-utilized elastic transmission algorithm eliminates the node-degree constraint in the initialization process by randomly generating an initial population within a reasonable range; during the crossover process, a reflexive operation is introduced for individuals that do not undergo crossover to further enrich the population diversity; in the mutation stage, the mutation amplitude is adaptively adjusted according to the degree value of the node, ensuring that nodes with larger degree values are adjusted towards higher resource allocation directions, while nodes with larger degree values usually require more resources due to their central role in the network; conversely, nodes with lower degrees have less demand for resources and are adjusted towards lower resource allocation directions.

[0033] In the transient routing and variable queue algorithm of deep reinforcement learning, two state matrices are constructed based on the environmental input. The first state matrix describes the transmission requirements of the traffic flow, including the source address, destination address, flow size, and transmission delay requirements. The second state matrix records the queue cache capacity of each node across domains; based on these two state matrices, decisions are made in two stages.

[0034] The first stage involves global routing selection decisions. A suitable end-to-end transmission path is selected according to the transmission delay requirements of the traffic flow. This path selection is carried out in two steps: first, a suitable transmission medium is selected, and then the specific route is determined according to the selected medium to ensure that the path meets the delay and reliability requirements of the traffic flow.

[0035] The second stage involves local queue scheduling decisions; a suitable sending queue is selected at the source node to determine the sending time of the traffic flow; for AVB and BE flows, a maximum waiting delay is selected to achieve time isolation queuing of differentiated traffic flows; if the local scheduling decision cannot meet the delay requirements or the cache capacity of the node is insufficient, a compensation strategy is triggered to attempt re-scheduling; within an acceptable delay range, traffic flows that fail to be scheduled according to the initial decision or arrive early due to upstream gating failure are re-scheduled to a relatively sub-optimal queue and bear an additional delay penalty.

[0036] After the scheduling decision is completed, the corresponding reward is obtained according to the decision; the current state, action, and reward information are stored in the experience replay pool, and then the process is switched to the next state to continue the decision-making process.

[0037] Embodiment 2

[0038] In this Embodiment 2, using the idea and characteristics of the intelligent collaborative network, a deterministic satellite-terrestrial integrated network (DetSTIN) architecture is provided to support deterministic communication, as Figure 1 shown. This architecture includes two parts: a deterministic technology framework and a satellite-terrestrial integrated network (STIN) entity. The deterministic technology framework provides hierarchical deterministic services for the STIN entity network to achieve smooth interconnection and integration of heterogeneous networks.

[0039] In this embodiment, the deterministic technology framework is responsible for making transmission scheduling decisions and supporting the STIN entity through deterministic transmission scheduling, including:

[0040] The Centralized Intelligent Control Center (CICC) is responsible for global management and control, determining peak resource allocation, and making global routing decisions;

[0041] The CICC is responsible for global management and control, and realizes centralized management of the entire network through the deterministic technology framework. It supervises the global time slot - cycle mapping to ensure that each time slot can be allocated and used according to a predetermined cycle, thus providing time guarantee for deterministic transmission;

[0042] The CICC determines peak resource allocation. According to the global resource status of the network and service requirements, it reasonably allocates peak resources to ensure that the network can still operate stably during peak service traffic periods and meet the requirements of deterministic transmission;

[0043] The CICC makes global routing decisions. By comprehensively considering factors such as network topology, link status, and service priority, it selects the optimal transmission path to achieve end - to - end (E2E) deterministic delay performance;

[0044] The Domain Controller (DC) works in coordination with the CICC and is responsible for control and management within the domain;

[0045] The DC receives the global routing decision and resource allocation instructions from the CICC and refines them to specific links and nodes within the domain to ensure the reasonable utilization of domain resources and the orderly transmission of services;

[0046] The DC is responsible for dynamically optimizing peak resource allocation, real - time monitoring the resource usage of links and nodes within the domain, and flexibly adjusting the resource allocation strategy according to the change of service traffic to cope with sudden service peaks and ensure the transmission performance and stability within the domain;

[0047] Based on the segmented delay quota allocated by the CICC, the DC makes precise delay allocation for each link within the domain and cross - domain links to ensure that each link can transmit data according to the predetermined delay requirements, thus realizing deterministic flow transmission within the domain.

[0048] In this embodiment, the Satellite - Terrestrial Integration Network (STIN) entity is responsible for the end - to - end transmission of differentiated service flows generated by fixed - mobile terminals, including:

[0049] Fixed-mobile terminals include a large number of devices equipped with dual-mode communication chips. Through the automatic switching function of the dual-mode communication chips, seamless switching of service flows between terrestrial networks and satellite networks is achieved, thus ensuring the efficient and accurate forwarding of service flows;

[0050] The terrestrial network includes a terrestrial communication network, a gateway station, and a centralized intelligent control center (CICC). Among them, the gateway station is responsible for establishing a satellite-terrestrial optical communication link with the satellite network to facilitate efficient data exchange between the terrestrial and satellite systems. In addition, as the global control and management center of DetSTIN, the CICC supports deterministic transmission scheduling and provides high-quality services between satellites and the ground;

[0051] The satellite network includes geostationary orbit (GEO) satellites, medium Earth orbit (MEO) satellites, and low Earth orbit (LEO) satellites. GEO satellites are equipped with advanced on-board monitoring and control capabilities. As logical sub-controllers, they receive peak resource allocation and global routing decisions from the ground CICC and formulate control commands based on these decisions, which are distributed to MEO and LEO satellites. MEO and LEO satellites perform deterministic forwarding of service flows through highly reliable inter-satellite optical links to achieve seamless and efficient flow transmission across satellites.

[0052] In this embodiment, based on the above architecture, a service flow transmission scheduling method under the deterministic satellite-terrestrial integrated network (DetSTIN) architecture is provided. This method includes two major steps: a resource adaptation step and a differential scheduling step.

[0053] As Figure 3 shown, the process of resource adaptation includes the following steps:

[0054] Step S101: Reserve peak cache resources between domains. Specifically, based on the overall service flow cache capacity requirement, predict and reserve peak cache resources based on historical data characteristics to ensure the stability and sufficiency of inter-domain transmission. Among them, sufficient cache resources are allocated to end-domain nodes to meet the cache requirements of the overall service flow, and resources are complementary between the terrestrial forwarding domain and the satellite forwarding domain to jointly support transmission.

[0055] Step S102: Initialization of in-domain cache resource configuration. Specifically, the Buffer Utilization baseD on Genetic algorithm for Elastic Transmission (BUDGET) is adopted, and the population is used to represent the set of all in-domain cache resource configuration schemes, where each individual in the population represents one of the schemes. Taking chromosomes to represent individuals, the genes (encoding) on the chromosomes represent the cache resource configuration amounts for a certain node.

[0056] Use chromosome C i to represent the individuals in the population, where LT is the length of the chromosome. The chromosome encoding adopts natural number encoding to associate the point positions with the network node indexes. The genes on the chromosome 1 ≤ K ≤ LT. It should be noted that when performing the in-domain cache resource configuration decision iteration based on BUDGET, represents the cache resource configuration decision for the k-th node in the domain in this scheme. For it is characterized by a random number within the range of (0.1C max , C max . In particular, the gateway node is fixed at C max . Among them, represents the total cache resource requirement of the overall traffic flow. The total in-domain cache resource configuration does not exceed the peak cache resources reserved in Step S101. Different in-domain cache resource configuration schemes correspond to different initial individuals, and the initial individuals form the initial population.

[0057] Individuals form a population where is the individual C in the population after the j-th iteration, i , and p is the number of individuals in the population. For the individual there is:

[0058]

[0059] When performing the in-domain cache resource configuration decision iteration based on BUDGET, this population corresponds to the potential solutions for the in-domain cache resource configuration decision. At the beginning of the algorithm, the population is initialized, that is,

[0060] Step S103: Optimization of in-domain cache resource configuration. Specifically, the Buffer Utilization baseD on Genetic algorithm for Elastic Transmission (BUDGET) is adopted to select the optimal individuals in the current population and perform crossover / reflection and mutation operations on the selected individuals to obtain potential solutions. Specifically, the following three processes are executed to iteratively approach the optimal solution.

[0061] Step S1031: Select the optimal individual according to the fitness. Specifically, use the fitness function Fitness as the index for selecting the optimal individual, measure each individual in the population by calculating its corresponding value, and exclude individuals with low fitness. The calculation formula of the fitness function is:

[0062]

[0063] s.t. τ≥τ req , η≥η req

[0064] where Λ, Γ, and Π are the weight coefficients of each index, with Λ>Γ>∏ and Λ + Γ + ∏ = 1. represents the normalized cache resource allocation overhead. The fitness directly reflects the network performance. Under the premise of maintaining a certain throughput rate and resource utilization rate, a lower resource allocation overhead corresponds to a higher fitness.

[0065] Individuals with high fitness have a greater chance of passing on their genes to the next generation. Therefore, the selection process adopts a roulette wheel algorithm based on probability. First, adopt the elite strategy and directly copy the top 5% of the individuals with the highest fitness to the offspring population, that is

[0066] Elites = Top σ (N1(SortDesc(fitness)))

[0067] Then perform parental selection. Use viability to evaluate the solution quality of the genetic iteration process, that is Obviously, the greater the viability, the greater the fitness value and the better the individual. And individuals with greater viability have more reproduction opportunities in the next generation, and vice versa. Select individuals that satisfy r ∈ [0, 1) as parental individuals.

[0068] Step S1032: Parental chromosome crossover / reflection. Specifically, once the parents are selected, randomly select a non-gateway node as the operation point for a probability-based operation. When the probability p c is satisfied, perform single-point crossover; otherwise, when the probability 1 - p c is satisfied, perform single-point reflection operation. In the case of crossover, the genes of offspring 1 are inherited from parent 1 before the operation point and from parent 2 after the operation point, while the genes of offspring 2 are the opposite. In the case of reflection, the genes of each offspring are mirrored at the operation point, generating two self-reflected offspring.

[0069] Step S1033: Gene mutation of offspring chromosomes. Specifically, for the generated offspring population, including the retained elite individuals and the individuals generated by parental chromosome crossover or reflexivity, gene mutation is performed with an adaptation probability p m The adaptive mutation is based on the population individuals and the average fitness to dynamically adjust the mutation rate. For individuals with lower fitness, a higher significant mutation probability is assigned, aiming to discover new potential solutions. The specific form of the mutation operator is:

[0070]

[0071] where represents the difference between the maximum mutation probability and the minimum mutation probability, At the same time, the magnitude of the mutation is negatively correlated with the node degree. The larger the node degree, the less resources are reduced or the more resources are increased per change. Through this adaptive mutation that differentiates the consideration of the node degree, the designed algorithm can converge quickly.

[0072] Step S104: Optimal solution search and iteration. Repeat Step S103 for elite retention / parent selection, chromosome crossover / reflexivity, and chromosome gene mutation until the iteration termination condition is reached, and the optimal solution for in-domain cache resource allocation, that is, the optimal resource adaptation decision, can be obtained.

[0073] Due to the adoption of the unique adaptive mutation operator technology, the resource adaptation algorithm with elite retention and innovative reflexivity operations is superior to the classical genetic algorithm in terms of population diversity, convergence speed, and convergence accuracy.

[0074] As Figure 4 shown, the process of differential scheduling includes the following steps:

[0075] Step S201: Perception of network status and differential traffic flow transmission requirements. Specifically, the in-domain DC perceives the transmission requirements of the traffic flow, specifically covering key information such as source address, destination address, flow size, and transmission delay requirements, and reports this data to the CICC. At the same time, the in-domain cache resource allocation situation is also reported to the CICC to achieve overall coordination and optimization management. The D3QN agent deployed in the CICC constructs two state matrices according to the environmental input. The first state matrix describes the transmission requirements of the traffic flow and can be expressed as:

[0076]

[0077] The second state matrix depicts the queue cache capacity of each node across domains. It can be expressed as:

[0078]

[0079] Among them, and respectively represent the remaining buffer capacity of the s-th queue of nodes v in the domain i and v j , and the remaining buffer capacity of the s-th queue of node v represents D Δ in the domain, where Δ ∈ {β, γ}. k The remaining buffer capacity of the s-th queue of node v

[0080] Step S202: Global routing selection decision. Specifically, the agent first selects a suitable transmission medium (i.e., terrestrial or satellite network) according to the transmission delay requirements of the service flow. On this basis, a flexible delay upper limit is divided for each domain, and a suitable one is selected from all feasible segment routes for the service flow. The segmented routes form a global route to ensure that the path meets the delay and reliability requirements of the service flow. On this basis, the end-to-end delay upper limit of the service flow is reasonably allocated to each domain. Within the domain, the segment route selection can be flexibly performed according to the delay requirements of the service flow, and a suitable route is screened out from all feasible segment routes. The global route is formed by splicing and integrating these segmented routes. For example, for time-sensitive service flows, segment routes with relatively fewer hops are preferentially selected to reduce delay; while for best-effort service flows, segment routes with more hops can be selected. Through this differential path selection strategy, the delay optimization of the service flow in a large-scale range can be achieved.

[0081] Step S203: Local queue scheduling decision. Specifically, the queue scheduling distinguishes between end-domain scheduling and forwarding-domain scheduling.

[0082] The end-domain uses the Cycling Queuing and Forwarding (CQF) mechanism for scheduling. Through the cyclic gating list, the two queues use a cyclic scheduling method. At the start of the cycle, the receiving gate of queue Q0 is opened first to prepare for receiving the service flow. At the same time, in this cycle, queue Q0 in the sending state transmits all the service flows in the queue to the downstream node as a whole. After a single cycle, the working states of the two queues are switched according to the gating list. This mechanism implicitly incorporates the link delay into the upstream and downstream cycles. The service flow sent by the upstream node queue Q c (c ∈ {0, 1}) at cycle Tξ will be received by queue Q (c+1)％2 of the downstream node and sent to the next hop at cycle T ξ+1 . The agent selects a suitable sending queue according to the delay requirements of the service flow and the network state to determine the sending time of the service flow, that is, to enter the next cycle of the current queue Q0 for sending, or to enter the next queue Q1 and wait for one cycle before sending.

[0083] The forwarding domain adopts a Wide Area Deterministic (DIP) mechanism for scheduling. The opening and closing of the sending gate and the receiving gate are controlled by a periodically rotated gating list to achieve the switching of the working states of m queues. In one cycle, only one queue is in the sending state, represented by 0, and the remaining M-1 queues are in the receiving state, represented by 1. Then queue Q m In transmission cycle T ξ The working state is:

[0084]

[0085] At regular nodes, a DIP mechanism with standard mapping of upstream and downstream node queues is adopted. For the traffic flow sent by the upstream node queue Q c (c∈{0,1,…,M-1}), when this traffic flow reaches the downstream node after the propagation delay , the working state of the queue has switched times. Therefore, this traffic flow enters the queue of the downstream node and waits before being sent to the next node. At the aggregation node, in view of the fact that when mixed traffic flows are mapped to the same queue of the downstream node, using an undifferentiated mapping method is very likely to cause the phenomenon of inverted delay requirements. Based on the standard DIP mechanism as the basic framework, a service time window configuration based on priority is constructed. For TT flows with low delay requirements, due to their strict requirements for delay, they are always exempt from secondary scheduling; while for AVB and BE flows, the agent selects a maximum waiting delay to achieve time isolation queuing of differentiated traffic flows. That is, for audio and video traffic flows with high bandwidth requirements, the agent decides whether they are postponed and can be postponed by at most one queue at most, and for best-effort flows, the agent decides whether they are postponed and can be postponed by at most two queues at most. By adding an upper limit constraint to the queuing offset, it can effectively prevent low-priority traffic flows from falling into the dilemma of being postponed infinitely, and prevent them from "starving" due to the long-term occupation of queue resources by high-priority traffic flows.

[0086] Step S204: Compensation strategy rescheduling. Specifically, if the local scheduling decision cannot meet the delay requirement or the buffer capacity of the node is insufficient, the DC will trigger a compensation strategy to try to reschedule. Within an acceptable delay range, traffic flows that fail to be scheduled according to the initial decision or arrive early due to upstream gating failure can be rescheduled to a relatively sub-optimal queue and bear an additional delay penalty. This ensures that even in the presence of scheduling deviations, the DC maintains a certain degree of flexibility to handle unexpected situations while minimizing transmission interruptions. It should be noted that the rescheduling process still considers the priority of the service flow and the current load status of the available queues to ensure that the impact on high-priority service flows is minimized. This strategy takes effect automatically to ensure that traffic flows are transmitted within the global delay requirement range without further intervention by the agent.

[0087] In summary, the deterministic satellite-terrestrial integration network (DetSTIN) architecture and service flow transmission scheduling method provided in this embodiment solve the problems of the lack of differentiated deterministic service concepts and loose adaptation between protocol stack levels in traditional network architectures by adopting a deterministic technology framework to support hierarchical deterministic services for satellite-terrestrial integration network (STIN) entities. At the same time, through resource adaptation and differentiated scheduling, the cache configuration overhead is minimized while ensuring network performance, and end-to-end low-latency and high-reliability transmission of a large number of differentiated service flows is achieved, obtaining both effects and efficiency.

[0088] Embodiment 3

[0089] In this Embodiment 3, a deterministic satellite-terrestrial integration network (Deterministic Satellite-terrestrial Network, DetSTIN) architecture is first provided, including: a deterministic technology framework and satellite-terrestrial integration network (Satellite-terrestrial Network, STIN) entities. The deterministic technology framework is used to provide hierarchical deterministic services, responsible for making transmission scheduling decisions, and supporting STIN entities through deterministic transmission scheduling to achieve smooth interconnection and integration of heterogeneous networks. STIN entities are used to connect network services and network nodes, responsible for end-to-end transmission of differentiated service flows generated by fixed-mobile terminals, and achieving low-latency and high-reliability transmission under the support of the deterministic technology framework, thereby ensuring the determinacy of the network service communication process.

[0090] Specifically, the deterministic technology framework includes:

[0091] The Centralized Intelligent Control Center (CICC) is responsible for global management and control, determining peak resource allocation, and formulating global routing decisions. The CICC is responsible for global management and control, and realizes centralized management of the entire network through a deterministic technology framework. It supervises the global time slot - cycle mapping to ensure that each time slot can be allocated and used according to a predetermined cycle, thus providing a time guarantee for deterministic transmission. The CICC determines peak resource allocation. According to the global resource status and service requirements of the network, it reasonably allocates peak resources to ensure that the network can still operate stably during peak service traffic periods and meet the requirements of deterministic transmission. The CICC formulates global routing decisions. By comprehensively considering factors such as network topology, link status, and service priority, it selects the optimal transmission path to achieve end - to - end deterministic delay performance. The Domain Controller (DC) works in coordination with the CICC and is responsible for control and management within the domain. The DC receives the global routing decisions and resource allocation instructions from the CICC and refines them to specific links and nodes within the domain to ensure the reasonable utilization of resources within the domain and the orderly transmission of services. The DC is responsible for dynamically optimizing peak resource allocation, monitoring the resource usage of links and nodes within the domain in real time, and flexibly adjusting the resource allocation strategy according to changes in service traffic to cope with sudden service peaks and ensure the transmission performance and stability within the domain. Based on the segmented delay quota allocated by the CICC, the DC performs precise delay allocation for each link within the domain and cross - domain links to ensure that each link can transmit data according to the predetermined delay requirements, thus realizing deterministic flow transmission within the domain.

[0092] Specifically, the STIN entity includes: The fixed-mobile terminal includes a large number of devices equipped with dual-mode communication chips. Through the automatic switching function of the dual-mode communication chips, seamless switching of the service flow between the terrestrial network and the satellite network is achieved, thereby ensuring the efficient and accurate forwarding of the service flow. The terrestrial network includes a terrestrial communication network, a gateway station, and a CICC. Among them, the gateway station is responsible for establishing a satellite-ground laser communication link with the satellite network to facilitate efficient data exchange between the ground and satellite systems. In addition, the CICC serves as the global control and management center of DetSTIN, supports deterministic transmission scheduling, and provides high-quality services between the satellite and the ground. The satellite network includes geosynchronous earth orbit (GEO) satellites, medium earth orbit (MEO) satellites, and low earth orbit (LEO) satellites. GEO satellites are equipped with advanced on-board monitoring and control capabilities. As logical sub-controllers, they receive peak resource allocation and global routing decisions from the ground CICC and formulate control commands based on these decisions, which are distributed to MEO and LEO satellites. MEO and LEO satellites perform deterministic forwarding of the service flow through highly reliable inter-satellite laser links to achieve seamless and efficient flow transmission across satellites.

[0093] For a concrete description, in this embodiment, DetSTIN is modeled as an undirected graph G(V, ε), where the node and link sets are respectively represented by V = {v0, v1, …, v n} and ε = {e∣e kj =(v k , v j ), e jl =(v j , v l ), …}.

[0094] In this embodiment, a service flow transmission scheduling method is implemented according to the above DetSTIN architecture, including two major steps: resource adaptation and differential scheduling.

[0095] The resource adaptation module is used to ensure that the cache capacity of the service flow transmission node meets the requirements. This module deploys the Buffer Utilization baseD on Genetic algorithm for Elastic Transmission (BUDGET). It reserves complementary satellite-ground cross-domain peak cache resources according to the total demand of the service flow, and performs adaptive optimization of cache resources within the domain. At the same time, it passes the node resource allocation result to the differential scheduling module. The differential scheduling module is used to perform two-dimensional intelligent scheduling of routing-queue for the service flow. This module deploys the Transient Routing And Varied quEue aLgorithm (TRAVEL). Based on the node resource allocation result generated by the resource adaptation module, it performs deterministic transmission scheduling to meet the differential cache capacity requirements and time requirements of the service flow.

[0096] Specifically, the resource adaptation module includes:

[0097] Inter-domain peak cache resource reservation driven by CICC. From the perspective of ensuring transmission, by analyzing the characteristics of cross-domain service flows and historical data, it reasonably predicts and reserves peak cache resources to ensure the stability and sufficiency of inter-domain transmission. In this module, CICC allocates sufficient cache resources for the end-domain nodes to fully meet the cache requirements of the service flow and ensure the reliability and integrity of data during the end-to-end transmission process. The resources between the ground forwarding domain and the satellite forwarding domain are allocated by CICC in a complementary manner to jointly support the transmission.

[0098] Intra-domain adaptive cache optimization controlled by DC. From the perspective of saving resources, it dynamically adjusts the cache resource allocation according to the real-time network conditions and intra-domain traffic changes, so as to achieve efficient utilization and rapid adaptation of resources. The intra-domain node caches are complementarily allocated by the DC deploying the BUDGET algorithm to jointly complete the multi-path transmission of the flow.

[0099] Specifically, the differential scheduling module includes:

[0100] Global routing planning formulated by CICC. From the perspective of large-scale delay optimization, according to the size and delay requirements of differential service flows, it first decides the forwarding medium, that is, each service flow is forwarded through the satellite network or the ground network. At the same time, the DC returns the intra-domain feasible paths to the CICC. The CICC performs segment routing selection and allocates segment delay quotas for the intra-domain and cross-domain links according to the routing selection result. Global determinism is achieved through segment determinism. Appropriate routing planning can reduce unnecessary node hops, thereby reducing the overall delay and maximizing the satisfaction of the end-to-end transmission requirements of the overall service flow;

[0101] Differentiated traffic flows: The traffic flow set generated by a large number of fixed-mobile terminals is:

[0102]

[0103] where f i represents the i-th traffic flow in the flow set of size , and are the source node and the destination node respectively. In addition, and represent the traffic flow size and the delay upper limit respectively, and is the unit benefit of end-to-end delay optimization. According to the size and delay requirements of the traffic flows, a large number of traffic flows are divided into time-triggered (TT) flows sensitive to delay, audio / video bridging (AVB) flows with relatively high bandwidth requirements, and best-effort (BE) flows that do not provide any guarantee for performance such as delay and reliability. For the traffic flow f i , its end-to-end actual delay needs to be less than the delay upper limit of this traffic flow. Usually, for the scheduling benefit brought by unit transmission delay optimization, there is ξ TT > ξ AVB > ξ BE , so the time priority of the three types of flows during transmission scheduling is usually TT > AVB > BE;

[0104] Global routing planning: When the mixed traffic flows are injected into DetSTIN, CICC selects the forwarding medium for each traffic flow. To represent, for differential scheduling, TT flows with strict delay requirements usually choose shorter paths, and BE flows with looser delay requirements can choose longer paths. Specifically, all traffic flows generated by air monitoring tasks and BE flows originating from the ground can be forwarded by satellites, denoted as while other ground flows are forwarded by ground nodes, denoted as thus providing flexibility in balancing network load and optimizing resource utilization. For the traffic flow f f selecting the path p i , its end-to-end actual delay is composed of the link delay and the node delay , that is

[0105]

[0106] Among them, e is the directed arc segment connecting adjacent nodes, and l is the global link formed by e.

[0107] Intra-domain segment routing: traffic flow f i The complete transmission path from the source end domain to the destination end domain through the terrestrial or satellite forwarding domain can be where the end domain D α The path segments (v0, v1, …, v λ ) and are usually fixed, while the segment routes (v β in the terrestrial forwarding domain D or the satellite forwarding domain D γ , v λ+1 , …, v λ+2 , …, v μ ) are flexible. Here, λ ∈ (0, |p f |) and μ ∈ (0, |p f |) represent node indices. Thus, the link delay can be measured by , where represents the length of e kj , represents the propagation speed, and the propagation speeds of different forwarding media are different, that is

[0108]

[0109] Local queue selection formulated by the DC. From the perspective of small-scale delay optimization, according to the size and delay requirements of differentiated traffic flows, traffic flows are differentially injected into determined queues at each node within the domain for intra-domain forwarding cycle decision-making, and the high-priority queue corresponds to an earlier forwarding cycle. Fine-grained queue selection can effectively utilize the reserved cache resources, flexibly adapt to mixed traffic flows, and improve the delay determinism and predictability of end-to-end transmission;

[0110] Local queue selection: Guide the traffic flow to be forwarded periodically at each node through a deterministic transmission scheduling mechanism. Essentially, it uses a queue management method with deterministic time slots, which can achieve highly flexible per-flow scheduling and effectively utilize the time-slot cache of the node. The delay of traffic flow f i at node v j consists of transmission delay queuing delay waiting delay , that is

[0111]

[0112] Among them, the transmission delay where represents the arc segment e j sent from node v jlBandwidth. Queue delay where f j Indicates that the queue is ranked at f i The previous business flow needs to wait for the previous business flow to be issued before it can be carried out i Considering the high bandwidth of laser and fiber optic communications, transmission delay and queuing delay can be close to nanoseconds. In addition, waiting delay Related to the intrinsic gating period of the queue.

[0113] Deterministic transmission scheduling mechanism: It is used for the queuing and forwarding process of business flows by each forwarding device in the network. By implementing a deterministic gating strategy, it ensures that the business flow is sent out in a predetermined period, thereby achieving precise time control. Specifically, M gated queues are set at the forwarding device port. Each queue has two working states, sending and receiving, which are controlled by the sending gate and receiving gate respectively. The switching of the queue working state is achieved by controlling the opening and closing of the sending gate and the receiving gate through a periodically rotating gating list. The time is divided into transmission cycles T of equal length. In one cycle, only one queue is in the sending state, represented by 0, and the remaining M-1 queues are in the receiving state, represented by 1. Then the queue Q m In the transmission period T ζ The working status is:

[0114]

[0115] Given this queue configuration, set the period duration accordingly To maintain send-receive alignment between the two mechanisms, i.e.

[0116]

[0117] This balance ensures that service flows can be seamlessly coordinated and forwarded within domains with different cycle lengths, thereby ensuring the determinism of service flow transmission.

[0118] End-domain deterministic transmission scheduling: In the end domain, a Cycling Queuing and Forwarding (CQF) mechanism with precise time synchronization is adopted, and two queues (M=2) are used to periodically switch between the sending and receiving states to alternately send business flows, avoiding queue congestion and uncertain delays. Through the cyclic gating list, the two queues adopt a cyclic scheduling method. At the beginning of the cycle, the receiving gate of queue Q0 is opened first to prepare for receiving business flows. At the same time, queue Q0 in the sending state transmits all business flows in the queue to the downstream node as a whole within the cycle. After a single cycle, the working state of the two queues is switched according to the gating list. This mechanism implicitly incorporates link delays into the upstream and downstream cycles. The upstream node queue Q c In period T ζThe outgoing traffic flow will be received at the queue Q of the downstream node (c+1)％2 and sent to the next hop in cycle T. ζ+1 Since the routing of traffic flow f i in the end domain is respectively (v0, v1, …, v λ ), and the transmission times are respectively

[0119]

[0120] Deterministic Transmission Scheduling in the Forwarding Domain: In the wide area, a Deterministic IP (DIP) mechanism that combines the concept of Differentiated Service (DiffServ) is adopted. Using multiple queues (M≥3), highly flexible per-flow relative queue scheduling for the massive and differentiated traffic flows converged in the forwarding domain is achieved, and the periodic caches of nodes are effectively utilized. These flows are scheduled according to their priorities to ensure that among the flows mapped to the same cycle, TT flows are prior to AVB flows, and AVB flows are prior to BE flows, thus realizing refined flow differentiation scheduling and effectively guaranteeing the transmission performance and quality of service requirements of different flow types. This mechanism introduces four key enhancements:

[0121] First, convert the traditional slot-based absolute queue scheduling to cycle-based relative queue scheduling. This eliminates the dependence on high-precision clock synchronization between nodes and allows upstream or downstream nodes to operate with misaligned cycle timings.

[0122] Second, allocate different types of traffic flows to relative queues with specific maximum waiting delays based on the per-hop cycle offset, achieving temporal isolation. The maximum waiting delay is measured by ψ∈{1, 2, …, M - 1}.

[0123] Third, divide the wide area nodes into regular nodes and aggregation nodes.

[0124] At regular nodes, adopt the DIP mechanism with standard mapping of upstream and downstream node queues. For the traffic flow sent from the upstream node queue Q c , when this traffic flow arrives at the downstream node after the propagation delay , the working state of the queue has switched or times. Therefore, this traffic flow enters the queue (or ) of the downstream node and waits (or ) before being sent to the next node. Since the link delay between adjacent nodes for the service flow may not be an integer multiple of the cycle, we consider the queuing delay as part of the total queuing waiting delay. From this, it can be known that traffic flow fi At node v j The waiting delay is as follows:

[0125]

[0126] At the aggregation node, combining the DiffServ concept, different priority traffic flows are mapped into the upstream and downstream queues with different maximum waiting delays. By setting the maximum waiting delay, different types of traffic flows are isolated in terms of time. Specifically, the TT flow enters the relative queue without deviation with a maximum waiting time of 1 DIP cycle, ensuring the stability and low-delay characteristics of the TT flow transmission. For the AVB flow, it can enter the relative queue with a maximum waiting time of 1 DIP cycle or 2 DIP cycles. While ensuring a certain transmission efficiency, it is given a relatively flexible scheduling space and can be adaptively scheduled according to the real-time network conditions to balance its resource occupancy and transmission timeliness. For the BE flow, it is arranged to enter the relevant queue with a maximum waiting time of two DIP cycles or three DIP cycles, thus allowing the BE flow to perform transmission scheduling within a relatively loose time limit. This differential scheduling ensures that high-priority flows experience the minimum delay, while maintaining the flexibility of low-priority flows (i.e., AVB and BE), and ensuring that the BE flow has a fair chance of forwarding and preventing the BE flow from being starved. Thus, the waiting delay of traffic flow f i At node v j is

[0127]

[0128] wherein, (or ) represents the relative queue mapped by traffic flow f i .[[ID=No.26]]

[0129] Fourth, combine the compensation policy to handle special cases. Within the acceptable delay range, traffic flows that fail to be scheduled according to the initial decision or arrive early due to upstream gating failure can be rescheduled to a relatively sub-optimal queue and bear an additional delay penalty. This ensures that even in the presence of scheduling deviations, the DC maintains a certain degree of flexibility to handle unexpected situations while minimizing transmission interruptions. It should be noted that the rescheduling process still considers the priority of the service flow and the current load status of the available queues, ensuring that the impact on high-priority service flows is minimized.

[0130] In this embodiment, the genetic-based buffer utilization elastic transmission algorithm (BUDGET) includes:

[0131] An improved genetic algorithm based on fitness, where the fitness comprehensively considers the cache resource configuration overhead the throughput rate τ and the resource utilization rate η, that is:

[0132]

[0133] such that \(v\geq\tau\) req , \(\eta\geq\eta\) req

[0134] Among them, \(\Lambda\), \(\Gamma\), and \(\Pi\) are the weight coefficients of each index, with \(\Lambda>\Gamma>\Pi\) and \(\Lambda+\Gamma+\Pi = 1\). represents the normalized cache resource configuration overhead. Inspired by the principles of biological evolution, the genetic algorithm simulates the concept of "survival of the fittest". During the population evolution process, through operations such as selection, crossover, and mutation, only individuals that adapt to the environment will be selected and retained. Therefore, it is suitable for global optimization decisions.

[0135] Different from the conventional genetic algorithm, BUDGET eliminates the node degree constraint in the initialization process by randomly generating the initial population within a reasonable range. During the crossover process, BUDGET introduces a reflexive operation for individuals that do not undergo crossover, further enriching the population diversity. In the mutation stage, the algorithm adaptively adjusts the mutation amplitude according to the degree value of the node, ensuring that nodes with larger degree values are adjusted towards higher resource allocation directions, while nodes with larger degree values usually require more resources due to their central role in the network. Conversely, nodes with lower degrees have less demand for resources and are adjusted towards lower resource allocation directions. In addition, the adaptive mutation operator is related to the fitness, and a larger mutation probability is assigned to low-fitness individuals with poor performance to encourage exploration. Combining the elite retention strategy increases the population diversity, improves the pertinence of exploration, and reduces the required number of evolutionary generations.

[0136] Population initialization: The in-domain cache resource configuration scheme is abstracted as an individual, and the cache configuration of each node in the domain is characterized by the genes on the individual chromosome. The chromosome encoding uses natural number encoding to associate the point position with the network node index. For each point on the chromosome, a random number is generated within the range of \((0.1C\) max , \(C\) max to represent the node cache resource amount, and the gateway node is fixed at \(C\) max . Among them, represents the overall traffic flow cache resource requirement. Different in-domain cache resource configuration schemes correspond to different initial individuals, and the initial individuals form the initial population.

[0137] Selection: The selection process adopts a roulette wheel algorithm based on probability. Individuals with higher fitness have a greater chance of passing on their genes to the next generation. Fitness directly reflects the network performance. Under the premise of maintaining a certain throughput rate and resource utilization rate, a lower resource allocation overhead corresponds to a higher fitness. An elitist strategy is adopted in the selection process. The top 5% of individuals with the highest fitness are directly copied into the offspring population, and then parental selection is carried out. This strategy aims to ensure that the genetic information of the optimal individuals is retained, thereby enhancing the stability and convergence speed of the algorithm.

[0138] Reflexive / Crossover: Once the parents are selected, a non-gateway node is randomly selected as the operating point. When the probability p c is met, single-point crossover is performed; otherwise, when the probability 1 - p c is met, a single-point reflexive operation is performed. In the case of crossover, the genes of offspring 1 are inherited from parent 1 before the operating point and from parent 2 after the operating point, while the genes of offspring 2 are the opposite. For reflexivity, the genes of each offspring are mirrored at the operating point, producing two self-reflected offspring.

[0139] Mutation: For the generated offspring population, including the retained elite individuals and those produced by the crossover or reflexivity of parental individuals, gene mutation occurs with an adaptation probability p m . Adaptive mutation dynamically adjusts the mutation rate according to the population individuals and the average fitness . For individuals with lower fitness, a higher significant mutation probability is assigned, aiming to discover new potential solutions. The specific form of the mutation operator is

[0140]

[0141] where, represents the difference between the maximum mutation probability and the minimum mutation probability, At the same time, the magnitude of the mutation is negatively correlated with the node degree. The larger the node degree, the less resources are reduced or the more resources are increased per change. Through this adaptive mutation that differentiates considering the node degree, the designed algorithm can converge quickly.

[0142] This resource adaptive algorithm achieves a good balance between exploration and exploitation. The adaptive mutation strategy plays a key role in exploring new solutions, while the elitist retention strategy emphasizes the exploitation of existing optimal solutions.

[0143] The transient routing and variable queue algorithm (TRAVEL) based on deep reinforcement learning described above includes:

[0144] Based on the deep reinforcement learning algorithm of Dueling Double Deep Q Network (D3QN), on the basis of the Deep Q Network (DQN), D3QN introduces the idea of Double DQN, forming a double-network structure where action selection and target Q-value calculation are separated. The structures of these two networks are the same but the parameters are different. Action selection is completed by the online network, while the calculation of the target Q-value is completed by the target network. This can more accurately estimate the Q-value and avoid the problem of overestimation.

[0145] The D3QN agent constructs two state matrices based on the environmental input, providing a basis for subsequent scheduling decisions. The first state matrix describes the transmission requirements of the traffic flow, including the source address, destination address, flow size, and transmission delay requirements. The second state matrix records the queue buffer capacity of each node across domains, which can be expressed as:

[0146]

[0147] where and respectively represent the remaining buffer capacity of the s-th queue of nodes v and v i in domain j , represents the remaining buffer capacity of the s-th queue of node v Δ in domain D k , Δ ∈ {β, γ}. The agent makes a decision a t executed in two stages based on these two state matrices.

[0148]

[0149] where δ(v j ) is the number of queues owned by node v j , and

[0150]

[0151] The first stage involves global routing selection decisions. The agent selects an appropriate end-to-end transmission path according to the transmission delay requirements of the traffic flow. This path selection is carried out in two steps: first, select an appropriate transmission medium (i.e., terrestrial or satellite network), and then determine the specific route according to the selected medium to ensure that the path meets the delay and reliability requirements of the traffic flow.

[0152] The second stage involves local queue scheduling decisions. The agent selects an appropriate transmission queue at the source node to determine the transmission time of the traffic flow. For AVB and BE flows, the agent selects a maximum waiting delay to achieve time isolation queuing for differentiated traffic flows. If the local scheduling decision cannot meet the delay requirements or the buffer capacity of the node is insufficient, the DC will trigger a compensation strategy to attempt to reschedule. Within an acceptable delay range, traffic flows that fail to be scheduled according to the initial decision or arrive early due to upstream gating failure can be rescheduled to a relatively sub-optimal queue and incur an additional delay penalty. This ensures that even in the presence of scheduling deviations, the DC maintains a certain degree of flexibility to handle unexpected situations while minimizing transmission interruptions. It is worth noting that the rescheduling process still considers the priority of the service flow and the current load status of the available queues to ensure that high-priority service flows are minimally affected. This strategy takes effect automatically to ensure that traffic flows are transmitted within the global delay requirements without further intervention from the agent.

[0153] After completing the scheduling decision, the agent obtains a corresponding reward according to its decision, that is

[0154]

[0155] This reward is directly related to the delay optimization of the traffic flow. Compared with BE flows, TT flows obtain a higher reward for unit delay optimization. A unified negative reward is given in the case of transmission failure. Therefore, the agent gives priority to the optimization of TT and AVB flows and places the delay optimization of BE flows at a lower priority. A well-designed reward function helps to accelerate the agent's learning process and enables it to converge to the optimal scheduling strategy efficiently.

[0156] Finally, the current state, action, and reward information are stored in the experience replay pool, and then the agent transitions to the next state to continue the decision-making process. Through iterative optimization, the agent gradually improves its scheduling performance for different traffic flows, enhancing the overall transmission performance of DetSTIN.

[0157] Embodiment 4

[0158] This Embodiment 4 provides a non-transitory computer-readable storage medium for storing computer instructions, which when executed by a processor, implement the above-mentioned differentiated traffic flow transmission scheduling method based on the satellite-terrestrial fusion network.

[0159] Embodiment 5

[0160] Embodiment 5 provides a computer device, including a memory and a processor, where the processor and the memory communicate with each other, the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the above-mentioned differential service flow transmission scheduling method based on the satellite-ground integrated network.

[0161] Embodiment 6

[0162] Embodiment 6 provides an electronic device, including: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device runs, the processor executes the computer program stored in the memory, so that the electronic device executes the instructions for implementing the above-mentioned differential service flow transmission scheduling method based on the satellite-ground integrated network.

[0163] In summary, the DetSTIN architecture provided by the embodiments of the present invention introduces the concept of determinism on the basis of the traditional STIN architecture, enabling each functional domain to have independent and collaborative deterministic guarantee capabilities. With the support of the deterministic technology framework, smooth interconnection and integration of heterogeneous networks are achieved, thereby ensuring the differential service support capabilities from the underlying network facilities of STIN entities to the entire architecture. The service flow transmission scheduling method provided by the embodiments of the present invention starts from actual service requirements and, through two modules of resource adaptation and differential scheduling, respectively meets the differential cache capacity requirements and time requirements of transmission scheduling, realizing end-to-end low-latency and high-reliability transmission of differential service flows. Differentially and concurrently supports applications and services with different deterministic requirements. While maintaining low network configuration overhead, it improves the effect and efficiency of service flow transmission scheduling, enhances the operability and resource utilization rate of the network, and makes up for the deficiency of the strict time scheduling mechanism in terms of scalability.

[0164] Although the specific implementation manners of the present invention have been described above in conjunction with the accompanying drawings, it is not a limitation on the protection scope of the present invention. Those skilled in the art should understand that, based on the technical solutions disclosed in the present invention, various modifications or deformations that can be made by those skilled in the art without creative efforts should be covered within the protection scope of the present invention.

Claims

1. A method for deterministic transmission scheduling of differentiated service flows in a satellite-ground integrated network, characterized in that Including: According to the total requirements of the service flow, based on the genetic-based cache utilization elastic transmission algorithm, calculate the reserved complementary satellite-ground cross-domain peak cache resources, and perform adaptive optimization of the cache resources within the domain to obtain the domain node resource allocation results; among them, the genetic-based cache utilization elastic transmission algorithm considers the cache resource configuration overhead The throughput τ and the resource utilization rate η, that is: s.t. τ ≥ τ req , η ≥ η req Among them, Λ, Γ, and Π are the weight coefficients of each index, with Λ > Γ > Π and Λ + Γ + Π = 1; represents the normalized cache resource configuration overhead; Based on the node resource allocation result, the transient routing and variable queue algorithm based on deep reinforcement learning is used to perform two-dimensional scheduling of routing-queue for service flows, and deterministic transmission scheduling is performed to meet the differentiated cache capacity requirements and time requirements of service flows; among them, the transient routing and variable queue algorithm based on deep reinforcement learning includes: on the basis of the deep Q network, the idea of Double DQN is introduced to form a dual-network structure in which action selection and target Q-value calculation are separated. The structures of these two networks are the same but the parameters are different. Action selection is completed by the online network, while the calculation of the target Q-value is completed by the target network.

2. The method for deterministic transmission scheduling of differentiated service flows in a satellite-ground integrated network according to claim 1, wherein The genetic-based cache utilization elastic transmission algorithm eliminates the node degree constraint in the initialization process by randomly generating an initial population within a reasonable range; in the crossover process, a reflexive operation is introduced for individuals that do not undergo crossover to further enrich the population diversity; in the mutation stage, the mutation amplitude is adaptively adjusted according to the degree value of the node, ensuring that nodes with a larger degree value are adjusted towards a higher resource allocation direction, and nodes with a larger degree value usually require more resources due to their central role in the network; conversely, nodes with a lower degree have less demand for resources and are adjusted towards a lower resource allocation direction.

3. The deterministic transmission scheduling method for differentiated service flows in a satellite-ground integrated network according to claim 1, wherein In the transient routing and variable queue algorithm of deep reinforcement learning, two state matrices are constructed according to the environmental input. The first state matrix describes the transmission requirements of the service flow, including the source address, destination address, flow size, and transmission delay requirements, and the second state matrix records the queue cache capacity of each node across domains. Based on these two state matrices, decisions are made in two stages.

4. The deterministic transmission scheduling method for differentiated service flows in the satellite-ground integrated network according to claim 3, wherein The first stage involves global routing selection decisions. A suitable end-to-end transmission path is selected according to the transmission delay requirements of the service flow. This path selection is carried out in two steps: first, a suitable transmission medium is selected, and then the specific route is determined according to the selected medium to ensure that the path meets the delay and reliability requirements of the service flow.

5. The method for deterministic transmission scheduling of differentiated service flows in a satellite-ground integrated network according to claim 4, wherein The second stage involves local queue scheduling decisions; a suitable sending queue is selected at the source node to determine the sending time of the service flow; for AVB and BE flows, a maximum waiting delay is selected to achieve time isolation queuing of differentiated service flows. If the local scheduling decision cannot meet the delay requirements or the cache capacity of the node is insufficient, a compensation strategy is triggered to attempt re-scheduling; within an acceptable delay range, service flows that fail to be scheduled according to the initial decision or arrive early due to upstream gating failure are re-scheduled to a relatively sub-optimal queue and bear an additional delay penalty.

6. The method for deterministic transmission scheduling of differentiated service flows in a space-ground integrated network according to claim 5, wherein After the scheduling decision is completed, the corresponding reward is obtained according to the decision; the current state, action, and reward information are stored in the experience replay pool, and then the next state is transitioned to continue the decision-making process.

7. A deterministic transmission scheduling system for differentiated service flows in a space-ground integrated network, characterized in that, Including: A resource adaptation module, which is used to calculate and reserve complementary satellite-ground cross-domain peak cache resources based on a genetic-based cache utilization elastic transmission algorithm according to the total requirements of the service flow, and perform adaptive optimization of the cache resources within the domain to obtain the domain node resource allocation result; among them, the genetic-based cache utilization elastic transmission algorithm takes into account the cache resource configuration overhead The throughput τ and the resource utilization rate η, that is: such that v ≥ τ req , η ≥ η req Among them, Λ, Γ, and ∏ are the weight coefficients of each index, with Λ > Γ > ∏ and Λ + Γ + Π = 1; represents the normalized cache resource configuration overhead; A differential scheduling module, which is used to perform two-dimensional routing-queue scheduling on service flows according to the node resource allocation results and based on the transient routing and variable queue algorithm of deep reinforcement learning, and perform deterministic transmission scheduling to meet the differentiated cache capacity requirements and time requirements of service flows; among them, the transient routing and variable queue algorithm based on deep reinforcement learning includes: on the basis of the deep Q network, introducing the idea of Double DQN to form a dual-network structure with separate action selection and target Q value calculation. The structures of these two networks are the same but the parameters are different. Action selection is completed by the online network, while the calculation of the target Q value is completed by the target network.

8. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the method for deterministic transmission scheduling of differentiated service flows in the satellite-ground integrated network as described in any one of claims 1-6 is implemented.

9. A computer device, characterized in that, It includes a memory and a processor, the processor and the memory communicate with each other, the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the method for deterministic transmission scheduling of differentiated service flows in the satellite-ground integrated network as described in any one of claims 1-6.

10. An electronic device, characterized in that, It includes: A processor, a memory, and a computer program; wherein, the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device runs, the processor executes the computer program stored in the memory, so that the electronic device executes instructions for implementing the method for deterministic transmission scheduling of differentiated service flows in the satellite-ground integrated network as described in any one of claims 1-6.

Images