CN114629543A - Satellite network adaptive traffic scheduling method based on deep supervised learning - Google Patents
Satellite network adaptive traffic scheduling method based on deep supervised learning Download PDFInfo
- Publication number
- CN114629543A CN114629543A CN202210106549.3A CN202210106549A CN114629543A CN 114629543 A CN114629543 A CN 114629543A CN 202210106549 A CN202210106549 A CN 202210106549A CN 114629543 A CN114629543 A CN 114629543A
- Authority
- CN
- China
- Prior art keywords
- network
- traffic
- satellite network
- path
- routing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 27
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 15
- 239000011159 matrix material Substances 0.000 claims description 33
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000003062 neural network model Methods 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 230000005540 biological transmission Effects 0.000 claims description 4
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 abstract description 10
- 230000008569 process Effects 0.000 abstract description 3
- 230000001133 acceleration Effects 0.000 abstract description 2
- 238000005457 optimization Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 1
- 230000002542 deteriorative effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/14—Relay systems
- H04B7/15—Active relay systems
- H04B7/185—Space-based or airborne stations; Stations for satellite systems
- H04B7/18521—Systems of inter linked satellites, i.e. inter satellite service
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L45/00—Routing or path finding of packets in data switching networks
- H04L45/02—Topology update or discovery
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L45/00—Routing or path finding of packets in data switching networks
- H04L45/12—Shortest path evaluation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L45/00—Routing or path finding of packets in data switching networks
- H04L45/22—Alternate routing
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L45/00—Routing or path finding of packets in data switching networks
- H04L45/50—Routing or path finding of packets in data switching networks using label swapping, e.g. multi-protocol label switch [MPLS]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Abstract
The invention discloses a satellite network adaptive traffic scheduling method based on deep supervised learning, which realizes the calculation of a fast routing path, ensures the minimum and maximum link utilization rate under the condition of delay and solves the problems of low speed and low throughput of the traditional algorithm. The invention uses a supervised learning algorithm, trains an intelligent routing model for each satellite node pair with non-zero flow demand, simplifies the complex routing calculation process into simple input and output, avoids multiple iterations during calculation so as to realize rapid calculation of routing paths, reduces forwarding delay due to the acceleration of the routing algorithm speed, ensures that the data packets discarded due to expiry of ttl are more likely to survive and successfully forwarded, and increases network throughput. The invention is provided with two stages of off-line training and on-line training, and the parameters are updated in a dynamic environment to select the optimal path, so that the invention has topology self-adaptability.
Description
Technical Field
The invention particularly relates to a satellite network self-adaptive traffic scheduling method based on deep supervised learning, and belongs to the field of satellite data transmission.
Background
Efficient networking is required among satellites. For an efficient self-organizing network of a small satellite constellation, how to provide an efficient and reliable traffic scheduling scheme for a satellite network is particularly important. In a low-orbit satellite network, satellite nodes move quickly, and the network topology changes frequently, so that the satellite network has the characteristics of poor expandability, low utilization rate of node link resources, high dynamic change of services and the like, and the flow scheduling becomes a difficult problem of the satellite network. How to process the network time-varying topology and the high-dynamic-change traffic service caused by the node movement is a key problem for solving the satellite network traffic scheduling.
Most of the traditional traffic scheduling methods are based on a shortest path algorithm, which determines the routing information of a service by adjusting link weights and configuring routing parameters, and takes some simple network parameters (such as path hop number, time delay and the like) as the optimization target of the algorithm, thereby realizing the routing optimization of the traffic flow and achieving load balancing. However, in a high dynamic service scenario, due to frequent topology changes, a single metric and an optimization target, congestion of part of key links is easily caused, and a problem of unbalanced network load is caused. Although the optimal path meeting the multiple constraint conditions can be found by the shortest path algorithm based on Lagrange relaxation, the heuristic algorithm needs multiple iterations to obtain the optimal path, the calculation cost is high, the timeliness is poor, and the optimization effect is not ideal under the high dynamic change environment of the network.
Disclosure of Invention
The invention aims to: aiming at the defects in the prior art, the satellite network adaptive traffic scheduling method based on deep supervised learning provided by the invention solves the problems in the prior art.
In order to achieve the purpose of the invention, the invention adopts the technical scheme that:
a satellite network adaptive traffic scheduling method based on deep supervised learning comprises the following steps:
s1, collecting traffic matrixes at multiple moments from the satellite network to obtain traffic demands among all node pairs in the satellite network at the multiple moments;
s2, calculating alternative path sets of all non-zero traffic demand node pairs in the satellite network by using a K shortest-path algorithm;
s3, obtaining the optimal data forwarding paths of all non-zero flow demand node pairs under each flow matrix by using a label data generation algorithm, and generating a label data set;
s4, performing off-line training on the intelligent routing model of each non-zero flow demand node pair by using a back propagation algorithm according to the label data set, and deploying the trained model in a satellite network;
s5, collecting a real-time traffic matrix at the current moment from the network;
s6, the satellite network calculates the optimal forwarding path of all nodes to the data route according to the real-time traffic matrix at the current moment by using the trained intelligent route model;
s7, the satellite network transmits data according to the calculated optimal path;
s8, detecting whether the topological structure of the satellite network is changed, if so, entering the step S9, otherwise, repeating the step S8;
and S9, performing online training, retraining the intelligent routing model of each node pair, and performing data transmission by using the updated intelligent routing model.
Further, in the step S1, the traffic matrix represents traffic demands between all pairs of source nodes and destination nodes in the satellite network, and the mth row and the nth column of the traffic matrix represent traffic demands from the node m to the node n.
Further, step S2 uses a K shortest-path algorithm to calculate K different loop-free candidate routes for each non-zero traffic demand node pair in the network, and the calculated candidate routes serve as the candidate route set of the node pair.
Further, in step S3, obtaining the optimal data forwarding path of all non-zero traffic demand node pairs under each traffic matrix by using a label data generation algorithm, and generating a label data set, specifically:
s31, initializing the sum of the flow of all links in the satellite network to be 0 for each flow matrix, and sequencing the D service flows to make a count value i equal to 1;
s32, placing the ith service into the network topology, namely from the alternative path set PdTaking out an alternative path p, adding the occupied bandwidth of the service to the sum of the flow of the links passed by the path p, and calculating the maximum link utilization ratio z of the current network; comparing the z values of all the alternative paths, and selecting the alternative path corresponding to the minimum value as the routing scheme of the service;
s33, checking whether the maximum link utilization z of the current network exceeds the preset value zmax(ii) a If yes, executing step S34, otherwise executing step S36;
s34, taking the placed last N services out of the network, wherein N is greater than 0, and subtracting the demand of the N services from the sum of the flow of the passing link;
s35, reordering the N taken-out services, putting the N taken-out services into the network again according to the current sequence, and executing the step S33;
s36, if i is not more than D, adding 1 to the value of i, executing step S32, otherwise, ending, and outputting the routing schemes of all the traffic services;
s37, representing the routing scheme of each service by using an optimal path label, wherein each traffic matrix has an optimal path label corresponding to the traffic matrix on each node pair required by the non-zero traffic service; when the routing strategies of the N traffic matrices are calculated, each node pair generates N pieces of training data to form a training database of the node pair, namely a label data set.
Further, in the step S4, the intelligent routing model is a BP fully-connected neural network; the input of the BP full-connection neural network comprises the network state of the SDN network, wherein the network state comprises network topology, a flow matrix, a bandwidth requirement and a time delay requirement; the BP full-connection neural network comprises an input layer, a full-connection layer and an output layer which are connected in sequence.
Further, offline training is performed on the neural network model of each node pair by adopting a back propagation algorithm, which specifically comprises the following steps:
will input quantity xiAnd the weight value wiIs multiplied by a given threshold biMaking a difference, and obtaining an output value y through the difference value through a transfer function fi(ii) a Meanwhile, comparing errors between the output value of the network and the expected value, and continuously adjusting and modifying link weights and thresholds among all layers of the network by utilizing a gradient descent method through the back propagation of the errors, so that the output result is closer to a real value and is converged to a preset target error function value E; changing the weight coefficient through a back propagation algorithm, randomly assigning an initial weight by a network, and then adjusting according to a network iteration result until a target error is met; and when the variation values of all the parameters are smaller than the iteration threshold or reach the maximum iteration times, finishing the model training.
Further, in step S9, the network controller monitors the change of the network topology in real time to detect whether the network topology changes, and when the network topology changes more than expected, the neural network model is retrained and deployed offline, and the model parameters are adjusted to adapt to a new environment, thereby implementing self-adaptation.
Compared with the prior art, the invention has the advantages that:
(1) the invention realizes the calculation of the fast routing path, maximizes the throughput under the condition of ensuring the delay and solves the problems of low speed and low throughput of the traditional algorithm.
(2) The invention uses the supervised learning algorithm, simplifies the route calculation process into simple input and output, avoids multiple iterations during calculation so as to realize the rapid calculation of the route path, reduces the forwarding delay due to the acceleration of the speed of the route algorithm, ensures that the data packet which is discarded due to the expiry of ttl has a higher probability of survival and successful forwarding, and increases the network throughput.
(3) The invention is provided with two stages of off-line training and on-line training, and the parameters are updated in a dynamic environment to select the optimal path, so that the invention has topology self-adaptability.
Drawings
Fig. 1 is a flowchart of a satellite network adaptive traffic scheduling method based on deep supervised learning according to the present invention;
FIG. 2 is a model diagram of a satellite network adaptive traffic scheduling method based on deep supervised learning;
FIG. 3 is a schematic diagram of a labeled training database;
fig. 4 is a schematic diagram of a BP neural network structure.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined in the appended claims, and all matters produced by the invention using the inventive concept are protected.
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
As shown in fig. 1, the present invention provides a method for adaptive traffic scheduling of a satellite network based on deep supervised learning, which includes the following steps:
and S1, collecting traffic matrixes at multiple moments from the network to obtain traffic demands among all node pairs at multiple moments.
In step S1, the traffic matrix is the traffic demand between all the source and destination node pairs in the network, for example, the mth row and the nth column of the traffic matrix represent the traffic demands from node m to node n.
And S2, calculating the alternative path set of all non-zero traffic demand node pairs in the network by using the K shortest path algorithm.
In step S2, a K shortest path algorithm is used to calculate K different loop-free candidate routes for each non-zero traffic demand node pair in the network, and the calculated K different loop-free candidate routes are used as the candidate route set of the node pair.
As shown in fig. 2, each node pair with non-zero traffic demand in the network maintains a network intelligent routing model for the node pair's traffic routing decisions. The network controller receives the updated network state (such as a traffic matrix, a link residual bandwidth, service end-to-end delay and the like) in real time, sends the updated network state to the network intelligent routing model of each node pair for routing decision, and receives the optimal path scheme of each node pair to form the network total routing scheme at the current moment so as to realize traffic scheduling.
And S3, obtaining the optimal data forwarding paths of all non-zero traffic demand node pairs under each traffic matrix by using a label data generation algorithm, and generating a label data set.
Further, the step S3 includes the following sub-steps:
s31, for each traffic matrix, initializing the sum of the traffic of all links in the network to 0, and sorting the D traffic flows, so that the count value i is equal to 1.
S32, placing the ith service into the network topology, namely from the alternative path set PdTaking out the alternative path p, adding the occupied bandwidth of the service to the sum of the flow of the links passed by the path p, and calculating the maximum link utilization rate z of the current network; and comparing the z values of all the alternative paths, and selecting the alternative path corresponding to the minimum value as the routing scheme of the service.
S33, checking whether the maximum link utilization z of the current network exceeds a preset value zmax. If so, go to step S34, otherwise go to step S36.
S34, taking the last N (N >0) pieces of business which are put in from the network, and subtracting the demand of the N pieces of business from the sum of the flow of the passed links.
S35, reordering the N taken-out services, putting the N taken-out services into the network again according to the current sequence, and executing the step S33.
And S36, if i is less than or equal to D, making i equal to i +1, executing a step S32, and if not, ending the step, and outputting the routing schemes of all the traffic.
S37, representing the routing scheme of each service by using an optimal path label, wherein each traffic matrix has an optimal path label corresponding to the traffic matrix on each node pair required by the non-zero traffic service; when the routing strategies of the N traffic matrices are calculated, each node pair generates N pieces of training data to form a training database of the node pair, namely a label data set.
After the traffic matrix is collected from the network, the optimal routing scheme of each service can be solved quickly through the steps. The routing scheme of each service is represented by an optimal path label, for example, when the candidate path set includes 5 candidate paths, the label is represented by a 3-bit binary number, that is, the first candidate path is represented by 000, the second candidate path is represented by 001, and so on. Thus, each traffic matrix has an optimal path label corresponding to the traffic matrix on each node pair with a non-zero traffic demand. When the routing strategies of the N traffic matrices are calculated, each node pair generates N pieces of training data to form a training database of the node pair, as shown in fig. 3.
And S4, performing off-line training on the intelligent routing model of each non-zero flow demand node pair by using a back propagation algorithm according to the label data set, and deploying the trained model in the satellite network.
As shown in fig. 4, the intelligent routing model is a BP fully-connected neural network; the input of the neural network model comprises network states of the SDN network, wherein the network states are network topology, a flow matrix, bandwidth requirements and delay requirements; the BP full-connection neural network comprises an input layer, a full-connection layer and an output layer which are connected in sequence.
Performing offline on the neural network model of each node pair by adopting a back propagation algorithmAnd (5) training. Will input quantity xiAnd the weight value wiIs multiplied by a given threshold biMaking a difference, and obtaining an output value y by the value through a transfer function fi. And simultaneously comparing errors between the output value of the network and the expected value, and continuously adjusting and modifying the link weight and the threshold value between each layer of the network by using a gradient descent method through the back propagation of the errors, so that the output result is closer to the true value and is converged to a preset target error function value E. And changing the weight coefficient through a back propagation algorithm, randomly assigning the initial weight by the network, and then adjusting according to the network iteration result until the target error is met. And when the variation values of all the parameters are smaller than the iteration threshold or reach the maximum iteration times, finishing the model training.
And S5, acquiring the real-time traffic matrix at the current moment from the network.
And S6, the satellite network calculates the optimal forwarding path of all nodes to the data route according to the real-time traffic matrix at the current moment by using the trained intelligent routing model.
And S7, the satellite network transmits data according to the calculated optimal path.
S8, detecting whether the topological structure of the satellite network is changed, if so, entering the step S9, otherwise, repeating the step S8.
And S9, performing on-line training, retraining the intelligent routing model of each node pair, and performing data transmission by using the updated intelligent routing model.
Detecting, by the network controller, whether a change in the network topology has occurred. When a large number of links in the network fail, if the original neural network model is still used for path prediction at this time, the obtained path may contain failed nodes or links, thereby deteriorating routing decision. Therefore, the network controller monitors the change condition of the network topology in real time, when the network topology is changed greatly, the neural network model is trained and deployed off line, the model parameters are adjusted to adapt to a new environment, and the self-adaptability is achieved.
The invention realizes the calculation of the fast routing path, maximizes the throughput under the condition of ensuring the delay and solves the problems of low speed and low throughput of the traditional algorithm.
Claims (7)
1. A satellite network adaptive traffic scheduling method based on deep supervised learning is characterized by comprising the following steps:
s1, collecting traffic matrixes at multiple moments from the satellite network to obtain traffic demands among all node pairs in the satellite network at the multiple moments;
s2, calculating alternative path sets of all non-zero traffic demand node pairs in the satellite network by using a K shortest path algorithm;
s3, obtaining the optimal data forwarding paths of all non-zero traffic demand node pairs under each traffic matrix by using a label data generation algorithm, and generating a label data set;
s4, performing off-line training on the intelligent routing model of each non-zero flow demand node pair by using a back propagation algorithm according to the label data set, and deploying the trained model in a satellite network;
s5, collecting a real-time flow matrix at the current moment from the network;
s6, the satellite network calculates the optimal forwarding path of all nodes to the data route according to the real-time traffic matrix at the current moment by using the trained intelligent route model;
s7, the satellite network transmits data according to the calculated optimal path;
s8, detecting whether the topological structure of the satellite network is changed, if so, entering the step S9, otherwise, repeating the step S8;
and S9, performing on-line training, retraining the intelligent routing model of each node pair, and performing data transmission by using the updated intelligent routing model.
2. The deep supervised learning based satellite network adaptive traffic scheduling method according to claim 1, wherein the method comprises the following steps: in the step S1, the traffic matrix represents traffic demands between all pairs of source nodes and destination nodes in the satellite network, and the mth row and the nth column of the traffic matrix represent traffic demands from the node m to the node n.
3. The deep supervised learning based satellite network adaptive traffic scheduling method according to claim 1, wherein the method comprises the following steps: step S2 uses K shortest path algorithm to calculate K different loop-free candidate routes for each non-zero traffic demand node pair in the network, and the calculated candidate routes serve as the candidate route set of the node pair.
4. The deep supervised learning based satellite network adaptive traffic scheduling method according to claim 2, wherein the method comprises the following steps: in step S3, obtaining the optimal data forwarding path of all non-zero traffic demand node pairs under each traffic matrix by using a label data generation algorithm, and generating a label data set, specifically:
s31, initializing the sum of the flow of all links in the satellite network to be 0 for each flow matrix, and sequencing the D service flows to make a count value i equal to 1;
s32, placing the ith service into the network topology, namely from the alternative path set PdTaking out an alternative path p, adding the occupied bandwidth of the service to the sum of the flow of the links passed by the path p, and calculating the maximum link utilization ratio z of the current network; comparing the z values of all the alternative paths, and selecting the alternative path corresponding to the minimum value as the routing scheme of the service;
s33, checking whether the maximum link utilization z of the current network exceeds the preset value zmax(ii) a If yes, executing step S34, otherwise executing step S36;
s34, taking the placed last N services out of the network, wherein N is greater than 0, and subtracting the demand of the N services from the sum of the flow of the passing link;
s35, reordering the N taken-out services, putting the N taken-out services into the network again according to the current sequence, and executing the step S33;
s36, if i is less than or equal to D, adding 1 to the value of i, executing step S32, otherwise, ending, and outputting the routing schemes of all the traffic services;
s37, representing the routing scheme of each service by using an optimal path label, wherein each traffic matrix has an optimal path label corresponding to the traffic matrix on each node pair required by the non-zero traffic service; when the routing strategies of the N traffic matrixes are calculated, each node pair generates N pieces of training data to form a training database of the node pair, namely a label data set.
5. The deep supervised learning based satellite network adaptive traffic scheduling method according to claim 4, wherein the method comprises the following steps: the intelligent routing model in the step S4 is a BP fully-connected neural network; the input of the BP full-connection neural network comprises the network state of the SDN network, wherein the network state comprises network topology, a flow matrix, a bandwidth requirement and a time delay requirement; the BP full-connection neural network comprises an input layer, a full-connection layer and an output layer which are connected in sequence.
6. The deep supervised learning based satellite network adaptive traffic scheduling method according to claim 5, wherein the method comprises the following steps: the neural network model of each node pair is trained off line by adopting a back propagation algorithm, and the method specifically comprises the following steps:
will input quantity xiAnd the weight value wiIs multiplied by a given threshold biMaking a difference, and obtaining an output value y through the difference value through a transfer function fi(ii) a Meanwhile, comparing errors between the output value of the network and the expected value, and continuously adjusting and modifying link weights and thresholds among all layers of the network by utilizing a gradient descent method through the back propagation of the errors, so that the output result is closer to a real value and is converged to a preset target error function value E; changing the weight coefficient through a back propagation algorithm, randomly assigning an initial weight by a network, and then adjusting according to a network iteration result until a target error is met; and when the variation values of all the parameters are smaller than the iteration threshold or reach the maximum iteration times, finishing the model training.
7. The deep supervised learning based satellite network adaptive traffic scheduling method according to claim 1, wherein the method comprises the following steps: in step S9, the network controller monitors the change of the network topology in real time, detects whether the network topology changes, and when the network topology changes more than expected, the neural network model is retrained and deployed offline, and the model parameters are adjusted to adapt to the new environment, thereby implementing self-adaptation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210106549.3A CN114629543B (en) | 2022-01-28 | 2022-01-28 | Satellite network self-adaptive flow scheduling method based on deep supervised learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210106549.3A CN114629543B (en) | 2022-01-28 | 2022-01-28 | Satellite network self-adaptive flow scheduling method based on deep supervised learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114629543A true CN114629543A (en) | 2022-06-14 |
| CN114629543B CN114629543B (en) | 2024-03-29 |
Family
ID=81898805
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210106549.3A Active CN114629543B (en) | 2022-01-28 | 2022-01-28 | Satellite network self-adaptive flow scheduling method based on deep supervised learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114629543B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115150335A (en) * | 2022-06-30 | 2022-10-04 | 武汉烽火技术服务有限公司 | Optimal flow segmentation method and system based on deep reinforcement learning |
| CN116708273A (en) * | 2023-07-28 | 2023-09-05 | 中国电子科技网络信息安全有限公司 | Method for generating stopping judgment classification model, network topology detection method and device |
| CN116827658A (en) * | 2023-07-17 | 2023-09-29 | 青岛启弘信息科技有限公司 | AI intelligent application security situation awareness prediction system and method |
| CN117674961A (en) * | 2023-11-20 | 2024-03-08 | 航天恒星科技有限公司 | Low orbit satellite network time delay prediction method based on space-time feature learning |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070076601A1 (en) * | 2004-04-16 | 2007-04-05 | Yufei Wang | Nodes for managing congestion and traffic flow by considering the minimization of link utilization values |
| CN109688056A (en) * | 2018-12-07 | 2019-04-26 | 南京理工大学 | Intelligent Network Control System and method |
| CN110611619A (en) * | 2019-09-12 | 2019-12-24 | 西安电子科技大学 | Intelligent routing decision method based on DDPG reinforcement learning algorithm |
| CN112187342A (en) * | 2020-09-30 | 2021-01-05 | 西安交通大学 | Satellite traffic routing method and system based on energy perception and load balancing |
| CN113472811A (en) * | 2021-08-23 | 2021-10-01 | 北京交通大学 | Heterogeneous service function chain forwarding protocol and method in intelligent fusion identification network |
-
2022
- 2022-01-28 CN CN202210106549.3A patent/CN114629543B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070076601A1 (en) * | 2004-04-16 | 2007-04-05 | Yufei Wang | Nodes for managing congestion and traffic flow by considering the minimization of link utilization values |
| CN109688056A (en) * | 2018-12-07 | 2019-04-26 | 南京理工大学 | Intelligent Network Control System and method |
| CN110611619A (en) * | 2019-09-12 | 2019-12-24 | 西安电子科技大学 | Intelligent routing decision method based on DDPG reinforcement learning algorithm |
| CN112187342A (en) * | 2020-09-30 | 2021-01-05 | 西安交通大学 | Satellite traffic routing method and system based on energy perception and load balancing |
| CN113472811A (en) * | 2021-08-23 | 2021-10-01 | 北京交通大学 | Heterogeneous service function chain forwarding protocol and method in intelligent fusion identification network |
Non-Patent Citations (6)
| Title |
|---|
| XIAONING ZHANG 等: "Local Fast Reroute With Flow Aggregation in Software Defined Networks", IEEE COMMUNICATIONS LETTERS, pages 785 - 788 * |
| 刘辰屹 等: "基于机器学习的智能路由算法综述", 计算机研究与发展, no. 04, pages 671 - 687 * |
| 李佳伟: "智慧标识网络域间流量工程机制研究", 中国博士学位论文全文数据库 信息科技辑, pages 139 - 6 * |
| 章小宁: "宽带通信网中Valiant负载平衡技术研究", 中国博士学位论文全文数据库 信息科技辑, pages 136 - 11 * |
| 蒋文娟;宗鹏;: "星座网络关键链路代价增量路由算法", 解放军理工大学学报(自然科学版), no. 03 * |
| 陈浩澜 等: "分布式星群网络组网技术研究", 通信技术, pages 1647 - 1657 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115150335A (en) * | 2022-06-30 | 2022-10-04 | 武汉烽火技术服务有限公司 | Optimal flow segmentation method and system based on deep reinforcement learning |
| CN115150335B (en) * | 2022-06-30 | 2023-10-31 | 武汉烽火技术服务有限公司 | Optimal flow segmentation method and system based on deep reinforcement learning |
| CN116827658A (en) * | 2023-07-17 | 2023-09-29 | 青岛启弘信息科技有限公司 | AI intelligent application security situation awareness prediction system and method |
| CN116827658B (en) * | 2023-07-17 | 2024-01-16 | 青岛启弘信息科技有限公司 | AI intelligent application security situation awareness prediction system and method |
| CN116708273A (en) * | 2023-07-28 | 2023-09-05 | 中国电子科技网络信息安全有限公司 | Method for generating stopping judgment classification model, network topology detection method and device |
| CN116708273B (en) * | 2023-07-28 | 2023-10-31 | 中国电子科技网络信息安全有限公司 | Method for generating stopping judgment classification model, network topology detection method and device |
| CN117674961A (en) * | 2023-11-20 | 2024-03-08 | 航天恒星科技有限公司 | Low orbit satellite network time delay prediction method based on space-time feature learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114629543B (en) | 2024-03-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114629543B (en) | Satellite network self-adaptive flow scheduling method based on deep supervised learning | |
| CN109818865B (en) | SDN enhanced path boxing device and method | |
| CN112437020B (en) | Data center network load balancing method based on deep reinforcement learning | |
| Kumar et al. | Dual reinforcement Q-routing: An on-line adaptive routing algorithm | |
| CN114143264B (en) | Flow scheduling method based on reinforcement learning under SRv network | |
| CN113328938A (en) | Network autonomous intelligent management and control method based on deep reinforcement learning | |
| CN116527567B (en) | Intelligent network path optimization method and system based on deep reinforcement learning | |
| CN109586785B (en) | Low-orbit satellite network routing strategy based on K shortest path algorithm | |
| CN114697229B (en) | Construction method and application of distributed routing planning model | |
| CN113518035B (en) | Route determining method and device | |
| Lei et al. | Congestion control in SDN-based networks via multi-task deep reinforcement learning | |
| CN105323166A (en) | Cloud computing-oriented routing method based on network efficiency priority | |
| CN113794638B (en) | SDN data center network elephant flow scheduling method based on differential evolution algorithm | |
| CN116390164A (en) | Low orbit satellite network trusted load balancing routing method, system, equipment and medium | |
| CN109246013A (en) | A kind of method for routing in FC-AE-1553 switching network | |
| CN115842768A (en) | SDN route optimization method based on time-space feature fusion of graph neural network | |
| Zeng et al. | Multi-agent reinforcement learning for adaptive routing: A hybrid method using eligibility traces | |
| Abou El Houda et al. | Cost-efficient federated reinforcement learning-based network routing for wireless networks | |
| CN112564712A (en) | Intelligent network coding method and equipment based on deep reinforcement learning | |
| CN115396366A (en) | Distributed intelligent routing method based on graph attention network | |
| CN113285832B (en) | NSGA-II-based power multi-mode network resource optimization allocation method | |
| CN117014355A (en) | TSSDN dynamic route decision method based on DDPG deep reinforcement learning algorithm | |
| Almasan et al. | Enero: Efficient real-time routing optimization | |
| An et al. | Dynamically split the traffic in software defined network based on deep reinforcement learning | |
| Wei et al. | G-Routing: Graph Neural Networks-Based Flexible Online Routing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |