CN114640568B - Network intelligent management and control architecture system based on deep reinforcement learning and operation method - Google Patents
Network intelligent management and control architecture system based on deep reinforcement learning and operation method Download PDFInfo
- Publication number
- CN114640568B CN114640568B CN202210536555.2A CN202210536555A CN114640568B CN 114640568 B CN114640568 B CN 114640568B CN 202210536555 A CN202210536555 A CN 202210536555A CN 114640568 B CN114640568 B CN 114640568B
- Authority
- CN
- China
- Prior art keywords
- network
- module
- data
- state information
- plane
- 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.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/04—Network management architectures or arrangements
- H04L41/042—Network management architectures or arrangements comprising distributed management centres cooperatively managing the network
-
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/08—Configuration management of networks or network elements
- H04L41/0803—Configuration setting
Abstract
The invention discloses a network intelligent management and control architecture system and an operation method based on deep reinforcement learning, which are applied to management and control in a network. The architecture system is composed of a data plane, a control plane and a management plane. The operation method comprises the following steps: the data plane detects various data in the network through a network telemetry technology, the control plane receives data uploaded by the data plane, online decision is made through a deep reinforcement learning technology, a configuration instruction is issued, and the data plane receives the instruction to process equipment in the network. And the management plane learns according to the network state data uploaded by each distributed control plane and shares the knowledge to each distributed control plane. The invention is based on the deep reinforcement learning technology, can realize intelligent management and control in the network, and effectively improves the utilization rate of resources in the network.
Description
Technical Field
The invention relates to a network intelligent management and control architecture system based on deep reinforcement learning and an operation method, and belongs to the technical field of management and control in a network.
Background
With the advent of the internet of everything age, various internet of things devices have exploded, and various applications such as VR, remote operations, car networking and the like have emerged, so that people have higher requirements on network performance such as low delay and high bandwidth, and the control and management of the network are related to whether the network can provide expected service quality for users.
Most of the current network control and management architectures are based on a terminal host or a centralized control framework, and the framework depends on a manual process, generates excessive communication and calculation overhead, cannot timely correspond to the dynamic change of a network, and has poor expansibility and robustness. A centralized control framework requires the collection and analysis of large amounts of network data even in response to a single network event, and therefore does not react in real-time in response to network dynamics. The existing network management and control scheme depends on a manual configuration flow of a network manager to a great extent, network operation and maintenance personnel need to carefully analyze network behaviors and design a corresponding control strategy (needing a few weeks at least), the current network becomes more and more complex and flexible, and the manually configured network has poor expandability and robustness and cannot meet the requirements of the current network. There is a need for a new network architecture to meet the differentiation requirements of the current networks.
In recent years, a deep reinforcement learning technique, which is one of artificial intelligence techniques, has been rapidly developed and widely used in the fields of natural language processing, image recognition, game strategy calculation, and the like. The strategies that the deep reinforcement learning model can learn are more and more complex, and the training and executing efficiency is higher and higher. Meanwhile, the development of programmable network hardware enables a deep reinforcement learning algorithm to be deployed in the network to analyze network data, and flexible processing can be executed in the network. At present, most of applications of a deep reinforcement learning technology in a network focus on aspects such as traffic classification and traffic prediction, and the deep reinforcement learning technology is not applied to management and control of the network.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the network intelligent management and control architecture system and the operation method based on the deep reinforcement learning are provided, the deep reinforcement learning is applied to network control and management, the adaptive capacity of the deep reinforcement learning is fully utilized, the network dynamic change can be sensitively sensed and timely responded, and the network strategy is continuously optimized through learning, so that the network management and control capacity is improved, and the user requirements are met.
The invention adopts the following technical scheme for solving the technical problems:
a network intelligent management and control architecture system based on deep reinforcement learning comprises a management plane, a plurality of distributed control planes and data planes which correspond to the control planes one by one; a bidirectional interface exists between the distributed control plane and the data plane corresponding to the distributed control plane, so that the data plane uploads data to the distributed control plane and the distributed control plane issues a configuration instruction to the data plane; a bidirectional interface exists between each distributed control plane and the management plane, so that the distributed control planes upload data to the management plane and the management plane shares knowledge to the distributed control planes;
the data plane comprises a network telemetry module, a data uploading module and a strategy execution module; the distributed control plane comprises a data platform and a controller, wherein the data platform comprises a data receiving module, a data storage module and a data preprocessing module, and the controller comprises an intelligent algorithm module, an online decision module and a decision issuing module; the management plane comprises a network data storage module, an intelligent algorithm training module and a network knowledge sharing module;
the network telemetry module is used for collecting network state information and transmitting the collected network state information to the data uploading module;
the data uploading module is used for uploading the network state information transmitted by the network telemetry module to a distributed control plane corresponding to the data plane where the data uploading module is located;
the data receiving module is used for receiving the network state information uploaded by the data uploading module;
the data storage module is used for storing the network state information received by the data receiving module;
the data preprocessing module is used for carrying out data cleaning on the network state information stored by the data storage module to obtain preprocessed network state information, transmitting the preprocessed network state information to the intelligent algorithm module and transmitting the preprocessed network state information to the network data storage module;
the intelligent algorithm module is used for loading a deep reinforcement learning model on line after receiving the preprocessed network state information transmitted by the data preprocessing module, starting an on-line decision module to make a corresponding configuration instruction according to the preprocessed network state information, and transmitting the configuration instruction to the decision issuing module by the on-line decision module;
the decision issuing module is used for issuing the configuration instruction transmitted by the online decision module to the decision executing module;
the decision execution module is used for processing the data packet transmitted in the network according to the configuration instruction issued by the decision issuing module;
the network data storage module is used for receiving and storing the preprocessed network state information transmitted by the data preprocessing module, and calling the intelligent algorithm training module to perform offline deep reinforcement learning algorithm training on the preprocessed network state information and generate a new deep reinforcement learning model;
and the network knowledge sharing module is used for deploying the new deep reinforcement learning model generated by the intelligent algorithm training module in the intelligent algorithm modules of the distributed control planes.
As a preferred aspect of the invention, the data plane is provided with capability support by a programmable network switch.
As a preferred solution of the present invention, the network state information collected by the network telemetry module includes network bandwidth, link utilization, micro-burst traffic, link congestion, and forwarding path.
As a preferred aspect of the present invention, the network telemetry module gathers network status information using network telemetry.
An operation method of the network intelligent management and control architecture system based on deep reinforcement learning specifically includes:
step 1, a network telemetry module in a data plane is used for collecting network state information, and the network state information is uploaded to a data platform in a distributed control plane corresponding to the data plane through a data uploading module;
step 2, the data platform receives the network state information and stores and preprocesses the network state information to obtain preprocessed network state information;
step 3, the data platform transmits the preprocessed network state information to the controller and simultaneously transmits the preprocessed network state information to a network data storage module in the management plane;
step 4, after the controller receives the preprocessed network state information, the intelligent algorithm module is used for loading the deep reinforcement learning model on line, the online decision module is started to make a corresponding configuration instruction according to the preprocessed network state information, and the configuration instruction is issued to the decision execution module in the data plane through the decision issuing module;
and 5, a network data storage module in the management plane receives and stores the preprocessed network state information, then an intelligent algorithm training module is called to perform offline deep reinforcement learning model training on the preprocessed network state information and generate a new deep reinforcement learning model, and meanwhile, a network knowledge sharing module is called to deploy the new deep reinforcement learning model on each distributed control plane to realize network control in a global coordination mode.
Compared with the prior art, the invention adopting the technical scheme has the following technical effects:
1. the invention introduces a network data telemetry technology to sense the change of the network bottom state in real time, processes and learns the network data through a deep reinforcement learning algorithm, and optimizes the management and control strategy in real time.
2. The invention realizes self-learning management and control, the self-adaptive capacity of the deep reinforcement learning can realize the analysis of network data and the design of a corresponding control strategy, and the learning capacity of the deep reinforcement learning model is also improved along with the accumulation of network data volume.
3. The data plane in the invention adopts programmable network hardware, and can dynamically reconfigure the network hardware through P4 and other high-level programming languages, thereby realizing flexible processing logic execution in the network.
Drawings
FIG. 1 is a general architecture diagram of a network intelligent management and control architecture system based on deep reinforcement learning according to the present invention;
FIG. 2 is a data plane architecture and workflow diagram of the present invention;
FIG. 3 is a control plane architecture and workflow diagram of the present invention;
FIG. 4 is a management plane architecture and a work flow diagram of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are illustrative only for the purpose of explaining the present invention, and are not to be construed as limiting the present invention.
In order to make the embodiments of the present invention better understood by those skilled in the art, techniques and terms involved in the present invention will be described below.
Deep reinforcement learning technology:
deep Reinforcement Learning (DRL) is an artificial intelligence method closer to human thinking, and can be controlled directly according to input information by combining the perception capability of Deep Learning and the decision-making capability of Reinforcement Learning. As the integration of deep learning and reinforcement learning, the deep reinforcement learning is more suitable for the dynamic environment of the network, and the strategy can be learned and optimized directly from the network data without following the preset rules.
Network telemetry:
the network remote measurement refers to the process of automatically and remotely collecting network multi-source heterogeneous state information, and storing, analyzing and using network measurement data. Upper layer telemetry data applications include event correlation analysis, historical data tracing, anomaly detection, performance monitoring, trend analysis, and the like. Compared with the traditional network measurement and software defined measurement, the network telemetry further plays a role of a data plane in the network measurement process, and changes a network measurement mode from a pull mode to a push mode. The network equipment actively pushes network state information remotely to realize a high-speed real-time network data acquisition function.
The data plane and the control plane may constitute a closed-loop control of the "measurement-learning-decision-action" process, facilitating autonomous control of the local area network. However, the network data perceived by each distributed data plane only contains local observation information of the network state, so the control plane only optimizes the control strategy in the local area, and therefore the mode can only adapt to the network dynamics locally and quickly, and the network as a distributed system can achieve the best performance by the cooperation of all nodes.
Aiming at the local problems brought by the distributed network system, the invention particularly provides a method for receiving network information and feedback information thereof uploaded by each distributed control plane by using a management plane, carrying out off-line training and modifying a deep reinforcement learning model, issuing the modified deep reinforcement learning model to the control plane for on-line decision, simplifying the training process of the control plane and realizing network control in a global coordination mode.
As shown in fig. 1, the present invention provides a network intelligent management and control architecture system based on deep reinforcement learning, which can be divided into three parts: the data plane, the control plane and the management plane are provided with bidirectional interfaces to realize data uploading and instruction issuing.
The data plane is supported by the programmable network switch, comprises a network telemetry module, a data uploading module and a strategy execution module, is responsible for measuring various network data, uploads the network data to the control plane, and carries out corresponding processing operation according to a configuration instruction issued by the control plane.
The network telemetry module utilizes network telemetry to collect network state information, and the programmable network switch processes data packets in the network according to configuration instructions.
The control plane consists of a data platform and a controller, learns the behavior of the network, and automatically generates a corresponding control strategy to provide computing power for the training process.
The data platform is responsible for data storage and data cleaning, the controller is responsible for loading a deep reinforcement learning algorithm on line, continuously learns network data, optimizes a control strategy and then feeds back an updated configuration instruction to the data plane.
The management plane is a centralized plane and is responsible for continuously collecting network information, learning global knowledge, sharing the knowledge to each distributed control plane, and modifying the learning process of the control strategy, so that the network performance is improved on the global level.
Based on the architecture system, the invention also provides an operation method of the network intelligent control architecture system based on deep reinforcement learning, and the operation mechanism is as follows:
step A: the network telemetry module collects network data and uploads the network data to a data platform in a control plane, wherein the network data comprises network running state information such as micro burst flow, link congestion and forwarding paths;
and B: the data platform receives and processes the network data, including data storage and data cleaning, and mass network data are converted into valuable and utilizable network state information after being processed;
and C: the data platform transmits the network state information to the controller; the data platform uploads the network state information to a management plane;
step D: the controller loads a depth strengthening algorithm, makes a corresponding configuration instruction aiming at the network state information and issues the configuration instruction to the data plane;
step E: the programmable network switch receives the configuration instruction and processes the transmission data packet in the network;
step F: and the management plane receives the global network state information and loads a deep reinforcement learning algorithm for training.
As shown in fig. 2, the data plane of the present invention includes a network telemetry module, a data upload module, and a policy enforcement module. The network telemetry module collects network state information (link congestion, flow burst, bandwidth and the like) by using a network telemetry technology, uploads the information to a control plane through the data upload module, and uploads the information to a management plane through the control plane; the policy enforcement module receives configuration instructions from the control plane.
As shown in fig. 3, the control plane includes two parts, namely a data platform and a controller, wherein the data platform includes a data receiving module, a data storage module and a data preprocessing module, the data receiving module receives network state information uploaded by the data plane and transmits the network state information to the data storage module for data storage, further, the network state information is transmitted to the data preprocessing module through the data storage module for data cleaning, and the data platform finally transmits the processed network state information to the controller; the controller comprises an intelligent algorithm module, an online decision module and a decision issuing module, when the controller receives network state information transmitted by the data platform, the intelligent algorithm module can load a deep reinforcement learning algorithm online, starts the online decision module to make a corresponding configuration instruction aiming at the network information, and finally issues the configuration instruction to the data plane through the decision issuing module.
As shown in fig. 4, the management plane includes a network data storage module, an intelligent algorithm training module, and a network knowledge sharing module. The data storage module receives and stores the network state information, then the intelligent algorithm training module is called to perform offline deep reinforcement learning algorithm training aiming at the network state information to generate a new deep reinforcement learning model, and meanwhile, the network knowledge sharing module is called to deploy the new algorithm model on each distributed control plane to realize network management and control in a global coordination mode.
And the physical terminal generates a calculation task and sends a task calculation request. A network switch (positioned in a data plane) for managing the physical terminal receives the task calculation request, calls a network telemetry module to collect network state information (network bandwidth, link utilization rate, micro-burst flow, link congestion and forwarding path) by utilizing a network telemetry technology, uploads the information to a control plane through a data upload module and uploads the information to a management plane through the control plane; a data receiving module in the control plane receives network state information uploaded by the data plane and transmits the network state information to a data storage module for data storage, further, the network state information is transmitted to a data preprocessing module through the data storage module for data cleaning, and the processed network state information is transmitted to a controller through the data platform; when the controller receives network state information transmitted by the data platform, the intelligent algorithm module can load a deep reinforcement learning algorithm on line, and starts an on-line decision module to make a corresponding configuration instruction aiming at the network state information, namely, the current network state information (network bandwidth, link utilization rate, micro-burst flow, link congestion and forwarding path) is used for selecting routing nodes, and which routing nodes participate in transmitting the calculation task; and finally, issuing the configuration command to the data plane through a decision issuing module. And the network switch corresponding to the data plane receives the configuration instruction and informs the routing nodes participating in the transmission of the calculation task according to the instruction information, meanwhile, the network switch responds to the task calculation request sent by the physical terminal, the physical terminal uploads the calculation task to the network in the form of a data packet, and the data packet is finally transmitted in the network under the help of the selected routing node.
Different deep reinforcement learning algorithms are stored in each distributed control plane, some algorithms can perform routing decision taking the shortest time delay as a target according to current network state information, some algorithms can perform routing decision taking the optimal link utilization rate as a target according to the current network state information, which means that each distributed control plane can collect different network state information and upload the network state information to a management plane, a data storage module receives and stores the network state information, then an intelligent algorithm training module is called to perform offline deep reinforcement learning algorithm training aiming at the network state information and generate a new deep reinforcement learning model, and further, a network knowledge sharing module is called to deploy the new algorithm model on each distributed control plane to realize network control in a global coordination mode.
The above embodiments are only for illustrating the technical idea of the present invention, and the technical idea of the present invention is not limited thereto, and any modifications made on the basis of the technical solution according to the technical idea of the present invention fall within the protective scope of the present invention.
Claims (5)
1. A network intelligent management and control architecture system based on deep reinforcement learning is characterized by comprising a management plane, a plurality of distributed control planes and data planes which correspond to the control planes one by one; a bidirectional interface exists between the distributed control plane and the data plane corresponding to the distributed control plane, so that the data plane uploads data to the distributed control plane and the distributed control plane issues a configuration instruction to the data plane; a bidirectional interface exists between each distributed control plane and the management plane, so that the distributed control planes upload data to the management plane and the management plane shares knowledge to the distributed control planes;
the data plane comprises a network telemetry module, a data uploading module and a decision execution module; the distributed control plane comprises a data platform and a controller, wherein the data platform comprises a data receiving module, a data storage module and a data preprocessing module, and the controller comprises an intelligent algorithm module, an online decision module and a decision issuing module; the management plane comprises a network data storage module, an intelligent algorithm training module and a network knowledge sharing module;
the network telemetry module is used for collecting network state information and transmitting the collected network state information to the data uploading module;
the data uploading module is used for uploading the network state information transmitted by the network telemetry module to a distributed control plane corresponding to the data plane where the data uploading module is located;
the data receiving module is used for receiving the network state information uploaded by the data uploading module;
the data storage module is used for storing the network state information received by the data receiving module;
the data preprocessing module is used for carrying out data cleaning on the network state information stored by the data storage module to obtain preprocessed network state information, transmitting the preprocessed network state information to the intelligent algorithm module and simultaneously transmitting the preprocessed network state information to the network data storage module;
the intelligent algorithm module is used for loading a deep reinforcement learning model on line after receiving the preprocessed network state information transmitted by the data preprocessing module, starting an on-line decision module to make a corresponding configuration instruction according to the preprocessed network state information, and transmitting the configuration instruction to the decision issuing module by the on-line decision module;
the decision issuing module is used for issuing the configuration instruction transmitted by the online decision module to the decision executing module;
the decision execution module is used for processing the data packet transmitted in the network according to the configuration instruction issued by the decision issuing module;
the network data storage module is used for receiving and storing the preprocessed network state information transmitted by the data preprocessing module, and calling the intelligent algorithm training module to perform offline deep reinforcement learning algorithm training on the preprocessed network state information and generate a new deep reinforcement learning model;
and the network knowledge sharing module is used for deploying the new deep reinforcement learning model generated by the intelligent algorithm training module in the intelligent algorithm modules of the distributed control planes.
2. The deep reinforcement learning-based network intelligence management and control architecture system of claim 1, wherein the data plane is supported by capabilities provided by a programmable network switch.
3. The deep reinforcement learning-based network intelligent management and control architecture system according to claim 1, wherein the network state information collected by the network telemetry module includes network bandwidth, link utilization, micro-burst traffic, link congestion, and forwarding paths.
4. The deep reinforcement learning-based network intelligent management and control architecture system according to claim 1, wherein the network telemetry module gathers network status information using network telemetry.
5. An operation method of the deep reinforcement learning-based network intelligent management and control architecture system according to any one of claims 1 to 4, wherein the operation method specifically comprises the following steps:
step 1, a network telemetry module in a data plane is used for collecting network state information, and the network state information is uploaded to a data platform in a distributed control plane corresponding to the data plane through a data uploading module;
step 2, the data platform receives the network state information and stores and preprocesses the network state information to obtain preprocessed network state information;
step 3, the data platform transmits the preprocessed network state information to the controller and simultaneously transmits the preprocessed network state information to a network data storage module in the management plane;
step 4, after receiving the preprocessed network state information, the controller online loads a deep reinforcement learning model by using an intelligent algorithm module, starts an online decision module to make a corresponding configuration instruction aiming at the preprocessed network state information, and issues the configuration instruction to a decision execution module in a data plane through a decision issuing module;
and 5, a network data storage module in the management plane receives and stores the preprocessed network state information, then an intelligent algorithm training module is called to perform offline deep reinforcement learning model training on the preprocessed network state information and generate a new deep reinforcement learning model, and meanwhile, a network knowledge sharing module is called to deploy the new deep reinforcement learning model on each distributed control plane to realize network control in a global coordination mode.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210536555.2A CN114640568B (en) | 2022-05-18 | 2022-05-18 | Network intelligent management and control architecture system based on deep reinforcement learning and operation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210536555.2A CN114640568B (en) | 2022-05-18 | 2022-05-18 | Network intelligent management and control architecture system based on deep reinforcement learning and operation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114640568A CN114640568A (en) | 2022-06-17 |
| CN114640568B true CN114640568B (en) | 2022-08-19 |
Family
ID=81953033
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210536555.2A Active CN114640568B (en) | 2022-05-18 | 2022-05-18 | Network intelligent management and control architecture system based on deep reinforcement learning and operation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114640568B (en) |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200225655A1 (en) * | 2016-05-09 | 2020-07-16 | Strong Force Iot Portfolio 2016, Llc | Methods, systems, kits and apparatuses for monitoring and managing industrial settings in an industrial internet of things data collection environment |
| CN109039942B (en) * | 2018-08-29 | 2022-08-12 | 南京优速网络科技有限公司 | Network load balancing system and balancing method based on deep reinforcement learning |
| CN111245688A (en) * | 2019-12-26 | 2020-06-05 | 的卢技术有限公司 | Method and system for intelligently controlling electrical equipment based on indoor environment |
| CN111564849B (en) * | 2020-05-15 | 2021-11-02 | 清华大学 | Two-stage deep reinforcement learning-based power grid reactive voltage control method |
-
2022
- 2022-05-18 CN CN202210536555.2A patent/CN114640568B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN114640568A (en) | 2022-06-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Lei et al. | Toward intelligent cooperation of UAV swarms: When machine learning meets digital twin | |
| CN108259367B (en) | Service-aware flow strategy customization method based on software defined network | |
| CN113395679B (en) | Resource and task allocation optimization system of unmanned aerial vehicle edge server | |
| CN109450790B (en) | Intelligent network service function chain supply method based on flow perception | |
| Wei et al. | Reinforcement learning-empowered mobile edge computing for 6G edge intelligence | |
| CN103177298A (en) | Smart energy network control method | |
| WO2022001119A1 (en) | Multi-agent system structure and control method therefor | |
| CN113341712B (en) | Intelligent hierarchical control selection method for unmanned aerial vehicle autonomous control system | |
| US20090228407A1 (en) | Distributed cognitive architecture | |
| CN116743635B (en) | Network prediction and regulation method and network regulation system | |
| CN115314355A (en) | Electric power communication network architecture system and method based on deterministic network | |
| CN114710437B (en) | Internet of things edge network routing system combining block chains | |
| Dai et al. | Routing optimization meets Machine Intelligence: A perspective for the future network | |
| CN114827284B (en) | Service function chain arrangement method and device in industrial Internet of things and federal learning system | |
| CN110806758B (en) | Unmanned aerial vehicle cluster autonomous level self-adaptive adjustment method based on scene fuzzy cognitive map | |
| Yao et al. | Power control in Internet of Drones by deep reinforcement learning | |
| Huang et al. | Machine learning for broad-sensed internet congestion control and avoidance: A comprehensive survey | |
| CN110601916A (en) | Flow sampling and application sensing system based on machine learning | |
| Zhang et al. | DSOQR: Deep reinforcement learning for online QoS routing in SDN-based networks | |
| CN114640568B (en) | Network intelligent management and control architecture system based on deep reinforcement learning and operation method | |
| CN116962409A (en) | Industrial Internet network architecture based on intelligent fusion of general sense calculation and virtual controller | |
| CN112260953A (en) | Multi-channel data forwarding decision method based on reinforcement learning | |
| CN116684472A (en) | Service deployment system and service deployment method for terminal-side computing network | |
| CN116436980A (en) | Real-time video task end network edge cooperative scheduling method and device | |
| CN115580885A (en) | Intelligent decision method and framework of unmanned aerial vehicle communication system |
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 |