CN117135693B - Real-time service distribution method based on federal learning under multi-AP environment - Google Patents
Real-time service distribution method based on federal learning under multi-AP environment Download PDFInfo
- Publication number
- CN117135693B CN117135693B CN202311408835.6A CN202311408835A CN117135693B CN 117135693 B CN117135693 B CN 117135693B CN 202311408835 A CN202311408835 A CN 202311408835A CN 117135693 B CN117135693 B CN 117135693B
- Authority
- CN
- China
- Prior art keywords
- real
- time
- time service
- delay
- period
- 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
- 238000000034 method Methods 0.000 title claims abstract description 19
- 230000006855 networking Effects 0.000 claims abstract description 19
- 230000006870 function Effects 0.000 claims abstract description 16
- 238000010801 machine learning Methods 0.000 claims abstract description 8
- 238000005457 optimization Methods 0.000 claims description 19
- 238000012790 confirmation Methods 0.000 claims description 5
- 238000004891 communication Methods 0.000 claims description 4
- 230000001934 delay Effects 0.000 claims description 3
- 238000010276 construction Methods 0.000 abstract description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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
- H04L41/16—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/02—Arrangements for optimising operational condition
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/08—Load balancing or load distribution
- H04W28/09—Management thereof
- H04W28/0958—Management thereof based on metrics or performance parameters
- H04W28/0967—Quality of Service [QoS] parameters
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Quality & Reliability (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The invention discloses a real-time service distribution method based on federal learning in a multi-AP environment, which relates to the field of wireless equipment networking, wherein local model construction is obtained by establishing a delay minimum function of a forwarding data packet and an average delay minimum function of the forwarding data packet of each real-time service in a period of time at an AP end, an optimized gradient result of a local model is obtained by combining delay data of the real-time service in a period of time under a local distribution strategy through a gradient descent machine learning method and is sent to a manager, global optimized gradient results are calculated by using federal learning, if the global optimized gradient results do not reach a preset range, the local real-time service distribution strategy of an AP is adjusted, the global optimized gradient is recalculated, and the real-time service is distributed by using the AP local real-time service distribution strategy of which the global optimized gradient results reach the preset range, so that the problem of low quality of QOS (quality) caused by unreasonable real-time service distribution is solved.
Description
Technical Field
The invention relates to the field of wireless equipment networking, in particular to a real-time service distribution method based on federal learning in a multi-AP environment.
Background
In the traditional Wifi6 scenario, although the link delay can reach millisecond level in the normal running state, when the wireless data service environment enters the crowded state, the longest system delay can still be very high, therefore, the wireless data service typified by Wifi7 is further optimized on the basis of the traditional QoS (Quality of Service) strategy: the key technologies such as multilink operation, multi-AP coordination and 320 MHz channel are introduced, so that the total transmission delay in the statistical sense is obviously reduced, but more strict QoS management is required for time delay sensitive application scenes with more strict time delay requirements, including but not limited to industrial Internet of things, AR/VR games, local emergency services and the like.
A specific QoS description model is defined under the Wi-Fi7 framework and used for identifying QoS traffic, several enhanced channel access methods are introduced on the basis of the specific QoS description model to prioritize the QoS traffic, and the methods of limiting Target Wakeup Time (TWT), triggering TXOP sharing priority access and the like are included.
Existing wireless networking (easy) implements multi-AP architecture management through a unique controller (manager).
Disclosure of Invention
The technical problems solved by the invention are as follows: the invention provides a real-time service distribution method based on federal learning in a multi-AP environment, which solves the problem of low quality of QOS caused by unreasonable real-time service distribution in the prior art.
The invention solves the technical problems by adopting the technical scheme that: the real-time service distribution method based on federal learning in the multi-AP environment is applied to a wireless networking environment for realizing multi-AP architecture management by using a unique manager, and comprises the following steps:
s1, at an AP end, aiming at real-time services, establishing a delay minimum function of a forwarding data packet and an average delay minimum function of the forwarding data packet of each real-time service in a period of time to obtain a local model;
s2, executing real-time service according to a local real-time service distribution strategy, acquiring real-time service delay data in a period of time, acquiring an optimized gradient result of a local model through a gradient descent machine learning method, and sending the optimized gradient result to a manager;
s3, receiving an optimized gradient result of each AP end local model at a manager end, and calculating a global optimized gradient result by utilizing federal learning;
and S4, if the global optimization gradient result does not reach the preset range, adjusting the local real-time service allocation strategy of the AP, and re-entering S2-S3 until the global optimization gradient result reaches the preset range, and allocating the real-time service by using the local real-time service allocation strategy of the AP at the moment.
Further, the minimum function of the delay of the forwarding data packet of the real-time service in a period of time is thatWhere Delay represents a Delay parameter, D represents a period of time, D represents a unit time in D, argmin represents a minimum Delay value of a forwarded packet per unit time D within a period of time D.
Further, the average delay minimization function of the forwarding data packet of the real-time service in a period of time is thatWherein Delay represents a Delay parameter, D represents a period of time, D represents a unit time in D, n represents a number of unit times D included in the period of time D, and argmin1 represents an average value of a sum of delays of forwarding the data packet in the period of time D.
Further, a lamda learning rate parameter is introduced in the gradient descent machine learning method in S2.
Further, in the wireless networking environment, if a new AP is added to the wireless networking environment, the configuration combination is performed under the wireless networking framework, and the election and the confirmation of a new manager are performed, and after the confirmation of the new manager in the wireless networking framework, the new manager exchanges data with multiple APs through a wired or wireless communication link.
Further, in S4, the adjusting the real-time traffic distribution policy local to the AP includes adjusting a priority of the real-time traffic at the AP.
Further, in S4, adjusting the weight of each AP, and allocating traffic to each AP according to the weight of each AP.
The invention has the beneficial effects that: the invention relates to a real-time service distribution method based on federal learning in a multi-AP environment, which comprises the steps of establishing a delay minimum function of a forwarding data packet of each real-time service in a period of time and an average delay minimum function of the forwarding data packet at an AP end to obtain a local model construction, combining delay data of the real-time service in a period of time under a local distribution strategy, obtaining an optimized gradient result of a local model through a gradient descent machine learning method, sending the optimized gradient result to a manager, receiving the optimized gradient result of the local model at the AP end at the manager end, calculating a global optimized gradient result by federal learning, adjusting the local real-time service distribution strategy of the AP if the global optimized gradient result does not reach a preset range, recalculating the global optimized gradient, and describing the QoS service quality of QoS by the global optimized gradient result, thereby converting the QoS service quality into the global optimized gradient; when the global optimization gradient result reaches a preset range, the QoS service quality reaches the standard, and the real-time service is distributed by using an AP local real-time service distribution strategy of which the global optimization gradient result reaches the preset range, so that the QoS service quality is improved by changing the real-time service distribution strategy, and the problem of low QOS quality caused by unreasonable real-time service distribution is solved.
Drawings
Fig. 1 is a flow chart of a real-time service allocation method based on federal learning in a multi-AP environment according to the present invention.
Detailed Description
The invention discloses a real-time service distribution method based on federal learning in a multi-AP environment, which aims to improve the service quality of QoS (quality of service) on the premise that a unique manager controls each AP node in a wifi wireless networking environment, and particularly as shown in a figure 1, the method comprises the following steps:
s1, at an AP end, aiming at real-time services, establishing a delay minimum function of a forwarding data packet and an average delay minimum function of the forwarding data packet of each real-time service in a period of time to obtain a local model;
specifically, the delay minimum function of the forwarding data packet of the real-time service in a period of time is thatWhere Delay represents a Delay parameter, D represents a period of time, D represents a unit time in D, argmin represents a minimum Delay value of a forwarded packet per unit time D within a period of time D.
The average delay minimization function of the forwarding data packet of the real-time service in a period of time is thatWherein Delay represents a Delay parameter, D represents a period of time, D represents a unit time in D, n represents a number of unit times D included in the period of time D, and argmin1 represents an average value of a sum of delays of forwarding the data packet in the period of time D.
S2, executing real-time service according to a local real-time service distribution strategy, acquiring real-time service delay data in a period of time, acquiring an optimized gradient result of a local model through a gradient descent machine learning method, and sending the optimized gradient result to a manager;
specifically, a lamda learning rate parameter is introduced into the gradient descent machine learning method so as to improve the convergence rate of the local model, and the optimized gradient result of the local model can represent that the delay of the real-time service is minimum and the average delay is minimum under the current real-time service allocation strategy of the AP local.
S3, receiving an optimized gradient result of each AP end local model at a manager end, and calculating a global optimized gradient result by utilizing federal learning;
specifically, the optimized gradient results of the AP end local model are aggregated by utilizing federal learning, global optimized gradient results are calculated, qoS service quality is characterized by the global optimized gradient results, namely, loss is minimized by the global optimized gradient, and therefore, the minimum delay and the minimum average delay in the global sense are achieved, and QoS service quality is guaranteed.
And S4, if the global optimization gradient result does not reach the preset range, adjusting the local real-time service allocation strategy of the AP, and re-entering S2-S3 until the global optimization gradient result reaches the preset range, and allocating the real-time service by using the local real-time service allocation strategy of the AP at the moment.
Specifically, the QoS service quality is described by using the global optimization gradient result, when the global optimization gradient result does not reach the preset range, the QoS service quality does not reach the standard, therefore, the real-time service allocation strategy of the AP local is required to be adjusted to update the delay data, so that the optimization of the global optimization gradient result is realized, until the global optimization gradient result reaches the preset range, namely, the QoS service quality reaches the standard, the AP real-time service allocation strategy is determined, the real-time service is allocated by the AP local real-time service allocation strategy at the moment, the standard QoS service quality can be obtained, and the adjustment of the AP local real-time service allocation strategy comprises the adjustment of the priority of the real-time service at the AP.
Particularly, when the global optimization gradient result is calculated through federal learning, weights can be given to the optimization gradient result of the local model of each AP end, and flow is distributed to each AP according to the weights of each AP, specifically, higher weights are given to some APs with more real-time services and large flow demand, and when the local optimization gradient of one AP is greatly different from the optimization gradient of the other APs, if the minimum delay and the average minimum delay of the one AP are smaller than the optimization gradient of the other APs by at least one order of magnitude, the one AP is given lower weight, and lower flow is distributed, so that delay data are changed by limiting the flow of the AP, and the global optimization gradient result is optimized.
In the invention, the data quantity required to be converged by federal learning is far smaller than a transmission sample, so that the communication cost of power consumption and bandwidth is reduced.
In the invention, in the wireless networking environment, if a new AP is added into the wireless networking environment, the framework combination is carried out under the wireless networking framework, the election and the confirmation of a new manager are carried out, after the new manager in the wireless networking framework is confirmed, the new manager exchanges data with multiple APs through a wired or wireless communication link, so that the steps S1-S4 are re-executed, and the real-time service is distributed by the AP local real-time service distribution strategy with the global optimal gradient result reaching the preset range. The basis for election and validation of the new manager is the same as the basis for selecting the unique manager for wireless networking.
Claims (7)
1. The real-time service distribution method based on federal learning in the multi-AP environment is applied to a wireless networking environment for realizing multi-AP architecture management by using a unique manager and is characterized by comprising the following steps:
s1, at an AP end, aiming at real-time services, establishing a delay minimum function of a forwarding data packet and an average delay minimum function of the forwarding data packet of each real-time service in a period of time to obtain a local model;
s2, executing real-time service according to a local real-time service distribution strategy, acquiring real-time service delay data in a period of time, acquiring an optimized gradient result of a local model through a gradient descent machine learning method, and sending the optimized gradient result to a manager;
s3, receiving an optimized gradient result of each AP end local model at a manager end, and calculating a global optimized gradient result by utilizing federal learning;
and S4, if the global optimization gradient result does not reach the preset range, adjusting the local real-time service allocation strategy of the AP, and re-entering S2-S3 until the global optimization gradient result reaches the preset range, and allocating the real-time service by using the local real-time service allocation strategy of the AP at the moment.
2. The method for distributing real-time traffic based on federal learning in a multi-AP environment according to claim 1, wherein a delay minimum function of a forwarding packet of real-time traffic over a period of time isWhere Delay represents a Delay parameter, D represents a period of time, D represents a unit time in D, argmin represents a minimum Delay value of a forwarded packet per unit time D within a period of time D.
3. The method for distributing real-time traffic based on federal learning in a multi-AP environment according to claim 1, wherein the average delay minimization function of the forwarding packets of the real-time traffic over a period of time isWherein Delay represents a Delay parameter, D represents a period of time, D represents a unit time in D, n represents a number of unit times D included in the period of time D, and argmin1 represents an average value of a sum of delays of forwarding the data packet in the period of time D.
4. The real-time traffic distribution method based on federal learning in a multi-AP environment according to claim 1, wherein a lamda learning rate parameter is introduced in the gradient descent machine learning method in S2.
5. The method for real-time traffic distribution based on federal learning in a multi-AP environment according to claim 1, wherein in a wireless networking environment, if a new AP is added to the wireless networking environment, the configuration is combined under a wireless networking framework, and the election and confirmation of a new manager are performed, and after the confirmation of the new manager in the wireless networking framework, the new manager exchanges data with the multi-AP through a wired or wireless communication link.
6. The method for real-time traffic distribution based on federal learning in a multi-AP environment according to claim 1, wherein in S4, said adjusting the real-time traffic distribution policy local to the AP comprises adjusting the priority of real-time traffic at the AP.
7. The method for real-time traffic distribution based on federal learning in a multi-AP environment according to claim 1, further comprising assigning weights to the optimized gradient results of the local model at each AP end when calculating the global optimized gradient results using federal learning, and distributing traffic to each AP according to the weights of each AP.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311408835.6A CN117135693B (en) | 2023-10-27 | 2023-10-27 | Real-time service distribution method based on federal learning under multi-AP environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311408835.6A CN117135693B (en) | 2023-10-27 | 2023-10-27 | Real-time service distribution method based on federal learning under multi-AP environment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117135693A CN117135693A (en) | 2023-11-28 |
| CN117135693B true CN117135693B (en) | 2024-01-23 |
Family
ID=88860457
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311408835.6A Active CN117135693B (en) | 2023-10-27 | 2023-10-27 | Real-time service distribution method based on federal learning under multi-AP environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117135693B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105636062A (en) * | 2016-01-25 | 2016-06-01 | 长江大学 | Cognitive radio network transmission learning method for moderate business services |
| CN111432270A (en) * | 2020-03-09 | 2020-07-17 | 重庆邮电大学 | Real-time service delay optimization method based on layered cache |
| CN112637883A (en) * | 2020-12-09 | 2021-04-09 | 深圳智芯微电子科技有限公司 | Federal learning method with robustness to wireless environment change in power Internet of things |
| CN114040425A (en) * | 2021-11-17 | 2022-02-11 | 中国电信集团系统集成有限责任公司 | Resource allocation method based on global resource availability optimization |
| CN115208604A (en) * | 2022-02-22 | 2022-10-18 | 长沙理工大学 | Method, device and medium for detecting AMI network intrusion |
| CN116582871A (en) * | 2023-07-07 | 2023-08-11 | 南京邮电大学 | Unmanned aerial vehicle cluster federal learning model optimization method based on topology optimization |
| CN116644820A (en) * | 2023-05-29 | 2023-08-25 | 中国重型机械研究院股份公司 | Knowledge collaboration method in complex heavy equipment customization production process |
| CN116862551A (en) * | 2023-06-15 | 2023-10-10 | 国网湖北省电力有限公司电力科学研究院 | New energy consumption price decision method considering user privacy protection |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20220141844A1 (en) * | 2020-10-29 | 2022-05-05 | Charter Communications Operating, Llc | Methods and apparatus for data traffic prioritization in small-cell wireless systems |
| US20230051166A1 (en) * | 2021-08-12 | 2023-02-16 | Wisys Technology Foundation, Inc. | Delay Sensitive Network Estimation System |
-
2023
- 2023-10-27 CN CN202311408835.6A patent/CN117135693B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105636062A (en) * | 2016-01-25 | 2016-06-01 | 长江大学 | Cognitive radio network transmission learning method for moderate business services |
| CN111432270A (en) * | 2020-03-09 | 2020-07-17 | 重庆邮电大学 | Real-time service delay optimization method based on layered cache |
| CN112637883A (en) * | 2020-12-09 | 2021-04-09 | 深圳智芯微电子科技有限公司 | Federal learning method with robustness to wireless environment change in power Internet of things |
| CN114040425A (en) * | 2021-11-17 | 2022-02-11 | 中国电信集团系统集成有限责任公司 | Resource allocation method based on global resource availability optimization |
| CN115208604A (en) * | 2022-02-22 | 2022-10-18 | 长沙理工大学 | Method, device and medium for detecting AMI network intrusion |
| CN116644820A (en) * | 2023-05-29 | 2023-08-25 | 中国重型机械研究院股份公司 | Knowledge collaboration method in complex heavy equipment customization production process |
| CN116862551A (en) * | 2023-06-15 | 2023-10-10 | 国网湖北省电力有限公司电力科学研究院 | New energy consumption price decision method considering user privacy protection |
| CN116582871A (en) * | 2023-07-07 | 2023-08-11 | 南京邮电大学 | Unmanned aerial vehicle cluster federal learning model optimization method based on topology optimization |
Non-Patent Citations (2)
| Title |
|---|
| "Blockchain-Enabled Clustered Federated Learning in Fog Computing Networks";Xiaoge Huang et al.;《IEEE》;全文 * |
| "无线传感器网络中高效能数据收集算法的研究";陈战胜;《中国博士学位论文电子期刊网》;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117135693A (en) | 2023-11-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100571175C (en) | A kind of cordless communication network bandwidth allocation methods and device | |
| Zhai et al. | Providing statistical QoS guarantee for voice over IP in the IEEE 802.11 wireless LANs | |
| CN101827027B (en) | Interlayer coordination-based home network QoS guarantee method | |
| CN103596224B (en) | Resource regulating method based on multistage-mapping under a kind of high-speed mobile environment | |
| CN109639470B (en) | VSAT satellite communication system bandwidth allocation method based on star networking | |
| CN103067985A (en) | Super-speed wireless local area network channel binding and allocation method based on channel quality | |
| Freitag et al. | Tuning of 802.11 e network parameters | |
| CN109905330B (en) | Dynamic weighted fair queue train network scheduling method based on queue length | |
| Lopez-Aguilera et al. | An asymmetric access point for solving the unfairness problem in WLANs | |
| CN103209494A (en) | Real-time video business resource distribution method based on importance mark | |
| CN104093009A (en) | Video transmission method in wireless self-organized network on basis of network utility | |
| CN114079977A (en) | 5G and TSN fusion network flow scheduling framework and resource allocation method | |
| Wan et al. | Adaptive scheduling for wireless video transmission in high-speed networks | |
| CN107659940B (en) | Multi-priority multi-channel media access control method based on channel perception | |
| CN117135693B (en) | Real-time service distribution method based on federal learning under multi-AP environment | |
| Zeng et al. | A novel scheduling scheme and MAC enhancements for IEEE 802.15. 3 high-rate WPAN | |
| CN101883400A (en) | Data transmission method and access control equipment in wireless local area network | |
| Borst et al. | Optimal path selection in multi-RAT wireless networks | |
| Zhou et al. | A cross-layer architecture for DiffServ in mobile ad-hoc networks | |
| CN1784872A (en) | WLAN scheduler | |
| Zhao et al. | Distributed MAC adaptation for WLAN QoS differentiation | |
| Kim et al. | Distributed access time control for per-station fairness in infrastructure WLANs | |
| Wang et al. | Improving the data scheduling efficiency of the ieee 802.16 (d) mesh network | |
| CN110290590B (en) | Distribution method of LTE system MAC layer resource block with improved weight | |
| CN113766514B (en) | Networking resource distribution method and device, electronic equipment and storage medium |
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 |