WO2022198634A1 - 一种部分重叠信道频谱共享的深度学习方法及系统 - Google Patents

一种部分重叠信道频谱共享的深度学习方法及系统 Download PDF

Info

Publication number
WO2022198634A1
WO2022198634A1 PCT/CN2021/083271 CN2021083271W WO2022198634A1 WO 2022198634 A1 WO2022198634 A1 WO 2022198634A1 CN 2021083271 W CN2021083271 W CN 2021083271W WO 2022198634 A1 WO2022198634 A1 WO 2022198634A1
Authority
WO
WIPO (PCT)
Prior art keywords
channel
channel state
state information
reinforcement learning
user equipment
Prior art date
Application number
PCT/CN2021/083271
Other languages
English (en)
French (fr)
Inventor
王璐
黄瑞锋
伍楷舜
Original Assignee
深圳大学
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 深圳大学 filed Critical 深圳大学
Priority to PCT/CN2021/083271 priority Critical patent/WO2022198634A1/zh
Priority to US18/281,418 priority patent/US20240121136A1/en
Publication of WO2022198634A1 publication Critical patent/WO2022198634A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00—Baseband systems
    • H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
    • H04L25/0202—Channel estimation
    • H04L25/024—Channel estimation channel estimation algorithms
    • H04L25/0254—Channel estimation channel estimation algorithms using neural network algorithms
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
    • H04W16/14—Spectrum sharing arrangements between different networks
    • 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
    • G06N3/092—Reinforcement learning
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/04—Arrangements for maintaining 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
    • H04W28/082—Load balancing or load distribution among bearers or channels
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L5/00—Arrangements affording multiple use of the transmission path
    • H04L5/003—Arrangements for allocating sub-channels of the transmission path
    • H04L5/0053—Allocation of signaling, i.e. of overhead other than pilot signals
    • H04L5/0057—Physical resource allocation for CQI
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
    • H04W16/18—Network planning tools
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/08—Testing, supervising or monitoring using real traffic

Definitions

  • the present invention relates to the field of communication technologies, and more particularly, to a deep learning method and system for spectral sharing of partially overlapping channels.
  • End devices such as tablets, smartphones, and heterogeneous Internet of Things (IoT) devices
  • IoT Internet of Things
  • DSM dynamic spectrum management
  • AI artificial intelligence
  • the purpose of the present invention is to overcome the above-mentioned defects of the prior art, and to provide a deep learning method and system for spectrum sharing of partially overlapping channels.
  • a deep learning method for spectral sharing of partially overlapping channels includes the following steps:
  • the base station In response to the received user transmission request, the base station inputs the channel state information CSI of multiple historical time slices into the trained channel prediction convolutional neural network model, and outputs the predicted channel state information CSI of the next time slice;
  • the channel state information CSI of the next time slice is input into the reinforcement learning model, and the channel allocation strategy of each user equipment in the collision domain of the base station is obtained, so as to realize the maximum throughput of simultaneous communication of each user equipment.
  • the reinforcement learning model is based on bandwidth efficiency. Performance is trained as a reward.
  • a deep learning system for spectral sharing of partially overlapping channels includes:
  • Channel state prediction unit used to input the channel state information CSI of multiple historical time slices into the trained channel prediction convolutional neural network model in response to the received user transmission request, and output the predicted channel state of the next time slice Information CSI;
  • Channel allocation unit used to input the channel state information CSI of the next time slice into the reinforcement learning model, and obtain the channel allocation strategy of each user equipment in the collision domain of the base station, so as to realize the maximum throughput of simultaneous communication of each user equipment, the Reinforcement learning models are trained with bandwidth efficiency performance as a reward.
  • the present invention has the advantage that, combined with deep learning, a concurrent spectrum access system architecture based on partially overlapping channels is proposed, which automatically learns additional coding redundancy from data on non-overlapping spectrums, And apply redundancy to data recovery on overlapping spectrums.
  • Fig. 1 is the execution flow chart of the communication system according to one embodiment of the present invention.
  • FIG. 2 is a schematic diagram of a communication system framework according to an embodiment of the present invention.
  • FIG. 3 is a flowchart of a deep learning method for spectral sharing of partially overlapping channels according to an embodiment of the present invention
  • FIG. 4 is a schematic diagram of a channel interleaving execution process according to an embodiment of the present invention.
  • FIG. 5 is a schematic diagram of a reinforcement learning model according to an embodiment of the present invention.
  • Reward-reward State-state; Environment-environment; Action-action; Agent-agent; Cloud-cloud.
  • the invention is based on the deep learning method of exploring partial overlapping channels based on 5G smart city spectrum sharing, innovatively uses the channel interleaving technology to realize overlapping channel sharing, uses the channel state historical time slice sequence to train the convolutional neural network, and predicts the channel state of the next time slice. Then, the predicted channel state information of the next time slice is used as the input of deep reinforcement learning to learn and predict the channel allocation strategy of the user equipment to maximize the throughput of the communication system.
  • the communication system architecture generally includes an access layer, a mobile edge layer, and a cloud center.
  • the access layer is used to implement concurrent transmission on overlapping channels.
  • the mobile edge layer includes multiple mobile edge base stations.
  • the cloud center is used to realize the learning of the partial overlapping channel (POC) allocation strategy.
  • the mobile edge base station senses the wireless channel state information of the surrounding multi-user equipment, and uses the channel state information of K time slices before the current moment as the input of the deep neural network to predict the channel state of the next time slice. information, and use this information as the input of the deep reinforcement learning model, with the goal of maximizing the network throughput, reinforcement learning of the channel allocation strategy of the user equipment.
  • the provided deep learning method for spectral sharing of partially overlapping channels includes the following steps.
  • Step S1 using the coding redundancy of error correction coding to improve the interleaver and deinterleaver of the signal transceiver of the user equipment and the base station, and evenly distribute the overlapping spectrum information in the non-overlapping spectrum.
  • step S1 includes the following sub-steps:
  • Step S101 obtaining the ratio of the overlapping channel bandwidth to the complete channel bandwidth according to the current channel state information
  • Step S102 according to the ratio of the overlapping channel bandwidth to the complete channel bandwidth, select an appropriate interleaving strategy, assuming that the channel is divided into n sub-channels;
  • Step S103 Process the signal according to the dynamically selected interleaving strategy, evenly distribute the signal of the overlapping part of the channel to n sub-channels, and perform interleaving coding by using the redundant information encoded by the ECC.
  • the channel interleaving process is: calculating the ratio of the overlapped channel bandwidth to the complete channel bandwidth, expressed as: Among them, N p represents the ratio of the overlapping bandwidth to the total channel bandwidth, C p represents the overlapping channel bandwidth, and C represents the total channel bandwidth; select an appropriate channel interleaving strategy, such as 1/2 interleaving, 1/4 interleaving, 1/8 interleaving, etc. .
  • an appropriate channel interleaving strategy such as 1/2 interleaving, 1/4 interleaving, 1/8 interleaving, etc. .
  • a 1/n interleaving strategy is adopted, that is, the channel is divided into n sub-channels with equal bandwidths, of which the n-th sub-channel is an overlapping channel, and each sub-channel is further divided into n partial channels.
  • all sub-channels of each sub-channel are Part of the channels are evenly distributed in all sub-channels after interleaving to achieve the purpose of channel interleaving.
  • ECC coding redundancy the disturbed information of overlapping channels can be recovered.
  • Step S2 the base station detects the current channel state and presents the sensed channel state information.
  • step S2 includes the following sub-steps:
  • Step S201 the base station (gNodeB) evaluates the current channel state according to the communication request of the user equipment or the channel sounding signal sent by itself;
  • Step S202 the base station presents the perceived current channel state information by means of a signal state information graph.
  • step S3 a convolutional neural network model is deployed at the base station, and the convolutional neural network model takes the current channel state information as an input, and the output is the channel state information at the next moment.
  • step S3 includes the following sub-steps:
  • Step S301 Embed the constructed convolutional neural network code for channel prediction into the base station system.
  • Step S302 the deep convolutional neural network includes two main building blocks, including high-dimensional CSI extraction and channel generation.
  • high-dimensional CSI extraction consists of several convolutional layers. The purpose is to extract high-dimensional features of the input CSI. Then, the extracted CSI features are input to the generation of channels, go through multiple fully connected layers, and output the generated channels as the final prediction result.
  • Step S4 the base station trains the convolutional neural network by using the collected continuous time series of the channel state information as training samples to obtain a channel prediction convolutional neural network model.
  • step S4 includes the following sub-steps:
  • Step S401 the CSI of each user equipment is collected by the base station, and the collection methods have autonomous detection and user equipment reports, etc., continuous CSI can be stored in time series, and the base station uses the CSI history to predict the channel state;
  • Step S402 the base station performs offline training, for example, using a sliding window of size K time slices as input, outputs the channel state information of the next time slice, and uses the similarity of channel state prediction as the prediction loss optimization model.
  • the K value can be set according to the requirements for training efficiency and training accuracy.
  • the offline training process can be performed in the cloud center or on the server.
  • Step S5 the reinforcement learning model is trained with the channel state information predicted by the channel prediction convolutional neural network model as the input, and the bandwidth efficiency performance as the reward.
  • step S5 includes the following sub-steps:
  • Step S501 sending the channel prediction information output by the channel prediction convolutional neural network training process into the deep reinforcement learning network, as the input of reinforcement learning;
  • Step S502 the channel allocation strategy is implemented using a convolutional neural network, the input is the CSI of the next time slice, the output is the channel allocation strategy, and the action space includes the allocation probabilities of all channels;
  • Step S503 the reinforcement learning model optimizes the channel allocation strategy by maximizing the environmental throughput as an incentive.
  • Step S6 in response to the transmission request sent by the user equipment, the base station inputs the current channel state information into the channel prediction convolutional neural network model, outputs the channel state information of the next time slot, and uses the output information as the input of the reinforcement learning model.
  • step S6 includes the following sub-steps:
  • Step S601 the user sends a communication request
  • Step S602 the base station receives the user request, and selects the first K time slices CSI at the current moment from the CSI history;
  • Step S603 the channel prediction convolutional neural network takes the CSI of the first K time slices at the current moment as input, and outputs the next time slice CSI;
  • Step S604 the channel prediction convolutional neural network sends the output channel prediction information to the reinforcement learning model.
  • Step S7 the reinforcement learning model accepts the channel state information of the next time slot of the channel prediction convolutional neural network model as an input, uses the output of the strategy network as an allocation strategy, and continues to use the feedback information for reinforcement learning.
  • step S7 includes the following sub-steps:
  • Step S701 the reinforcement learning model takes the channel state information of the next time slice as the input and the output of the policy network as the channel allocation strategy, so as to realize the channel allocation strategy of maximizing the throughput of the simultaneous communication of the multi-user equipment
  • UEs user equipments
  • S total represents the total number of blocks within the channel.
  • CSI i is the channel state
  • s i and Pi represent the blocks and overlapping parts allocated to UE i
  • ri i represents the achievable data rate of UE i under a certain overlapping part Pi .
  • the goal is to maximize overall throughput, as defined by system utilities
  • a deep q-learning network is used to find the optimal policy under different system states, that is, the optimal POC allocation under different channel states.
  • Each terminal CSI is input to the DQN as an input.
  • the agent's action is the POC weight assigned to each terminal (ie the overlap assigned to each terminal).
  • the action space includes all appropriate POC weight assignments.
  • System rewards are considered to be the defined system utility.
  • Figure 5 depicts the architecture of DQN, which takes the CSI of each UE as the system input and the approximation of the POC weight of each UE as the system output. During training, experience replay is employed to reduce the correlation between training samples.
  • the agent on the central controller collects all CSI from the terminals and takes different actions (eg, choosing different POC weights) to obtain the q value.
  • An action is chosen if it maximizes the value of q in the long run.
  • step S702 the feedback information of the actual allocation stage is still used as a learning sample, and the reinforcement learning model is continued to be optimized.
  • the feedback information (such as channel state information and corresponding system throughput, etc.) collected in the actual allocation stage can be used as a sample for continued learning.
  • the present invention also provides a deep learning system for spectral sharing of partially overlapping channels, which is used to implement one or more aspects of the above method.
  • the system includes a preprocessing module, a channel interleaving module, a channel deinterleaving module, a channel state prediction module and a channel allocation module.
  • Preprocessing module The base station continuously perceives the channel state information in an active and passive manner, divides the time-continuous channel state information into time-slice continuous channel state information, and stores the sampled K-time slice sliding window as historical channel state information.
  • the K-slice sliding window refers to that the base station stores the channel state information of K time slices before the current moment.
  • a sliding window for storing K time slices is maintained, the time slice window slides with time, and the data in the window is kept up to date, that is, the channel state information of K time slices before the current moment is saved.
  • Channel interleaving module Through a simple calculation method of overlapping channel ratio: (where N p represents the ratio of the overlapping bandwidth to the total channel bandwidth, C p represents the overlapping channel bandwidth, and C represents the total channel bandwidth), select an appropriate channel interleaving strategy, such as: 1/2 interleaving, 1/4 interleaving, 1/ 8 interweaving, etc.
  • an appropriate channel interleaving strategy such as: 1/2 interleaving, 1/4 interleaving, 1/ 8 interweaving, etc.
  • a 1/n interleaving strategy is adopted, that is, the channel is divided into n sub-channels with equal bandwidths, of which the n-th sub-channel is an overlapping channel, and each sub-channel is further divided into n partial channels.
  • all sub-channels of each sub-channel are Part of the channels are evenly distributed in all sub-channels after interleaving to achieve the purpose of channel interleaving.
  • ECC coding redundancy the disturbed information of overlapping channels can be recovered.
  • Channel deinterleaving module The interleaving strategy of each transmission channel interleaving module is acquired by the channel deinterleaving module through additional information, and the deinterleaving module performs the reverse process to reform some of the channels rearranged due to channel interleaving, and restore the atomic channel. arrangement.
  • Channel state prediction module The base station obtains the channel state information of the K historical time slices before the current moment through the K-time slice sliding window, and uses the channel state information of the K-time slice as the input of the convolutional neural network, and outputs the next moment. channel state information to realize the channel prediction of the next time slice.
  • the channel state information of K time slices are relatively independent K inputs, and the K relatively independent inputs are input into the channel prediction convolution network as a whole.
  • Channel allocation module According to the predicted channel state information of the next time slice as input, in order to maximize the overall throughput as the incentive, train the reinforcement learning model, and output the allocation probability of all assignable sub-channels.
  • the deep reinforcement learning implementation method for partially overlapping channel sharing is a new technical solution for implementing overlapping channel sharing by utilizing the channel interleaving technology.
  • This scheme uses the channel state history time slice sequence to train the convolutional neural network, predicts the channel state information of the next time slice, and then uses the predicted channel state information of the next time slice as the input of deep reinforcement learning to learn to predict the channel allocation of user equipment. strategy to maximize network throughput.
  • the present invention may be a system, method and/or computer program product.
  • the computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of the present invention.
  • a computer-readable storage medium may be a tangible device that can hold and store instructions for use by the instruction execution device.
  • the computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing.
  • Non-exhaustive list of computer readable storage media include: portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM) or flash memory), static random access memory (SRAM), portable compact disk read only memory (CD-ROM), digital versatile disk (DVD), memory sticks, floppy disks, mechanically coded devices, such as printers with instructions stored thereon Hole cards or raised structures in grooves, and any suitable combination of the above.
  • RAM random access memory
  • ROM read only memory
  • EPROM erasable programmable read only memory
  • flash memory static random access memory
  • SRAM static random access memory
  • CD-ROM compact disk read only memory
  • DVD digital versatile disk
  • memory sticks floppy disks
  • mechanically coded devices such as printers with instructions stored thereon Hole cards or raised structures in grooves, and any suitable combination of the above.
  • Computer-readable storage media are not to be construed as transient signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (eg, light pulses through fiber optic cables), or through electrical wires transmitted electrical signals.
  • the computer readable program instructions described herein may be downloaded to various computing/processing devices from a computer readable storage medium, or to an external computer or external storage device over a network such as the Internet, a local area network, a wide area network, and/or a wireless network.
  • the network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and/or edge servers.
  • a network adapter card or network interface in each computing/processing device receives computer-readable program instructions from a network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing/processing device .
  • the computer program instructions for carrying out the operations of the present invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state setting data, or instructions in one or more programming languages.
  • Source or object code written in any combination including object-oriented programming languages, such as Smalltalk, C++, Python, etc., and conventional procedural programming languages, such as the "C" language or similar programming languages.
  • the computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server implement.
  • the remote computer may be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (eg, using an Internet service provider through the Internet connect).
  • LAN local area network
  • WAN wide area network
  • custom electronic circuits such as programmable logic circuits, field programmable gate arrays (FPGAs), or programmable logic arrays (PLAs)
  • FPGAs field programmable gate arrays
  • PDAs programmable logic arrays
  • Computer readable program instructions are executed to implement various aspects of the present invention.
  • These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer or other programmable data processing apparatus to produce a machine that causes the instructions when executed by the processor of the computer or other programmable data processing apparatus , resulting in means for implementing the functions/acts specified in one or more blocks of the flowchart and/or block diagrams.
  • These computer readable program instructions can also be stored in a computer readable storage medium, these instructions cause a computer, programmable data processing apparatus and/or other equipment to operate in a specific manner, so that the computer readable medium on which the instructions are stored includes An article of manufacture comprising instructions for implementing various aspects of the functions/acts specified in one or more blocks of the flowchart and/or block diagrams.
  • Computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other equipment to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other equipment to produce a computer-implemented process , thereby causing instructions executing on a computer, other programmable data processing apparatus, or other device to implement the functions/acts specified in one or more blocks of the flowcharts and/or block diagrams.
  • each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more functions for implementing the specified logical function(s) executable instructions.
  • the functions noted in the blocks may occur out of the order noted in the figures. For example, two blocks in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved.
  • each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations can be implemented in dedicated hardware-based systems that perform the specified functions or actions , or can be implemented in a combination of dedicated hardware and computer instructions. It is well known to those skilled in the art that implementation in hardware, implementation in software, and implementation in a combination of software and hardware are all equivalent.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Power Engineering (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Biophysics (AREA)
  • Computing Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本发明公开了一种部分重叠信道频谱共享的深度学习方法及系统。该方法包括：响应于接收到的用户传输请求，基站将多个历史时间片的信道状态信息CSI输入到经训练的信道预测卷积神经网络模型，输出预测的下一个时间片的信道状态信息CSI；将所述下一时间片的信道状态信息CSI输入强化学习模型，获得基站碰撞域中各用户设备的信道分配策略，以实现各用户设备同时通信的最大化吞吐量，该强化学习模型以带宽效率性能作为奖励经训练获得。本发明对于通信网络具有高普适性、高带宽利用率和高吞吐量。

Description

一种部分重叠信道频谱共享的深度学习方法及系统 技术领域
本发明涉及通信技术领域，更具体地，涉及一种部分重叠信道频谱共享的深度学习方法及系统。
背景技术
终端设备，如平板电脑、智能手机和异构物联网(IoT)设备，正在成为5G智能城市中主要的带宽消耗组件。为终端设备设计的新应用越来越多，如交互游戏、导航、自然语言处理、人脸识别、增强现实等，都需要大量的频谱资源。据预测，到2025年，联网物联网设备将达到416亿台，每年产生79.4ZB的数据。随着各种创新但急需带宽的应用程序的出现，有效进行频谱管理具有重要意义。
针对频谱稀缺问题和静态频谱分配策略利用不足之间的矛盾，有研究提出了从固定频谱分配向动态频谱管理(DSM)的范式转变。在DSM中，以前不允许传输的未授权用户现在可以与授权用户一起访问授权频谱。接入方式可以是机会频谱接入，也可以是并发频谱接入。在前一种模式下，未经授权的用户只能在未激活的情况下访问已授权的频谱。在后一种模式下，非授权用户和授权用户可以共存，只要他们的传输不超过授权用户的干扰裕度。
传统的DSM很少采用部分重叠信道(POC)实现并发频谱访问。为并发传输分配合适的POC需要复杂的功率控制和干扰消除算法，这对于有硬件约束的物联网设备来说是不切实际的。此外，由于物联网设备的环境是高度动态的，很难对POC分配的完整和准确的通道信息进行测量。然而，物联网设备的激增导致它们在相对较小的地理区域内高度密集地部署。因此，POC的使用成为必然，在5G智慧城市中充分利用频谱效率展现出巨大潜力。
近年来，人工智能(AI)取得了重大成就，并被应用到DSM中，以应对各种技术挑战。人工智能技术不需要建立基于完整准确信息的DSM模型，而是可以从周围环境中学习或探索访问策略，并根据动态环境定期调整访问策略。近年来的研究表明，人工智能技术可以有效地提高系统的鲁棒性和频谱效率。尽管人们对基于人工智能的DSM的兴趣激增，但在POC分配方面仍存在一些挑战。例如，如何利用部分重叠信道传输的特性来进行并发的频谱访问，同时具有轻量级的计算开销，这仍然是一个值得关注的问题。此外，联网物联网设备的环境是高度动态的，因此POC分配体系结构也需要灵活性和适应性。
发明内容
本发明的目的是克服上述现有技术的缺陷，提供一种部分重叠信道频谱共享的深度学习方法及系统。
根据本发明的第一方面，提供一种部分重叠信道频谱共享的深度学习方法。该方法包括以下步骤：
响应于接收到的用户传输请求，基站将多个历史时间片的信道状态信息CSI输入到经训练的信道预测卷积神经网络模型，输出预测的下一个时间片的信道状态信息CSI；
将所述下一时间片的信道状态信息CSI输入强化学习模型，获得基站碰撞域中各用户设备的信道分配策略，以实现各用户设备同时通信的最大化吞吐量，该强化学习模型以带宽效率性能作为奖励经训练获得。
根据本发明的第二方面，提供一种部分重叠信道频谱共享的深度学习系统。该系统包括：
信道状态预测单元：用于响应于接收到的用户传输请求，将多个历史时间片的信道状态信息CSI输入到经训练的信道预测卷积神经网络模型，输出预测的下一个时间片的信道状态信息CSI；
信道分配单元：用于将所述下一时间片的信道状态信息CSI输入强化学习模型，获得基站碰撞域中各用户设备的信道分配策略，以实现各用户设备同时通信的最大化吞吐量，该强化学习模型以带宽效率性能作为奖励 经训练获得。
与现有技术相比，本发明的优点在于，结合深度学习，提出了一种基于部分重叠信道的并发频谱访问系统架构，该架构自动从非重叠频谱上的数据中学习额外的编码冗余，并将冗余应用于重叠频谱上的数据恢复。
通过以下参照附图对本发明的示例性实施例的详细描述，本发明的其它特征及其优点将会变得清楚。
附图说明
被结合在说明书中并构成说明书的一部分的附图示出了本发明的实施例，并且连同其说明一起用于解释本发明的原理。
图1是根据本发明一个实施例的通信系统执行流程图；
图2是根据本发明一个实施例的通信系统框架示意图；
图3是根据本发明一个实施例的部分重叠信道频谱共享的深度学习方法的流程图；
图4是根据本发明一个实施例的信道交织执行过程示意图；
图5是根据本发明一个实施例的强化学习模型示意图；
附图中，Reward-奖励；State-状态；Environment-环境；Action-动作；Agent-代理；Cloud-云。
具体实施方式
现在将参照附图来详细描述本发明的各种示例性实施例。应注意到：除非另外具体说明，否则在这些实施例中阐述的部件和步骤的相对布置、数字表达式和数值不限制本发明的范围。
以下对至少一个示例性实施例的描述实际上仅仅是说明性的，决不作为对本发明及其应用或使用的任何限制。
对于相关领域普通技术人员已知的技术、方法和设备可能不作详细讨论，但在适当情况下，所述技术、方法和设备应当被视为说明书的一部分。
在这里示出和讨论的所有例子中，任何具体值应被解释为仅仅是示例性的，而不是作为限制。因此，示例性实施例的其它例子可以具有不同的 值。
应注意到：相似的标号和字母在下面的附图中表示类似项，因此，一旦某一项在一个附图中被定义，则在随后的附图中不需要对其进行进一步讨论。
本发明是基于5G智慧城市频谱共享探索部分重叠通道的深度学习方法，创新地利用信道交织技术实现重叠信道共享，利用信道状态历史时间片序列训练卷积神经网络，预测下一时间片的信道状态信息，再以预测的下一时间片的信道状态信息作为深度强化学习输入，学习预测用户设备信道分配策略，以最大化通信系统吞吐量。
结合图1和图2所示，通信系统架构总体上包括接入层、移动边缘层和云中心，接入层用于实现重叠信道上的并发传输，移动边缘层包括多个移动边缘基站，用于实现信道状态预测，云中心用于实现部分重叠信道(POC)分配策略的学习。简言之，在本发明中，移动边缘基站感知周围多用户设备的无线信道状态信息，利用当前时刻之前K个时间片的信道状态信息作为深度神经网络的输入，预测下一时间片的信道状态信息，并以该信息作为深度强化学习模型的输入，以最大化网络吞吐量为目标，强化学习用户设备信道分配策略。
具体地，参见图3所示，所提供的部分重叠信道频谱共享的深度学习方法包括以下步骤。
步骤S1，利用纠错编码的编码冗余，改进用户设备和基站的信号收发器的交织器和解交织器，将重叠频谱信息均匀分布于非重叠频谱中。
在一个实施例中，步骤S1包括以下子步骤：
步骤S101，根据当前的信道状态信息，获取重叠信道带宽与完整信道带宽的比例；
步骤S102，根据重叠信道带宽与完整信道带宽的比例，选择合适的交织策略，假设将信道划分成n个子信道；
步骤S103，根据动态选择的交织策略对信号进行处理，将信道重叠部分信号均匀的分布于n个子信道，利用ECC编码的冗余信息进行交织编码。
例如，结合图4所示，信道交织过程是：计算重叠信道带宽与完整信道带宽的比例，表示为：
Figure PCTCN2021083271-appb-000001
其中N p表示重叠部分带宽占信道总带宽的比例，C p表示重叠信道带宽，C表示信道总带宽；选择合适的信道交织策略，如1/2交织，1/4交织，1/8交织等。假设采用1/n交织策略，即将信道划分成n个带宽相等的子信道，其中第n个子信道为重叠信道，同时将每个子信道再划分成n个部分信道，交织前每一个子信道的所有部分信道均匀的分散于交织后所有子信道中，达到信道交织的目的，利用ECC编码冗余，能够恢复重叠信道受干扰的信息。
步骤S2，基站探测当前的信道状态并呈现感知到的信道状态信息。
在一个实施例中，步骤S2包括以下子步骤：
步骤S201，基站(gNodeB)根据用户设备的通信请求或自身发出的信道探测信号评估当前信道状态；
步骤S202，基站通过信号状态信息图的方式呈现感知到的当前信道状态信息。
步骤S3，在基站部署卷积神经网络模型，该卷积神经网络模型以当前信道状态信息作为输入，输出为下一时刻的信道状态信息。
在一个实施例中，步骤S3包括以下子步骤：
步骤S301，将构建好的用于信道预测的卷积神经网络代码嵌入到基站系统中。
步骤S302，深度卷积神经网络包含两个主要的构建模块，包括高维CSI提取和信道生成。
具体来说，高维CSI提取包含几个卷积层。目的是提取输入CSI的高维特征。然后，将提取的CSI特征输入信道的产生，经过多次全连接层，并输出生成的通道作为最终的预测结果。
步骤S4，基站通过采集的信道状态信息的连续时间序列作为训练样本，训练卷积神经网络，获得信道预测卷积神经网络模型。
在一个实施例中，步骤S4包括以下子步骤：
步骤S401，每个用户设备的CSI由基站采集，采集方式有自主探测和用户设备报告等，连续CSI会以时间序列存储，基站使用CSI历史来预 测信道状态；
步骤S402，基站进行离线训练，例如使用大小为K个时间片的滑动窗口作为输入，输出下一个时间片的信道状态信息，以信道状态预测的相近性作为预测损失优化模型。
需说明的是，K值可根据对训练效率和训练精度的要求进行设置。离线训练过程可在云中心或服务器上进行。
步骤S5，以信道预测卷积神经网络模型预测的信道状态信息作为输入，以带宽效率性能作为奖励，训练强化学习模型。
在一个实施例中，步骤S5包括以下子步骤：
步骤S501，将信道预测卷积神经网络训练过程输出的信道预测信息送入深度强化学习网络，作为强化学习的输入；
步骤S502，信道分配策略使用卷积神经网络实现，输入为下一时间片的CSI，输出为信道分配策略，动作空间包含了所有信道的分配概率；
步骤S503，该强化学习模型以最大化环境吞吐量作为激励，优化信道分配策略。
步骤S6，响应于用户设备发出的传输请求，基站将当前信道状态信息输入到信道预测卷积神经网络模型，输出下一时隙的信道状态信息，并将该输出信息作为强化学习模型的输入。
在一个实施例中，步骤S6包括以下子步骤：
步骤S601，用户发出通信请求；
步骤S602，基站接收到用户请求，从CSI历史中选取当前时刻的前K个时间片CSI；
步骤S603，信道预测卷积神经网络以当前时刻的前K个时间片的CSI作为输入，输出下一个时间片CSI；
步骤S604，信道预测卷积神经网络将输出的信道预测信息送往强化学习模型。
步骤S7，强化学习模型接受信道预测卷积神经网络模型的下一时隙信道状态信息作为输入，将策略网络的输出作为分配策略，并将反馈信息继续用于强化学习。
在一个实施例中，步骤S7包括以下子步骤：
步骤S701，强化学习模型以下一时间片的信道状态信息作为输入，以策略网络的输出作为信道分配策略，实现多用户设备同时通信的最大化吞吐量的信道分配策略
具体地，为了在不同的用户信道状态下找到最优的对讲分配，首先提出问题。假设在一个gNodeB的碰撞域中有n个用户设备(UE)。S total表示通道内的块总数。CSI i是的信道状态，s i和Pi表示分配给UE i的块和重叠部分，用r i表示UE i在一定重叠部分P i下的可达数据率。目标是最大化总体吞吐量，由系统实用程序定义
Figure PCTCN2021083271-appb-000002
这里使用深度q学习网络(DQN)来寻找不同系统状态下的最优策略，即不同信道状态下的最优POC分配。每个终端CSI被输入到DQN作为输入。代理的动作是分配给每个终端的POC权重(即分配给每个终端的重叠部分)。因此，动作空间包括所有适当的POC权重分配。系统奖励被认为是定义的系统效用。图5描述了DQN的架构，它以每个UE的CSI作为系统输入，以每个UE的POC权值的近似作为系统输出。在训练过程中，采用经验重放来降低训练样本之间的相关性。对于该实施例的DQN，中央控制器上的代理从终端收集所有CSI，并采取不同的操作(例如，选择不同的POC权重)来获得q值。如果某一行动能够长期带来q值的最大化，那么它便是所选择的行动。
步骤S702，将实际分配阶段的反馈信息，仍作为学习样本，继续优化强化学习模型。
为增强强化学习模型的适用性和精确性，优选地，可将实际分配阶段采集的反馈信息(如信道状态信息和对应的系统吞吐量等)作为继续学习的样本。
相应地，本发明还提供一种部重叠信道频谱共享的深度学习系统，用于实现上述方法的一个方面或多个方面。例如，该系统包括预处理模块、信道交织模块、信道解交织模块、信道状态预测模块和信道分配模块。
预处理模块：基站通过主动和被动的方式连续感知信道状态信息，将时间连续的信道状态信息划分成时间片连续的信道状态信息，采样K-时间 片滑动窗口的形式存储为历史信道状态信息。
例如，对于预处理模块，K-时间片滑动窗口指的是基站存储当前时刻之前的K个时间片的信道状态信息。维护一个存储K个时间片的滑动窗口，该时间片窗口随时间滑动，保持窗口内数据是最新的，即保存当前时刻之前的K个时间片的信道状态信息。
信道交织模块：通过简单的重叠信道占比计算方法：
Figure PCTCN2021083271-appb-000003
(其中N p表示重叠部分带宽占信道总带宽的比例，C p表示重叠信道带宽，C表示信道总带宽)，选择合适的信道交织策略，如：1/2交织，1/4交织，1/8交织等。假设采用1/n交织策略，即将信道划分成n个带宽相等的子信道，其中第n个子信道为重叠信道，同时将每个子信道再划分成n个部分信道，交织前每一个子信道的所有部分信道均匀的分散于交织后所有子信道中，达到信道交织的目的，利用ECC编码冗余，能够恢复重叠信道受干扰的信息。
信道解交织模块：每一次传输信道交织模块的交织策略通过额外的信息被信道解交织模块获取，解交织模块执行反向的过程，重整因信道交织被重排的部分信道，恢复原子信道的排列。
信道状态预测模块：基站通过K-时间片滑动窗口获取当前时刻之前的K个历史时间片的信道状态信息，以K-时间片的信道状态信息共同作为卷积神经网络的输入，输出下一时刻的信道状态信息，实现下一时间片的信道预测。
例如，对于信道状态预测模块，K个时间片的信道状态信息是相对独立的K个输入，将K个相对独立的输入作为一个整体输入信道预测卷积网络。
信道分配模块：根据预测的下一时间片的信道状态信息作为输入，以最大化总体吞吐量为激励，训练强化学习模型，输出所有可分配子信道的分配概率。
综上所述，本发明提供的部分重叠信道共享的深度强化学习实现方法，是利用信道交织技术实现重叠信道共享的新技术方案。该方案利用信道状态历史时间片序列训练卷积神经网络，预测下一时间片的信道状态信息， 再以预测的下一时间片的信道状态信息作为的深度强化学习输入，学习预测用户设备信道分配策略，最大化网络吞吐量。
为进一步验证本发明的效果，在PHY层(物理层)和MAC层(介质访问控制层)提出了一个系统级的案例研究作为一个图解的架构。验证结果表明所提出的部分重叠信道共享的深度强化学习实现方法具有高性能、高普适性、高带宽利用率和高吞吐量。
本发明可以是系统、方法和/或计算机程序产品。计算机程序产品可以包括计算机可读存储介质，其上载有用于使处理器实现本发明的各个方面的计算机可读程序指令。
计算机可读存储介质可以是可以保持和存储由指令执行设备使用的指令的有形设备。计算机可读存储介质例如可以是――但不限于――电存储设备、磁存储设备、光存储设备、电磁存储设备、半导体存储设备或者上述的任意合适的组合。计算机可读存储介质的更具体的例子(非穷举的列表)包括：便携式计算机盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、静态随机存取存储器(SRAM)、便携式压缩盘只读存储器(CD-ROM)、数字多功能盘(DVD)、记忆棒、软盘、机械编码设备、例如其上存储有指令的打孔卡或凹槽内凸起结构、以及上述的任意合适的组合。这里所使用的计算机可读存储介质不被解释为瞬时信号本身，诸如无线电波或者其他自由传播的电磁波、通过波导或其他传输媒介传播的电磁波(例如，通过光纤电缆的光脉冲)、或者通过电线传输的电信号。
这里所描述的计算机可读程序指令可以从计算机可读存储介质下载到各个计算/处理设备，或者通过网络、例如因特网、局域网、广域网和/或无线网下载到外部计算机或外部存储设备。网络可以包括铜传输电缆、光纤传输、无线传输、路由器、防火墙、交换机、网关计算机和/或边缘服务器。每个计算/处理设备中的网络适配卡或者网络接口从网络接收计算机可读程序指令，并转发该计算机可读程序指令，以供存储在各个计算/处理设备中的计算机可读存储介质中。
用于执行本发明操作的计算机程序指令可以是汇编指令、指令集架构 (ISA)指令、机器指令、机器相关指令、微代码、固件指令、状态设置数据、或者以一种或多种编程语言的任意组合编写的源代码或目标代码，所述编程语言包括面向对象的编程语言—诸如Smalltalk、C++、Python等，以及常规的过程式编程语言—诸如“C”语言或类似的编程语言。计算机可读程序指令可以完全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中，远程计算机可以通过任意种类的网络—包括局域网(LAN)或广域网(WAN)—连接到用户计算机，或者，可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。在一些实施例中，通过利用计算机可读程序指令的状态信息来个性化定制电子电路，例如可编程逻辑电路、现场可编程门阵列(FPGA)或可编程逻辑阵列(PLA)，该电子电路可以执行计算机可读程序指令，从而实现本发明的各个方面。
这里参照根据本发明实施例的方法、装置(系统)和计算机程序产品的流程图和/或框图描述了本发明的各个方面。应当理解，流程图和/或框图的每个方框以及流程图和/或框图中各方框的组合，都可以由计算机可读程序指令实现。
这些计算机可读程序指令可以提供给通用计算机、专用计算机或其它可编程数据处理装置的处理器，从而生产出一种机器，使得这些指令在通过计算机或其它可编程数据处理装置的处理器执行时，产生了实现流程图和/或框图中的一个或多个方框中规定的功能/动作的装置。也可以把这些计算机可读程序指令存储在计算机可读存储介质中，这些指令使得计算机、可编程数据处理装置和/或其他设备以特定方式工作，从而，存储有指令的计算机可读介质则包括一个制造品，其包括实现流程图和/或框图中的一个或多个方框中规定的功能/动作的各个方面的指令。
也可以把计算机可读程序指令加载到计算机、其它可编程数据处理装置、或其它设备上，使得在计算机、其它可编程数据处理装置或其它设备上执行一系列操作步骤，以产生计算机实现的过程，从而使得在计算机、其它可编程数据处理装置、或其它设备上执行的指令实现流程图和/或框图 中的一个或多个方框中规定的功能/动作。
附图中的流程图和框图显示了根据本发明的多个实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段或指令的一部分，所述模块、程序段或指令的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个连续的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以用执行规定的功能或动作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。对于本领域技术人员来说公知的是，通过硬件方式实现、通过软件方式实现以及通过软件和硬件结合的方式实现都是等价的。
以上已经描述了本发明的各实施例，上述说明是示例性的，并非穷尽性的，并且也不限于所披露的各实施例。在不偏离所说明的各实施例的范围和精神的情况下，对于本技术领域的普通技术人员来说许多修改和变更都是显而易见的。本文中所用术语的选择，旨在最好地解释各实施例的原理、实际应用或对市场中的技术改进，或者使本技术领域的其它普通技术人员能理解本文披露的各实施例。本发明的范围由所附权利要求来限定。

Claims (10)

  1. 一种部分重叠信道频谱共享的深度学习方法，包括以下步骤：
    响应于接收到的用户传输请求，基站将多个历史时间片的信道状态信息CSI输入到经训练的信道预测卷积神经网络模型，输出预测的下一个时间片的信道状态信息CSI；
    将所述下一时间片的信道状态信息CSI输入强化学习模型，获得基站碰撞域中各用户设备的信道分配策略，以实现各用户设备同时通信的最大化吞吐量，该强化学习模型以带宽效率性能作为奖励经训练获得。
  2. 根据权利要求1所述的方法，其中，根据以下步骤训练所述信道预测卷积神经网络模型：
    采集目标用户设备的信道状态信息CSI，并以时间序列进行存储；
    以信道状态预测的相近性作为损失，训练卷积神经网络，在训练过程中，使用大小为K个时间片的滑动窗口作为输入，输出下一个时间片的信道状态信息，获得满足优化目标的卷积神经网络，作为所述信道预测卷积神经网络模型，其中K是大于等于2的整数。
  3. 根据权利要求1所述的方法，其中，所述信道预测卷积神经网络模型包括多个卷积层和多个全连接层，所述多个卷积层用于提取信道状态信息CSI的高维特征，并经过所述多个全连接层，输出预测的下一个时间片的信道状态信息。
  4. 根据权利要求1所述的方法，其中，根据以下步骤训练所述强化学习模型：
    以所述信道预测卷积神经网络模型训练过程输出的信道状态信息CSI作为强化学习模型的输入，动作空间包含所有信道的分配概率，并且该强化学习模型以最大化环境吞吐量作为激励，优化信道分配策略。
  5. 根据权利要求1所述的方法，其中，将所述下一时间片的信道状态信息CSI输入强化学习模型，获得基站碰撞域中各用户设备的信道分配策略，以实现各用户设备同时通信的最大化吞吐量包括：
    假设在一个基站的碰撞域中有n个用户设备UE，S total表示通道内的块总数，CSI i是信道状态，s i和Pi表示分配给UE i的块和重叠部分，r i表示UE i 在一定重叠部分P i下的可达数据率，目标是最大化总体吞吐量
    Figure PCTCN2021083271-appb-100001
    使用所述强化学习模型来寻找不同信道状态下的最优分配策略，其中，每个用户设备的信道状态信息CSI被输入到所述强化学习模型作为输入，代理的动作是分配给每个用户设备的部分重叠信道权重，动作空间包括所有可用的部分重叠信道权重分配。
  6. 根据权利要求1所述的方法，其中，对于所述基站和所述用户设备，信道交织过程包括以下步骤：
    根据当前的信道状态信息CSI，获取重叠信道带宽与完整信道带宽的比例；
    根据重叠信道带宽与完整信道带宽的比例，选择交织策略，以将信道划分成n个子信道；
    根据所选择的交织策略对信号进行处理，将信道重叠部分信号均匀的分布与n个子信道，利用纠错编码的冗余信息进行交织编码。
  7. 根据权利要求1所述的方法，还包括：所述强化学习模型将策略网络的输出作为信道分配策略，同时将反馈信息继续用于强化学习。
  8. 一种部分重叠信道频谱共享的深度学习系统，包括：
    信道状态预测单元：用于响应于接收到的用户传输请求，将多个历史时间片的信道状态信息CSI输入到经训练的信道预测卷积神经网络模型，输出预测的下一个时间片的信道状态信息CSI；
    信道分配单元：用于将所述下一时间片的信道状态信息CSI输入强化学习模型，获得基站碰撞域中各用户设备的信道分配策略，以实现各用户设备同时通信的最大化吞吐量，该强化学习模型以带宽效率性能作为奖励经训练获得。
  9. 一种计算机可读存储介质，其上存储有计算机程序，其中，该程序被处理器执行时实现根据权利要求1至7中任一项所述方法的步骤。
  10. 一种电子设备，包括存储器和处理器，在所述存储器上存储有能够在处理器上运行的计算机程序，其特征在于，所述处理器执行所述程序时实现权利要求1至7中任一项所述的方法的步骤。
PCT/CN2021/083271 2021-03-26 2021-03-26 一种部分重叠信道频谱共享的深度学习方法及系统 WO2022198634A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2021/083271 WO2022198634A1 (zh) 2021-03-26 2021-03-26 一种部分重叠信道频谱共享的深度学习方法及系统
US18/281,418 US20240121136A1 (en) 2021-03-26 2021-03-26 Deep learning method and system for spectrum sharing of partially overlapping channels

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2021/083271 WO2022198634A1 (zh) 2021-03-26 2021-03-26 一种部分重叠信道频谱共享的深度学习方法及系统

Publications (1)

Publication Number Publication Date
WO2022198634A1 true WO2022198634A1 (zh) 2022-09-29

Family

ID=83396277

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/083271 WO2022198634A1 (zh) 2021-03-26 2021-03-26 一种部分重叠信道频谱共享的深度学习方法及系统

Country Status (2)

Country Link
US (1) US20240121136A1 (zh)
WO (1) WO2022198634A1 (zh)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111126562A (zh) * 2019-11-22 2020-05-08 上海交通大学 基于神经网络的目标算法拟合方法、终端以及应用
CN111711666A (zh) * 2020-05-27 2020-09-25 梁宏斌 一种基于强化学习的车联网云计算资源优化方法
US20200382354A1 (en) * 2019-08-16 2020-12-03 Intel Corporation Spectrum sharing between fifth generation new radio and long term evolution in licensed and unlicensed bands

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200382354A1 (en) * 2019-08-16 2020-12-03 Intel Corporation Spectrum sharing between fifth generation new radio and long term evolution in licensed and unlicensed bands
CN111126562A (zh) * 2019-11-22 2020-05-08 上海交通大学 基于神经网络的目标算法拟合方法、终端以及应用
CN111711666A (zh) * 2020-05-27 2020-09-25 梁宏斌 一种基于强化学习的车联网云计算资源优化方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
TANG, FENGXIAO ET AL.: "On a Novel Deep-Learning-Based Intelligent Partially Overlapping Channel Assignment in SDN-IoT", IEEE COMMUNICATIONS MAGAZINE, vol. 56, no. 9, 17 September 2018 (2018-09-17), XP055517135, DOI: 10.1109/MCOM.2018.1701227 *

Also Published As

Publication number Publication date
US20240121136A1 (en) 2024-04-11

Similar Documents

Publication Publication Date Title
WO2022199132A1 (zh) 一种部分重叠信道频谱共享的深度学习方法及系统
US11410046B2 (en) Learning-based service migration in mobile edge computing
Li et al. PyramidFL: A fine-grained client selection framework for efficient federated learning
Park et al. Communication-efficient and distributed learning over wireless networks: Principles and applications
Li et al. Improved cooperative spectrum sensing model based on machine learning for cognitive radio networks
CN116546221A (zh) 使用神经网络压缩图像
Iyer et al. A survey on semantic communications for intelligent wireless networks
WO2022039815A1 (en) Hash-based attribute prediction for point cloud coding
US20210241094A1 (en) Rank selection in tensor decomposition based on reinforcement learning for deep neural networks
Balieiro et al. Adaptive spectrum sensing for cognitive radio based on multi‐objective genetic optimisation
Xu et al. Stochastic adaptive event‐triggered control and network scheduling protocol co‐design for distributed networked systems
Yue et al. Coding for distributed fog computing in internet of mobile things
Yarinezhad et al. Exact and approximate algorithms for clustering problem in wireless sensor networks
Jo et al. Deep reinforcement learning‐based joint optimization of computation offloading and resource allocation in F‐RAN
Goh et al. Partial Offloading MEC Optimization Scheme using Deep Reinforcement Learning for XR Real-Time M&S Devices
WO2022198634A1 (zh) 一种部分重叠信道频谱共享的深度学习方法及系统
Ateya et al. Lightweight deep learning-based model for traffic prediction in fog-enabled dense deployed iot networks
KR20230061423A (ko) 기계 학습 파이프라인의 분산 리소스 인식 훈련
Luan et al. Correntropy induced joint power and admission control algorithm for dense small cell network
Chen et al. Energy‐efficiency fog computing resource allocation in cyber physical internet of things systems
Liao et al. Joint mission planning and spectrum resources optimization for multi‐UAV reconnaissance
Guerra et al. Simple role‐based rendezvous algorithm for cognitive ad hoc radio networks
Hu et al. Communication-Efficient Coded Computing for Distributed Multi-Task Learning
Khani et al. Deep reinforcement learning‐based resource allocation in multi‐access edge computing
Liu et al. Adaptive federated pruning in hierarchical wireless networks

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 31/01/2024)