CN111162852B - Ubiquitous power Internet of things access method based on matching learning - Google Patents

Ubiquitous power Internet of things access method based on matching learning Download PDF

Info

Publication number
CN111162852B
CN111162852B CN201911406929.3A CN201911406929A CN111162852B CN 111162852 B CN111162852 B CN 111162852B CN 201911406929 A CN201911406929 A CN 201911406929A CN 111162852 B CN111162852 B CN 111162852B
Authority
CN
China
Prior art keywords
data
channel
model
time slot
time
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
Application number
CN201911406929.3A
Other languages
Chinese (zh)
Other versions
CN111162852A (en
Inventor
周振宇
潘超
廖海君
刘念
汪中原
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
North China Electric Power University
Original Assignee
North China Electric Power University
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 North China Electric Power University filed Critical North China Electric Power University
Priority to CN201911406929.3A priority Critical patent/CN111162852B/en
Publication of CN111162852A publication Critical patent/CN111162852A/en
Application granted granted Critical
Publication of CN111162852B publication Critical patent/CN111162852B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/391—Modelling the propagation channel
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/382—Monitoring; Testing of propagation channels for resource allocation, admission control or handover
    • 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
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/70—Services for machine-to-machine communication [M2M] or machine type communication [MTC]
    • 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)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • Electromagnetism (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

The invention discloses a ubiquitous power Internet of things access method based on matching learning, which comprises the following steps of: s1, constructing a system model; s2, refining the model to obtain a task/data transmission model, an energy consumption model, a time delay model and a service reliability model; s3, maximizing long-term throughput and determining an optimization problem; s4, converting the optimization problem based on the virtual queue theory and the lyapunov optimization theory; and S5, realizing the optimal channel selection through learning and matching, and further realizing the maximization of throughput. The invention realizes the optimization on channel selection by learning and matching, thereby realizing the maximization of throughput; based on the MAB theory, the lyapunov optimization theory and the matching theory, energy perception and service reliability perception are combined with machine learning, and therefore the maximum and the optimum of energy utilization rate and service reliability are achieved.

Description

Ubiquitous power Internet of things access method based on matching learning
Technical Field
The invention relates to an electric power Internet of things, in particular to a ubiquitous electric power Internet of things access method based on matching learning.
Background
The fourth industrial revolution aims at realizing interconnection response, intellectualization and self-optimization of manufacturing processes and systems through seamless integration of advanced manufacturing technologies and industrial internet of things. In this new model, billions of Machine Type Devices (MTDs) will be deployed for continuously performing various tasks such as monitoring, billing, and protection, etc. However, the contradiction between resource-limited MTD and computationally intensive tasks has become a bottleneck in providing reliable services.
Shifting the compute-intensive tasks from resource-limited MTDs to powerful edge servers provides a promising solution to accommodate the rapidly increasing computing demands. In conventional cloud computing, a remote cloud server is usually located far away from the MTD, and data transmission over a long distance causes many problems, including connection instability, network congestion, and intolerable delay. In contrast, edge computing, which shifts computing power from a remote cloud to an edge server in a wireless access network, is a promising paradigm for reducing latency, alleviating congestion, and extending battery life.
However, although edge computing provides a promising way to take advantage of the rich computing resources of edge servers, its performance may be severely impacted due to limited spectrum resources, limited capacity batteries, and context-imperceptible issues. First, in order to deliver large numbers of tasks from MTDs to edge servers in real-time, channel selection must be dynamically optimized according to time-varying context parameters such as Channel State Information (CSI), Energy State Information (ESI), server load, and service reliability requirements. Traditional centralized optimization methods rely on a common assumption that there is a central node, such as a base station, that has complete information of all context parameters. This assumption is difficult to implement in practice, considering the high cost of the signalling overhead for collecting the entire network information. Therefore, a distributed optimization approach, where each MTD individually optimizes its local information-based channel selection strategy, is preferable. However, when the number of MTDs far exceeds the number of available channels, and multiple MTDs contend for the same channel, selection conflicts frequently occur, thereby coupling channel selection strategies between different MTDs. Second, due to the limited battery capacity, the MTD will cease to be used when the battery energy is depleted. Thus, short-term channel selection strategies also couple long-term energy budgets. Last but not least, industrial applications often need to guarantee a certain service reliability. How to meet stringent reliability requirements with limited resources and information presents another area of difficulty.
The matching theory provides a flexible, low-complexity and efficient tool for solving the combination problems of channel selection, task selection, server selection and the like. However, it requires full knowledge of Global State Information (GSI) to construct a preference list, which specifies the basic matching criteria. There have been several attempts to optimize research by channel selection based on matching and game theory. However, they rely on uncertain context parameters, which can suffer severe performance loss if the actual probability distribution of the uncertain factors is not consistent with the assumed statistical model.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a ubiquitous power Internet of things access method based on matching learning, which can realize the optimal channel selection through learning and matching and further realize the maximization of throughput.
The purpose of the invention is realized by the following technical scheme: a ubiquitous power Internet of things access method based on matching learning comprises the following steps:
s1, constructing a system model;
s2, refining the model to obtain a task or data transmission model, an energy consumption model, a time delay model and a service reliability model;
s3, maximizing long-term throughput and determining an optimization problem;
s4, converting the optimization problem based on the virtual queue theory and the lyapunov optimization theory;
and S5, realizing the optimal channel selection through learning and matching, and further realizing the maximization of throughput.
Preferably, the ubiquitous power internet of things access method further comprises the step of ubiquitous power internet of things access: after the channel selection is optimized in step S5, the ubiquitous power internet of things is accessed through the channel, and data transmission is completed.
Further, the system model constructed in step S1 includes:
setting a base station service cell which comprises a base station and an edge server which are distributed in a similar way, wherein the base station provides connection service for K Machine Type Devices (MTD) in the cell, and the edge server provides calculation service; wherein the K MTDs are defined as: m ═ M1,…,mk,…,mK}；
A total of J orthogonal subchannels defined as: c ═ C1,…,cj,…,cJ}；
The model adopts a time slot model, the time slot model divides the whole optimization time into T time slots with the length of tau, and all the time slots are defined as: t {1, …, T, …, T };
in the time slot model, the information of the channel remains unchanged in each time slot; the information of the channel changes between different time slots;
in each time slot, each MTD independently determines the channel selection strategy; specifically, each MTD may face J +1 choices, i.e., selecting one of J subchannels or staying dormant.
Further, the process of refining the task or data transmission model in step S2 includes:
each MTD generates new data in each slot, which is temporarily stored in memory local to the MTD, thus forming a data queue at each MTD:
Qk(t+1)＝max{Qk(t)-Uk(t)，0}+γkAk(t)+Yk(t)
wherein Qk(t) is the current time slot device mkAmount of data stored on memory, Uk(t) amount of data offloaded to the edge server for the current timeslot, Ak(t) and γkAre respectively a device mkThe rate at which data is generated and the average size of the generated data, the product of the two terms representing the amount of newly generated data, Yk(t) represents the amount of data that needs to be retransmitted due to the presence of errors;
each time slot MTD can unload certain data to the edge server, the unloading rate of the data is determined by the channel selected by the current time slot of the MTD, and the signal-to-noise ratio and the transmission rate of the information are obtained according to the selected channel j:
the signal-to-noise ratio is:
Figure GDA0002631913350000031
the data transmission rate is:
Figure GDA0002631913350000032
wherein2Is the noise power, PTXIs the transmission power, Hk,j,tIndicates that at the t-th time slot, device mkAnd a sub-channel c between the base stationsjChannel gain of, BjRepresenting the transmission bandwidth, thereby obtaining a device mkThroughput of (2):
Figure GDA0002631913350000033
after the throughput is known, the current time slot device m is obtainedkAmount of data transferred to edge server:
Figure GDA0002631913350000034
wherein xk,j,tRecording the channel selection condition of each device in each time slot when xk,j,tWhen 1, it stands for device mkChannel j is selected within time slot t;
acquiring the transmission error rate according to the signal-to-noise ratio:
Figure GDA0002631913350000035
obtaining the data volume needing to be retransmitted according to the obtained error rate:
Figure GDA0002631913350000036
further, the process of refining the energy consumption model in step S2 includes:
device mkThe energy consumed for selecting the channel j to transmit data in the tth time slot is as follows:
Figure GDA0002631913350000041
wherein P isTXWhich is representative of the power of the transmission,
Figure GDA0002631913350000046
represents the total amount of data to be transmitted divided by the data transmission rate, i.e., the time it takes to transmit all the data; wherein the total amount of data to be transmitted is the device mkThe amount of data stored on the memory;
Figure GDA0002631913350000047
the length of one time slot is expressed, namely, data transmission can be carried out in only one time slot; therefore, it is not only easy to use
Figure GDA0002631913350000048
Representing the actual transmission time; pTXAnd
Figure GDA0002631913350000049
the multiplication represents the energy consumed by transmission; if the J +1 th channel is selected, it represents the device mkMaintaining a sleep state at a current time slot, the sleep state not transmitting data, so energy consumption is 0;
obtaining a device m based on a restriction of battery capacitykLong-term energy constraint:
Figure GDA0002631913350000042
wherein Ek,maxRepresentative apparatus mkThe capacity of the battery is constrained.
The process of refining and obtaining the time delay model in step S2 includes:
dividing the time delay into two parts on the whole, wherein the first part is transmission time delay which is time consumed in data transmission; the second part is to calculate the time delay, which is the time consumed in data processing, and the details are as follows:
Figure GDA0002631913350000043
the concrete model of the transmission delay is as follows:
Figure GDA0002631913350000044
when the equipment mkWhen the mobile terminal is in a dormant state, data transmission is not carried out, so that the transmission delay is infinite;
the specific model for calculating the time delay is as follows:
Figure GDA0002631913350000045
wherein λ isk,tPresentation apparatus mkThe calculation intensity of the transmitted data is in the unit of each bit of the CPU period, and the calculated equipment m is obtained according to the calculation intensitykThe transmitted data has a total requirement of zk,j,tλk,tCPU cycle; xik,tRepresentative apparatus mkAvailable computing resources; when the equipment mkIn the sleep state, there is no data transmission, and from the viewpoint of service reliability, there is no data transmission that meets the requirement of the service, so the calculation delay when the device is in sleep is defined as infinity here.
The process of refining and obtaining the service reliability model in step S2 includes:
assuming a delay requirement of d, if this indicates a task offload failure, then device m is available in T slotskNumber of successful task unloads:
Figure GDA0002631913350000051
wherein II { x } represents: if x is true, then II { x } ═ 1, whereas if x is false, then II (x) ═ 0; (ii) a
Based on the successful uninstallation times, the service reliability is defined as:
Figure GDA0002631913350000052
wherein eta iskFor the proposed service reliability requirement, the device data is successfully unloaded during the whole period of time TThe ratio should be greater than ηk；
Further, the optimization problem constructed in the step S3 is P1:
Figure GDA0002631913350000053
Figure GDA0002631913350000054
Figure GDA0002631913350000055
Figure GDA0002631913350000056
Figure GDA0002631913350000057
wherein the optimized variable is the long-term throughput of all devices; constraint C1And C2Representing a constraint on channel selection that requires that each device be able to select only one channel at a time, and that a channel be selected by only one device at a time; constraint C2And C3Energy constraints and service reliability constraints.
Further, the step S4 includes:
firstly, based on the theory of virtual queue, the above C is2And C3Two constraints are converted into a virtual queue:
Figure GDA0002631913350000061
Figure GDA0002631913350000062
these two queues represent the energy constraint and the service reliability constraint, respectively;
then, based on the lyapunov optimization theory and in combination with the established virtual queue, converting the long-term energy efficiency optimization problem into a problem of maximizing energy efficiency and service reliability and minimizing energy consumption in each time slot to obtain an optimization problem P2:
Figure GDA0002631913350000063
s.t.C1～C2.
i.e. C in the optimization problem P12And C3The constraints are transformed into optimization objectives, and perception of energy consumption and service reliability is achieved.
Further, the step S5 includes:
the first step is as follows: for any equipment mkIt is connected to channel cjBelong to the group C ═ C1,…,cj,…,cJTemporarily matching, and observing the throughput, energy consumption and time delay performance of the system; the apparatus mkBelong to M ═ M1,…,mk,…,mK}；
The second step is that: first of all, device mkEstimating its preference degree for the jth sub-channel according to the following formula and constructing its preference list
Figure GDA0002631913350000064
Figure GDA0002631913350000065
Figure GDA0002631913350000066
Figure GDA0002631913350000067
Wherein
Figure GDA0002631913350000071
Indicating that as of the current time slot t,optimizing variable thetak,j,tAverage value of (d);
Figure GDA0002631913350000072
presentation apparatus mkSelecting channel c as of time slot t-1jThe number of times of (c); rhok,jPresentation apparatus mkSelecting channel cjThe cost to be paid;
then the device mkTransmitting the preference list to an edge server, and then matching all MTDs with channels based on a pricing matching theory, wherein each MTD selects a channel of the MTD;
the third step: solving an optimization problem P2 according to the channel selected in the second step;
the fourth step: let t be t +1, and calculate and update according to steps S1-S5 to obtain a new value
Figure GDA0002631913350000073
Uk(t)，Yk(t)，Qk(t+1),Nk(t +1) and Fk(t +1), performing channel optimization again; by analogy, the value of t is added with 1 every time, and continuous learning is carried out, so that the optimal channel selection is realized, and the maximum throughput is realized.
The invention has the beneficial effects that: the invention applies machine learning to channel selection, so that the optimal decision of channel selection can be completed without global information, and the cost is lower; the optimization on channel selection is realized through learning and matching, and further the maximization of throughput is realized; based on the MAB theory, the lyapunov optimization theory and the matching theory, energy perception and service reliability perception are combined with machine learning, and therefore the maximum and the optimum of energy utilization rate and service reliability are achieved.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a schematic diagram of a system model constructed according to the present invention.
Detailed Description
The technical solutions of the present invention are further described in detail below with reference to the accompanying drawings, but the scope of the present invention is not limited to the following.
As shown in fig. 1, a ubiquitous power internet of things access method based on matching learning includes the following steps:
s1, constructing a system model:
as shown in fig. 2: a base station and an edge server are contained in a base station service cell and are distributed closely. Wherein, the base station provides connection service for K MTDs in the cell, and the edge server provides calculation service. Wherein the K MTDs are defined as: m ═ M1,…,mk,…,mK}；
A total of J orthogonal subchannels defined as: c ═ C1,…,cj,…,cJ}
Here, a slot model is used, which divides the entire optimization time into T slots of length τ, and defines all slots as: t ═ 1, …, T, …, T }
In the time slot model, the information of the channel remains unchanged in each time slot; and the information of the channel changes between different time slots.
In each time slot, each MTD independently determines the channel selection strategy; specifically, each MTD is faced with J +1 choices, i.e., selecting one of J subchannels or staying dormant (idle, not transmitting data)
S2, refining the model to obtain a task or data transmission model, an energy consumption model, a time delay model and a service reliability model;
A. task or data transfer model
Each MTD generates new data in each slot, which is temporarily stored in memory local to the MTD, thus forming a data queue at each MTD:
Qk(t+1)＝max{Qk(t)-Uk(t)，0}+γkAk(t)+Yk(t)
wherein Qk(t) is the current time slot device mkAmount of data stored on memory, Uk(t) offloading to edge for current timeslotAmount of data on the Server, Ak(t) and γkAre respectively a device mkThe rate at which data is generated and the average size of the generated data, the product of the two terms representing the amount of newly generated data, Yk(t) represents the amount of data that needs to be retransmitted due to the presence of errors.
Each time slot MTD can unload certain data to the edge server, the unloading rate (transmission rate) of the data is determined by the channel selected by the current time slot of the MTD, and the signal-to-noise ratio and the transmission rate of the information are obtained according to the selected channel j:
the signal-to-noise ratio is:
Figure GDA0002631913350000081
the data transmission rate is:
Figure GDA0002631913350000082
wherein2Is the noise power, PTXIs the transmission power, Hk,j,tIndicates that at the t-th time slot, device mkAnd a sub-channel c between the base stationsjChannel gain of, BjRepresenting the transmission bandwidth, thereby obtaining a device mkThroughput of (2):
Figure GDA0002631913350000083
after the throughput is known, the current time slot device m can be obtainedkAmount of data transferred to edge server:
Figure GDA0002631913350000084
wherein xk,j,tRecording the channel selection condition of each device in each time slot when xk,j,tWhen 1, it stands for device mkChannel j is selected within time slot t.
Meanwhile, the bit error rate of transmission can be obtained according to the signal-to-noise ratio:
Figure GDA0002631913350000091
the BPSK modulation method is taken as an example, and corresponding formulas are different for different modulation methods.
And obtaining the data volume needing to be retransmitted according to the obtained error rate:
Figure GDA0002631913350000092
B. energy consumption model
Device mkThe energy consumed for selecting the channel j to transmit data in the tth time slot is as follows:
Figure GDA0002631913350000093
wherein P isTXWhich is representative of the power of the transmission,
Figure GDA0002631913350000096
represents the total amount of data to be transmitted divided by the data transmission rate, i.e., the time it takes to transmit all the data; wherein the total amount of data to be transmitted is the device mkThe amount of data stored on the memory;
Figure GDA0002631913350000097
the length of one time slot is expressed, namely, data transmission can be carried out in only one time slot; therefore, it is not only easy to use
Figure GDA0002631913350000098
Representing the actual transmission time; pTXAnd
Figure GDA0002631913350000099
the multiplication represents the energy consumed by transmission; if the J +1 th channel is selected, it represents the device mkMaintaining a sleep state in which no number is transmitted in a current time slotAccordingly, the energy consumption is 0;
on the former basis, we can get a device m due to the limited capacity of the batterykLong-term energy constraint:
Figure GDA0002631913350000094
wherein Ek,maxRepresentative apparatus mkCapacity constraint of battery
C. Time delay model
The time delay is divided into two parts as a whole, wherein the first part is transmission time delay which is time consumed in data transmission; the second part is the calculated latency, which is the time consumed in data processing. The method comprises the following specific steps:
Figure GDA0002631913350000095
the concrete model of the transmission delay is as follows:
Figure GDA0002631913350000101
when the equipment mkWhen the mobile terminal is in the dormant state, data transmission is not performed, so that the transmission delay is infinite.
The specific model for calculating the time delay is as follows:
Figure GDA0002631913350000102
wherein λ isk,tPresentation apparatus mkThe calculation intensity of the transmitted data is in the unit of each bit of the CPU cycle, so that the calculated device m can be obtainedkThe transmitted data has a total requirement of zk,j,tλk,tCPU cycle; xik,tRepresentative apparatus mkAvailable computing resources; when the equipment mkWhen the mobile terminal is in the dormant state, no data is transmitted, and the requirement of meeting the service is not met when no data is transmitted from the viewpoint of service reliability, so the mobile terminal is in the dormant stateThe calculated delay when the device is asleep is defined as infinity.
D. Service reliability model
Assuming a delay requirement of d, if this indicates a task offload failure, then device m is available in T slotskNumber of successful task unloads:
Figure GDA0002631913350000103
wherein II { x } represents: if x is true, then II { x } ═ 1, whereas if x is false, then II (x) ═ 0;
based on the successful uninstalling times, the service reliability is defined as:
Figure GDA0002631913350000104
wherein eta iskFor the proposed service reliability requirements. The proportion of successful unloading of the device data should be greater than eta for the entire period of time Tk；
S3, maximizing long-term throughput, and determining an optimization problem P1:
Figure GDA0002631913350000111
Figure GDA0002631913350000112
Figure GDA0002631913350000113
Figure GDA0002631913350000114
Figure GDA0002631913350000115
wherein the optimized variable isLong term throughput of all devices; constraint C1And C2 represents the constraints on channel selection that require that each device be able to select only one channel at a time, and that a channel be selected by only one device at a time; constraint C2And C3The previously mentioned energy constraint and service reliability constraint. The energy constraint and service reliability constraint require that the energy consumed does not exceed the limits of the energy constraint and that the reliability of the service meets the requirements set forth;
s4, converting the optimization problem based on the theory of the virtual queue and the lyapunov optimization theory:
first, based on the theory of virtual queues, we will refer to C above2And C3Two constraints are converted into a virtual queue:
Figure GDA0002631913350000116
Figure GDA0002631913350000117
these two queues represent the energy constraints and service reliability constraints previously proposed, respectively.
Then, based on the lyapunov optimization theory, in combination with the virtual queues established previously, we transform the long-term energy efficiency optimization problem into a problem of maximizing energy efficiency and service reliability and minimizing energy consumption at each time slot (i.e., transform one long-term optimization problem into a plurality of short-term optimization problems, that is, implement optimization at each time slot to implement optimization at a long time). We therefore get the following problem P2 to be solved:
Figure GDA0002631913350000118
s.t.C1～C2.
it can be seen that through the theory of virtual queue correlation, we solve problem C in P12And C3Constraint is transformed into ourAmong the optimization objectives, a perception of energy consumption and service reliability is thus achieved.
S5, realizing the optimal channel selection through learning and matching, and further realizing the maximization of throughput:
the step S5 includes:
the first step is as follows: for any equipment mkIt is connected to channel cjBelong to the group C ═ C1,…,cj,…,cJTemporarily matching, and observing the throughput, energy consumption and time delay performance of the system; the apparatus mkBelong to M ═ M1,…,mk,…,mK}；
The second step is that: first of all, device mkEstimating its preference degree for the jth sub-channel according to the following formula and constructing its preference list
Figure GDA0002631913350000121
Figure GDA0002631913350000122
Figure GDA0002631913350000123
Figure GDA0002631913350000124
Wherein
Figure GDA0002631913350000125
Indicating the optimization variable theta until the current time slot tk,j,tAverage value of (d);
Figure GDA0002631913350000126
presentation apparatus mkSelecting channel c as of time slot t-1jThe number of times of (c); rhok,jPresentation apparatus mkSelecting channel cjThe cost to be paid;
then the device mkTransmitting the preference list to an edge serviceAll the MTDs are matched with the channels based on a pricing matching theory, and each MTD selects the channel of the MTD;
the third step: solving an optimization problem P2 according to the channel selected in the second step;
the fourth step: let t be t +1, and calculate and update according to steps S1-S5 to obtain a new value
Figure GDA0002631913350000127
Uk(t)，Yk(t)，Qk(t+1),Nk(t +1) and Fk(t +1), performing channel optimization again;
by analogy, the value of t is added with 1 every time, and continuous learning is carried out, so that the optimal channel selection is realized, and the maximum throughput is realized.
The ubiquitous power Internet of things access method further comprises the step of ubiquitous power Internet of things access: after the channel selection is optimized in step S5, the ubiquitous power internet of things is accessed through the channel, and data transmission is completed.
In conclusion, the learning algorithm is used for channel selection, service reliability sensing, energy sensing and context sensing are integrated, each device can dynamically select a channel for data transmission, and the optimal overall network throughput can be achieved without global information; the MAB problem, the lyapunov optimization theory and the matching theory are combined with machine learning, so that the effects of energy perception and service reliability perception are achieved, and the network energy utilization rate and the service reliability are improved.
The foregoing is a preferred embodiment of the present invention, it is to be understood that the invention is not limited to the form disclosed herein, but is not to be construed as excluding other embodiments, and is capable of other combinations, modifications, and environments and is capable of changes within the scope of the inventive concept as expressed herein, commensurate with the above teachings, or the skill or knowledge of the relevant art. And that modifications and variations may be effected by those skilled in the art without departing from the spirit and scope of the invention as defined by the appended claims.

Claims (2)

1. A ubiquitous power Internet of things access method based on matching learning is characterized in that: the method comprises the following steps:
s1, constructing a system model;
the system model constructed in step S1 includes:
setting a base station service cell which comprises a base station and an edge server which are distributed in a similar way, wherein the base station provides connection service for K Machine Type Devices (MTD) in the cell, and the edge server provides calculation service; wherein the K MTDs are defined as: m ═ M1,…,mk,…,mK}；
A total of J orthogonal subchannels defined as: c ═ C1,…,cj,…,cJ}；
The model adopts a time slot model, the time slot model divides the whole optimization time into T time slots with the length of tau, and all the time slots are defined as: t {1, …, T, …, T };
in the time slot model, the information of the channel remains unchanged in each time slot; the information of the channel changes between different time slots;
in each time slot, each MTD independently determines the channel selection strategy; each MTD faces J +1 choices, i.e., selects one of the J subchannels or stays dormant;
s2, refining the model to obtain a task or data transmission model, an energy consumption model, a time delay model and a service reliability model;
the process of refining and obtaining the task or data transmission model in the step S2 includes:
each MTD generates new data in each slot, which is temporarily stored in memory local to the MTD, thus forming a data queue at each MTD:
Qk(t+1)＝max{Qk(t)-Uk(t)，0}+γkAk(t)+Yk(t)
wherein Qk(t) is the current time slot device mkAmount of data stored on memory, Uk(t) is whenAmount of data that is pre-slotted to be offloaded to an edge server, Ak(t) and γkAre respectively a device mkThe rate at which data is generated and the average size of the generated data, the product of the two terms representing the amount of newly generated data, Yk(t) represents the amount of data that needs to be retransmitted due to the presence of errors;
each time slot MTD can unload certain data to the edge server, the unloading rate of the data is determined by the channel selected by the current time slot of the MTD, and the signal-to-noise ratio and the transmission rate of the information are obtained according to the selected channel j:
the signal-to-noise ratio is:
Figure FDA0002661587260000011
the data transmission rate is:
Figure FDA0002661587260000021
wherein2Is the noise power, PTXIs the transmission power, Hk,j,tIndicates that at the t-th time slot, device mkAnd a sub-channel c between the base stationsjChannel gain of, BjRepresenting the transmission bandwidth, thereby obtaining a device mkThroughput of (2):
Figure FDA0002661587260000022
after the throughput is known, the current time slot device m is obtainedkAmount of data transferred to edge server:
Figure FDA0002661587260000023
wherein xk,j,tRecording the channel selection condition of each MTD in each time slot when xk,j,tWhen 1, it stands for device mkChannel j is selected within time slot t;
acquiring the transmission error rate according to the signal-to-noise ratio:
Figure FDA0002661587260000024
obtaining the data volume needing to be retransmitted according to the obtained error rate:
Figure FDA0002661587260000025
the process of refining and obtaining the energy consumption model in the step S2 includes:
device mkThe energy consumed for selecting the channel j to transmit data in the tth time slot is as follows:
Figure FDA0002661587260000026
wherein P isTXWhich is representative of the power of the transmission,
Figure FDA0002661587260000027
represents the total amount of data to be transmitted divided by the data transmission rate, i.e., the time it takes to transmit all the data; wherein the total amount of data to be transmitted is the device mkThe amount of data stored on the memory;
Figure FDA0002661587260000028
the length of one time slot is expressed, namely, data transmission can be carried out in only one time slot; therefore, it is not only easy to use
Figure FDA0002661587260000029
Representing the actual transmission time; pTXAnd
Figure FDA00026615872600000210
the multiplication represents the energy consumed by transmission; if the J +1 th channel is selected, it represents the device mkA sleep state is maintained in the current slot, the sleep state does not transmit data, so the energy consumption is0；
Obtaining a device m based on a restriction of battery capacitykLong-term energy constraint:
Figure FDA0002661587260000031
wherein Ek,maxRepresentative apparatus mkCapacity constraints of the battery;
the process of refining and obtaining the time delay model in step S2 includes:
dividing the time delay into two parts on the whole, wherein the first part is transmission time delay which is time consumed in data transmission; the second part is to calculate the time delay, which is the time consumed in data processing, and the details are as follows:
Figure FDA0002661587260000032
the concrete model of the transmission delay is as follows:
Figure FDA0002661587260000033
when the equipment mkWhen the mobile terminal is in a dormant state, data transmission is not carried out, so that the transmission delay is infinite;
the specific model for calculating the time delay is as follows:
Figure FDA0002661587260000034
wherein λ isk,tPresentation apparatus mkThe calculation intensity of the transmitted data is in the unit of each bit of the CPU period, and the calculated equipment m is obtained according to the calculation intensitykThe transmitted data has a total requirement of zk,j,tλk,tCPU cycle; xik,tRepresentative apparatus mkAvailable computing resources; when the equipment mkWhen the mobile terminal is in the dormant state, no data is transmitted, and the requirement for meeting the service is not met when no data is transmitted from the viewpoint of service reliabilityTherefore, the computation delay when the device is sleeping is defined herein as infinity;
the process of refining and obtaining the service reliability model in step S2 includes:
assume a delay requirement of dk,tIf, if
Figure FDA0002661587260000035
Indicating a task offload failure, device m may be obtained in T slotskNumber of successful task unloads:
Figure FDA0002661587260000036
wherein II { x } represents: if x is true, then II { x } ═ 1, whereas if x is false, then II (x) ═ 0;
based on the successful uninstallation times, the service reliability is defined as:
Figure FDA0002661587260000041
wherein eta iskFor the proposed service reliability requirement, the proportion of successful device data offload should be greater than η for the entire period of time Tk；
S3, maximizing long-term throughput and determining an optimization problem;
the optimization problem constructed in the step S3 is P1:
Figure FDA0002661587260000042
s.t.
Figure FDA0002661587260000043
Figure FDA0002661587260000044
Figure FDA0002661587260000045
Figure FDA0002661587260000046
wherein the optimized variable is the long-term throughput of all devices; constraint C1And C2Representing a channel selection constraint that requires that each MTD be able to select only one channel at a time, and that a channel be selected by only one device at a time; constraint C2And C3The energy consumed by the energy constraint and the service reliability constraint requirements cannot exceed the limit of the energy constraint, and the service reliability is required to meet the proposed requirements;
s4, converting the optimization problem based on the virtual queue theory and the lyapunov optimization theory;
the step S4 includes:
firstly, based on the theory of virtual queue, the above C is2And C3Two constraints translate into two virtual queues:
Figure FDA0002661587260000047
Figure FDA0002661587260000048
these two queues represent the energy constraint and the service reliability constraint, respectively;
then, based on the lyapunov optimization theory and in combination with the established virtual queue, converting the long-term energy efficiency optimization problem into a problem of maximizing energy efficiency and service reliability and minimizing energy consumption in each time slot to obtain an optimization problem P2:
Figure FDA0002661587260000051
wherein theta isk,j,tTo optimize the variables, it is a weighted sum of the latter three terms; vk，αk，βkThe weights of the three terms are respectively used for adjusting the latter three terms to be the same order of magnitude; n is a radical ofk(t)，Fk(t) virtual queues representing energy constraints and service reliability constraints, respectively; p2 will optimize C in problem P12And C3Two constraints are converted into an optimization target, and the perception of energy consumption and service reliability is realized;
s5, realizing the optimal channel selection through learning and matching, and further realizing the maximization of throughput;
the step S5 includes:
the first step is as follows: for any equipment mkIt is connected to channel cjTemporarily matching, and observing the throughput, energy consumption and time delay performance of the system; the apparatus mkBelong to M ═ M1,…,mk,…,mKC, the channel cjBelong to the group C ═ C1,…,cj,…,cJ}；
The second step is that: first of all, device mkEstimating its preference degree for the jth sub-channel according to the following formula and constructing its preference list
Figure FDA0002661587260000052
Figure FDA0002661587260000053
Figure FDA0002661587260000054
Figure FDA0002661587260000055
Wherein
Figure FDA0002661587260000056
Indicating the optimization variable theta until the current time slot tk,j,tAverage value of (d);
Figure FDA0002661587260000057
presentation apparatus mkSelecting channel c as of time slot t-1jThe number of times of (c); rhok,jPresentation apparatus mkSelecting channel cjThe cost to be paid;
then the device mkTransmitting the preference list to an edge server, and then matching all MTDs with channels based on a pricing matching theory, wherein each MTD selects a channel of the MTD;
the third step: solving an optimization problem P2 according to the channel selected in the second step;
the fourth step: let t be t +1, and calculate and update according to steps S1-S5 to obtain a new value
Figure FDA0002661587260000061
Uk(t)，Yk(t)，Qk(t+1),Nk(t +1) and Fk(t +1), performing channel optimization again; by analogy, the value of t is added with 1 every time, and continuous learning is carried out, so that the optimal channel selection is realized, and the maximum throughput is realized.
2. The ubiquitous power internet of things access method based on matching learning of claim 1, wherein the ubiquitous power internet of things access method comprises the following steps: the method further comprises the following steps of ubiquitous power Internet of things access: after the channel selection is optimized in step S5, the ubiquitous power internet of things is accessed through the channel, and data transmission is completed.
CN201911406929.3A 2019-12-31 2019-12-31 Ubiquitous power Internet of things access method based on matching learning Active CN111162852B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911406929.3A CN111162852B (en) 2019-12-31 2019-12-31 Ubiquitous power Internet of things access method based on matching learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911406929.3A CN111162852B (en) 2019-12-31 2019-12-31 Ubiquitous power Internet of things access method based on matching learning

Publications (2)

Publication Number Publication Date
CN111162852A CN111162852A (en) 2020-05-15
CN111162852B true CN111162852B (en) 2020-10-27

Family

ID=70559678

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911406929.3A Active CN111162852B (en) 2019-12-31 2019-12-31 Ubiquitous power Internet of things access method based on matching learning

Country Status (1)

Country Link
CN (1) CN111162852B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113162658B (en) * 2021-02-25 2023-05-02 华北电力大学 Task unloading method based on price increasing quota matching in power line communication
CN114375010B (en) * 2021-06-28 2022-10-14 山东华科信息技术有限公司 Power distribution Internet of things system based on SDN and matching theory

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7839830B2 (en) * 2006-04-11 2010-11-23 Nec Laboratories America, Inc. Uplink scheduler for cellular packet access
CN109905864B (en) * 2019-02-28 2020-11-03 华北电力大学 Power Internet of things oriented cross-layer resource allocation scheme
CN109905334B (en) * 2019-03-01 2021-06-08 华北电力大学 Access control and resource allocation method for power Internet of things mass terminal
CN109922160B (en) * 2019-03-28 2021-07-06 全球能源互联网研究院有限公司 Terminal secure access method, device and system based on power Internet of things
CN110189169A (en) * 2019-05-21 2019-08-30 国网北京市电力公司 System based on ubiquitous electric power Internet of Things
CN110444001A (en) * 2019-08-29 2019-11-12 上海戎磐网络科技有限公司 A kind of ubiquitous electric power data transmission of internet of things system

Also Published As

Publication number Publication date
CN111162852A (en) 2020-05-15

Similar Documents

Publication Publication Date Title
CN110928654B (en) Distributed online task unloading scheduling method in edge computing system
CN113242568B (en) Task unloading and resource allocation method in uncertain network environment
CN111182509B (en) Ubiquitous power Internet of things access method based on context-aware learning
CN112004239A (en) Computing unloading method and system based on cloud edge cooperation
CN109600178B (en) Optimization method for energy consumption, time delay and minimization in edge calculation
CN111586696A (en) Resource allocation and unloading decision method based on multi-agent architecture reinforcement learning
US11831708B2 (en) Distributed computation offloading method based on computation-network collaboration in stochastic network
CN112416603B (en) Combined optimization system and method based on fog calculation
CN111556516B (en) Distributed wireless network task cooperative distribution method facing delay and energy efficiency sensitive service
CN111162852B (en) Ubiquitous power Internet of things access method based on matching learning
CN114697333B (en) Edge computing method for energy queue equalization
WO2023179010A1 (en) User packet and resource allocation method and apparatus in noma-mec system
CN109639833A (en) A kind of method for scheduling task based on wireless MAN thin cloud load balancing
CN111556576B (en) Time delay optimization method based on D2D _ MEC system
CN104767833A (en) Cloud transferring method for computing tasks of mobile terminal
Yao et al. Cost-efficient tasks scheduling for smart grid communication network with edge computing system
CN114116061B (en) Workflow task unloading method and system in mobile edge computing environment
CN113271221B (en) Network capacity opening method and system and electronic equipment
CN108650710B (en) Resource allocation method of wireless energy supply communication network based on hybrid multiple access
CN111294886B (en) Mobile edge calculation method and device based on wireless energy drive
CN116302404B (en) Resource decoupling data center-oriented server non-perception calculation scheduling method
CN111580943A (en) Task scheduling method oriented to multi-hop unloading in low-delay edge calculation
CN114615705B (en) Single-user resource allocation strategy method based on 5G network
CN116170844A (en) Digital twin auxiliary task unloading method for industrial Internet of things scene
CN113207150B (en) Active and passive hybrid unloading method and device based on backscatter communication

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