CN108415763A - A kind of distribution method of edge calculations system - Google Patents
A kind of distribution method of edge calculations system Download PDFInfo
- Publication number
- CN108415763A CN108415763A CN201810140988.XA CN201810140988A CN108415763A CN 108415763 A CN108415763 A CN 108415763A CN 201810140988 A CN201810140988 A CN 201810140988A CN 108415763 A CN108415763 A CN 108415763A
- Authority
- CN
- China
- Prior art keywords
- task
- energy consumption
- formula
- submodel
- computing terminal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/48—Program initiating; Program switching, e.g. by interrupt
- G06F9/4806—Task transfer initiation or dispatching
- G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
- G06F9/485—Task life-cycle, e.g. stopping, restarting, resuming execution
- G06F9/4856—Task life-cycle, e.g. stopping, restarting, resuming execution resumption being on a different machine, e.g. task migration, virtual machine migration
- G06F9/4862—Task life-cycle, e.g. stopping, restarting, resuming execution resumption being on a different machine, e.g. task migration, virtual machine migration the task being a mobile agent, i.e. specifically designed to migrate
- G06F9/4875—Task life-cycle, e.g. stopping, restarting, resuming execution resumption being on a different machine, e.g. task migration, virtual machine migration the task being a mobile agent, i.e. specifically designed to migrate with migration policy, e.g. auction, contract negotiation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/48—Program initiating; Program switching, e.g. by interrupt
- G06F9/4806—Task transfer initiation or dispatching
- G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
- G06F9/4881—Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
- G06F9/4893—Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues taking into account power or heat criteria
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0283—Price estimation or determination
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- Development Economics (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a kind of distribution methods of edge calculations system, include the following steps：S1. computing terminal receives the mission bit stream of the task to be allocated of scheduler dispatches；S2. the cost for executing the task and firm offer are calculated according to preset Offer Model, and is offered to task described in scheduler application according to described；S3. the scheduler receives the quotation of each computing terminal, assigns the task to computing terminal by preset Task Assignment Model so that the sum of quotation of each task is optimal.The present invention has the privacy that can protect computing terminal, meets the randomness of task arrival, ensures the authenticity of ROA schemes, to ensure the advantages that system award summation is optimal.
Description
Technical field
The present invention relates to technical field of the computer network more particularly to a kind of distribution methods of edge calculations system.
Background technology
The arrival in Internet of Things (Internet of Things, IoTs) epoch will will produce a large amount of initial data.Pass through
Big data analysis can therefrom obtain very valuable information.It is substantially exceeded however, analyzing the required calculating of these data
The computing capability of IoT equipment.Therefore, it is desirable to cloud computing and service to be expanded to the position of generation data, i.e. network
Edge.
It in order to solve this problem, can be by the way that computation-intensive analysis task to be moved to the movement near IoT equipment
Intelligent cell (Mobile Devices, MDs) is realized.These MDs have used the interconnection of small range alternating-current technique, form one
A new architecture element referred to as the high in the clouds ad-hoc, the framework element get up mobile computing and cloud computing fusion.Total comes
It says, edge calculations provide intelligent Service nearby, disclosure satisfy that the demand of the low reaction delay of Internet of Things application.
Currently there is the research that many technical concerns unload edge calculations calculating task.A kind of edge meter is invented by Hohai University
Calculate the loading commissions migration algorithm (patent No. of optimised power consumption in environment：CN201710043453.6), Beijing University of Technology invents
Indoor wireless positioning method (the patent No. based on edge calculations and Bayes posterior probability model：
CN201610426115.6)；Bid and auction for MDs to calculating task also have numerous studies literature research, such as document
12 (A.L.Jin, W.Song, and W.Zhuang, " Auction-based resource allocation for sharing
Cloudlets in mobile cloud computing, " IEEE Trans.Emerg.Topics Comput., to
Appear, DOI:10.1109/TETC.2015.2487865.) described, the target of auction is exactly maximize triumph MDs whole
Body is worth.
The centralization for needing MDs to send privacy to third party system or other MDs has only been used in existing technology
Optimization method；Or define a static models, it is assumed that the arrival situation of calculating task is known in advance, cannot protect that MDs's is hidden
The randomness that private and calculating task reaches.
Invention content
The technical problem to be solved in the present invention is that：For technical problem of the existing technology, the present invention provides one
Kind can protect the privacy of computing terminal, meet the randomness of task arrival, ensure ROA (Rewards-optimal
Auction, reward optimum auction) scheme authenticity, to ensure the optimal edge calculations system of system award summation point
Method of completing the square.
In order to solve the above technical problems, technical solution proposed by the present invention is：A kind of distribution method of edge calculations system,
Include the following steps：
S1. computing terminal receives the mission bit stream of the task to be allocated of scheduler dispatches；
S2. the cost for executing the task and firm offer are calculated according to preset Offer Model, and according to described
It offers to task described in scheduler application；
S3. the scheduler receives the quotation of each computing terminal, is assigned the task to by preset Task Assignment Model
Computing terminal so that the sum of quotation of each task is optimal.
Further, the mission bit stream include the input data amount of task, output data quantity, CPU calculating cycles number and
Task is rewarded.
Further, the Offer Model includes CPU speed submodel, energy consumption submodel and cost estimation submodel；
The CPU speed submodel, which is used to be calculated according to the CPU calculating cycles number, determines the CPU speed for executing the task
Degree；
The energy consumption submodel is used to be calculated according to the CPU speed, input data amount, output data quantity of the task and execute
The energy consumption of the task；
The cost estimation submodel, which is used to reward estimation according to the CPU speed, energy consumption and task, executes the task
Cost, and according to the cost determine task offer.
Further, it is to meet task execution to require most to calculate determining CPU speed by the CPU speed submodel
Low CPU speed.
Further, the CPU speed submodel meets as shown in formula (1)：
In formula (1),To calculate the CPU speed of determination,It can be completed in a time slot for computing terminal described
The set of the CPU speed of task, is expressed as For preset speed limit, αnAnd βnFor the systematic parameter of predetermined computing terminal CPU.
Further, the energy consumption submodel includes communication energy consumption submodel and execution energy consumption submodel；
The communication energy consumption submodel according to the input data amount, output data quantity for being calculated as executing the task
And it carries out communicating required communication energy consumption；
The execution energy consumption submodel is used to be calculated according to the CPU calculating cycles number and CPU speed and determine described in execution
The execution energy consumption of required by task；
The energy consumption submodel executes the energy consumption of the task according to the communication energy consumption and execution energy consumption calculation.
Further, the communication energy consumption submodel meets as shown in formula (2)：
dn,k(t)=Cn,rxIk(t)+Cn,rxOk(t) (2)
In formula (2), dn,k(t) it is the communication energy consumption being calculated, Cn,rxThe data of a unit are received for computing terminal
Energy consumption, Ik(t) it is the reception data volume of the task, Cn,rxThe energy consumption of the data of a unit, O are sent for computing terminalk(t)
The output data quantity of the task is executed for computing terminal；
The execution energy consumption submodel meets as shown in formula (3)：
zn,k(t)=cn(t)ln,k(t) (3)
In formula (3), zn,k(t) it is the execution energy consumption being calculated, cn(t) it is handled per unit time for the CPU of computing terminal
The energy consumption of process, ln,k(t)=φk(t)/rn(t) it is the time for executing the task and needing, φk(t) it is that the CPU calculates week
Issue, rn(t) it is CPU speed；
It calculates and executes shown in the energy consumption such as formula (4) of the task：
Pn,k(t)=zn(t)+dn,k(t) (4)
In formula (4), Pn,k(t) it is energy consumption, zn,k(t) it is to execute energy consumption, dn,k(t) it is communication energy consumption.
Further, the cost estimation submodel meets shown in formula (5)：
vn,k(t)=Pn,k(t)(En(t)-Λn)-Vwk(t) (5)
In formula (5), vn,k(t) it is the cost for calculating gained, Pn,k(t) it is energy consumption, En(t) it is energy queue, ΛnFor energy
Buffering area, V are preset weights, wk(t) it is the preset task reward.
Further, in the step S3 shown in preset Task Assignment Model such as formula (6),
In formula (6), J (t) is the set of task, Jn,k(t) it is victor's index, Jn,k(t)=1 it indicates to distribute task k
It is executed to terminal n, otherwise is 0, bn,k(t) it is quotation of the terminal n to task k, K (t) is the set of all tasks of t time slots, and N is
Computing terminal number.
Further, further include distribution step S4, scheduler is after assigning the task to computing terminal, and to described
The corresponding remuneration of computing terminal, the remuneration determine according to preset distribution model, the preset distribution mould
Shown in type such as formula (7),
In formula (7), pn,k(t) it is that task k is distributed to the remuneration that computing terminal n is distributed to after computing terminal n,For
Minimum value non-participating in terminal n, being determined by formula (6),To be not involved in scheduling in terminal n and task k
In the case of, the minimum value that formula (6) determines, K (t) is the set of all tasks of t time slots.
Compared with the prior art, the advantages of the present invention are as follows：The present invention is by each computing terminal according to task to be allocated
Information respectively calculates the cost for oneself completing task, and according to the cost firm offer, to scheduler application task, by dispatching
Device to whole computing terminals to the quotation of each task on the basis of, select optimal scheme to carry out task point in global scope
Match so that task allocation plan can ensure ROA (Rewards-optimal in the randomness for meeting task arrival
Auction, reward optimum auction) scheme authenticity on the basis of, ensure system award summation it is optimal；Also, computing terminal
It only needs the quotation of oneself uploading to scheduler, the privacy of can effectively protect computing terminal.
Description of the drawings
Fig. 1 is specific embodiment of the invention flow diagram.
Fig. 2 is the relationship comparison diagram of the system award of three kinds of algorithms of different in the specific embodiment of the invention at any time.
Fig. 3 is the time averaging MD benefits of the specific embodiment of the invention and time averaging system award figure.
Fig. 4 is that specific embodiment of the invention l-G simulation test obtains the sum of system award schematic diagram.
Specific implementation mode
Below in conjunction with Figure of description and specific preferred embodiment, the invention will be further described, but not therefore and
It limits the scope of the invention.
As shown in Figure 1, the distribution method of the edge calculations system of the present embodiment, step are：S1. computing terminal receives scheduling
The mission bit stream for the task to be allocated that device is sent；
S2. the cost for executing the task and firm offer are calculated according to preset Offer Model, and according to described
It offers to task described in scheduler application；S3. the scheduler receives the quotation of each computing terminal, is distributed by preset task
Model assigns the task to computing terminal so that the sum of quotation of each task is optimal.
In the present embodiment, the mission bit stream includes the input data amount, output data quantity, CPU calculating cycles of task
Number and task reward.The Offer Model includes CPU speed submodel, energy consumption submodel and cost estimation submodel；The CPU
Speed submodel, which is used to be calculated according to the CPU calculating cycles number, determines the CPU speed for executing the task；The energy consumption submodule
Type is used to calculate the energy consumption for executing the task according to the CPU speed, input data amount, output data quantity of the task；It is described
Cost estimation submodel, which is used to be rewarded according to the CPU speed, energy consumption and task, estimates the cost for executing the task, and according to
The cost determines that task is offered.
In the present embodiment, it is illustrated by taking the limbic system of scheduler and N number of computing terminal (MD) there are one having as an example,
N number of computing terminal is denoted as respectively：MDn, n={ 1,2 ..., N }.Limbic system is run with the time slot of preset unit-sized, time slot
It is denoted as t ∈ { 1,2 ..., T }.Computing terminal has energy module, and such as solar panel, piezoelectric device can be by environmental energy
Be converted to electric energy be stored in rechargeable battery it is middle for using in the future.Scheduler accumulates calculating task from IoT equipment, and will appoint
Business information is broadcast to neighbouring MDn。MDnTheir bid is simultaneously sent to scheduler by the cost of calculation processing task, and scheduler is determined
The victor of fixed each task, is then sent to mission bit stream the MD of triumphn.Calculating task is to delay-sensitive, and the time limit is not
It can exceed that the length of a time slot, if task is not handled in current time, it is possible to be dropped or by addition
Computing unit handle.In the present embodiment, using scheduler as the auctioneer of task, using computing terminal as the throwing of task
Mark person, the cost of computing terminal calculation processing task are submitted a tender to scheduler, and scheduler is according to the bid of each computing terminal, selection
Middle target computing terminal executes task, and gives winning bidder's remuneration (rewarding) for executing task.
In the present embodiment, consider the calculating task based on batch and stream, it then follows MapReduce and Google
Dataflow programming models.In t-th of time slot, available calculating task constitutes a set, with K (t)=| K (t) | table
Show.In view of MDnGenerally existing, it is assumed that MDnNumber be more than calculating task number, i.e. N>K(t).Victor's index
By Jn,k(t) it indicates, if MDnTask k is won in time slot t, then value is 1, and otherwise value is 0.In a time slot,
Each MD can at most win a task, and each task also only needs a MD to be calculated, that is, meet shown in formula (8)：
In formula (8), Jn,k(t) it is victor's index, K (t) is the set of all tasks of t time slots, and N is computing terminal number.
In the present embodiment, computing terminal scales (Dynamic voltage and by using dynamic electric voltage and frequency
Frequence scaling, DVFC) ability, MDnCPU speed can be adjusted to take set rn={ rn,1,rn,2,…,rn,max}
In one of the available discrete velocity that provides meet formula (9) in the CPU speed of time slot t that is, for computing terminal n：
In formula (9), rn(t) be computing terminal n in the CPU speed of time slot t, rnFor the CPU sets of speeds of computing terminal.
If handling the required CPU calculating cycles number φ of task kk(t) it indicates, then the time of calculating task k can
To be expressed as ln,k(t)=φk(t)/rn(t), and the time of calculating task k is no more than the length of a time slot, that is, meets
Shown in formula (10)：
ln,k(t)≤τ (10)
In formula (10), ln,k(t) CPU time that task k needs is handled for computing terminal n, τ is the size of unit time slot.By
This, it may be determined that computing terminal n can complete the set of the CPU speed of task k in a time slotIt is expressed asMeet As meet
Task k is completed in a time slot executes desired minimum CPU speed.
In the present embodiment, it is to meet task execution requirement to calculate determining CPU speed by the CPU speed submodel
Minimum CPU speed.CPU speed submodels meet as shown in formula (1)：
In formula (1),To calculate the CPU speed of determination,It can be completed in a time slot for computing terminal described
The set of the CPU speed of task, is expressed as For preset speed limit, αnAnd βnFor the systematic parameter of predetermined computing terminal CPU.
In the present embodiment, the object function as shown in formula (11) is set,
In formula (11), f (rn,k(t)) it is object function, ΛnFor the size in the energy snubber area of computing terminal, En(t) it is meter
Calculate the size of the energy queue of terminal, αnAnd βnFor the systematic parameter of predetermined computing terminal CPU, φk(t) it is that processing is appointed
The business required CPU calculating cycles numbers of k, rn(t) it is the CPU speed of computing terminal.
In the present embodiment, it is possible to measure the maximum capacity that buffering area refers to computing terminal battery, energy queue refers to computing terminal
The current remaining capacity of battery, the electricity=maximum capacity for remaining capacity+can supplement.
To formula (10) derivation, and enable it that can be obtained for 0,
It is available
IfJust represent σn,k(t) it is available CPU speed, then CPU speed controls are most
Good solution is There are following three kinds of situations：
If 1,WhenWhen, the energy expenditure monotonic increase of computing terminal, so
If 2,The convex optimum results of CPU speed controls are rn,iOr rn,i+1, because
This,
If 3,Monotone decreasing, so
In the present embodiment, the energy consumption submodel includes communication energy consumption submodel and execution energy consumption submodel；It is described logical
News energy consumption submodel be used to be calculated as executing the task according to the input data amount, output data quantity and needed for being communicated
Communication energy consumption；The execution energy consumption submodel, which is used to be calculated according to the CPU calculating cycles number and CPU speed, determines execution institute
State the execution energy consumption of required by task；The energy consumption submodel is according to the communication energy consumption and executes the energy consumption calculation execution task
Energy consumption.
In the present embodiment, the communication energy consumption submodel meets as shown in formula (2)：
dn,k(t)=Cn,rxIk(t)+Cn,rxOk(t) (2)
In formula (2), dn,k(t) it is the communication energy consumption being calculated, Cn,rxThe data of a unit are received for computing terminal
Energy consumption, Ik(t) it is the reception data volume of the task, Cn,rxThe energy consumption of the data of a unit, O are sent for computing terminalk(t)
The output data quantity of the task is executed for computing terminal；The execution energy consumption submodel meets as shown in formula (3)：
zn,k(t)=cn(t)ln,k(t) (3)
In formula (3), zn,k(t) it is the execution energy consumption being calculated, cn(t) it is handled per unit time for the CPU of computing terminal
The energy consumption of process, ln,k(t)=φk(t)/rn(t) it is the time for executing the task and needing, φk(t) it is that the CPU calculates week
Issue, rn(t) it is CPU speed；It calculates and executes shown in the energy consumption such as formula (4) of the task：
Pn,k(t)=zn(t)+dn,k(t) (4)
In formula (4), Pn,k(t) it is energy consumption, zn,k(t) it is to execute energy consumption, dn,k(t) it is communication energy consumption.
In the present embodiment, the energy consumption c of the CPU of computing terminal processing procedures per unit timen(t) meet formula (12),
cn(t)=anrn(t)3+βn (12)
In formula (12), αnAnd βnFor the systematic parameter of predetermined computing terminal CPU, rn(t) be computing terminal n when
The CPU speed of gap t.
In the present embodiment, if MDnTask k has been won in time slot t, then its benefit by the payment received by it and
The cost structure of reason task.We use pn,k(t) and vn,k(t) payment and the cost that receive are indicated respectively, then MDnBenefit just
It can be indicated with the difference of the two, as shown in formula (13)：
In formula (13), Un(t) it is the MD being calculatednBenefit, pn,k(t) it is that computing terminal completes what task k can be received
Payment, vn,k(t) cost that task k is paid, J are completed by computing terminaln,k(t) it is victor's index.
In the present embodiment, progressive award of the system award summation dependent on the task of all completions, is represented by such as formula
(14) shown in：
In formula (14), R (t) is system award summation, Jn,k(t) it is victor's index, ωk(t) it is the prize of each task
It encourages.
The decision variable of system award summation includes the speed and task victor's index of CPU, with Γ (t)=r (t),
J (t) } indicate this decision variable.As a result, definition reward summation optimization (Sum-of-Rewards Optimization,
SRO) problem, target are exactly to maximize the long-term reward summation of limbic system, which is represented by as shown in formula (15)：
In formula (15), R (t) is system award summation, and IE [] is to calculate mathematic expectaion.
When handling SRO problems, on the one hand, need to keep long-term average reward summation appointing dependent on current and future
Business distribution；On the other hand, the maximum value for summation being rewarded to obtain network can cause contradiction, if because taking centralized approach meeting
Exposure MDnPrivacy, and take decentralized method that can make MDnSubmitted a tender in a manner of maximizing himself benefit rather than
It is submitted a tender in a manner of system award summation.
In the present embodiment, the cost estimation submodel meets shown in formula (5)：
vn,k(t)=Pn,k(t)(En(t)-Λn)-Vwk(t) (5)
In formula (5), vn,k(t) it is the cost for calculating gained, Pn,k(t) it is energy consumption, En(t) it is energy queue, ΛnFor energy
Buffering area, V are preset weights, wk(t) it is the preset task reward.
In the present embodiment, computing terminal collects energy in the energy resource of surrounding, such as solar radiation, human motion
Deng for the battery charging with capacity.Computing terminal MDnEnergy snubber area size be denoted as Λn, the length of energy queue is denoted as
En(t).So, the length of the energy queue of whole computing terminals constitutes set E (t)={ E1(t),E2(t),…,EN(t)}.It is fixed
One random process η of justicen(t), it is used to indicate in time slot t, computing terminal MDnThe ceiling capacity obtained, ηn(t) on existing
Limit ηmax, because of the η in a time slotn(t) it remains unchanged, but when across period, it is possible to change.Computing terminal
More energy can be collected as far as possible from environment to charge to their battery, but the length of energy queue is no more than energy
The size of buffering area.Therefore, in a time slot t, MDnThe energy of collection is represented by：en(t)=min (Λn-En(t),ηn
(t)).The energy that computing terminal is collected is as input, using the energy of computing terminal consumption as output, so that it may to be calculated
Terminal random parameter queue formula as shown in formula (16)：
En(t+1)=En(t)-Pn(t)+en(t) (16)
In formula (16), En(t) and En(t+1) energy queue when being respectively time slot t and t+1, Pn(t) it is in a time slot
The energy that can be consumed, en(t) it is the energy that can be supplemented in a time slot.Since the energy of consumption is no more than available energy
Amount, that is, meet Pn(t)≤En(t)。
In the present embodiment, by following process to formula (5) into line justification.The cost estimate of calculating task can influence by
MDnThe reward summation of acquisition, and MDnNeed the benefit of comprehensive assessment processing task.The present embodiment is optimized by Liapunov
Technology designs the cost function of calculating task.It defines liapunov function and weighs whole computing terminal MDnEnergy possess
Amount is as shown in formula (17)：
In formula (17), L (t) is the energy ownership being calculated,The electricity that can be supplemented for computing terminal.Wherein,ΛnFor the size in the energy snubber area of computing terminal, En(t) it is the big of the energy queue of computing terminal
It is small.
It can be released by formula (13), if L (t) is smaller, show that the energy of battery storage is more, closer to full electric shape
State.It is measured shown in the variation such as formula (18) of liapunov function when across period by introducing Liapunov offset：
Δ (t)=IE [L (t+1)-L (t) | E (t)] (18)
In formula (18), Δ (t) is the Liapunov deviant being calculated, and IE [] is to calculate mathematic expectaion, L (t) and L
(t+1) be respectively time slot t and t+1 energy ownership, En(t) it is the size of the energy queue of computing terminal.
In the present embodiment, in order to maximize reward summation, R (t) weightings in formula (14) are merged into Δ (t), to obtain
Decreasing effect function (drift-minus-reward) function must be deviated, as shown in formula (19)：
Δ ν (t)=Δ (t)-VR (t) (19)
In formula (19), Δ (t) is Liapunov deviant, and V is preset weights, and R (t) is system award summation.Add
Weights V indicates R (t) proportion in formula (19), and V is bigger, illustrates that the ratio in formula shared by R (t) is bigger, and system is to executing
The reward that task can obtain is more sensitive, becomes relative insensitivity to the stabilization of energy queue.Δ (t) is smaller, illustrates energy
The variation of former and later two time slots of queue is smaller, and energy queue is more stable.- VR (t) is small, illustrate that VR (t) is big or V is very big or
R (t) is very big, or both very big.Therefore, maximized system award summation R (t) in order to obtain, it is necessary to minimize Δ
(t).As long as minimizing Δ ν (t), it will be able to while ensureing energy string stability so that R (t) is maximum.
In the present embodiment, formula (19) is a quadratic function equation about two variables, is calculated to simplify
Journey is derived as follows：
The definition of each parameter is same as described above in formula (20).
Formula (16) both sides square are obtained into formula (21),
The definition of each parameter is same as described above in formula (21).
Formula (21) is updated to formula (20), formula (22) can be obtained：
The definition of each parameter is same as described above in formula (22).
Formula (22) is updated in formula (19), formula (23) is obtained：
In formula (23),PmaxTo execute the maximum energy consumption of task, ηmaxFor in the present embodiment
Defined random process ηn(t) definition of maximum value, remaining parameter is same as described above.
In the present embodiment, the maximum energy consumption P of task is executedmaxIt can be calculated from historic task, such as some time slot
To four tasks s1, s2, s3, s4, the information provided by task, we can calculate the energy consumption of each task, p1, p2,
P3, p4, wherein maximum is Pmax。
Formula (23) is exactly the linear equation after simplifying, and in order to maximize reward summation, needs to minimize inequality (23)
The right.The right and the simplification that formula (14) is substituted into formula (23), can be obtained formula (24)：
In formula (24), the definition of each parameter is same as described above.
It therefore deduces that, the right for minimizing inequality (23) namely minimizes in formula (24)By it with linear structure, it can be deduced that computing terminal MDnProcessing is appointed
The cost of business k, as shown in formula (5).Cost function shown in formula (5) contains two parts, and a part is related to energy, i.e.,
Pn,k(t)(En(t)-Λn)；Another part is related to weight reward, i.e. Vwk(t).With MDnThe energy expenditure of processing task k
Increase, cost function is also increasing, because the increase of energy expenditure can make MDnThe probability for handling Future direction reduces.By (En
(t)-Λn) regard P asn,k(t) weight can also determine that V is bigger, and R (t) is bigger, and system award summation is bigger.
In the present embodiment, in the step S3 shown in preset Task Assignment Model such as formula (6),
In formula (6), J (t) is the set of task, Jn,k(t) it is victor's index, Jn,k(t)=1 it indicates to distribute task k
It is executed to terminal n, otherwise is 0, bn,k(t) it is quotation of the terminal n to task k, K (t) is the set of all tasks of t time slots, and N is
Computing terminal number.Specific can be used calculates such as the methods of Hungary Algorithm to handle the distribution of Task Assignment Model.Task point
With model using formula (8) as constraints, i.e., in a time slot, each MD can at most win a task, each task
Also a MD is only needed to be calculated.
In the present embodiment, further include distribution step S4, scheduler is given after assigning the task to computing terminal
The corresponding remuneration of computing terminal, the remuneration determine that the preset remuneration divides according to preset distribution model
With shown in model such as formula (7),
In formula (7), pn,k(t) it is that task k is distributed to the remuneration that computing terminal n is distributed to after computing terminal n,For
Minimum value non-participating in terminal n, being determined by formula (6),To be not involved in scheduling in terminal n and task k
In the case of, the minimum value that formula (6) determines, K (t) is the set of all tasks of t time slots.
In the present embodiment, by specific emulation experiment to the method (SRO optimization algorithms) and the prior art of the present invention
In energy consumption optimal algorithm and greedy algorithm carried out contrast test verification, be provided with 10 computing terminals (MD), all MD tool
There is identical systematic parameter：αn=0.34, βn=0.35 (identical as Google Nexus), the energy consumption for sending unit data are h
(t)×308.6×10-9W/bit, in order to ensure that constant transmission rate, h (t) change with the change of channel conditions, h (t) ∈
[0.5,2.0].The energy consumption for receiving unit data is：200×10-9The CPU adjustable-speeds ranging from r of W/bit, MDn=0.1,
0.2,0.4,0.8,1,1.44}GHz。
Task handled by system is a series of task of homogeneys, and input data amount magnitude range is：2000Kb,
4000Kb, 6000Kb, 8000Kb }.All tasks are divided into two classes：Data processing task and decision generic task.Data processing class is appointed
The computation complexity of business is 1000cycles/bit, and output data is the half of input data, Decision Classes task computation complexity
For 3000cycles/bit, output data size is 500Kb.In addition, the length of each time slot τ is 60s, energy harvesting (EH)
Speed is 20mW.
By l-G simulation test, the system award relationship at any time that three kinds of algorithms obtain is as shown in Figure 2, it is found that is
It unites starting stage of operation, system award rapid increase over time then settles out again.It is obtained by SRO algorithms
The system award obtained is higher than the system award by energy consumption optimal algorithm and greedy algorithm acquisition.This is because energetic optimum algorithm
Execution task energy consumption is only considered, the task of part high energy consumption cannot may execute always, and greedy algorithm only considered
The reward that current task can obtain is executed, this makes the premature energy for running out of itself of terminal, even if occurring prize in the future
The higher task of value is encouraged, due to the limitation of energy, terminal can not also execute the task.And the cost valuation functions of SRO algorithms are comprehensive
It closes and considers the currently available battery capacity of terminal, executes the energy expenditure of task and execute the reward that task can obtain.
In the present embodiment, different value, the maximum of computing terminal are taken respectively in preset weights V and computing terminal quantity N
The upper limit η of energymaxTake when different value to the present invention SRO algorithms carry out simulation calculation respectively, obtain in varied situations when
Between average MD benefits and time averaging system award figure it is as shown in Figure 3.In Fig. 3 (a), ordinate indicate system award when
Between average (2000 time slots), by taking the point of V=0.8 in black lines as an example, which indicates in V=0.8, the item of N=5
Under part, the average value for the system award that 5 MD terminals obtain in 2000 time slots.In Fig. 3 (b), ordinate indicates terminal effect
The time of benefit is average (2000 time slots), and by taking the point of V=0.8 in black lines as an example, which indicates in V=0.8, most
In the case that big energy harvesting (EH) rate is 20, the time of the sum of all MD terminals (N=10) benefits is average.Pass through figure 3 above
(a) it can be found that V is when between 0.2~2.0, system award can increase as the quantity of MD increases, because the quantity of MD is got over
It is more, it is meant that more tasks can be executed by system, so as to obtain more system awards.In order to verify SRO algorithms
Validity, we have taken a very big V (V=100), by Fig. 3 (a) it can be found that theoretical optimal system award
The gap of (corresponding three lines in Fig. 3 (a)) is about between (three lines in Fig. 3 (a)) obtains system award with SRO algorithms
It is very close between the system award that 1.2% to 2.4%, SRO algorithm obtain and theoretially optimum value.It can be sent out by Fig. 3 (b)
Existing, energy harvesting rate is faster, and the benefit of MD is also bigger, this is because MD, which can possess more energy, goes execution task.This
Outside, by comparing two figures (N=10) it can be found that system award is more than the benefit of MD, this is also meaned that, scheduler can be
Profit among fetching portion between task producer and terminal.
In the present embodiment, in the case of preset weights V differences, l-G simulation test is carried out to the SRO algorithms of the present invention
It is as shown in Figure 4 to obtain the sum of system award.Ordinate indicates the sum of the system award of all terminals in figure, and each point indicates working as
Under preceding time slot, the sum of the system award of all MD acquisitions.It can be found that in the starting stage of system operation, due to final energy
Limitation, can not executing any task or being able to carry out for task is seldom, and system award maintains a very low level.Through
After a period of time, terminal has collected enough energy, is able to carry out more tasks, and system award is consequently increased.Most
Eventually, system award reaches stable.In the case of V biggers, system award can also reach the value of a bigger.
Above-mentioned only presently preferred embodiments of the present invention, is not intended to limit the present invention in any form.Although of the invention
Disclosed above with preferred embodiment, however, it is not intended to limit the invention.Therefore, every without departing from technical solution of the present invention
Content, technical spirit any simple modifications, equivalents, and modifications made to the above embodiment, should all fall according to the present invention
In the range of technical solution of the present invention protection.
Claims (10)
1. a kind of distribution method of edge calculations system, which is characterized in that include the following steps：
S1. computing terminal receives the mission bit stream of the task to be allocated of scheduler dispatches；
S2. the cost for executing the task and firm offer are calculated according to preset Offer Model, and according to the quotation
To task described in scheduler application；
S3. the scheduler receives the quotation of each computing terminal, and calculating is assigned the task to by preset Task Assignment Model
Terminal so that the sum of quotation of each task is optimal.
2. the distribution method of edge calculations system according to claim 1, it is characterised in that：The mission bit stream includes appointing
Input data amount, output data quantity, CPU calculating cycles number and the task reward of business.
3. the distribution method of edge calculations system according to claim 2, it is characterised in that：The Offer Model includes
CPU speed submodel, energy consumption submodel and cost estimation submodel；
The CPU speed submodel, which is used to be calculated according to the CPU calculating cycles number, determines the CPU speed for executing the task；
The energy consumption submodel is used to be calculated described in execution according to the CPU speed, input data amount, output data quantity of the task
The energy consumption of task；
The cost estimation submodel be used for according to the CPU speed, energy consumption and task reward estimation execute the task at
This, and determine that task is offered according to the cost.
4. the distribution method of edge calculations system according to claim 3, it is characterised in that：Pass through CPU speed
It is to meet the minimum CPU speed of task execution requirement that model, which calculates determining CPU speed,.
5. the distribution method of edge calculations system according to claim 4, it is characterised in that：The CPU speed submodel
Meet as shown in formula (1)：
In formula (1),To calculate the CPU speed of determination,For computing terminal the task can be completed in a time slot
CPU speed set, be expressed as For preset speed limit, αnAnd βnFor the systematic parameter of predetermined computing terminal CPU.
6. the distribution method of edge calculations system according to claim 3, it is characterised in that：The energy consumption submodel includes
It communicates energy consumption submodel and executes energy consumption submodel；
The communication energy consumption submodel be used to be calculated as executing the task according to the input data amount, output data quantity and into
Communication energy consumption needed for row communication；
The execution energy consumption submodel, which is used to be calculated according to the CPU calculating cycles number and CPU speed, determines the execution task
Required execution energy consumption；
The energy consumption submodel executes the energy consumption of the task according to the communication energy consumption and execution energy consumption calculation.
7. the distribution method of edge calculations system according to claim 6, it is characterised in that：The communication energy consumption submodel
Meet as shown in formula (2)：
dn,k(t)=Cn,rxIk(t)+Cn,rxOk(t) (2)
In formula (2), dn,k(t) it is the communication energy consumption being calculated, Cn,rxThe energy consumption of the data of a unit is received for computing terminal,
Ik(t) it is the reception data volume of the task, Cn,rxThe energy consumption of the data of a unit, O are sent for computing terminalk(t) it is to calculate
Terminal executes the output data quantity of the task；
The execution energy consumption submodel meets as shown in formula (3)：
zn,k(t)=cn(t)ln,k(t) (3)
In formula (3), zn,k(t) it is the execution energy consumption being calculated, cn(t) be computing terminal CPU processing procedures per unit time
Energy consumption, ln,k(t)=φk(t)/rn(t) it is the time for executing the task and needing, φk(t) it is the CPU calculating cycles number,
rn(t) it is CPU speed；
It calculates and executes shown in the energy consumption such as formula (4) of the task：
Pn,k(t)=zn(t)+dn,k(t) (4)
In formula (4), Pn,k(t) it is energy consumption, zn,k(t) it is to execute energy consumption, dn,k(t) it is communication energy consumption.
8. the distribution method of edge calculations system according to claim 7, it is characterised in that：The cost estimation submodel
Meet shown in formula (5)：
vn,k(t)=Pn,k(t)(En(t)-Λn)-Vwk(t) (5)
In formula (5), vn,k(t) it is the cost for calculating gained, Pn,k(t) it is energy consumption, En(t) it is energy queue, ΛnFor energy snubber
Area, V are preset weights, wk(t) it is the preset task reward.
9. the distribution method of edge calculations system according to claim 8, it is characterised in that：It is preset in the step S3
Shown in Task Assignment Model such as formula (6),
In formula (6),J (t) is the set of task, Jn,k(t) it is victor's index, Jn,k(t)=1 it indicates task k distributing to terminal
N is executed, otherwise is 0, bn,k(t) it is quotation of the terminal n to task k, K (t) is the set of all tasks of t time slots, and N is to calculate eventually
Hold number.
10. the distribution method of edge calculations system according to claim 9, it is characterised in that：It further include distribution step
Rapid S4, scheduler give the corresponding remuneration of the computing terminal after assigning the task to computing terminal, and the remuneration is according to pre-
If distribution model determine, shown in the preset distribution model such as formula (7),
In formula (7), pn,k(t) it is that task k is distributed to the remuneration that computing terminal n is distributed to after computing terminal n,For at end
In the case of holding n non-participating, the minimum value determined by formula (6),To be not involved in the feelings of scheduling in terminal n and task k
Under condition, the minimum value that formula (6) determines, K (t) is the set of all tasks of t time slots.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810140988.XA CN108415763B (en) | 2018-02-11 | 2018-02-11 | Distribution method of edge computing system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810140988.XA CN108415763B (en) | 2018-02-11 | 2018-02-11 | Distribution method of edge computing system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108415763A true CN108415763A (en) | 2018-08-17 |
| CN108415763B CN108415763B (en) | 2020-12-11 |
Family
ID=63128502
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810140988.XA Active CN108415763B (en) | 2018-02-11 | 2018-02-11 | Distribution method of edge computing system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108415763B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109121151A (en) * | 2018-11-01 | 2019-01-01 | 南京邮电大学 | Distributed discharging method under the integrated mobile edge calculations of cellulor |
| CN109343904A (en) * | 2018-09-28 | 2019-02-15 | 燕山大学 | A kind of mist calculating dynamic offloading method based on Lyapunov optimization |
| CN109714797A (en) * | 2019-02-18 | 2019-05-03 | 南京邮电大学 | A kind of mobile edge network resource allocation methods based on Game Theory |
| CN109710336A (en) * | 2019-01-11 | 2019-05-03 | 中南林业科技大学 | The mobile edge calculations method for scheduling task of joint energy and delay optimization |
| CN110113140A (en) * | 2019-03-08 | 2019-08-09 | 北京邮电大学 | A kind of mist calculates the calculating discharging method in wireless network |
| CN110913279A (en) * | 2018-09-18 | 2020-03-24 | 北京悉见科技有限公司 | Processing method for augmented reality and augmented reality terminal |
| CN111443990A (en) * | 2020-03-25 | 2020-07-24 | 中南大学 | Edge calculation task migration simulation system |
| CN112929915A (en) * | 2021-02-20 | 2021-06-08 | 中南大学 | Dynamic data unloading method and system for mobile edge calculation |
| CN113301158A (en) * | 2021-05-25 | 2021-08-24 | 东北大学 | Resource allocation method based on auction theory under mobile edge computing environment |
| CN113626107A (en) * | 2021-08-20 | 2021-11-09 | 中南大学 | Mobile computing unloading method, system and storage medium |
| CN113703970A (en) * | 2021-08-13 | 2021-11-26 | 北京信息科技大学 | Auction mechanism-based server resource allocation method, device, equipment and medium |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1283817A (en) * | 1999-06-18 | 2001-02-14 | 凤凰技术有限公司 | Method and appts for automatic matching, setting and spreading small microsoft windows applied program for calculating equipment |
| US20090157936A1 (en) * | 2007-12-13 | 2009-06-18 | Texas Instruments Incorporated | Interrupt morphing and configuration, circuits, systems, and processes |
| CN101918933A (en) * | 2007-12-05 | 2010-12-15 | 生命力有限公司 | System and method for intelligently allocating client requests to server centers |
| US20120296974A1 (en) * | 1999-04-27 | 2012-11-22 | Joseph Akwo Tabe | Social network for media topics of information relating to the science of positivism |
| CN103853618A (en) * | 2014-03-06 | 2014-06-11 | 南京理工大学 | Resource allocation method with minimized cloud system cost based on expiration date drive |
| CN105225016A (en) * | 2015-10-29 | 2016-01-06 | 华东师范大学 | A kind of in the cloud computing system of renewable energy supply based on the energy distributing method of cooperative game |
| CN105426241A (en) * | 2015-11-16 | 2016-03-23 | 北京航空航天大学 | Cloud computing data center based unified resource scheduling energy-saving method |
| CN105474166A (en) * | 2013-03-15 | 2016-04-06 | 先进元素科技公司 | Methods and systems for purposeful computing |
| US20170060574A1 (en) * | 2015-08-27 | 2017-03-02 | FogHorn Systems, Inc. | Edge Intelligence Platform, and Internet of Things Sensor Streams System |
| CN107295109A (en) * | 2017-08-16 | 2017-10-24 | 重庆邮电大学 | Task unloading and power distribution joint decision method in self-organizing network cloud computing |
| CN107317700A (en) * | 2017-06-09 | 2017-11-03 | 湖北理工学院 | Vehicle-mounted edge calculations node selects system and method |
-
2018
- 2018-02-11 CN CN201810140988.XA patent/CN108415763B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120296974A1 (en) * | 1999-04-27 | 2012-11-22 | Joseph Akwo Tabe | Social network for media topics of information relating to the science of positivism |
| CN1283817A (en) * | 1999-06-18 | 2001-02-14 | 凤凰技术有限公司 | Method and appts for automatic matching, setting and spreading small microsoft windows applied program for calculating equipment |
| CN101918933A (en) * | 2007-12-05 | 2010-12-15 | 生命力有限公司 | System and method for intelligently allocating client requests to server centers |
| US20090157936A1 (en) * | 2007-12-13 | 2009-06-18 | Texas Instruments Incorporated | Interrupt morphing and configuration, circuits, systems, and processes |
| CN105474166A (en) * | 2013-03-15 | 2016-04-06 | 先进元素科技公司 | Methods and systems for purposeful computing |
| CN103853618A (en) * | 2014-03-06 | 2014-06-11 | 南京理工大学 | Resource allocation method with minimized cloud system cost based on expiration date drive |
| US20170060574A1 (en) * | 2015-08-27 | 2017-03-02 | FogHorn Systems, Inc. | Edge Intelligence Platform, and Internet of Things Sensor Streams System |
| CN105225016A (en) * | 2015-10-29 | 2016-01-06 | 华东师范大学 | A kind of in the cloud computing system of renewable energy supply based on the energy distributing method of cooperative game |
| CN105426241A (en) * | 2015-11-16 | 2016-03-23 | 北京航空航天大学 | Cloud computing data center based unified resource scheduling energy-saving method |
| CN107317700A (en) * | 2017-06-09 | 2017-11-03 | 湖北理工学院 | Vehicle-mounted edge calculations node selects system and method |
| CN107295109A (en) * | 2017-08-16 | 2017-10-24 | 重庆邮电大学 | Task unloading and power distribution joint decision method in self-organizing network cloud computing |
Non-Patent Citations (1)
| Title |
|---|
| 施巍松: ""边缘计算：万物互联时代新型计算模型"", 《计算机研究与发展》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110913279A (en) * | 2018-09-18 | 2020-03-24 | 北京悉见科技有限公司 | Processing method for augmented reality and augmented reality terminal |
| CN109343904A (en) * | 2018-09-28 | 2019-02-15 | 燕山大学 | A kind of mist calculating dynamic offloading method based on Lyapunov optimization |
| CN109343904B (en) * | 2018-09-28 | 2021-12-10 | 燕山大学 | Lyapunov optimization-based fog calculation dynamic unloading method |
| CN109121151B (en) * | 2018-11-01 | 2021-06-11 | 南京邮电大学 | Distributed unloading method under small cell integrated mobile edge calculation |
| CN109121151A (en) * | 2018-11-01 | 2019-01-01 | 南京邮电大学 | Distributed discharging method under the integrated mobile edge calculations of cellulor |
| CN109710336A (en) * | 2019-01-11 | 2019-05-03 | 中南林业科技大学 | The mobile edge calculations method for scheduling task of joint energy and delay optimization |
| CN109714797A (en) * | 2019-02-18 | 2019-05-03 | 南京邮电大学 | A kind of mobile edge network resource allocation methods based on Game Theory |
| CN110113140A (en) * | 2019-03-08 | 2019-08-09 | 北京邮电大学 | A kind of mist calculates the calculating discharging method in wireless network |
| CN111443990A (en) * | 2020-03-25 | 2020-07-24 | 中南大学 | Edge calculation task migration simulation system |
| CN111443990B (en) * | 2020-03-25 | 2023-04-07 | 中南大学 | Edge calculation task migration simulation system |
| CN112929915A (en) * | 2021-02-20 | 2021-06-08 | 中南大学 | Dynamic data unloading method and system for mobile edge calculation |
| CN112929915B (en) * | 2021-02-20 | 2022-08-02 | 中南大学 | Dynamic data unloading method and system for mobile edge calculation |
| CN113301158A (en) * | 2021-05-25 | 2021-08-24 | 东北大学 | Resource allocation method based on auction theory under mobile edge computing environment |
| CN113703970A (en) * | 2021-08-13 | 2021-11-26 | 北京信息科技大学 | Auction mechanism-based server resource allocation method, device, equipment and medium |
| CN113626107A (en) * | 2021-08-20 | 2021-11-09 | 中南大学 | Mobile computing unloading method, system and storage medium |
| CN113626107B (en) * | 2021-08-20 | 2024-03-26 | 中南大学 | Mobile computing unloading method, system and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108415763B (en) | 2020-12-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108415763A (en) | A kind of distribution method of edge calculations system | |
| Shin et al. | Auction-based charging scheduling with deep learning framework for multi-drone networks | |
| Zhang et al. | Near-optimal and truthful online auction for computation offloading in green edge-computing systems | |
| Lin et al. | A survey on computation offloading modeling for edge computing | |
| US10317924B2 (en) | Collaborative load balancing within a community of energy nodes | |
| Toosi et al. | An auction mechanism for cloud spot markets | |
| CN102185708B (en) | Grid resource distribution method based on Nash equilibrium | |
| CN103927229A (en) | Scheduling Mapreduce Jobs In A Cluster Of Dynamically Available Servers | |
| Wang et al. | Autonomous PEV charging scheduling using Dyna-Q reinforcement learning | |
| Hayakawa et al. | Online mechanisms for charging electric vehicles in settings with varying marginal electricity costs | |
| CN111694664A (en) | Calculation unloading distribution method of edge server | |
| US11861702B2 (en) | Method and apparatus for renewable energy allocation based on reinforcement learning | |
| Liu et al. | RLC: A reinforcement learning-based charging algorithm for mobile devices | |
| Liu et al. | Efficient auction mechanism for edge computing resource allocation in mobile blockchain | |
| Peng et al. | A decentralized collaborative approach to online edge user allocation in edge computing environments | |
| Lu et al. | Auction design for cross-edge task offloading in heterogeneous mobile edge clouds | |
| Qian et al. | Optimal recharge scheduler for drone-to-sensor wireless power transfer | |
| Zhang et al. | A smart cross-system framework for joint allocation and scheduling with vehicle-to-grid regulation service | |
| Hajiesmaili et al. | Crowd-sourced storage-assisted demand response in microgrids | |
| Wang et al. | Distributed game-theoretical route navigation for vehicular crowdsensing | |
| Mao et al. | Efficient satellite scheduling based on improved vector evaluated genetic algorithm | |
| CN117639043A (en) | Electric automobile dispatching method and device under wind power cooperation, terminal equipment and medium | |
| Luong et al. | Optimal auction for effective energy management in UAV-assisted vehicular metaverse synchronization systems | |
| Xu et al. | Basic: Distributed task assignment with auction incentive in uav-enabled crowdsensing system | |
| Tightiz et al. | Novel deep deterministic policy gradient technique for automated micro-grid energy management in rural and islanded areas |
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 |