CN106161624A - The band redundant data method for uploading of effectiveness perception in the mobile cloud of cooperation - Google Patents

Abstract

The present invention provides the band redundant data method for uploading of effectiveness perception in a kind of mobile cloud that cooperates, in the t+D feedback delay of system (D be) individual upload procedure, all mobile terminals receive about all states of each mobile terminal in the t upload procedure and decision information, the core thinking of the distributed related optimization used is that each mobile terminal in each upload procedure finds optimal solution in each uploads strategy, and energy expenditure constraint is converted into string stability sex chromosome mosaicism.So that each mobile terminal can independently carry out decision-making in the case of not knowing future channel.

Description

The band redundant data method for uploading of effectiveness perception in the mobile cloud of cooperation

Technical field

The present invention relates to the data uploading method in mobile cloud that cooperates, especially relate to sense of efficacy in the mobile cloud of a kind of cooperation The band redundant data method for uploading known.

Background technology

Along with the fast development of the widely available of mobile terminal Yu wireless communication technology, increasing mobile terminal by with In fields such as emergency processing and health monitorings.Problem of uploading with redundant data receives more and more attention and interest.In association How make band redundant data in mobile cloud to upload a challenge of problem is at a connection breaking, the uncertain wireless mobile of bandwidth Realize effective in network, the data of low energy consumption are uploaded.

1. introduction

Recent years, mobile terminal develops rapidly.Cisco prediction mobile terminal user in 2016 will be more than 4,500,000,000 people. Along with the universal of mobile terminal and the development of information technology, mobile terminal changes our life in all fields.In order to more Mending the limited battery capacity of mobile terminal and deficiency that computing capability is brought, mobile cloud can utilize resource unlimited in cloud to prop up Hold the application of resource-intensive in mobile terminal.By by applying with load shedding to the method for cloud data center, increasing The computing capability of mobile device, such as memory capacity.

Upload the challenge of data.Increasing mobile terminal is applied to emergency processing, such as disaster emergency response and army Thing is taken action.The mobile terminal being equipped with built-in sensors (such as camera and mike) can be used to the figure in collection process disaster area As video data.But, the communication network of extreme environment is dynamically change, and discontinuity connects blocks up with uncertain communication Plug makes stable link and teledata upload to become actual.In these cases, the mobile cloud of cooperation moves eventually for strengthening The computing capability of end provides a kind of new method, and the mobile cloud of cooperation refers to one group of neighbouring mobile terminal, these mobile terminals Between mutually share part computing capability.

The present situation of the mobile cloud of cooperation.Great majority are all absorbed in the storage promoting mobile terminal about the research moving cloud that cooperates Ability.Can be transferred in adjacent terminals by large datas such as the image of a mobile terminal generation, videos.Data are transmitted And be stored in neighbouring terminal, it is possible to reduce the pressure that remote data center produces due to disconnecting.Owing to using height Bandwidth connects and the spacing of terminal is shorter, compares and uploads data to remote data center, and the mobile terminal near utilization is permissible Substantially reduce cost improving performance.For the uncertain mobility of unstability and mobile terminal tackling wireless connections, The data storage of redundancy is way popular in the mobile cloud of cooperation, and many numbers evidence is stored in several terminal to protect by this method The reliability of card data, especially in emergency processing is applied.

But, in few people's cloud mobile to cooperation, band redundant data is uploaded problem and is studied so far.Those are needed Want reliable Distributed Storage and upload data to the application of data center (in disaster response application and Shan Li, desert The application of search work): on the one hand, the mobile cloud computing of cooperation utilizes other mobile terminal to store the mass data such as video, image, On the other hand the mobile cloud that cooperates needs to upload data to remote data center and carries out data mining, backup.When dynamically uploading channel Being in poor communications status, data can be copied to several closing in mobile terminal and store.After the some time, the mobile cloud of cooperation In each mobile terminal can decide whether to upload data.How to formulate and upload strategy efficiently for cooperation movement For cloud, it it is a critically important and an open question.Solve this problem and face following challenge: first, in disaster response and army In the poor environment of thing action, use center type scheduling scheme can cause unacceptable communication and meter each mobile terminal Calculate loss.Further, center type method can cause single point failure problem.Second, owing to the uncertain mobility of mobile terminal is with many Jumping the dynamic of network, a mobile terminal is difficult to know in time the situation of other mobile terminals.Two or more mobile terminals Identical data may be uploaded simultaneously, waste limited power resources and cause the redundancy uploading data.Finally, unstable Network facility environment also can be made dynamically to change with uncontrollable.Off-line optimization method can not be used in reality to solve Arraign is inscribed.

2. related work

The mobile cloud of cooperation strengthens storage be one in industry, the problem of the most awfully hot door.Moon proposes one Plant the storage Enhancement Method in efficient self-sustaining network.Abolfazli devises a kind of framework coming from service, utilizes around Mobile terminal reduces storage and limits.In order to ensure the reliability of data, need to produce mass of redundancy data.Pheonix is as one Plant distributed storage agreement, use at least two backup to ensure the reliability of data in single-hop networks.Chen provides one Plant the framework of k out of n, as long as it is available for demonstrating n data terminal from there being k, always can successfully fetch, process number According to.Work before allows to use the mobile terminal connected to strengthen storage capacity.But, how to upload these data to cooperation Problem in mobile cloud is the most undecided.

The cloud data center uploading data to far-end brings the biggest energy burden to mobile terminal.Deliver a letter on especially When road situation is poor.Many researchers is made great efforts that energy-conservation data are uploaded problem and is positioned in dynamic network solution.Lombardo Markov model is utilized to devise the energy-conservation delivery protocol of applicable uploading channel situation.Data transmission problems is converted by Xiang For Random Discrete dynamic problem, to optimizing data throughout and energy consumption simultaneously.Fang utilizes Lyapunov optimization The theoretical Optimum Operation for data transmission provides a kind of online algorithm.This algorithm can be while avoiding the occurrence of mistake Reduce power consumption to minimum, and the information of the channel conditions before not using.These existing researchs are absorbed in single mobile whole The power saving of end.In the present invention, we in the dynamic network that multiple mobile terminals form to the upload procedure that is mutually related It is optimized.

In mobile sensor field, some existing researchs are attempted solving distributed data collection problem.But, these grind Study carefully and ignore some redundant datas and take up room the phenomenon the most not played a role.Deliver a letter it addition, these researchs are not investigated The concrete condition in road.Being different from mobile sensor field, the Energy-saving Data transmission of mobile terminal is absorbed in our work.On deliver a letter System performance is had a significant impact by road.Redundant data is uploaded in dynamic network and is carried out optimization by we, ensures mobile with this Terminal uploads data in an efficient way.

Summary of the invention

Providing the band redundant data method for uploading of effectiveness perception in a kind of mobile cloud that cooperates in the present invention, we will cooperate In mobile cloud, the data of band redundancy are uploaded problem and are modeled as the maximization of utility problem of a power consumption constraint, to realize certain Under power consumption constraint, maximize the valid data amount uploaded.It is proposed that a kind of online distributed method makes each mobile terminal Optimal Decision-making can be independently made in the case of the status information being independent of other-end and following oneself state information.Strictly The effectiveness of theory analysis and a large amount of simulation results shows the method and superiority.Its technical scheme is as described below:

The band redundant data method for uploading of effectiveness perception in a kind of mobile cloud that cooperates, the mobile cloud of cooperation comprises N number of mobile whole End, represents the t upload procedure of cooperation cloud, d with t_i(t)=(d_i1(t),d_i2(t),...,d_iK(t)) represent that t is uploaded During, whether data block k is stored in i-th mobile terminal, and K is the number of different types of data block, ω_iT () represents t The channel status of i-th mobile terminal, ω in upload procedure_iT the channel of () numerical value the biggest expression i-th mobile terminal is in good Good state；Uploading channel when can use, the mobile cloud of cooperation carries out a secondary data and uploads；

Comprise the following steps:

(1) the channel status ω of each mobile terminal is observed_iT (), each mobile terminal decides whether according to channel status Upload its data；Observation is used for representing energy expenditure constraint queue Q (t), described Q (t)=(Q₁(t),...,Q_N(t)), described Q_iT () is the virtual queue of each mobile terminal definition, queue Q (t) is the most stable, represents that energy expenditure is the least；

(2) in the t+D upload procedure, D is the feedback delay of system, and all mobile terminals receive about the t upload procedure In all states and decision information, each mobile terminal in this upload procedure exists One strategy of middle selectionDescribedRepresent all possible set of strategies Close,Represent that a kind of optimal strategy can make the upper limit of drift-penaltyHave Minima, whereinRepresent perform this strategy time i-th terminal estimate energy consumption,Represent and perform estimating of this strategy Effectiveness, parameter V isWeight, control the balance between energy consumption and effectiveness；

(3) optimum performing to determine uploads decision-making RepresentIn i-th Individual terminal uploads decision-making；

(4) at the end of the t time is uploaded, information ω (t-D) about channel status and decision-making α (t-D) are received, the most right Q (t) is updated.

Further, in step (1), described Q_iT () is updated the t time upload procedure by this equation following:

Q_i(t+1)=max{Q_i(t)+p_i(t-D)-c_i,0}

Wherein, c_iIt is the average energy consumption binding occurrence of i-th mobile terminal, p_i(0)=0, p_i(-1)=...=p_i(- D)=0, each mobile terminal maintains queue Q (t), utilizes feedback information to update team after upload procedure terminates each time Row, when virtual queue is stablized, namelyThe most just disclosure satisfy that energy expenditure retrains, its In, E represents and seeks mathematic expectaion, and definition Lyapunov function is:

$L\left(t\right)=\frac{1}{2}\underset{i=1}{\overset{N}{\Σ}}{Q}_i{\left(t\right)}^2---\left(1\right)$

The Lyapunov function representation degree of congestion of Q (t) queue, the change the least expression mobile terminal of L (t) virtual Queue has strong stability, in other words, it is possible to meet energy expenditure constraint.

Further, in step (2), ω is worked as_iOptimal strategy when () increases tIn i-th, i.e. i-th terminal Decision-makingAlso with ω_iT () increases, the following form of optimal strategy:

${\widehat{\mathrm{\&alpha;}}}_i\left({\mathrm{\&omega;}}_i\right(t\left)\right)=\{\begin{array}{ccc}0& if& {\mathrm{\&omega;}}_i\left(t\right)<{\mathrm{\&omega;}}_i^{*}\left(t\right)\\ 1& if& {\mathrm{\&omega;}}_i\left(t\right)\&GreaterEqual;{\mathrm{\&omega;}}_i^{*}\left(t\right)\end{array}---\left(2\right)$

ω^*T () represents threshold status, when channel status is inferior to ω^*(t)(ω_i(t) ＜ ω^*(t)) time, determining of mobile terminal Plan value is 0, does not uploads；When channel status is better than ω^*(t)(ω_i(t)≥ω^*(t)) time, the decision value of mobile terminal is 1, namely upload data.Thus this problem is converted into each mobile terminal seeking ω^*(t), described ω^*T () has the individual possibility of | Ω | Value, therefore, the quantity of efficient solution is reduced to

Further, in step (2), the V for given > 0, in each upload procedure:

$E\&lsqb;\mathrm{\&Delta;}(t+D)-Vu\left(t\right)|Q\left(t\right)\&rsqb;\&le;A(1+2D)-\underset{i=1}{\overset{N}{\&Sigma;}}{c}_i{Q}_i\left(t\right)+E\&lsqb;\underset{i=1}{\overset{N}{\&Sigma;}}{p}_i\left(t\right)Q_i\left(t\right)-Vu\left(t\right)|Q\left(t\right)\&rsqb;---\left(3\right)$

This formula is obtained by mathematical derivation, does not has concrete physical meaning, only represents that the upper limit of inequality left side item is The right item of formula；E represents the effectiveness asking mathematic expectaion, u (t) to represent the t time upload procedure；Parameter V is the weight of u (t)；

Wherein

Utilizing drift to add and penalize technology, in each upload procedure, each mobile terminal exists Middle selection one, minimizes the right item of inequality (3), to solve the maximization of utility problem of power consumption constraint.

Further, Utilization strategies is worked asTime, each mobile terminal i is at strategyUnder estimate energy consumptionWith estimate benefit u^(m)T () can be approximated to be:

${\tilde{p}}_i^{\left(m\right)}\left(t\right)=\frac{1}{S}\underset{s=1}{\overset{S}{\&Sigma;}}{\hat{p}}_i\left({{\hat{a}}_i}^{\left(m\right)}\right({\mathrm{\&omega;}}_i\left(t-D-s\right)),{\mathrm{\&omega;}}_i(t-D-s\left)\right)---\left(4\right)$

${\tilde{u}}^{\left(m\right)}\left(t\right)=\frac{1}{S}\underset{s=1}{\overset{S}{\&Sigma;}}\hat{u}\left({\hat{a}}^{\left(m\right)}\right({\mathrm{\&omega;}}_i\left(t-D-s\right)),d(t-D-s\left)\right)---\left(5\right)$

Wherein, S is a positive number, represents sample size.

Further, for any one channel of mobile terminal, for arbitrary V > 0 and S > 0, S is to represent that sample is big Little positive number, parameter V (>=0) it is the balance in order to control between string stability and effectiveness, less V means that cooperation moves Dynamic cloud tends to the lowest energy consumption of hold queue rather than collects more valid data block, then has:

The gap of the effectiveness that best practice and this method are tried to achieve is:

${\stackrel{\&OverBar;}{u}}^{opt}-\frac{1}{T}\underset{t=0}{\overset{T-1}{\&Sigma;}}E\&lsqb;u\left(t\right)\&rsqb;\&le;\frac{A(1+2D)}{V}+\frac{E\&lsqb;L\left(D\right)\&rsqb;}{VT}+O(1/\sqrt{S})---\left(6\right)$

Wherein,Represent the best practice maximum average utility under power consumption constraint,Represent this method The effectiveness tried to achieve；WhereinT represents and always uploads number of times, and L (D) represents the Lyapunov functional value that parameter is D,Expression comprisesThe multinomial of item；

This method ensure that the average energy consumption of each mobile terminal meets:

$\frac{1}{T}\underset{t=0}{\overset{T-1}{\&Sigma;}}E\&lsqb;p_i\left(t\right)\&rsqb;\&le;c_i+O\left(\sqrt{\frac{V}{T}}\right),\&ForAll;i---\left(7\right)$

Wherein,Represent the average energy consumption of i-th mobile terminal, c_iRepresent power consumption constraint value, Expression comprisesThe multinomial of item；

Collaborative cloud computing can carry out the decision-making of elasticity by adjusting the size of V between effectiveness and energy loss；Separately Outward, bigger S value can also make solution closer to optimal value；But, it takes longer for simultaneously calculating and bigger Memory space.

The band redundant data method for uploading of effectiveness perception in the mobile cloud of described cooperation so that each mobile terminal is not being known Can independently carry out decision-making in the case of road future channel, for reducing the complexity of algorithm, carry out rigorous theory analysis, reduced The search volume of algorithm.

Accompanying drawing explanation

Fig. 1 a is showing the increase along with V, and effectiveness improves, constantly close to the schematic diagram of optimal value；

Fig. 1 b is showing the increase along with V, the schematic diagram that average energy consumption improves constantly；

Fig. 2 is that average energy loss is decreased obviously under the value of different V, and understands showing of fast approaching power dissipation constraints It is intended to；

Fig. 3 a shows along with D becomes big, that system utility value declines schematic diagram；

Fig. 3 b shows the decline of energy loss, consumes and shows when D becomes big with the ratio of effectiveness, the effect of the energy consumption of system Rate declines；

Fig. 4 a shows the average utility of 900 upload procedure；

Fig. 4 b describes the average energy consumption of each upload procedure；

Fig. 5 a is when single mobile terminal channel status changes in distribution, and significant change does not occur in average utility；

Fig. 5 b is when single mobile terminal channel status changes in distribution, and significant change does not occur in average energy consumption；

Fig. 5 c shows the average decision-making in 600 upload procedure of terminal 1 and terminal 8, terminal 1 after 300 times due to letter Road state is deteriorated, it is intended to do not upload；Terminal 8 improves due to channel status after 300 times, it is intended to upload；

Fig. 5 d shows the average accumulated decision-making in 600 upload procedure of terminal 1 and terminal 8；

Fig. 6 a shows that the value of utility of four kinds of algorithms compares under different power consumption constraint；

Fig. 6 b shows that the average energy loss-rate of four kinds of algorithms is relatively under different power consumption constraint；

Fig. 7 is operating procedure flow chart.

Detailed description of the invention

The mobile cloud of cooperation is made up of one group of mobile terminal, and these mobile terminals are worked in coordination with each other by shared storage capacity and deposited The common data produced of storage.In order to ensure stable, store reliably, same number is deposited according to can be copied in multiple mobile terminal Storage.Above-mentioned uploading problem to tackle, we devise a kind of based on the relevant distribution on line formula scheduling optimized of distribution Method.This method allows each mobile terminal at the shape of the channel status and other mobile terminals being independent of self future Independently make under state and upload decision-making.

Main work is as follows: with regard to from the point of view of the Document Knowledge that we grasp, this is first by band redundancy in mobile for cooperation cloud Data upload problem and be modeled as the maximization of utility problem of a power consumption constraint, to realize under certain power consumption constraint, The valid data amount that bigization is uploaded.Second, a kind of distribution on line formula Optimization Framework is proposed, to help each mobile terminal to formulate Independent uploads strategy.In During Process of Long-term Operation, strict theory analysis demonstrates our algorithm close to optimum.? After, by substantial amounts of experiment, show the effect of this optimization framework.

3. problem modeling

In our study, it is assumed that the mobile cloud of cooperation comprises N number of mobile terminal.Mobile cooperation cloud is repeated in data Pass.We represent the t upload procedure of cooperation cloud with t.In the middle of twice upload procedure, some mobile terminals produce a large amount of note The record image of disaster scene, video data.In order to strengthen the storage capacity of a mobile terminal, and improve the reliable of data Property, these data are divided into the data block of several formed objects, and these data blocks are copied in multiple mobile terminal distributed Storage.d_i(t)=(d_i1(t),d_i2(t),...,d_iK(t)) represent, in the t upload procedure, whether data block k is stored in i-th Individual mobile terminal, K is the number of different types of data block.d_ikT ()=1 represents that kth data block is stored in i-th data terminal In, d_ikT ()=0 represents not storage.Represent these d_iThe vector of (t),

Uploading channel when can use, the mobile cloud of cooperation carries out a secondary data and uploads.It is in diverse location in view of mobile terminal, The channel conditions of different mobile terminal may be different.ω_iT () represents the channel shape of i-th mobile terminal in the t upload procedure State, ω_iT the channel of () numerical value the biggest expression i-th mobile terminal is in good state.Therefore ω_iT ()=| Ω |-1 represents Channel status is best, and ω_iT ()=0 represents that channel status is worst.Represent these d_iThe vector of (t).

The t upload procedure, the situation that mobile terminal is observed according to oneself decides whether to upload data block.Binary variable α_iT { 0,1} is for representing the decision of i-th mobile terminal for () ∈.α_iT ()=1 represents that i-th mobile terminal determines the t upload procedure In upload data block, α_iT ()=0 represents does not uploads.It is the vector of α (t), wherein So, the valid data block count table in the t upload procedure is shown as:

$u\left(t\right)=\hat{u}\left(\mathrm{\&alpha;}\right(t),d(t\left)\right)=\underset{j=1}{\overset{K}{\&Sigma;}}min\{\underset{i=1}{\overset{N}{\&Sigma;}}{d}_{ij}\left(t\right){\mathrm{\&alpha;}}_i\left(t\right),1\}---\left(1\right)$

Owing to each data block has multiple backup in different mobile terminal storages.When two or more terminals are uploaded same During the backup of individual data block, only one of which backup effectively, other backup invalidation.Therefore we can obtain the effect as shown in (1) Use function.If we assume that mobile terminal determines to upload data, just uploading all data in this upload procedure.Accordingly, it is capable to Consume and determined that pi (t) is the energy expenditure of i-th terminal in the t upload procedure by channel status.Therefore we have a following formula:

$p_i\left(t\right)=\hat{p}\left({\mathrm{\&alpha;}}_i\right(t),{\mathrm{\&omega;}}_i(t\left)\right)---\left(2\right)$

Energy expenditure formulaSufficient research has been obtained in experience energy consumption model.Therefore, we are not Detail is given in this part.But, no matter what energy expenditure function concrete form is, it is evident that when uploading data, poor Channel status can cause higher energy loss.According to ω_iT the definition of (), can sum up It is variable ω_iThe nonincreasing function of (t).

The target that cooperation moves cloud main is to make repeatedly valid data block average in upload procedure reach maximum quantity.But It is, it is contemplated that limited battery capacity that energy loss must limit within limits.Therefore we provide energy constraint Maximization of utility problem:

$max:\stackrel{\&OverBar;}{u}=\underset{T\&RightArrow;\mathrm{\&infin;}}{lim}\frac{1}{T}\underset{t=0}{\overset{t-1}{\&Sigma;}}u\left(t\right)---\left(3\right)$

$s.t.:\stackrel{\&OverBar;}{p_i}=\underset{T\&RightArrow;\mathrm{\&infin;}}{lim}\frac{1}{T}\underset{t=0}{\overset{t-1}{\&Sigma;}}{p}_i\left(t\right)\&le;c_i,\&ForAll;i---\left(4\right)$

Decisions are distributed.(5)

Wherein, u (t) represents the effectiveness of the t time upload procedure,Represent the average utility of T upload procedure, p_iT () represents I-th mobile terminal energy expenditure in the t time upload procedure,Represent the flat of i-th mobile terminal in T upload procedure All energy expenditure, c_iIt it is the binding occurrence of the average energy consumption of i-th mobile terminal.Formula (5) represents that each mobile terminal is being worked as Not knowing status information and the decision situation of other-end in front decision making process, the decision-making of uploading of each mobile terminal is distributed 's.

4. distribution on line formula Optimization Framework

In reality, the maximization of utility problem of solution energy constraint is one and is rich in challenging work.Due to terminal Mobility and upload the unstability of channel, the environment residing for terminal is highly dynamic and uncertain, this make from Line optimization method becomes impossible.Worse mobile terminal cannot know in time the situation of current time other-end with And upload decision-making, mobile terminal can upload useless redundant data block, consumes again energy simultaneously.In order to solve these problems, We make use of distributed related optimization, devises online distributed scheduling algorithm, so that each mobile terminal can Independent Optimal Decision-making is made with the state observed according to self.

4.1 complexity beta prunings

In a upload procedure, each mobile terminal observes its channel status, and decides whether according to channel status Upload its data.We define,Represent that i-th terminal is according to determining strategyIt is made whether to upload Decision-making.So, the strategy of all terminals may be constructed one group of vector

$\widehat{\mathrm{\&alpha;}}\left(\mathrm{\&omega;}\right)=\{{\widehat{\mathrm{\&alpha;}}}_1\left({\mathrm{\&omega;}}_1\right),{\widehat{\mathrm{\&alpha;}}}_2\left({\mathrm{\&omega;}}_2\right),...,{\widehat{\mathrm{\&alpha;}}}_N\left({\mathrm{\&omega;}}_N\right)\}---\left(6\right)$

Theorem 1: optimal decision strategyIt is variable ω_iNon-decreasing function.

Prove: given two channel status ω and γ, ω ＜ γ.Assuming that the optimum under both channel status determines decision-making MeetWe are by finding a new strategy, and this strategy is possible not only to meet non-decreasing, and not Lose optimality, prove theorem 1.

BecauseSoProvide following two new strategies:

${{\widehat{\mathrm{\&alpha;}}}_i}^{low}\left({\mathrm{\&omega;}}_i\right)=\left\{\begin{array}{ccc}{\widehat{\mathrm{\&alpha;}}}_i\left({\mathrm{\&omega;}}_i\right)& if& {\mathrm{\&omega;}}_i\&NotElement;\{\mathrm{\&omega;},\mathrm{\&gamma;}\}\\ 0& if& {\mathrm{\&omega;}}_i\&Element;\{\mathrm{\&omega;},\mathrm{\&gamma;}\}\end{array}\right.---\left(7\right)$

${{\widehat{\mathrm{\&alpha;}}}_i}^{high}\left({\mathrm{\&omega;}}_i\right)=\left\{\begin{array}{ccc}{\widehat{\mathrm{\&alpha;}}}_i\left({\mathrm{\&omega;}}_i\right)& if& {\mathrm{\&omega;}}_i\&NotElement;\{\mathrm{\&omega;},\mathrm{\&gamma;}\}\\ 1& if& {\mathrm{\&omega;}}_i\&Element;\{\mathrm{\&omega;},\mathrm{\&gamma;}\}\end{array}\right.---\left(8\right)$

Obviously, the two strategy is satisfied by non-decreasing.Assume that i-th mobile terminal is in the general of channel status ω with γ Rate is pr respectively_i(ω) and pr_i(γ).So, we define a kind of new randomized policy

With Probability p r_i(γ)/(pr_i(ω)+pr_i(γ)) for determining strategy

With Probability p r_i(ω)/(pr_i(ω)+pr_i(γ)) for determining strategy

We useRepresent N-dimensional vector α, wherein,Represent the uploading of other-end in addition to i-th terminal Decision-making.So New Policy can be calculated as follows with energy consumption with the utility function of i-th terminal under former strategy:

AssumeIt is chosen to be New Policy, and ω_i(t)=ω, then, the utility function under former strategy isUtility function under New Policy isUnder former strategy The energy consumption of i terminal isEnergy consumption under New Policy is

AssumeIt is chosen to be New Policy, and ω_i(t)=γ, then, the utility function under former strategy isUtility function under New Policy isUnder former strategy The energy consumption of i terminal isEnergy consumption under New Policy is

In other cases, u (t)=u'(t), p_i(t)=p_i'(t)。

Thus, ask expectation to obtain with the utility function of old strategy and the difference of energy consumption New Policy:

$\begin{array}{c}E\&lsqb;u\left(t\right)-u^{\&prime;}\left(t\right)\&rsqb;=r_i\left(\mathrm{\&omega;}\right)\frac{{\mathrm{pr}}_i\left(\mathrm{\&gamma;}\right)}{{\mathrm{pr}}_i\left(\mathrm{\&omega;}\right)+{\mathrm{pr}}_i\left(\mathrm{\&gamma;}\right)}.\left(\hat{u}\right(\&lsqb;1,a_i^{-}\left(t\right)\&rsqb;,d\left(t\right))-\hat{u}(\&lsqb;0,a_i^{-}\left(t\right)\&rsqb;,d\left(t\right)\left)\right)\\ +{\mathrm{pr}}_i\left(\mathrm{\&gamma;}\right)\frac{{\mathrm{pr}}_i\left(\mathrm{\&omega;}\right)}{{\mathrm{pr}}_i\left(\mathrm{\&omega;}\right)+{\mathrm{pr}}_i\left(\mathrm{\&gamma;}\right)}.\left(\hat{u}\right(\&lsqb;0,a_i^{-}\left(t\right)\&rsqb;,d\left(t\right))-\hat{u}(\&lsqb;1,a_i^{-}\left(t\right)\&rsqb;,d\left(t\right)\left)\right)\\ =0\end{array}---\left(9\right)$

$\begin{array}{c}E\&lsqb;p_i\left(t\right)-p_i^{\&prime;}\left(t\right)\&rsqb;={\mathrm{pr}}_i\left(\mathrm{\&omega;}\right)\frac{{\mathrm{pr}}_i\left(\mathrm{\&gamma;}\right)}{{\mathrm{pr}}_i\left(\mathrm{\&omega;}\right)+{\mathrm{pr}}_i\left(\mathrm{\&gamma;}\right)}.\left(\hat{p}\right(1,\mathrm{\&omega;})-\hat{p}(0,\mathrm{\&omega;}\left)\right)\\ +{\mathrm{pr}}_i\left(\mathrm{\&gamma;}\right)\frac{{\mathrm{pr}}_i\left(\mathrm{\&omega;}\right)}{{\mathrm{pr}}_i\left(\mathrm{\&omega;}\right)+{\mathrm{pr}}_i\left(\mathrm{\&gamma;}\right)}.\left(\hat{p}\right(0,\mathrm{\&gamma;})-\hat{p}(1,\mathrm{\&gamma;}\left)\right)\end{array}---\left(10\right)$

When uploading decision-making and being 0, do not upload data, then energy consumption is also 0, therefore,Examine Consider, function(representing any expression formula) is ω_iThe nonincreasing function of (t), then,So, E [p_i(t)-p_i'(t)]≥0。

Therefore, it has been found that new strategy not only meets descending, and does not increase the energy consumption of mobile terminal, does not drops The effectiveness of low system.

Theorem 1 illustrates to work as ω_iOptimal strategy when () is bigger tAlso increase.There is most a following form of strategy:

${\widehat{\mathrm{\&alpha;}}}_i\left({\mathrm{\&omega;}}_i\right(t\left)\right)=\left\{\begin{array}{ccc}0& if& {\mathrm{\&omega;}}_i\left(t\right)<{\mathrm{\&omega;}}_i^{*}\left(t\right)\\ 1& if& {\mathrm{\&omega;}}_i\left(t\right)\&GreaterEqual;{\mathrm{\&omega;}}_i^{*}\left(t\right)\end{array}\right.---\left(11\right)$

Therefore problem is converted into as each mobile terminal seeking threshold value channel statusWhen mobile terminal has | Ω | individual During feasible solution, the quantity of efficient solution is reduced toDue to mobile terminal in the mobile cloud of cooperation in an area Limited amount, therefore this complexity is acceptable.

4.2 distribution on line formula dispatching algorithms

The core thinking of distributed related optimization is that each mobile terminal in each upload procedure existsMiddle searching optimal solution.We assume that, individual at the t+D feedback delay of system (D be) In upload procedure, all mobile terminals receive about all states in the t upload procedure and decision information.This is assumed Reality can be realized by piggybacking technology.

First, energy expenditure constraint is converted into string stability sex chromosome mosaicism by us.To each mobile terminal, define a void Intend queue Q_i(t), Q (t)=(Q₁(t),...,Q_N(t))。

Q_iT () is updated the t time upload procedure by this equation following:

Q_i(t+1)=max{Q_i(t)+p_i(t-D)-c_i,0} (12)

Wherein, p_i(0)=0, p_i(-1)=...=p_i(-D)=0.Each mobile terminal maintains queue Q (t).Each time Upload procedure utilizes feedback information to update queue after terminating.When virtual queue is stablized, namely The most just can meet energy expenditure constraint.We define Lyapunov function:

$L\left(t\right)=\frac{1}{2}\underset{i=1}{\overset{N}{\&Sigma;}}{Q}_i{\left(t\right)}^2---\left(13\right)$

The Lyapunov function representation degree of congestion of Q (t) queue.L (t) changes the virtual team of less expression mobile terminal Row have strong stability, in other words, can meet energy expenditure constraint.For making Lyapunov be not at blocked state, we define D-slot lyapunov drift is:

Δ (t+D)=L (t+D+1)-L (t+D) (14)

From the point of view of intuitively, reduce above-mentioned D-slot lyapunov drift, can be with hold queue stability.In order to make function (1) Maximizing, we use drift to add and penalize technology.It makes the maximization of utility problem of power consumption constraint be converted in each upload procedure Make the problem that the upper bound of following formula minimizes:

E[Δ(t+D)-Vu(t)|Q(t)] (15)

-Vu (t) can be considered penalty, because former problem target is to make effectiveness u (t) maximize ,-Vu (t) should use up Measure little, i.e. punishment is tried one's best little.Parameter V (>=0) it is the balance in order to control between string stability and effectiveness.Less V The mobile cloud that means to cooperate tends to hold queue stable (the lowest energy consumption) rather than collects more valid data block.With Lower lemma gives the upper bound of formula (15).

Lemma 1: the V for given > 0, in each upload procedure:

$E\&lsqb;\mathrm{\&Delta;}(t+D)-Vu\left(t\right)|Q\left(t\right)\&rsqb;\&le;A(1+2D)-\underset{i=1}{\overset{N}{\&Sigma;}}{c}_i{Q}_i\left(t\right)+E\&lsqb;\underset{i=1}{\overset{N}{\&Sigma;}}{p}_i\left(t\right)Q_i\left(t\right)-Vu\left(t\right)|Q\left(t\right)\&rsqb;---\left(16\right)$

Wherein,Being a constant, Q (t) is virtual queue, and D is that system feedback postpones, c_iFor i-th eventually The average energy consumption binding occurrence of end.

Adding penalize expression formula to minimize drift, our method is to reduce the right of inequality, is ensureing Q (t) with this While stability, the lower limit of utility function is maximized, it means that constraint (4) can be met.Finally, utilize drift to add to penalize Technology, in each upload procedure, each mobile terminal existsMiddle selection one, comes The right item of littleization inequality (16), to solve the maximization of utility problem of power consumption constraint.

Because mobile terminal do not know other terminal in the channel status of current upload procedure and decision-making, mobile terminal can not P on the right of calculation equation_i(t) and u (t).But, feedback mechanism makes information ω (t-D) about channel status and decision-making α (t-D) at the end of the t time is uploaded, become available.Work as Utilization strategiesTime,And u^(m)T () can be approximated to be:

${\tilde{p}}_i^{\left(m\right)}\left(t\right)=\frac{1}{S}\underset{s=1}{\overset{S}{\&Sigma;}}{\hat{p}}_i\left({{\hat{a}}_i}^{\left(m\right)}\right({\mathrm{\&omega;}}_i\left(t-D-s\right)),{\mathrm{\&omega;}}_i(t-D-s\left)\right)---\left(17\right)$

${\tilde{u}}^{\left(m\right)}\left(t\right)=\frac{1}{S}\underset{s=1}{\overset{S}{\&Sigma;}}\hat{u}\left({\hat{a}}^{\left(m\right)}\right({\mathrm{\&omega;}}_i\left(t-D-s\right)),d(t-D-s\left)\right)---\left(18\right)$

Wherein, S is the positive number representing sample size.

The false code of distribution on line formula dispatching algorithm is as it is shown in fig. 7, the Chinese implication of correspondence is as described below:

1: to each mobile terminal i ∈ 1 ..., N}:

2: observe the channel status ω of this mobile terminal_i(t) and quene state Q (t)；

3: from set of strategiesA kind of strategy of middle selection makes

Minimize；

4: perform to upload decision-making

5: receive about ω (t-D) and the feedback information of α (t-D) and according to formula (6) renewal Q (t).

4.3 performance evaluation

The performance gap that following theorem two gives optimal case and our algorithm is given between scheme.

Theorem 2: for any one channel of mobile terminal, for arbitrary V > 0 and S > 0, Wo Menyou:

The gap of the effectiveness that best practice and our algorithm provide is:

${\stackrel{\&OverBar;}{u}}^{opt}-\frac{1}{T}\underset{t=0}{\overset{T-1}{\&Sigma;}}E\&lsqb;u\left(t\right)\&rsqb;\&le;\frac{A(1+2D)}{V}+\frac{E\&lsqb;L\left(D\right)\&rsqb;}{n}+O(1/\sqrt{S})---\left(19\right)$

Represent the maximum average utility under power consumption constraint；

Our algorithm ensure that the average energy consumption of each mobile terminal meets:

$\frac{1}{T}\underset{t=0}{\overset{T-1}{\&Sigma;}}E\&lsqb;p_i\left(t\right)\&rsqb;\&le;c_i+O\left(\sqrt{\frac{V}{T}}\right),\&ForAll;i---\left(20\right)$

Proving: in each upload procedure, drift adds penalizes technology to provide the method minimized on the right of equation.Utilizing essence When really related scheduling algorithm solves, Wo Menyou:

$E\&lsqb;\mathrm{\&Delta;}(t+D)-VU\left(t\right)|Q\left(t\right)\&rsqb;\&le;A(1+2D)-V{\stackrel{\&OverBar;}{u}}^{opt}---\left(21\right)$

Wherein, following several formulas are formulation process, and meaning of parameters is consistent, and described above.

Calculate the expectation of inequality above, Wo Menyou:

$E\&lsqb;\mathrm{\&Delta;}(t+D)-VE\&lsqb;u\left(t\right)\&rsqb;\&rsqb;\&le;A(1+2D)-V{\stackrel{\&OverBar;}{u}}^{opt}---\left(22\right)$

Calculate 0 ..., all of t in T-1}, Wo Menyou:

$E\&lsqb;L(T+D)-E\&lsqb;L\left(D\right)\&rsqb;\&rsqb;-V\underset{t=0}{\overset{T-1}{\&Sigma;}}E\&lsqb;u\left(t\right)\&rsqb;\&le;AT(1+2D)-VT{\stackrel{\&OverBar;}{u}}^{opt}---\left(23\right)$

Because Ε [L (T+D)] >=0, so recalculating inequality above, Wo Menyou:

${\stackrel{\&OverBar;}{u}}^{opt}-\frac{1}{T}\underset{t=0}{\overset{T-1}{\&Sigma;}}E\&lsqb;u\left(t\right)\&rsqb;\&le;\frac{A(1+2D)}{V}+\frac{E\&lsqb;L\left(D\right)\&rsqb;}{VT}---\left(24\right)$

Our algorithm employs Delay Feedback and carrys out approximate exact value.Approximate data and the gap of exact link allocation algorithm It isTherefore, inequality (19) is set up.Recalculate inequality (23), Wo Menyou:

E[L(t+D)]≤(V+CV)T (25)

Wherein B=E [L (D)]+A (1+2D), C is defined as meeting constant

According to the definition of L (t), Wo Menyou:

$E\&lsqb;\underset{i=1}{\overset{N}{\&Sigma;}}{Q}_i{(T+D)}^2\&rsqb;\&le;2(B+CV)T---\left(26\right)$

By Jensen ' s inequality:

$\frac{E\&lsqb;\underset{i=1}{\overset{N}{\&Sigma;}}{Q}_i(T+D)\&rsqb;}{T}\&le;\sqrt{\frac{2(B+CV)}{T}}---\left(27\right)$

In conjunction with above-mentioned two inequality, we can sum up inequality (20).

Gap between the average energy consumption upper limit and the optimal case of the algorithm that theorem 2 gives our design.? Formula (19) is as long as illustrating and selecting sufficiently large V, and average utility can be arbitrarily close to optimal value.But such as inequality (20) institute Showing, the V of the biggest quantity can produce higher energy expenditure.Collaborative cloud computing can be by adjusting the size of V, at effectiveness and energy The decision-making of elasticity is carried out between amount loss.It addition, bigger S value can also make solution closer to optimal value.But, simultaneously it Take longer for calculates and bigger memory space.

5. Performance Evaluation

In order to assess the performance of our proposed framework, we complete a series of emulation experiment in this section.At me Experiment in, a collaborative cloud computing comprises 8 mobile terminals.Between uploading at twice, mobile terminal at most produces 20 kinds Different types of data block.In order to ensure the reliability of data, each data block is replicated to three backups, is stored in these and moves In dynamic terminal.Having 4 kinds of state representation uploading channel is Ω=(0,1,2,3).In each upload procedure, channel status with Equal probabilities randomly generates in Ω.Can be obtained by experience energy consumption model, transmission energy and channel status are inversely proportional to.Therefore different letters The loss uploading energy under road state is { 60,30,12,6}J.The mean consumption of each mobile terminal is constrained to 8J.Acquiescence feedback Delay is D=2, and sample size is S=30.

The checking of 5.1 energy consumptions-utility balance

Parameter V as the control variable between effectiveness and energy loss, is first verified that the effectiveness of parameter V by us.? In Fig. 1 a and Fig. 1 b, in Fig. 1 a, abscissa is the change of V-value, and vertical coordinate is average utility value；In Fig. 1 b, abscissa is the change of V-value Changing, vertical coordinate is average energy consumption.

Fig. 1 a shows the increase along with V, and effectiveness improves, constantly close to optimal value.But, lifting process is along with effectiveness Slow down gradually close to optimal value.This results show theorem two, shows that to promote average utility permissible along with O (1/V) Reach optimal value.But, the lifting of effectiveness adds the energy burden of mobile terminal.Fortunately when V < when 600, average energy Loss is less than constraint c=8J.Even if as V=1200, average energy loss is only high than constraint by 1.67%.In a word, result Between display effectiveness and energy consumptionBalance, is consistent with theorem two.In order to be issued to more at energy constraint High effectiveness, we arrange V=600 in following experiment.

We verify whether energy expenditure meets constraint under different V further.Fig. 2 can be seen that average energy is lost It is decreased obviously under the value of different V, and can fast approaching power dissipation constraints.In Fig. 2, abscissa is 400 upload procedure, Vertical coordinate is the energy consumption in each upload procedure.V-value is the least, and decrease speed is the fastest.Working as V=1, when 100, average energy loss exists Below constraint is rapidly dropped to after uploading several times.Meanwhile, working as V=600, the consumption of 1200 is also at t > after 100 close to constraint. These results show that bigger V can reach higher effectiveness, but average energy loss also extended close to the time retrained ?.

The impact of 5.2 feedback delays

Fig. 3 a and Fig. 3 b shows the performance impact of parameter D.Can find out that parameter D has for effectiveness in both figures the most aobvious Write impact.In Fig. 3 a, abscissa is that system feedback postpones, and vertical coordinate is average utility.In Fig. 3 b, abscissa is that system feedback is prolonged Late, left side vertical coordinate is average energy consumption, and right side vertical coordinate is energy consumption/effectiveness ratio, represents the efficiency of energy consumption.

When feedback delay D is extended, average utility declines substantially.Can sum up intuitively and reduce prolonging of the mobile cloud of cooperation Feedback can be obviously improved its performance late.Having benefited from piggybacking technology, feedback information can be passed at short notice. Although Fig. 3 b shows the decline of energy loss, consume and show when D becomes big, under the efficiency of the energy consumption of system with the ratio of effectiveness Fall.This means that cooperatively move cloud consumes more multi-energy when uploading data.

The impact of 5.3 channel conditions

TABLE 1:Distribution of Channel State

In order to check impact and the suitability of our algorithm of channel conditions, We conducted a series of experiment, its The distribution situation of middle channel status changes.The number of times of upload procedure increases to 900 times, and is divided into three phases.First In the stage (t < 300), the channel status of all terminals is all equiprobable.Stage two is 300 < t < 600, as shown in Table 1, and may Property changes in distribution be type one (Type 1) be distributed (poor channel situation).At stage three, t > 600, channel conditions becomes class Type two (Type 2) distribution (preferable channel status).

Fig. 4 a and Fig. 4 b shows the average utility of 900 upload procedure and average energy loss.Fig. 4 a abscissa is 900 Secondary upload procedure, vertical coordinate is the average utility in each upload procedure；Fig. 4 b abscissa is 900 upload procedure, vertical coordinate For the average energy consumption in each upload procedure.

The independent emulation experiment average more than 100 times has all been carried out in upload procedure each time.The vertical line of Fig. 4 a shows Having shown once channel state variations, the mobile cloud of cooperation just can produce new optimal value.This result shows that our algorithm can To adapt to uncertain communication environment.It addition, the average utility that second stage reaches is less than the first stage, therefore we are permissible Find that channel status is the bottleneck that system utility promotes.Fig. 4 b describes the average energy consumption of each upload procedure.Work as channel During situation fluctuation, two appearance of fluctuating significantly.But, but after upload procedure several times, energy expenditure is the most again close to about Bundle.

In view of different mobile terminal to upload channel conditions different, we next prove algorithm for one mobile eventually The situation that end channel status changes.600 upload procedure are divided into two stages.In the first phase (t < 300), all The channel status of mobile terminal is evenly distributed on Ω.In the stage 2 (t > 300), the channel status of mobile terminal 1 becomes type One distribution, the channel state variations of mobile terminal 8 is type two distribution.Other mobile terminal maintains identical probability and divides Cloth.

In Fig. 5 a to Fig. 5 d, Fig. 5 a abscissa represents 600 upload procedure, and vertical coordinate represents putting down in each upload procedure All effectiveness；Fig. 5 b abscissa represents 600 upload procedure, and vertical coordinate represents the average energy consumption in each upload procedure；Fig. 5 c is horizontal 600 upload procedure of coordinate representation, vertical coordinate terminal 1 and the average decision-making of terminal 8；Fig. 5 d abscissa represents and is transmitted through on 600 times Journey, vertical coordinate terminal 1 and the cumulative mean decision-making of terminal 8.

It can be seen that when channel status changes in distribution, average utility and average energy consumption are all from Fig. 5 a and Fig. 5 b Significant change does not occur.In order to study the reason of this phenomenon, we depict the average decision-making of mobile terminal 1 and 8.On each It is transmitted through journey and repeats 100 independent experiments.The definition of α i (t) is pointed out, the meansigma methods of decision value means the most greatly mobile terminal Tend to upload data in this upload procedure.In the stage one, the average decision-making of terminal 1 and terminal 8 consistent.But In the stage two, the channel status of terminal 1 becomes weaker, and average decision-making is the most wavy, is reduced in a lot of upload procedure Less than 0.5.Contrary, in the stage two, the average decision-making of terminal 8 is maintained at more than 0.95.In order to clearly indicate this trend, We depict the accumulated value of the average decision-making of each upload procedure in Figure 5The curve of terminal 1 is at t=300 Time show obvious downward trend, and terminal 8 rises.These experimental results show that we have well adapting to property by algorithm. Although mobile terminal does not knows the channel status of current time other-end when uploading, rely on distributed related optimization, Our algorithm still can allow distributed terminal collaborative work.

5.4 Performance comparision

In order to show that (distributed being correlated with uploads decision-making distributed correlated upload to our algorithm Decision, DCUD) performance boost, we compare with three comparator algorithms.Greedy algorithm: mobile terminal is at average energy Determine to upload data when amount loss is less than constraint.OPERA: provide a kind of online algorithm and be devoted to reduce energy loss, fall The drop probabilities of low data.Unlike other algorithm, CENTER ALGORITHM CENTRAL based on greedy algorithm is that one is known The center type dispatching method of all mobile terminal situations.

Fig. 6 a and Fig. 6 b shows the Performance comparision of four kinds of algorithms under different-energy retrains, and result shows our algorithm There is obvious advantage, especially in the case of this energy constraint of emergency processing is very strong.In Fig. 6 a, abscissa represents different Power consumption constraint value, vertical coordinate represents the average utility of algorithms of different；In Fig. 6 b, abscissa represents different power consumption constraint values, vertical The average energy consumption of coordinate representation algorithms of different.Wherein in Fig. 6 a and Fig. 6 b, each figure includes four block diagrams, each column Scheme vertical four posts and represent DCUD, GREEDY, OPERA, CENTRAL respectively.

Such as Fig. 6 a and Fig. 6 b, as energy constraint c_iWhen=1, effectiveness improves 77.5% than greedy algorithm.Work as energy constraint Time more loose, our algorithm is still good than other reference algorithms except centralization algorithm.But CENTRAL algorithm have employed collection The dispatching method of Chinese style, it is assumed that can know the information of all mobile terminals, chapter 1 is it is stated that this is the most this Assume unactual.

6. sum up

The storage capacity that collaborative mobile cloud utilizes mobile terminal around to promote this mobile terminal, is extension mobile terminal A kind of effective mechanism of computing capability.A main challenge is how to upload band redundancy in collaborative mobile cloud The data of backup are to the cloud data center of far-end.Herein this problem is modeled as the maximization of utility problem of power consumption constraint.It is subject to The inspiration of distributed related optimization, we devise distribution on line formula dispatching method so that each mobile terminal is not Decision-making can be independently carried out in the case of knowing future channel.We, for reducing the complexity of algorithm, have carried out rigorous theory and have divided Analysis, reduces the search volume of algorithm.Theory analysis shows that our algorithm is permissible on the premise of meeting power dissipation constraints Close to optimum average potency.We conducted a series of emulation experiment, illustrate having of distribution on line formula optimized algorithm Effect property and superiority.

Claims (6)

1. the band redundant data method for uploading of effectiveness perception in the mobile cloud that cooperates, it is characterised in that: the mobile cloud of cooperation comprises N Individual mobile terminal, represents the t upload procedure of cooperation cloud, d with t_i(t)=(d_i1(t),d_i2(t),...,d_iK(t)) represent the In t upload procedure, whether data block k is stored in i-th mobile terminal, and K is the number of different types of data block, ω_i(t) table Show the channel status of i-th mobile terminal, ω in the t upload procedure_iThe letter of (t) numerical value the biggest expression i-th mobile terminal Road is in good state；Uploading channel when can use, the mobile cloud of cooperation carries out a secondary data and uploads；

Comprise the following steps:

(1) the channel status ω of each mobile terminal is observed_iT (), each mobile terminal decides whether to upload it according to channel status Data；Observation is used for representing energy expenditure constraint queue Q (t), described Q (t)=(Q₁(t),...,Q_N(t)), described Q_i(t) The virtual queue defined for each mobile terminal, queue Q (t) is the most stable, represents that energy expenditure is the least；

(2) in the t+D upload procedure, D is the feedback delay of system, and all mobile terminals receive about in the t upload procedure All states and decision information, each mobile terminal in this upload procedure exists One strategy of middle selectionDescribedRepresent all possible set of strategies Close,Represent that a kind of optimal strategy can make the upper limit of drift-penaltyHave Minima, whereinRepresent perform this strategy time i-th terminal estimate energy consumption,Represent and perform estimating of this strategy Effectiveness, parameter V isWeight, control the balance between energy consumption and effectiveness；

(3) optimum performing to determine uploads decision-makingRepresentMiddle i-th Terminal uploads decision-making；

(4) at the end of the t time is uploaded, information ω (t-D) about channel status and decision-making α (t-D) are received, again to Q (t) It is updated.

The band redundant data method for uploading of effectiveness perception in the mobile cloud of cooperation the most according to claim 1, it is characterised in that: In step (1), described Q_iT () is updated the t time upload procedure by this equation following:

Q_i(t+1)=max{Q_i(t)+p_i(t-D)-c_i,0}

Wherein, c_iIt is the average energy consumption binding occurrence of i-th mobile terminal, p_i(0)=0, p_i(-1)=...=p_i(-D)=0, Each mobile terminal maintains queue Q (t), utilizes feedback information to update queue, work as void after upload procedure terminates each time When intending string stability, namelyThe most just disclosure satisfy that energy expenditure retrains, wherein, E represents Seeking mathematic expectaion, defining Lyapunov function is:

$L\left(t\right)=\frac{1}{2}\underset{i=1}{\overset{N}{\&Sigma;}}{Q}_i{\left(t\right)}^2---\left(1\right)$

The Lyapunov function representation degree of congestion of Q (t) queue, the virtual queue of the change the least expression mobile terminal of L (t) There is strong stability, in other words, it is possible to meet energy expenditure constraint.

The band redundant data method for uploading of effectiveness perception in the mobile cloud of cooperation the most according to claim 1, it is characterised in that: In step (2), work as ω_iOptimal strategy when () increases tIn i-th, i.e. the decision-making of i-th terminalAlso With ω_iT () increases, the following form of optimal strategy:

${\widehat{\mathrm{\&alpha;}}}_i\left({\mathrm{\&omega;}}_i\right(t\left)\right)=\left\{\begin{array}{cc}0& {\mathrm{if\&omega;}}_i\left(t\right)<{\mathrm{\&omega;}}_i^{*}\left(t\right)\\ 1& {\mathrm{if\&omega;}}_i\left(t\right)\&GreaterEqual;{\mathrm{\&omega;}}_i^{*}\left(t\right)\end{array}\right.---\left(2\right)$

ω^*T () represents threshold status, when channel status is inferior to ω^*(t)(ω_i(t) ＜ ω^*(t)) time, the decision value of mobile terminal It is 0, does not uploads；When channel status is better than ω^*(t)(ω_i(t)≥ω^*(t)) time, the decision value of mobile terminal is 1, also It is exactly to upload data.Thus this problem is converted into each mobile terminal seeking ω^*(t), described ω^*T () has that | Ω | is individual may be taken Value, therefore, the quantity of efficient solution is reduced to

The band redundant data method for uploading of effectiveness perception in the mobile cloud of cooperation the most according to claim 1, it is characterised in that: In step (2), the V for given > 0, in each upload procedure:

$E\&lsqb;\mathrm{\&Delta;}(t+D)-Vu\left(t\right)|Q\left(t\right)\&rsqb;\&le;A(1+2D)-\underset{i=1}{\overset{N}{\&Sigma;}}{c}_i{Q}_i\left(t\right)+E\&lsqb;\underset{i=1}{\overset{N}{\&Sigma;}}{p}_i\left(t\right)Q_i\left(t\right)-Vu\left(t\right)|Q\left(t\right)\&rsqb;---\left(3\right)$

This formula is obtained by mathematical derivation, does not has concrete physical meaning, only represents that the upper limit of inequality left side item is inequality The right item；E represents the effectiveness asking mathematic expectaion, u (t) to represent the t time upload procedure；Parameter V is the weight of u (t)；

Wherein

Utilizing drift to add and penalize technology, in each upload procedure, each mobile terminal exists Middle selection one, minimizes the right item of inequality (3), to solve the maximization of utility problem of power consumption constraint.

The band redundant data method for uploading of effectiveness perception in the mobile cloud of cooperation the most according to claim 4, it is characterised in that: Work as Utilization strategiesTime, each mobile terminal i is at strategyUnder estimate energy consumptionWith estimate benefit u^(m) T () can be approximated to be:

${\tilde{p}}_i^{\left(m\right)}\left(t\right)=\frac{1}{S}\underset{s=1}{\overset{S}{\&Sigma;}}\widehat{p_i}\left({\widehat{a_i}}^{\left(m\right)}\right({\mathrm{\&omega;}}_i\left(t-D-s\right)),{\mathrm{\&omega;}}_i(t-D-s\left)\right)---\left(5\right)$

${\tilde{u}}^{\left(m\right)}\left(t\right)=\frac{1}{S}\underset{s=1}{\overset{S}{\&Sigma;}}\hat{u}\left({\hat{a}}^{\left(m\right)}\right({\mathrm{\&omega;}}_i\left(t-D-s\right)),d(t-D-s\left)\right)---\left(5\right)$

Wherein, S is a positive number, represents sample size.

The band redundant data method for uploading of effectiveness perception in the mobile cloud of cooperation the most according to claim 1, it is characterised in that: For any one channel of mobile terminal, for arbitrary V > 0 and S > 0, S is the positive number representing sample size, parameter V (>= 0) being the balance in order to control between string stability and effectiveness, the less V mobile cloud that means to cooperate tends to hold queue Stablize the lowest energy consumption rather than collect more valid data block, then having:

The gap of the effectiveness that best practice and this method are tried to achieve is:

${\stackrel{\&OverBar;}{u}}^{opt}-\frac{1}{T}\underset{t=0}{\overset{T-1}{\&Sigma;}}E\&lsqb;u\left(t\right)\&rsqb;\&le;\frac{A(1+2D)}{V}+\frac{E\&lsqb;L\left(D\right)\&rsqb;}{VT}+O(1/\sqrt{S})---\left(6\right)$

Wherein,Represent the best practice maximum average utility under power consumption constraint,Represent that this method is tried to achieve Effectiveness；WhereinT represents and always uploads number of times, and L (D) represents the Lyapunov functional value that parameter is D,Expression comprisesThe multinomial of item；

This method ensure that the average energy consumption of each mobile terminal meets:

$\frac{1}{T}\underset{t=0}{\overset{T-1}{\&Sigma;}}E\&lsqb;p_i\left(t\right)\&rsqb;\&le;c_i+O\left(\sqrt{\frac{V}{T}}\right),\&ForAll;i---\left(7\right)$

Wherein,Represent the average energy consumption of i-th mobile terminal, c_iRepresent power consumption constraint value,Represent CompriseThe multinomial of item；

Collaborative cloud computing can carry out the decision-making of elasticity by adjusting the size of V between effectiveness and energy loss；It addition, more Big S value can also make solution closer to optimal value；But, what it took longer for simultaneously calculates and bigger storage Space.