CN102752810B - Task negotiation method of HGA (hybrid genetic algorithm)-based wireless sensor network node - Google Patents

Abstract

The invention relates to a task negotiation method of an HGA (hybrid genetic algorithm)-based wireless sensor network node. The method sequentially comprises the following steps of: 1, before the task negotiation begins, representing requests of a tenderer in the negotiation as an issue of ZA, and representing the resources of a bidder as an issue of TA; 2, after the task negotiation begins, generating a strategy generation offer according to the HGA by the TA, and then sending the strategy generation offer to the ZA; and 3, when sending and receiving processes finish, evaluating the received offer by the ZA according to an evaluation strategy and performing corresponding treatment according to the evaluation result. The method has the advantages that the task negotiation problem of the wireless sensor network node is converted into the negotiation problem of an Agent issue, and independent negotiation of the task is realized; 2, the negotiation result enables the utility sum of negotiating each node to be similar to the highest social benefits; 3, the negotiation efficiency is improved; and 4, the negotiation flexibility and finishing rate are improved.

Description

Based on the wireless sensor network node task machinery of consultation of HGA

Technical field

The present invention relates to the mechanics of communication of wireless sensor network node, specifically a kind of based on HGA(genetic algorithm) the machinery of consultation of wireless sensor network node task.

Background technology

Along with the development of wireless sensor network technology, its application is also in continuous expansion, expand to other field by initial military field, such as disaster early warning and relief can be completed, task that traditional system such as family health care detects, space exploration cannot complete.Because wireless sensor network generally just seldom manually carries out managing and intervening after deployment substantially, this just needs wireless sensor network can carry out consulting from main task.Equipment major part in wireless sensor network is all wireless device, their network capacity, computing capability and electricity have certain restriction, task between device node is consulted normally when to counter-party information, omniscient, computational resource are limited and have and carry out when strict time restriction, and therefore good task negotiation mechanism should be able to fast and effeciently search out optimal solution or approximate optimal solution in the negotiation space of a larger complexity.The special research for negotiation efficiency is not also had in the research that the current task about wireless sensor network node is consulted.

Agent technology is the result of distributed computing technology and Artificial Intelligence Development.Agent has independence, interactivity, collaborative and these features intelligent.Agent has the computational resource and the Behavior-Based control mechanism of local in self that belong to himself, can not have in extraneous direct operated situation, according to its internal state and the environmental information perceived, is determining and control the behavior of self.Simultaneously can effectively with other Agent collaborative works, be well suited for the negotiations process for representing between node in wireless sensor network and node.

Summary of the invention

Technical problem to be solved by this invention is, provides a kind of and can simplify the negotiations process of wireless sensor network node task, the wireless sensor network node task machinery of consultation of raising negotiation efficiency.

Method of the present invention in turn includes the following steps:

Step one, before task consults to start, is first the subject under discussion of ZA by consulting the requirement representation of Tender Based side, is the subject under discussion of TA by the resource representation of tenderer simultaneously; The respective described demand of node in the wireless sensor network that participation task is consulted and resource conversion be Agent at many subjects under discussion value vector spatially: $\stackrel{\&RightArrow;}{V}=<{v^{i_k}}_1,{v^{i_k}}_2,\mathrm{\&Lambda;},{v^{i_k}}_n>;$

In above formula, represent subject under discussion i _ka value, Ω _krepresent the codomain of this subject under discussion, represent many subjects under discussion vector; For wireless sensor network, can using energy loss as one of subject under discussion;

Step 2, after negotiation starts, TA utilizes HGA according to proposal generation strategy proposing offers, then sends to ZA;

Step 3, when transmission with after receiving course terminates, ZA processes according to the result assessed the proposal row assessment received accordingly according to assessment strategy.

In above-mentioned steps two, the flow process of described HGA is:

A. produce initial population, wherein comprise 1 individuality making Agent effectiveness the highest and from initial population set n-1 of random selecting individual;

B. fitness individual in population is calculated, and according to fitness size to the individuality sequence in population; Described fitness is determined by fitness function fit, and its definition mode is:

1) utility function is defined.

In order to evaluate the effectiveness participating in each Agent consulted, definition Agent a is at many subject under discussion vectors on utility function be:

$U_a\left(\stackrel{\&RightArrow;}{\mathrm{offer}}\right)=\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_a^{i_k}\&times;{\mathrm{\&Phi;}}_a^{i_k}.$

Wherein, represent that Agent a is at subject under discussion i _kon value, represent that Agenta is at subject under discussion i _kon minimum value, represent that Agent a is at subject under discussion i _kon maximum occurrences, represent that Agenta is at subject under discussion i _knormalization result.If value is set, then represent with the mean value of set element;

represent subject under discussion i _mweights, and

2) fitness function is defined

Fitness function fit is defined as follows:

$\mathrm{fit}=\mathrm{\&alpha;}\&times;\mathrm{TP}\left(t\right)=\frac{U\left(\stackrel{\&RightArrow;}{\mathrm{offer}}\right)}{U\left({\stackrel{\&RightArrow;}{\mathrm{offer}}}_{\max }\right)}+(1-\mathrm{\&alpha;}\&times;\mathrm{TP}\left(t\right))\&times;(1-\frac{\mathrm{dist}(\stackrel{\&RightArrow;}{\mathrm{offer}},{\stackrel{\&RightArrow;}{\mathrm{offer}}}_{\mathrm{opponent}})}{\mathrm{dist}({\stackrel{\&RightArrow;}{\mathrm{offer}}}_{\max },{\stackrel{\&RightArrow;}{\mathrm{offer}}}_{\min })})$

Wherein, α is balance factor, and between zero and one, TP (t) is time factor to α value, and TP (t) ∈ [0,1], T ^deadlinefor task consults deadline, β is for proposing the factor, and as β >1, then Agent is eager to compromise; When β=1, then the linear compromise of Agent; As β <1, Agent compromises hardly in time; for many subject under discussion vectors that effectiveness is maximum, for many subject under discussion vectors that effectiveness is minimum; represent corresponding many subject under discussion vectors that Agent a proposes with many subject under discussion vectors that the other side Agent proposes between distance; represent the maximum many subjects under discussion vector of effectiveness ( with many subject under discussion vectors that effectiveness is minimum between distance;

C. the proposal as individual can meet the requirement of the other side, then export the proposal that the highest individuality of this fitness is consulted as epicycle, HGA operates end, otherwise produces population of future generation;

D. produce population of future generation, its step comprises:

D-1. in proportion r by individual replicate the highest for fitness in this generation in the next generation, r value 0.1;

D-2. judge whether individual amount of future generation suits the requirements, if so, then calculate fitness individual in population of new generation, and according to fitness size to the individuality sequence in new population, otherwise according to crossing-over rate P _cindividuality is intersected, according to aberration rate P _mcarry out individual variation;

Wherein, crossing-over rate P _cfor:

$P_c=\left\{\begin{array}{c}{P}_{c1}-(P_{c1}-P_{c2})\&times;\frac{({\mathrm{fit}}_{l\arg e}-{\mathrm{fit}}_{\mathrm{avg}})}{{\mathrm{fit}}_{\max }-{\mathrm{fit}}_{\mathrm{avg}}},{\mathrm{fit}}_{l\arg e}\&GreaterEqual;{\mathrm{fit}}_{\mathrm{avg}}\\ P_{c1},{\mathrm{fit}}_{\mathrm{lag}}<{\mathrm{fit}}_{\mathrm{avg}}\end{array}\right.$

Fit in above formula _maxand fit _avgbe respectively maximum adaptation degree and the average fitness of colony, fit _largefor fitness larger in individuality will be intersected.

Aberration rate P _mfor:

$P_m=\left\{\begin{array}{c}{P}_{m1}-(P_{m1}-P_{m2})\&times;\frac{({\mathrm{fit}}_{\max }-\mathrm{fit})}{{\mathrm{fit}}_{\max }-{\mathrm{fit}}_{\mathrm{avg}}},\mathrm{fit}\&GreaterEqual;{\mathrm{fit}}_{\mathrm{avg}}\\ P_{m1},{\mathrm{fit}}_{\mathrm{lag}}<{\mathrm{fit}}_{\mathrm{avg}}\end{array}\right.$

Wherein P _c1, P _c2, P _m1, P _m2be and be less than 1 constant being greater than 0.

D-3. accept to worsen with certain probability P and separate, thus make algorithm have the stronger global optimization ability escaped local extremum and avoid Premature Convergence. for the individual j in previous generation population, entering the received probability of corresponding individual j ' in new population intersected or obtained after variation is:

$P(j\&DoubleRightArrow;j^{\&prime;})=\left\{\begin{array}{c}\begin{array}{cc}1,& \mathrm{fit}\left(j^{\&prime;}\right)\&le;\mathrm{fit}\left(j\right)\end{array}\\ \exp \left(\frac{\mathrm{fit}\left(j^{\&prime;}\right)-\mathrm{fit}\left(j\right)}{S}\right),\mathrm{fit}\left(j^{\&prime;}\right)\&le;\mathrm{fit}\left(j\right)\end{array}\right.$

Wherein, S is the annealing regulation factor to `; Fit (j) is the fitness of the individual j in previous generation population; Fit (j ') is the fitness of the corresponding individual j ' of the novel species masses;

D-4. d-2 to d-3 is repeated, until population at individual number suits the requirements.

In described step 3, for consult in negotiation time to reach, consult in negotiation time not reach and negotiation time to but three kinds of situations are not reached in negotiation, specify corresponding processing mode; Concrete assessment strategy is:

$E=\left\{\begin{array}{c}\mathrm{accept},U\left({\stackrel{\&RightArrow;}{\mathrm{offer}}}_{\mathrm{oppent}}\right)\&GreaterEqual;U\left(\stackrel{\&RightArrow;}{\mathrm{offer}}\right)\\ \mathrm{offer}\left(\mathrm{next}\right),\mathrm{otherwise}\\ \mathrm{accept}\left({\mathrm{fit}}_{\max }\right),t>T^{\mathrm{deadline}}\end{array}\right.$

In above formula, the effectiveness that in the negotiation of expression task, the other side proposes, expression task is our effectiveness of proposing in consulting, and t represents the time of negotiation, T ^deadlinerepresent the deadline of consulting.

By this assessment strategy, ZA assesses the proposal received, and then consults to reach as met its demand; Otherwise go to step two to enter and consult next time; When negotiation time arrives but when consulting unsuccessful, for ensureing that task completes, then in selecting the other side to propose, the maximum proposal of fitness is as negotiation result.

The advantage of the inventive method is: 1. the negotiation problem task negotiation problem of wireless sensor network node being converted to Agent subject under discussion, makes full use of the independence of Agent, interactivity, collaborative and the intelligent autonomous negotiating realizing task; 2. devise and consider the fitness function that task consults each node effectiveness, consult to obtain result and make the effectiveness summation of consulting each node be similar to the highest social benefit; 3. devise HGA algorithm, can automatically adjust crossing-over rate and aberration rate.And can accept to worsen solution on certain probability.Optimal solution or approximate optimal solution can be fast and effeciently searched out in the negotiation space of a larger complexity.Improve the efficiency of negotiation; 4. for consult in negotiation time to reach, consult in negotiation time not reach and negotiation time to but three kinds of situations are not reached in negotiation, define corresponding operation, improve flexibility and the completion rate of negotiation.

Accompanying drawing explanation

Fig. 1 is the overview flow chart of the inventive method;

Fig. 2 is the flow chart of HGA algorithm in the inventive method.

Embodiment

See Fig. 1 and Fig. 2, in embodiments of the present invention, when certain node in wireless sensor network needs to carry out task negotiation with other node, step is as follows:

Step one, before task consults to start, the respective demand of node in the wireless sensor network that participation task is consulted and resource conversion be Agent at many subjects under discussion value vector spatially:

$\stackrel{\&RightArrow;}{V}=<{v^{i_k}}_1,{v^{i_k}}_2,\mathrm{\&Lambda;},{v^{i_k}}_n>;$

In above formula, represent subject under discussion i _ka value

Step 2, after negotiation starts, TA utilizes HGA according to proposal generation strategy proposing offers (Offer), then sends to ZA.The product process of offer is as follows:

1. produce initial population, wherein comprise 1 the highest chromosome (individuality) of effectiveness and from initial population set random selecting n-1 bar individuality.

2. calculate fitness fit individual in population, and according to fitness size to the individuality sequence in population.

3. the offer as individual can meet the requirement of the other side, then export the offer that the highest individuality of this fitness is consulted as epicycle, HGA operates end, otherwise produces population of future generation.

4. the step producing population of future generation comprises:

4.1 in proportion r by individual replicate the highest for fitness in this generation in the next generation,

4.2 judge that individual amount of future generation suits the requirements, and if so, then calculate fitness individual in population of new generation, and according to fitness size to the individuality sequence in new population,

Otherwise according to crossing-over rate P _cindividuality is intersected, according to aberration rate P _mcarry out individual variation.

Wherein, crossing-over rate P _cwfor: $P_c=\left\{\begin{array}{c}{P}_{c1}-(P_{c1}-P_{c2})\&times;\frac{({\mathrm{fit}}_{l\arg e}-{\mathrm{fit}}_{\mathrm{avg}})}{{\mathrm{fit}}_{\max }-{\mathrm{fit}}_{\mathrm{avg}}},{\mathrm{fit}}_{l\arg e}\&GreaterEqual;{\mathrm{fit}}_{\mathrm{avg}}\\ P_{c1},{\mathrm{fit}}_{\mathrm{lag}}<{\mathrm{fit}}_{\mathrm{avg}}\end{array}\right.$

Fit in above formula _maxand fit _avgbe respectively maximum adaptation degree and the average fitness of colony, fit _largefor fitness larger in individuality will be intersected.

Aberration rate P _mfor: $P_m=\left\{\begin{array}{c}{P}_{m1}-(P_{m1}-P_{m2})\&times;\frac{({\mathrm{fit}}_{\max }-\mathrm{fit})}{{\mathrm{fit}}_{\max }-{\mathrm{fit}}_{\mathrm{avg}}},\mathrm{fit}\&GreaterEqual;{\mathrm{fit}}_{\mathrm{avg}}\\ P_{m1},{\mathrm{fit}}_{\mathrm{lag}}<{\mathrm{fit}}_{\mathrm{avg}}\end{array}\right.$

Wherein P _c1, P _c2, P _m1, P _m2be and be less than 1 constant being greater than 0.

4.3 separate by accepting to worsen with certain probability P, thus make algorithm have the stronger global optimization ability escaped local extremum and avoid Premature Convergence. for the individual j in previous generation population, entering the received probability of corresponding individual j ' in new population intersected or obtained after variation is:

$P(j\&DoubleRightArrow;j^{\&prime;})=\left\{\begin{array}{c}\begin{array}{cc}1,& \mathrm{fit}\left(j^{\&prime;}\right)\&le;\mathrm{fit}\left(j\right)\end{array}\\ \exp \left(\frac{\mathrm{fit}\left(j^{\&prime;}\right)-\mathrm{fit}\left(j\right)}{S}\right),\mathrm{fit}\left(j^{\&prime;}\right)\&le;\mathrm{fit}\left(j\right)\end{array}\right.$

Wherein, S is the fitness of annealing regulation factor .fit (j) for the individual j in previous generation population to `; Fit (j ') is the fitness of the corresponding individual j ' of the novel species masses;

4.4 repeat 4.2 and 4.3, until population at individual number suits the requirements.The parameter used in HGA algorithm is as shown in table 1.

Table 1

Parameter factors	Value
		Balance factor α (ZA)	0.8
Balance factor α (TA)	0.5
		Propose factor-beta	0.6
Consult td deadline eadline	500
		Population Size n	80
Individual replicate ratio r	0.1
		Crossing-over rate constant P c1	0.9
Crossing-over rate constant P c2	0.6
		Aberration rate constant P m1	0.1
Aberration rate constant P m2	0.05
		Annealing regulation factor S	0.95

Step 3, when transmission with after receiving course terminates, ZA to assess the Offer received according to assessment strategy and processes accordingly according to the result assessed.Concrete be exactly for consult in negotiation time to reach, consult in negotiation time do not reach and negotiation time to but three kinds of situations are not reached in negotiation, specify corresponding processing mode, guarantee finish the work through consultation.Concrete assessment strategy is:

$E=\left\{\begin{array}{c}\mathrm{accept},U\left({\stackrel{\&RightArrow;}{\mathrm{offer}}}_{\mathrm{oppent}}\right)\&GreaterEqual;U\left(\stackrel{\&RightArrow;}{\mathrm{offer}}\right)\\ \mathrm{offer}\left(\mathrm{next}\right),\mathrm{otherwise}\\ \mathrm{accept}\left({\mathrm{fit}}_{\max }\right),t>T^{\mathrm{deadline}}\end{array}\right.$

By this assessment strategy, ZA assesses the offer received, and then consults to reach as met its demand; Otherwise go to step two and enter next round negotiation; When negotiation time arrives but when consulting unsuccessful, for ensureing that task completes, then in the proposal selecting the other side to propose, the maximum proposal of fitness is as negotiation result.

Claims (3)

1. a wireless sensor network node task machinery of consultation, is characterized in that: the method in turn includes the following steps:

Step one, before task consults to start, is first the subject under discussion of ZA by consulting the requirement representation of Tender Based side, is the subject under discussion of TA by the resource representation of tenderer simultaneously; The respective described demand of node in the wireless sensor network that participation task is consulted and resource conversion be Agent at many subjects under discussion value vector spatially:

In above formula, represent subject under discussion i _kvalue, Ω _krepresent the codomain of this subject under discussion, represent many subjects under discussion vector; For wireless sensor network, can using energy loss as one of subject under discussion;

Step 2, after negotiation starts, TA utilizes HGA according to proposal generation strategy proposing offers (Offer), then sends to ZA;

Step 3, when transmission with after receiving course terminates, ZA to assess the proposal received (Offer) according to assessment strategy and processes accordingly according to the result assessed;

In step 2, the flow process of described HGA is:

A. produce initial population, wherein comprise 1 individuality making Agent effectiveness the highest and from initial population set n-1 of random selecting individual;

B. fitness individual in population is calculated, and according to fitness size to the individuality sequence in population;

C. the proposal as individual can meet the requirement of the other side, then export the proposal that the highest individuality of this fitness is consulted as epicycle, HGA operates end, otherwise produces population of future generation;

D. the step producing population of future generation comprises:

D-1. in proportion r by individual replicate the highest for fitness in this generation in the next generation, r value 0.1;

D-2. judge whether individual amount of future generation suits the requirements, if so, then calculate fitness individual in population of new generation, and according to fitness size to the individuality sequence in new population, otherwise according to crossing-over rate P _cindividuality is intersected, according to aberration rate P _mcarry out individual variation;

Wherein, crossing-over rate P _cfor:

Fit in above formula _maxand fit _avgbe respectively maximum adaptation degree and the average fitness of colony, fit _largefor fitness larger in individuality will be intersected;

Aberration rate P _mfor:

Wherein P _c1, P _c2, P _m1, Pm2 is and is less than 1 constant being greater than 0;

D-3. accept to worsen with certain probability P and separate, thus make algorithm have the stronger global optimization ability escaped local extremum and avoid Premature Convergence. for the individual j in previous generation population, entering the received probability of corresponding individual j ' in new population intersected or obtained after variation is:

Wherein, S is the annealing regulation factor; Fit (j) is the fitness of the individual j in previous generation population; Fit (j ') is the fitness of the corresponding individual j ' of the novel species masses;

D-4. d-2 to d-3 is repeated, until population at individual number suits the requirements.

2. wireless sensor network node task according to claim 1 machinery of consultation, is characterized in that: described fitness is determined by fitness function fit, and its definition mode is:

1) utility function is defined;

In order to evaluate the effectiveness participating in each Agent consulted, definition Agent a is at many subject under discussion vectors on utility function be:

Wherein, represent that Agent a is at subject under discussion i _kon value, represent that Agent a is at subject under discussion i _kon minimum value, represent that Agent a is at subject under discussion i _kon maximum occurrences, represent that Agent a is at subject under discussion i _knormalization result; If value is set, then represent with the mean value of set element;

represent subject under discussion i _mweights, and

2) fitness function is defined

Fitness function fit is defined as follows:

Wherein, α is balance factor, and α value between zero and one; TP (t) is time factor, and TP (t) ∈ [0,1], T ^deadlinefor task consults deadline; β is for proposing the factor, and as β >1, then Agent is eager to compromise, when β=1, then and the linear compromise of Agent, as β <1, Agent compromises hardly in time; for many subject under discussion vectors that effectiveness is maximum, for many subject under discussion vectors that effectiveness is minimum; represent corresponding many subject under discussion vectors that Agent a proposes with many subject under discussion vectors that the other side Agent proposes between distance; represent many subject under discussion vectors that effectiveness is maximum with many subject under discussion vectors that effectiveness is minimum between distance.

3. wireless sensor network node task according to claim 1 machinery of consultation, it is characterized in that: in described step 3, for consult in negotiation time to reach, consult in negotiation time not reach and negotiation time to but three kinds of situations are not reached in negotiation, specify corresponding processing mode; Concrete assessment strategy is:

By this assessment strategy, ZA assesses the proposal received, and then consults to reach as met its demand;

Otherwise go to step two to enter and consult next time; Arrive when negotiation time but when consulting unsuccessful, for ensureing that task completes, then select to propose that the maximum proposal of fitness is as negotiation result.