CN104219154A - Resource optimization method under network coding environment based on ant colony optimization algorithm - Google Patents

Abstract

The invention discloses a resource optimization method under a network coding environment based on an ant colony optimization algorithm. The resource optimization method is used for optimizing network coding resources as much as possible on a network topological graph by use of the ant colony optimization algorithm while guaranteeing a transmission rate unchanged. The resource optimization method is mainly characterized by comprising a distributed multi-dimensional maintenance mode of pheromones, an inspiring factor proposed for profiles, an updating mode of local punishment and reward pheromones and the like. Study and experiments indicate that the resource optimization method under the network coding environment based on the ant colony optimization algorithm is capable of finding out the optimal solutions in all test cases; meanwhile, compared with other algorithms capable of well solving the network coding resource optimization problem at present, the resource optimization method has remarkable advantages in both performance and efficiency, and therefore, the feasibility and the usability of the resource optimization method are proven.

Description

A kind of based on method for optimizing resources under the network code environment of ant colony optimization algorithm

Technical field

The present invention relates to the method for ant colony optimization algorithm coding nodes resource optimization under network code environment, belong to multimedia communication and network transmission technology field.

Background technology

Traditional Internet Transmission interior joint can not do any operation to the data flow received, and transfer of data adopts the mode storing/forward to carry out.But, adopt the theoretical upper bound that can not ensure that multicast rate can reach maximum flow minimum cut theorem and determines in this way.2000, the people such as Ahlswede proposed the concept of network code first, proved, in multicast network, to utilize network coding technique, and multicast rate can reach the upper limit that maximum flow minimum cut theorem is determined.Because network code can reduce data transmission times, improve network throughput and network data transmission efficiency, become a focus of research field in recent years.

But after introducing network code, node needs to carry out extra encoding operation (mathematical operation complicated in finite field), can bring the expense of the resources such as calculating, storage.In initial research, all nodes in network are all taken as coding nodes and carry out encoding operation.Have research to point out subsequently, not all coding nodes all necessarily needs to carry out encoding operation, only needs wherein a part of node to carry out encoding operation and just can ensure peak transfer rate.Like this, how under the prerequisite ensureing network transmission speed, reduce encoding operation as much as possible, thus the expense that minimizing network code brings, become the research direction that in network code research field one is important, the i.e. proposition of network code resource optimization (Network Coding Resource Minimization, NCRM) problem.

Present stage, network code method for optimizing resources has following two types:

1, based on the method for greedy algorithm

C.Fragouli et al. and M.Langberg et al. proposes two kinds of methods based on greedy algorithm respectively and solves this problem, but greedy algorithm is easily absorbed in local optimum, once selects improperly just may cause very undesirable result.Generally, effect of optimization is unsatisfactory.

2, based on the method for evolution algorithm

It is a NP-hard problem (also just meaning that the above-mentioned method based on greedy algorithm is difficult to solve this problem well) that Kim et al. demonstrates network code resource optimize problem, and proposes two kinds are solved problem method based on genetic algorithm.Subsequently, Xing et al. adopts quantum derivative algorithm, Incremental Learning Algorithm based on population, compact genetic algorithm and the evolution algorithm based on path code to solve network code resource optimize problem respectively.Domestic scholar Deng Liang etc. and Shao's magnitude also give oneself solution respectively by genetic algorithm.These methods all belong to evolution algorithm, as everyone knows, evolution algorithm is a class based on the random search algorithm of natural evolution and selection, seldom use due to algorithm pattern or substantially do not use itself some characteristics of dealing with problems, so evolution algorithm has very strong robustness and adaptability, be applicable to various optimization field.But also just because of this, evolution algorithm effectively cannot use the information of local message or problem itself, search is become blindly, causes result or deterioration of efficiency.

In general, although there is multiple method for network code resource optimization, effect of optimization and efficiency can't be entirely satisfactory, particularly in network application, the consumption for time and resource is particularly valued.Ant colony optimization algorithm is just used to when proposing solve path configuration problem (traveling salesman problem), and this algorithm can utilize global information and local message well.And network code resource optimize problem also can be understood as the multiple set of paths meeting data rate from starting point to specific terminal of structure, and the problem making coding nodes the least possible.Thus, the present invention adopts ant colony optimization algorithm to solve this problem, is intended to be optimized from effect and efficiency simultaneously.

Summary of the invention

In order to overcome the shortcoming of prior art, the present invention adopts ant colony optimization algorithm to solve the problem of network code resource optimization.

First 1, two basic elements using ant colony optimization algorithm to solve this problem are described, the structure of pheromones τ and a factor of heuristic η and maintenance:

(1) pheromones is used to provide the guidance of overall importance to ant group, so for this problem, the value of pheromones is correlated with the number of coding nodes.In addition, due to the particularity of this problem, a limit in network may be selected once by the ant in ant group, repeatedly or not selects, if use individual traditional pheromones table, will cause the covering of pheromones, thus cannot instruct ant clearly.The present invention is directed to this particularity, have employed a kind of distribution, the pheromones maintenance mode of multidimensional, every corresponding pheromones table of ant, only has different iterations, ant just shared same the pheromones table of same position.

(2) effect of heuristic factor is to provide local tutorial message, the present invention proposes a kind of heuristic factor for ant, using the limit in present case lower network topology by the number of times selected as heuristic factor.After ant group Successful construct path collection before, 1 is added to selected time of every bar limit number attribute that this path is concentrated, in ant group afterwards ant structure path time will with reference to this attribute, after using figure decomposition, each potential coding nodes only has one to go out limit, if this potential coding nodes have be greater than 1 enter limit, then illustrate that this node needs coding, so ant can with reference to this attribute as heuristic factor, the heuristic factor that choosing is as far as possible large, ensure that present node is only selected one as far as possible and entered limit, namely ensure that present node does not do encoding operation as far as possible.

The concrete means that the present invention realizes its goal of the invention are:

A kind of based on method for optimizing resources under the network code environment of ant colony optimization algorithm, construct multiple set of paths meeting data rate from starting point to specific terminal, and make coding nodes the least possible, with when ensureing that transmission rate is constant, optimized network coding resource as far as possible, comprises following treatment step:

Step 1, input initial primary topology G (V, E), wherein comprised a source node S, d receiving node, calculated the maximum rate R of this topology by maximum flow minimum cut theorem;

Step 2, use figure decompose (list of references 1M.Kim, V.Aggarwal, U.O Reilly, M.M é dard, and W.Kim, " Genetic representations for evolutionary minimization of network coding resources; ": Springer, 2007, pp.21-31.) method, potential coding nodes in input network topology structure is decomposed, makes to be convenient to judge the transmission means of data in potential coding nodes, thus judge that whether this potential coding nodes is the coding nodes on practical significance; Topological structure new after decomposing, as input, makes ant group find path on this topological structure;

Step 3, initialization ant group Pheromone Matrix, other parameters of ant group and iterations I;

Step 4, generate d ant group according to the number d of receiving node, for each group specifies a receiving node, there be R ant in each ant group.Ant group sequence number is made to be k, k:=1;

Step 5, kGe ant group build path collection, make the sequence number of ant be n, n:=1;

Step 6, n-th ant build path, if path construction success, jump to step 7, otherwise jump to step 8;

Step 7, the path built is joined in taboo list, select this path again, n=n+1 to prevent the ant with group; If n<=R, jump to step 6, otherwise jump to step 9;

Step 8, to be emptied by taboo list, perform pheromones local and punish update mechanism, the path of concentrating current path has performed the pheromones local updating strategy of punishment effect, is again selected by ant to prevent unaccommodated path; Jump to step 5 after Pheromone update, this group ant rebuilds path collection;

Step 9, the path collection using this group ant successfully to build carry out pheromones local updating reward mechanism.This group path set pair heuristic factor is used to upgrade; K=k+1, if k<=d, jumps to step 5, otherwise, jump to step 10;

Step 10, all set of paths be a solution; Calculate the coding nodes number of the program, compare with globally optimal solution, be better than global optimum and then upgrade globally optimal solution by current solution, and perform global information element renewal rewards theory.I=I-1, if I>0, jumps to step 4, otherwise algorithm terminates, and exports globally optimal solution.

In actual process: in the process of step 3, step 6, step 8, step 9 and step 10, every corresponding pheromones table of ant, only have different iterations, the ant of same position just shares same pheromones table.

In step 6 and step 9, using the limit in present case lower network topology by the number of times selected as heuristic factor; After ant group Successful construct path collection before, 1 is added to selected time of every bar limit number attribute that this path is concentrated; Each potential coding nodes only has one to go out limit, if this potential coding nodes have be greater than 1 enter limit, representing this potential coding nodes is actual coding node.Namely adopt this attribute as heuristic factor; The heuristic factor that choosing is as far as possible large.

During step 6 n-th ant build path, if ant can only be transferred on another one node on one node, then directly redirect; If there is multiple node to select, then use formula

$\mathrm{np}=\left\{\begin{array}{cc}\underset{u\&Element;{\mathrm{allowed}}_i}{\arg \max }{\left[\mathrm{\&tau;}(t_k,n,(i,u))\right]}^{\mathrm{\&alpha;}}{\left[\mathrm{\&eta;}(i,u)\right]}^{\mathrm{\&beta;}}& \mathrm{if\; q}\&le;q_0\\ \mathrm{\&psi;}& \mathrm{otherwise}\end{array}\right.$

Wherein τ (t _k, n, (i, u)) and the value of representative information element on limit (i, u), η (i, u) is the value of heuristic factor on limit (i, u).We can see, have two other parameters, receiving node t in the function of pheromones τ _kwith ant sequence number n.These two parameters maintenance mode that is distributed to the pheromones above mentioned, multidimensional is relevant, is used to locating information element table.Parameter alpha and β be used for scaling information element and inspire element relative importance.Q is random number, q ₀for constant, if q is greater than q ₀then the value of ψ draws according to the probability transfer formula as follows based on roulette:

$p(i,j)\text{=}\left\{\begin{array}{cc}\frac{{\left[\mathrm{\&tau;}(t_k,n,(i,j))\right]}^{\mathrm{\&alpha;}}{\left[\mathrm{\&eta;}(i,j)\right]}^{\mathrm{\&beta;}}}{\underset{u\&Element;{\mathrm{allowed}}_i}{\mathrm{\&Sigma;}}{\left[\mathrm{\&tau;}(t_k,n,(i,u))\right]}^{\mathrm{\&alpha;}}{\left[\mathrm{\&eta;}(i,u)\right]}^{\mathrm{\&beta;}}},& j\&Element;{\mathrm{allowed}}_i\\ 0,& \mathrm{otherwise}\end{array}\right.$

If path construction success, jumps to step 7, otherwise jumps to step 8.

In the local updating mode of step 8 with step 9 pheromones:

Punishment update mechanism in pheromones local is selected by ant again for preventing unaccommodated path, and update mode is shown in following formula:

τ(t _k,n,(i,j))＝τ(t _k,n,(i,j))-Δτ _l

Update mechanism is rewarded in pheromones local, and this strategy plays award effect for by the path collection (the infeasible path collection that solution space is a large amount of in comprising) successfully built; Update mode is shown in following formula:

τ(t _k,n,(i,j))＝τ(t _k,n,(i,j))+Δτ _l

Wherein Δ τ _lvalue is 1/|V ' |, wherein | V ' | be the node number after figure decomposition.

Compared with prior art, beneficial effect of the present invention: utilize ant colony optimization algorithm to solve coding resource optimization problem, the effect of resource optimization is better than or is equal to the method proposed at present, and the efficiency optimized is better than current existing method.Utilize the present invention, while use network coding technique ensures maximum data rate, spend the less time can be optimized current network topology network code resource, reduce the computing node needing additional computational resources, not only can increase efficiency of transmission, and play the effect of green energy conservation.For network application, significant.

Accompanying drawing explanation

Examples of the present invention will be described by way of reference to the accompanying drawings, wherein:

Fig. 1 is the flow chart of the ant colony optimization algorithm that the present invention uses;

Fig. 2 is that figure decomposes illustrated example;

Fig. 3 is ant population spikes path example 1;

Fig. 4 is ant population spikes path example 2;

Fig. 5 is the example for unreasonable path;

Fig. 6 is the example of simulating scenes;

Fig. 7 is experiment test topological diagram parameter list;

Fig. 8 is the experimental result table of comparisons.

Embodiment

Below in conjunction with accompanying drawing, enforcement of the present invention is described in further detail:

The present invention utilizes ant colony optimization algorithm setting up performance advantage superior in path, solves network code resource optimize problem, and its main handling process is shown in shown in accompanying drawing 1.

Below in conjunction with step 2 in accompanying drawing 2 and step, figure decomposition method is described.Each node m having multiple-input and multiple-output is potential coding nodes, such Node Decomposition is two set of intermediate nodes by we, input set of intermediate nodes and output set of intermediate nodes, the corresponding input intermediate node in each input limit of origin node, each exports corresponding one of limit and exports intermediate node, and then input intermediate node is connected respectively between two with output intermediate node.As shown in Figure 2, Fig. 2 (a) is former topological structure, wherein node4 and node7 is potential coding nodes, Fig. 2 (b) illustrates the topological structure after decomposition, node4 is broken down into input set of intermediate nodes: node4_i_1 and node4_i_2, and exports set of intermediate nodes: node4_o_1 and node4_o_2.Node7 also in like manner.Like this, how the topological structure after being decomposed can clearly exhibition information transmit when flowing through coding nodes.In addition, after decomposing, any output intermediate node only has one to export limit, if this node has many input limits like this, just shows that this node needs to carry out encoding operation, and whether easy equally decision node is the process of coding nodes.

Process below in conjunction with pair ant group build path of step 3-step 10 in accompanying drawing 3 and step makes an explanation.Topological diagram after Fig. 3 (a) illustrates and is decomposed, former figure is shown in Fig. 2 (a).This figure has a source node node1, two receiving node node8 and node9, simultaneously the data rate R=2 of this figure.Like this, according to the quantity of receiving node, our ant all living creatures Cheng Liangge ant group.Again according to data rate, then there are two ants in each ant group.Ant group 1 is responsible for finding from node1 to node8 two independently paths, and ant group 2 is responsible for two independent pathways from node1 to node9.As shown in Fig. 3 (b)-(c), two ants of ant group 1 construct path p respectively ₁(1,8)=1->2->8 and p ₂(1,8)=1->3->4_i_2->4 _ o_1->5->7_i_1-GreatT.GreaT.G T7_o_1->8.Meanwhile, two ants of ant group 2 construct path p respectively ₁(1,9)=1->2->4_i_1->4 _ o_1->5->7_i_1-GreatT.GreaT.G T7_o_2->9 and p ₂(1,9)=1->3->9.Then the path of ant group 1 integrates as Paths (1,8)={ p ₁(1,8), p ₂(1,8) }, the path of ant group 2 integrates as Paths (1,9)={ p ₁(1,9), p ₂(1,9) }.Finally, a total solution has built, i.e. Solution ₁(G _d)={ Paths (1,8), Paths (1,9) }, as shown in Fig. 3 (d).The coding nodes number of this solution is 1, and coding nodes is node4_o_1.Accompanying drawing 4 and Fig. 3 in like manner, but illustrate the building process of final solution for optimum (0 coding nodes).

Below in conjunction with the necessity that the special interpretation procedure 8 of accompanying drawing 5 upgrades with local message element in step 9, owing to there is a large amount of illegal paths in this problem, particularly for an ant group, last the irrational Path selection of ant will cause after ant cannot set up path independent of path before, as shown in Fig. 5 (a), if the ant 1 of ant group 1 constructs path p ₁(1,8)=1->2->4_i_1->4 _ o_1->5->7_i_1-GreatT.GreaT.G T7_o_1->8, in any case ant 2 all cannot be constructed from node1 to node8 independent of p ₁the path collection of (1,8), as shown in Fig. 5 (b) He (c).So before re-constructing, need to use the penalty mechanism of pheromones local updating to avoid similar p ₁the path of (1,8) is selected again.Simultaneously, as described, relative to feasible path collection, in solution space, there is more substantial infeasible path collection above, so use the reward mechanism of pheromones local updating to reward feasible path collection, ant can be more effectively guided to set up out feasible path collection.

Embodiment

The present invention is based on to verify validity and the feasibility that ant colony optimization algorithm solves network code resource optimize problem method, having carried out emulation experiment.

1, the present invention tests 14 groups of topological diagrams altogether, wherein the topological diagram of 4 groups of set forms and 10 groups of random topology figure.The topological diagram (Fix1-4) of set form is n-copy topological diagram, and accompanying drawing 6 (b) gives the pattern of 3-copy, wherein accompanying drawing 6 (a) the source figure that is n-copy.The directed networks topological diagram that 10 groups of random topology figure (Rnd1-10) are made up of the node that 20-60 is not waited.The parameter of each topological diagram is provided by Fig. 7.

2, the present invention and classical way compare, need test and the method that compares as follows:

● genetic algorithm (GA): list of references 1

● the inventive method.

3, the performance index that compare of the present invention are as follows:

● mean value and variance (Mean and standard deviation, SD) test mean value and the variance of operation 50 suboptimum result.This index embodies the overall performance of algorithm.

● success rate (Successful ratio, SR) tests the success rate that optimal solution is found out in operation for 50 times, embodies the overall search capability of algorithm.

● average calculation times (Average computational time, ACT) tests the time of operation 50 average cost, embodies time complexity and the availability of algorithm.

4, optimum configurations, is arranged in ant colony optimization algorithm, α=0.8, β=4, ρ=0.1 and q ₀=0.6.Other algorithms carry out assignment by the parameter provided in list of references.

5, for topological diagram Fix-1, the specific implementation step of described ant colony optimization algorithm optimized network coding resource problem is as follows:

A. initial primary topology G (V, E) is inputted, Fix-1.Wherein comprise a source node, 4 receiving nodes, calculated the maximum rate R=2 of this topology by maximum flow minimum cut theorem.Adopt the method that figure decomposes, potential coding nodes in input network topology structure is decomposed.Topological structure new after decomposing is as input.

B. initialization ant group Pheromone Matrix, generates d*R=4*2=8 and opens pheromones table, the initial value τ of pheromones on every paths ₀=1/|V ' |=1/49, wherein | and V ' | be the node number after figure decomposition.Other parameter alpha=0.8 of ant group, β=4, ρ=0.1 and q ₀=0.6 and iterations I=100.

C. generate 4 ant groups, for each group specifies a receiving node, there are 2 ants in each ant group.Ant group sequence number is made to be k, k:=1.

D. kGe ant group build path collection, makes the sequence number of ant be n, n:=1.

E. n-th ant build path, if ant can only be transferred on another one node on one node, then directly redirect; If there is multiple node to select, then formula is used to select.

$\mathrm{np}=\left\{\begin{array}{cc}\underset{u\&Element;{\mathrm{allowed}}_i}{\arg \max }{\left[\mathrm{\&tau;}(t_k,n,(i,u))\right]}^{\mathrm{\&alpha;}}{\left[\mathrm{\&eta;}(i,u)\right]}^{\mathrm{\&beta;}}& \mathrm{if\; q}\&le;q_0\\ \mathrm{\&psi;}& \mathrm{otherwise}\end{array}\right.$

$p(i,j)\text{=}\left\{\begin{array}{cc}\frac{{\left[\mathrm{\&tau;}(t_k,n,(i,j))\right]}^{\mathrm{\&alpha;}}{\left[\mathrm{\&eta;}(i,j)\right]}^{\mathrm{\&beta;}}}{\underset{u\&Element;{\mathrm{allowed}}_i}{\mathrm{\&Sigma;}}{\left[\mathrm{\&tau;}(t_k,n,(i,u))\right]}^{\mathrm{\&alpha;}}{\left[\mathrm{\&eta;}(i,u)\right]}^{\mathrm{\&beta;}}},& j\&Element;{\mathrm{allowed}}_i\\ 0,& \mathrm{otherwise}\end{array}\right.$

If path construction success, jumps to f, otherwise jumps to g.

F. the path built is joined in taboo list, n=n+1.If n<=2, jump to e, otherwise jump to h.

G. emptied by taboo list, the path of concentrating current path has performed the pheromones local updating strategy of punishment effect, and update mode is shown in following formula:

τ(t _k,n,(i,j))＝τ(t _k,n,(i,j))-Δτ _l

Δτ _l＝1/|V’|＝1/49。Jump to d after Pheromone update, this group ant rebuilds path collection.

H. the path collection using this group ant successfully to build has carried out the pheromones local updating of award effect:

τ(t _k,n,(i,j))＝τ(t _k,n,(i,j))+Δτ _l

After Pheromone update, this group path set pair heuristic factor is used to upgrade.K=k+1, if k<=4, jumps to d, otherwise, jump to i.

I. all ant groups path collection all successfully constructs, all set of paths be a solution.Calculate the coding nodes number of the program, compare with globally optimal solution, be better than global optimum and then upgrade globally optimal solution by current solution, and perform global information element renewal rewards theory, more new formula is as follows:

τ(t _k,n,(i,j))＝(1-ρ)τ(t _k,n,(i,j))+ρΔτ _g

Wherein Δ τ _g=1/ coding nodes number.I=I-1, if I>0, jumps to c, otherwise algorithm terminates, and exports globally optimal solution.

6, the comparative result of each index is shown in Fig. 8, optimal value is marked in each table with black matrix.Weigh the present invention from mean value, variance and success rate and be much better than classic algorithm GA, can optimal solution be found out under all use-cases.Be aided with the comparison of average calculation times again, can find, computational speed of the present invention is also much better than classic algorithm, demonstrates and utilizes the characteristic of ant colony optimization algorithm to be more better than adopting evolution algorithm to solve this problem.Embody feasibility of the present invention and availability, particularly in this network application of network code.

Claims (4)

1. one kind based on method for optimizing resources under the network code environment of ant colony optimization algorithm, construct multiple set of paths meeting data rate from starting point to specific terminal, and make coding nodes the least possible, with when ensureing that transmission rate is constant, optimized network coding resource as far as possible, comprises following treatment step:

Step 1, input initial primary topology G (V, E), wherein comprised a source node S, d receiving node, calculated the maximum rate R of this topology by maximum flow minimum cut theorem;

Step 2, the method using figure to decompose, decompose potential coding nodes in input network topology structure, make to be convenient to judge the transmission means of data in potential coding nodes, thus judge that whether this potential coding nodes is the coding nodes on practical significance; Topological structure new after decomposing, as input, makes ant group find path on this topological structure;

Step 3, initialization ant group Pheromone Matrix, other parameters of ant group and iterations I;

Step 4, generate d ant group according to the number d of receiving node, for each group specifies a receiving node, there be R ant in each ant group: make ant group sequence number be k, k:=1;

Step 5, kGe ant group build path collection, make the sequence number of ant be n, n:=1;

Step 6, n-th ant build path, if path construction success, jump to step 7, otherwise jump to step 8;

Step 7, the path built is joined in taboo list, select this path again, n=n+1 to prevent the ant with group; If n<=R, jump to step 6, otherwise jump to step 9;

Step 8, to be emptied by taboo list, perform pheromones local and punish update mechanism, the path of concentrating current path has performed the pheromones local updating strategy of punishment effect, is again selected by ant to prevent unaccommodated path; Jump to step 5 after Pheromone update, this group ant rebuilds path collection;

Step 9, the path collection using this group ant successfully to build carry out pheromones local updating reward mechanism; This group path set pair heuristic factor is used to upgrade; K=k+1, if k<=d, jumps to step 5, otherwise, jump to step 10;

Step 10, all set of paths be a solution; Calculate the coding nodes number of the program, compare with globally optimal solution, be better than global optimum and then upgrade globally optimal solution by current solution, and perform global information element renewal rewards theory; I=I-1, if I>0, jumps to step 4, otherwise algorithm terminates, and exports globally optimal solution.

2. the method for the solution network code resource optimization based on ant colony optimization algorithm according to claim 1, it is characterized in that, in the process of step 3, step 6, step 8, step 9 and step 10, every corresponding pheromones table of ant, only have different iterations, the ant of same position just shares same pheromones table.

3. the method for the solution network code resource optimization based on ant colony optimization algorithm according to claim 1, is characterized in that, in step 6 and step 9, using the limit in present case lower network topology by the number of times selected as heuristic factor; After ant group Successful construct path collection before, 1 is added to selected time of every bar limit number attribute that this path is concentrated; Each potential coding nodes only has one to go out limit, if this potential coding nodes have be greater than 1 enter limit, representing this potential coding nodes is actual coding node; Namely adopt this attribute as heuristic factor; The heuristic factor that choosing is as far as possible large.

4. the method for the solution network code resource optimization based on ant colony optimization algorithm according to claim 1, is characterized in that, in the local updating mode of step 8 with step 9 pheromones:

Punishment update mechanism in pheromones local is selected by ant again for preventing unaccommodated path, and update mode is shown in following formula:

τ(t _k,n,(i,j))＝τ(t _k,n,(i,j))-Δτ _l

Update mechanism is rewarded in pheromones local, and this strategy plays award effect for by the path collection (the infeasible path collection that solution space is a large amount of in comprising) successfully built; Update mode is shown in following formula:

τ(t _k,n,(i,j))＝τ(t _k,n,(i,j))+Δτ _l

Wherein Δ τ _lvalue is 1/|V ' |, wherein | V ' | be the node number after figure decomposition.