CN101478802A - Self-organizing QoS routing method based on ant colony algorithm - Google Patents

Abstract

The invention relates to a wasp algorithm-based self-organization QoS routing method and the method comprises the following steps: A. receiving the data message transmitted by a neighborhood router; B. judging whether the message is a unicast message according to the destination address of the data message; C. entering a unicast QoS routing method; D. initializing a router register; E. examining path evaluation value J<path> and finding the path meeting the demand of users; F. passing to the step K; G. entering the multicast QoS routing method; H. initializing the router register and constructing multicast trees; I. multicast tree cost allocation; J. calculating the evaluation value JT of the multicast tree and finding the present feasible multicast trees; K. retransmitting the data message to the next hop router according to the calculated present feasible multicast tree. The method has the advantages of effectively routing and retransmitting the data on the basis of the QoS request, improving routing success rate, and having higher superiority than traditional network models, and the routing method of the design has good performance and practicability.

Description

A kind of self-organizing QoS routing method based on ant colony algorithm

Technical field

The invention belongs to networking technology area, be specifically related to a kind of self-organizing QoS routing method based on ant colony algorithm.

Background technology

Along with the progressively maturation of self-organizing network, network organization becomes more diverse, and network size enlarges day by day, and service quality (QoS) assurance problem also becomes more and more important.QoS is exactly the measurable predefined service quality based on end to end performance of a cover that network must satisfy during to end data for the user transmission end, and index comprises time delay, delay variation, available bandwidth and grouping packet loss etc.Because self-organizing network has multi-hop, dynamic topology, distributed, provisional, characteristics such as link bandwidth is limited, energy constraint, till now not the ripe method for routing of a cover can be in self-organizing network stabilizing effective transmission QoS information.

Summary of the invention

Deficiency at prior art exists the invention provides a kind of self-organizing QoS routing method based on ant colony algorithm.This method based on:

1. ant colony algorithm

Bee colony (BeeHive) algorithm mainly relies on honeybee agent to carry out pathfinding.For agent number in the limiting network and raising pathfinding efficient,, honeybee agent is divided into two classes: short distance honeybee agent and long apart from honeybee agent according to agent life cycle length.

(1) region of search (Foraging Region)

According to certain standard network topology is divided into several regions, each zone is exactly a region of search, and stipulating has a Centroid in each region of search.Every node that arrives in given jumping figure from Centroid all belongs to same region of search.Under network topology one stable condition, stipulate that any one node belongs to a certain region of search at least, and can only belong to a region of search at the most.

(2) search band (Foraging Zone)

The search band is at node, and for the arbitrary node in the network, its search band is the set that arrives node from this node in given jumping figure, thereby the search band of each node is different.

Honeybee is actual to carry out information exchange in honeycomb, here, realize information exchange by three routing tables at each node definition, three routing tables are respectively IFR (Inter Foraging Region) routing table and region of search member FRM (Foraging Region Member) routing table between interior IFZ (the Intra Foraging Zone) routing table of search band, region of search.

(3) ant colony algorithm basic thought

(a) network is organized into the fixedly division that is referred to as region of search.A division comes from the characteristic of network topology.Each region of search has a representation node.At present, the node of ip address minimum is elected as representation node in the territory.If this node failure, then election is next has the node of bigger ip address as representation node.

(b) each node also has a special search band, and the node of forming the search band is all nodes that can arrive from present node short distance honeybee.

(c) each non-representation node periodically sends a short distance honeybee, and the copy of broadcasting it is to each neighbor node.

(d) when the particular copy of certain honeybee arrives a website, the routing iinformation its change there, copy is continued inundation then, but copy can not be sent to its website of process.This process is continuing, until the life cycle of honeybee arrives, perhaps works as this honeybee and is received at this website, and then it is with death.

(e) representation node only send long apart from honeybee, and it will be received by neighbours and as d) in situation propagated.Just, their life cycle is to be subjected to long restriction apart from honeybee.

(f) each honeybee collects and carries routing information in the process of flight, and routing information is retained in each its node of visiting, and honeybee is distributed routing iinformation by priority queue.

(g) like this, each node is just being safeguarded a current routing iinformation that arrives other node, representation node in its search band and in the arrival region of search.This method makes that node (its destination node is outside its search band) can be with the path transmission of a packet towards the representation node in territory, destination node place.

When (h) transmitting a packet, next-hop node is to select in a kind of mode of probability, according to neighbours' mass measurement.Thus, make not all bag all along the present feasible multicast tree, this will help maximum system performance.This is a principle that comes from the honeybee behavior: a honeybee can only maximize the benefit of its colony, if it monitors the food of thirsting for most that " dancing " sought widely always.

2. worldlet model

Famous Stanley Milgram experiment is found, just can connect any two people in the society by average 6 person-times acquaintance's transmission, and this phenomenon is called the Small-World phenomenon.The experiment of Milgram has disclosed two discoveries: (1) short chain effect ubiquity; (2) people can find short chain.(2) individual discovery explanation when network presents certain topological structure, only utilizes local message just can find short chain effectively.This is found to be distributed route opportunity is provided.

The model that Watt and Strogatz propose (being called for short the WS model) is a kind of Small-World model commonly used: N node is distributed on the annulus, and during initial condition, each node has k connection, connects respectively to nearest k point.Then, adjust the connection of each node successively, change the terminal of connection randomly with Probability p, but the company of avoiding is to node itself.

Note D (i j) is beeline between node i and the j, and the computing formula of average path distance L is as follows:

$L=\frac{1}{n(n-1)/2}\underset{1\&le;i,j\&le;n}{\mathrm{\&Sigma;}}D(i,j)$

When p ≈ 0, $L\&RightArrow;\frac{n}{2k},$ This moment, network topology was rule state; When 0.001≤p≤0.01, $L\&RightArrow;\frac{\ln n}{\ln k},$ This moment, node not only was connected with the adjacent node existence, had also set up the shortcut of minority with remote node, and these shortcut have effectively shortened L just, make whole network present the worldlet feature.In the worldlet network model, comparative maturity be Jon Kleinberg model.Two kinds of limits have been defined in this model: a large amount of local limits (neighbour) and a spot of long-range limit (distant relative).In the two-dimentional Grid model of n node (

Be one dimension node number), model has proved that the distributed routing algorithm chain length upper bound is

The feature of worldlet network: (1) does not consider network topology structure, and very little average distance is arranged between node; (2) bigger acquaintance coefficient is arranged.Certain node, in its all neighbours, backfence even limit number accounts for the subnet that comprises this node and connects limit sum ratio, is called the acquaintance coefficient.It has reflected the backfence acquaintance degree of certain node, and it and network size have nothing to do and be permanent in 1.In sociology, a people not only can utilize the acquaintance to come acquisition of information in the process of seeking another person, also can collect more information by the attribute of company under the acquaintance or corporations.Such as, someone wishes to look for the people to help to compile a program, and he can inquire whether they can programme to the acquaintance, can inquire also whether they are familiar with the student of (being equivalent to a Virtual Organization) of department of computer science simultaneously.

3. the design of self-organizing QoS routing method

3.1 self organizing network model

Self organizing network model can be expressed as a undirected connected graph G (V, E), wherein V has the biological nature behavior and has to transmit and the set of the intelligent node of disposal ability, represents the router in the network; E is the set on limit, the link in the expression network, as shown in Figure 3.

3.2 Mathematical Modeling

3.2.1 clean culture Mathematical Modeling

In the unicast QoS routing issue, optimization aim is: under the situation that satisfies user QoS constraint, effectiveness and the user and the network provider effectiveness sum of maximization user, Virtual network operator make user and network provider effectiveness reach doulbe-sides' victory.

Then have:

UU _P→max{UU _P}

UN _P→max{UN _P}

${\mathrm{UU}}_P+{\mathrm{UN}}_P\&RightArrow;\max \{{\mathrm{UU}}_P+{\mathrm{UN}}_P\}$

$s.t.\underset{e_l\&Element;P_{\mathrm{sd}}}{\min }\left\{{\mathrm{abw}}_l\right\}\&GreaterEqual;\mathrm{bw}\_r_l$ (the available bandwidth minimum value on the link 1=〉 lower band)

$\underset{e_l\&Element;P_{\mathrm{sd}}}{\mathrm{\&Sigma;}}{\mathrm{dl}}_l\&le;\mathrm{dl}\_r_h$ (delay on the link 1 and＜=postpone the upper limit)

$\underset{e_l\&Element;P_{\mathrm{sd}}}{\mathrm{\&Sigma;}}{\mathrm{jt}}_l\&le;\mathrm{jt}\_r_h$ (delay jitter on the link 1 and＜=the delay jitter upper limit)

$1-\underset{e_l\&Element;P_{\mathrm{sd}}}{\mathrm{\&Pi;}}(1-{\mathrm{ls}}_l)\&le;\mathrm{ls}\_r_h$ (product of the accuracy on the 1-link 1)＜=error rate upper limit)

Pay _P≤ p (the required multicast tree expense ratio＜=user who shares of group membership p is willing to pay and uses the upper limit)

UU _PBe the user's effectiveness on path or the multicast tree, UN _PThen be network provider effectiveness.

3.2.2 multicast Mathematical Modeling

In multicast QoS route, optimization aim is: under the prerequisite that satisfies all group membership's end-to-end QoS constraints, and effectiveness and the user and the network provider effectiveness sum of maximization user, Virtual network operator.

Then have: UU _P→ max{UU _P(UUp level off to UUp maximum)

UN _P→max{UN _P}

UU _P+UN _P→max{UU _P+UN _P}

Make and satisfy $\underset{e_l\&Element;P_{\mathrm{sd}}}{\min }\left\{{\mathrm{abw}}_l\right\}\&GreaterEqual;\mathrm{bw}\_r_l$ (the available bandwidth minimum value on the link 1=〉 lower band)

$\underset{e_l\&Element;P_{\mathrm{sd}}}{\mathrm{\&Sigma;}}{\mathrm{dl}}_l\&le;\mathrm{dl}\_r_h$ (delay on the link 1 and＜=postpone the upper limit)

$\underset{e_l\&Element;P_{\mathrm{sd}}}{\mathrm{\&Sigma;}}{\mathrm{jt}}_l\&le;\mathrm{jt}\_r_h$ (delay jitter on the link 1 and＜=the delay jitter upper limit)

$1-\underset{e_l\&Element;P_{\mathrm{sd}}}{\mathrm{\&Pi;}}(1-{\mathrm{ls}}_l)\&le;\mathrm{ls}\_r_h$ (product of the accuracy on the 1-link 1)＜=error rate upper limit)

Pay _P≤ p (the required multicast tree expense ratio＜=user who shares of group membership p is willing to pay and uses the upper limit)

UU _PBe the user's effectiveness on path or the multicast tree, UN _PThen be network provider effectiveness

UU _T→max{UU _T}

UN _T→max{UN _T}

UU _T+UN _T→max{UU _T+UN _T}

$s.t.\underset{e_l\&Element;P_{\mathrm{sd}}}{\min }\left\{\mathrm{ab}{w}_l^d\right\}\&GreaterEqual;\mathrm{bw}\_r_l^d$ (the available bandwidth interval value on the link 1=〉 bandwidth interval value)

$\underset{e_l\&Element;P_{\mathrm{sd}}}{\mathrm{\&Sigma;}}d{l}_l^d\&le;\mathrm{dl}\_r_h^d$ (the delay interval value on the link 1 and＜=postpone interval value)

$\underset{e_l\&Element;P_{\mathrm{sd}}}{\mathrm{\&Sigma;}}j{t}_l^d\&le;\mathrm{jt}\_r_h^d$ (the delay jitter interval value interval value on the link 1 and＜=delay jitter interval value)

$1-\underset{e_l\&Element;P_{\mathrm{sd}}}{\mathrm{\&Pi;}}(1-l{s}_l^d)\&le;\mathrm{ls}\_r_h^d$ (product of the accuracy interval value on the 1-link 1)＜=error rate interval value)

$P{\mathrm{ay}}_p^d\&le;p^d$ (the required multicast tree expense interval value ratio＜=user who shares of group membership p is willing to pay and uses interval value)

UU _TUser's effectiveness on path or the multicast tree, UN _TThen be network provider effectiveness.

3.2.3 worldlet behavior

(1) the worldlet behavior has following characteristic

(a) the node number of degrees high changeability of distributing can obtain short path and the height coefficient that clusters;

(b) when node number of degrees changeability when not being very high, it can not cause the worldlet behavior separately;

(c) the local connection of preference can cause the worldlet behavior, especially in the network topology of router level;

Self-organizing network has the biological nature behavior, and node is dynamic, and the limit also can be set up or be removed, so the node number of degrees are variable.In addition, internodal in the self-organizing network all is local alternately, and just so-called preference is local to be connected.But, become the tangible network of worldlet behavior to need a process by network evolution at random, may not directly utilize the worldlet characteristics to carry out route in the meantime.And also have a bit, in the existing worldlet network model, what the connection between the node referred to all is direct connection, that is to say the limit that physical connection is directly arranged with it that the changeability of the node number of degrees also is meant.Here, produce the worldlet behavior and obtain corresponding advantages, the worldlet limit of mentioning in the self-organizing behavior design thereby can serve as direct-connected limit with indirectly connected limit.

For example, the directly continuous node of node A has A ₁, A ₂, A ₃, A ₄, but it is several with A to be that the route of source node has been set up, and its destination node is respectively D ₁, D ₂, D ₃, from A to D ₁, D ₂, D ₃Just there is the worldlet limit respectively.When calculating the number of degrees of A, these three worldlet limits can be counted.When route, also can dynamically worldlet there be the limit optionally to see optional next jumping limit of A as, have so just realized a kind of changeability of the node number of degrees.

Certainly, though directly can produce very favourable outcome to stretch as a limit, also brought some problems simultaneously, the complexity of handling a stretch footpath and an actual limit of processing after all in the process of route is incomplete same.

This mode also can cause another problem sometimes, though the path of finding by the worldlet limit is that through step number seldom the actual jumping figure in path may be a lot, is not a very excellent path just also, as shown in Figure 4.

Have the worldlet limit among Fig. 4 between S and the M, have the worldlet limit between M and the D, then S only needs through twice pathfinding to M to D.In fact but will jump through 8, and S → A → B → D in fact only needs 3 to jump, though will be through three pathfindings.

Cause occurring problem recited above for fear of repeatedly introducing the pathfinding of worldlet limit, will be limited in routing procedure here one time, only can in the end one jump introducing worldlet limit.Just only introduce the worldlet limit that can directly arrive destination node, but also will limit the length on worldlet limit at present node.

Like this, just can reduce the probability that above-described problem takes place greatly.And,, then also can eliminate gradually by optimised algorithm if performance is bad for introducing the path that the worldlet limit obtains.In addition, in the dynamic environment of self-organizing network, it is often more meaningful than optimum to reach suboptimum fast.

(2) in self-organizing network, realize the worldlet behavior

(a) utilize the worldlet limit

Along with the operation of algorithm, node can accumulate down serves the worldlet limit.Along with increasing of worldlet limit, reached at the node that node is known is also just more and more, then in the routing procedure of back, just can directly utilize the reachability information on worldlet limit to accelerate routing procedure.

As shown in table 1 by the worldlet path record that present node sets out.

The tabulation of table 1 worldlet limit

Initialization: each node all has such worldlet routing table, when initial, shows to be sky.

Add: finished by algorithm, the every paths that successfully finds of algorithm just joins this path relevant information in the source node worldlet routing table, and value is 0 wherein " to be used counting ".

Counting: after the arrival destination node was successfully adopted on a worldlet limit, just " being used number of times " value with the worldlet limit added 1.

Upgrade: every time Δ t _SwpScan the value of " being used number of times " in table, delete those values and be 0 row, " being used number of times " value zero clearing of each row that will remain then.

(b) node route success rate

Be meant the shared percentage of number of times successful in the total routing times of node.The node that success rate is higher, more believable when route.

All set two counters at each node, rcs (successful routing times counter), rct (the total routing times counter of node).When node was used in route, rct just increased 1, and after the success of this time route, rcs just increases 1.Here, ageing for guaranteeing, rcs, rct value are every some cycles Δ t _RcWith regard to zero clearing.

The route success rate formula that calculates i node is as follows:

$I_i=\frac{{\mathrm{rcs}}_i}{{\mathrm{rct}}_i}$

(c) according to address structure: the node with identical ip addresses prefix probably can be positioned at the zone of closing on, and the same prefix part is long more, may be positioned at nearer zone more, and this specific character with the address is called the phase recency here.For example: 202.118.1.64 and 202.118.1.206, identical prefix part is 202.118.1, so the two is positioned at very near zone, the phase recency of the two is higher.

Calculate the address of optional next-hop node and the recency mutually of this route destination address, the phase recency is high more, and then next-hop node might find destination node more, this also similar familiarity that is next-hop node to this route destination address.

Following method is taked in the calculating of two address phase recencies: 32 bit address are scanned relatively binary digit from left to right by turn, if identical just the continuation scanned, if difference just finishes, write down identical figure place M _Ik, phase recency V then _iFor:

$V_i=\frac{M_{\mathrm{ik}}}{32}$

Method of the present invention is as follows:

Steps A: receive the data message that neighbor router sends;

Step B: whether the destination address according to data message is that unicast address judges whether type of message is unicast message;

It is characterized in that: according to step B is unicast message, execution in step C then, otherwise execution in step G;

Step C: the QoS method for routing that enters the clean culture modeling;

Step D: the initialization router register, send forward direction bee colony agent pathfinding, call ant colony algorithm, wherein agent is responsible for collecting record current network state information;

Step e: examination path evaluation of estimate J _Path, find the path that meets user's request;

Step F: forward step K to;

Step G: enter the QoS method for routing of multicast modeling, given multicast request: $R(v_s,v_d,{\mathrm{\&Delta;}}_{\mathrm{bw}}^d,{\mathrm{\&Delta;}}_{\mathrm{dl}}^d,{\mathrm{\&Delta;}}_{\mathrm{jt}}^d,{\mathrm{\&Delta;}}_{\mathrm{ls}}^d,p^d),$ Be its structure multicast tree, wherein a v _sRepresent source node, v _dRepresent destination node,

Represent between the bandwidth demand confining region,

Represent between the delay requirement confining region,

Represent between delay jitter demand confining region,

Represent between the error rate desired region p ^dRepresentative of consumer is willing to pay and is used the upper limit;

Step H: the initialization router register makes up multicast tree;

Step I: multicast tree cost allocation;

After forming multicast tree, because selected limit selects for use the user shared, so also naturally by selecting for use the user to share jointly, the principle of sharing is rate: the part that the high more then user of expense that the user takies required pair of this paths alone shares in multicast tree is paid is big more

The setting source node to the set of each group membership's the required paying in path is:

W _p={ pay ₁, pay ₂..., pay _N-1, pay _nPay in the formula _nBe the required paying in each group membership's path, n=1,2 ..., N, N belongs to natural number, then i group membership v _iRequired multicast tree expense ratio of sharing is:

${\mathrm{per}}_d=\frac{{\mathrm{pay}}_i}{\underset{k\&Element;\{\mathrm{1,2},...\}}{\mathrm{\&Sigma;}}{\mathrm{pay}}_k}$ I ∈ n in the formula, k ∈ n;

Step J: calculate multicast tree evaluation of estimate J _T, find out the present feasible multicast tree;

Step K: data message forwarding is arrived next hop router according to the present feasible multicast tree that calculates.

Calling ant colony algorithm among the step D comprises:

Step D1: the initialization router register, maximum iteration time TIN is set, iterations IN=0 is set, set each iteration and send honeybee agent quantity BN, the honeybee agent quantity bn=0 that has sent, path evaluation of estimate J _Path=∞, feasible path are empty, and the initialization network is divided into several territory with network, and each zone is exactly a region of search, and stipulating has a representation node in each region of search, and the node of ip address minimum is elected as representation node in the territory.If this node failure, then the next node with bigger ip address of election is as representation node, and every node that arrives in given jumping figure from representation node all belongs to same region of search;

Step D2: judge at source node whether destination node belongs to the search band of source node, if, then set to send the short distance honeybee, if, then set and send longly apart from honeybee, if do not know, then set and long and shortly respectively account for half ratio apart from honeybee and send;

Step D3:IN ← IN+1, source node is with interval delta _tSend BN honeybee agent, agent of every transmission is with regard to bn ← bn+1, each honeybee agent of initialization, wherein: honeybee agent sequence number seq=bn;

Step D4: forward direction agent record path, the node that lives through is preserved, write down relevant network state information simultaneously, forward direction honeybee agent is after source node is issued, and the node of each process in network is act of execution by the following step;

Step D41: judge whether current honeybee agent jumping figure transfinites, if, then honeybee agent death;

Step D42: honeybee agent is that destination node is explored the way with territory, destination node place representation node earlier if source node and destination node not in same territory, are then explored the way; After arriving this territory representation node, be that destination node is explored the way with the actual purpose node again;

Step D43: if the direct next-hop node of present node has this route destination node, then investigate with this node between the limit whether satisfy the user and ask constraint, as satisfied, then honeybee agent jumps to destination node, forwards step D46 to;

Step D44: realize the worldlet limit at present node;

Step D45: carry out next at present node and jump selection;

Step D451:, route total degree counter is increased by 1 at present node;

Step D452: realize the worldlet behavior;

Step D453: if next jumping is selected successfully to return, then honeybee agent proceeds to this node;

Step D454: select failure, honeybee agent death if next is jumped;

Step D46: forward direction honeybee agent task is finished;

Step D5: when forward direction agent arrived destination node, one of triggering for generating was inherited the back to agent of forward direction agent relevant information, and the back oppositely arrives source node to the information of agent along original route from destination node, upgrades the information in the routing table simultaneously;

Step D51: in destination node, according to forward direction honeybee agent information, according to formula $J_P=\frac{{\mathrm{\&alpha;}}_{\mathrm{up}}}{{\mathrm{UU}}_P}+\frac{{\mathrm{\&alpha;}}_{\mathrm{np}}}{{\mathrm{UN}}_P},$ Calculating path evaluation of estimate J _P, and give the back of generation to agent, and wherein, α _Up, α _NpBe inclination weights to the user, 0＜α _Up, α _Np＜1, α _Up+ α _Np=1, give the back to honeybee agent with the path that forward direction honeybee agent finds;

Step D52: apart from honeybee,, upgrade the IFZ routing table for long, after flying over territory, destination node place representation node, also will upgrade the IFR routing table oppositely not flying to before the representation node in territory, destination node place;

Step D53: the back by the routing information of preserving, is oppositely jumped to upstream node to agent, upgrades IFZ or IFR table then; At present node, honeybee increases by 1 with route number of success counter, if honeybee agent has introduced the worldlet limit at present node, then upgrades worldlet limit table;

Step D54: judge whether the back arrives source node to honeybee agent, if no show then jumps to step D53;

Step D55: the back is compared to evaluation of estimate and the current path evaluation of estimate that honeybee agent carries the path,, then new route is saved as the present feasible path if agent carries the evaluation of estimate in path less than the current path evaluation of estimate;

Step D56: finish to the agent task back, calls ant colony algorithm and finish.

Step D44 realizes that at present node the worldlet limit comprises:

Step D441: judge whether current agent is using the worldlet limit, if then directly take out the worldlet limit as next-hop node; Otherwise, jump to step D443;

Step D442: judge that can this next-hop node satisfy the user and ask constraint, if can, honeybee agent selects this node as next-hop node, preserve the QoS information and the cost information on limit between present node and this node, honeybee agent finishes in the present node behavior, proceeds to this next-hop node then; Otherwise, honeybee agent death;

Step D443: investigate the sequence number of honeybee agent, judge that whether this agent is this iteration last in sending, if not, then jump to step D45;

Step D444: investigate present node worldlet limit table, whether have the worldlet limit that arrives destination node, if there is no, then jump to step D45;

Step D445: deposit the worldlet limit in honeybee agent, and honeybee agent use worldlet limit is set, honeybee agent writes down this node address then, jumps to D441.

Step D452 realizes that the worldlet behavior comprises:

If the sequence number of honeybee agent is an even number, then select next-hop node at random; Otherwise, calculate and select next to jump, calculate the user satisfaction value computing formula of corresponding sides: $W_l={\mathrm{\&alpha;}}_{\mathrm{bw}}\&CenterDot;g_{\mathrm{bw}}^l+{\mathrm{\&alpha;}}_{\mathrm{dl}}\&CenterDot;g_{\mathrm{dl}}^l+{\mathrm{\&alpha;}}_{\mathrm{jt}}\&CenterDot;g_{\mathrm{jt}}^l+{\mathrm{\&alpha;}}_{\mathrm{ls}}\&CenterDot;g_{\mathrm{ls}}^l$ In the formula, α _Bw, α _Dl, α _JtAnd α _LsBe respectively the weight coefficient of each qos parameter, α _Bwα _Dlα _Jtα _Ls∈ [01], and α _Bw+ α _Dl+ α _Jt+ α _Ls=1, W _lBe the numerical value between 0 and 1,

Expression bandwidth satisfaction,

Expression time-delay satisfaction,

Expression delay jitter satisfaction,

Represent the error rate satisfaction, calculate the route success rate of i node: $I_i=\frac{{\mathrm{rcs}}_i}{{\mathrm{rct}}_i},$ Rcs is successfully the routing times counter in the formula, and rct is that the total routing times counter of node is, when node is used in route, rct just increases 1, and after the success of this time route, rcs just increases 1, ageing for guaranteeing, rcs, rct value are every some cycles Δ t _RcWith regard to zero clearing; Calculate the progress of thinking of i node and this route destination address: 32 bit address are scanned relatively binary digit from left to right by turn,,, write down identical figure place M if difference just finishes if identical just the continuation scanned _Ik, phase recency then $V_i=\frac{M_{\mathrm{ik}}}{32}:$ M _TkRepresent two 32 ip address to do binary system relatively, the number that figure place is identical, span are calculated next bar and are selected probability: pro between 0-32 _j=P _Ij* W _j* I _j* V _j, wherein: P _IjBe corresponding N in the present node pheromones table _jPheromone concentration, W _jUser satisfaction value, the I of expression node j _jRoute success rate, the V of expression node j _jThe phase recency of expression node j is according to pro _j, select corresponding pro _jThe maximum next-hop node of value is jumped as next and is selected.

Step e examination path evaluation of estimate J _Path, find the path that meets user's request to comprise:

Step e 1: at the source node Δ that waits for a period of time _tAfter, examination path evaluation of estimate J _Path

Step e 2: if J _Path＜∞, then algorithm successfully finds the path that meets user's request, at source node current surviving path is added the worldlet table, goes to step K;

Step e 3: if J _PathWhether=∞ then judges IN less than TIN, if, then go to step D3, otherwise routing failure.

Step H initialization router register makes up multicast tree and comprises:

Step H1: the initialization router register is provided with iterations CN, iterations cn=0, multicast tree evaluation of estimate J _Tree=∞, the present feasible multicast tree is for empty;

Step H2: all group memberships by the descending arrangement of bandwidth request, are set pending group membership and gather W={v ₁, v ₂..., v _N-1, v _n, v _nBe pending group membership, n= ₁, ₂..., N, N belongs to natural number, processed group member set A=Φ, feasible path set P=Φ, wherein Φ is null set;

Step H3: get v _d∈ W utilizes the singlecast router algorithm based on the ant network, by v _sTo v _dSeek one and satisfy member v _dThe feasible path P of request constraint _dIf, do not find, then this time make up the multicast tree failure, jump to step J3;

Step H4: with node v _dFrom the W deletion, and add A;

Step H5: check feasible path P _dOn whether contain the group membership who is arranged in W, if do not have, then jump to step H7;

Step H6: if P _dOn contain the group membership who is arranged in W, judge that can these group memberships' request constraint at P _dBe met, the node that can be satisfied is deleted from W, and adds A;

Step H7: with feasible path P _dAdd set P;

Step H8:, then jump to step H1 if set W is not empty;

Step H9: the path among the P in the feasible path is combined into a multicast tree, and in the process of Cheng Shu, each group membership's request constraint all will be satisfied, and ring can not occur, otherwise this makes up the multicast tree failure, jumps to step J3.

Step J calculates multicast tree evaluation of estimate J _T, find out the present feasible multicast tree and comprise:

Step J1: calculate multicast tree evaluation of estimate J according to formula _T

Step J2: if J _T＜J _Tree, J _TreeBe the evaluation of estimate of present feasible multicast tree, then replace present feasible multicast tree, J with the multicast tree that calculates _Tree=J _T

Step J3:cn ← cn+1 is if cn＜CN jumps to step H;

Step J4: judge J _TreeWhether＜∞ sets up, if set up, and then algorithm success, otherwise algorithm failure.

The singlecast router algorithm of ant network is among the step H3:

Step (1): the initialization router register, send forward direction ant agent pathfinding, call ant algorithm, wherein agent is responsible for collecting record current network state information;

Step (2): examination path evaluation of estimate J _Path, find the path that meets user's request;

Step (1-1): the initialization router register, maximum iterations TIN is set, iterations IN=0, the ant agent quantity AN that the each iteration of initialization sends, the agent that has sent counts an=0, path evaluation of estimate J _Path=∞, feasible path are empty;

Step (1-2): IN ← IN+1, source node is with interval delta _tSend AN forward direction ant agent pathfinding, agent of every transmission is with regard to an ← an+1, each forward direction agent of initialization, wherein: ant agent sequence number seq=an;

Step (1-3): forward direction agent record path, the node that lives through is preserved, write down relevant network state information simultaneously, forward direction ant agent is after source node is issued, and the node of each process in network is act of execution by the following step;

Step (1-3-1): judge whether current ant agent jumping figure transfinites, if, then ant agent death;

Step (1-3-2): if the direct next-hop node of present node has this route destination node, then investigate with this node between the limit whether satisfy the user and ask constraint, as satisfied, then ant agent jumps to destination node, forwards step (1-3-5) to;

Step (1-3-3): realize worldlet at present node;

Step (1-3-4): carry out next at present node and jump selection;

Step (1-3-4-1):, route total degree counter is increased by 1 at present node;

Step (1-3-4-2): realize the worldlet behavior;

Step (1-3-4-3): if next jumping is selected successfully to return, then ant agent proceeds to this node;

Step (1-3-4-4): select failure, ant agent death if next is jumped;

Step (1-3-5): forward direction ant agent task is finished;

Step (1-4): when forward direction agent arrived destination node, one of triggering for generating was inherited the back to agent of forward direction agent relevant information, and the back oppositely arrives source node to the information of agent along original route from destination node, upgrades the information in the routing table simultaneously;

Step (1-4-1): in destination node, according to forward direction ant agent information, according to formula $J_P=\frac{{\mathrm{\&alpha;}}_{\mathrm{up}}}{{\mathrm{UU}}_P}+\frac{{\mathrm{\&alpha;}}_{\mathrm{np}}}{{\mathrm{UN}}_P}$ (wherein, α _Up, α _NpBe inclination weights to the user, 0＜α _Up, α _Up＜1, α _Up+ α _Np=1) calculating path evaluation of estimate J _p, and give the back of generation to agent; Give the back to ant agent with the path that forward direction ant agent finds;

Step (1-4-2): the back by the routing information of preserving, is oppositely jumped to upstream node to agent, then the plain table of lastest imformation; At present node, ant increases by 1 with route number of success counter, if ant agent has introduced the worldlet limit at present node, then upgrades worldlet limit table;

Step (1-4-3): judge whether the back arrives source node to ant agent, if no show then jumps to step (1-4-2);

Step (1-4-4): the back is compared to evaluation of estimate and the current path evaluation of estimate that ant agent carries the path,, then new route is saved as the present feasible path if agent carries the evaluation of estimate in path less than the current path evaluation of estimate;

Step (1-4-5): finish to the agent task back, calls ant group algorithm and finish.

Step step (1-3-3) realizes that at present node the worldlet limit comprises:

Step (1-3-3-1): judge whether current agent is using the worldlet limit, if then directly take out the worldlet limit as next-hop node; Otherwise, jump to step (1-3-3-3);

Step (1-3-3-2): judge that can this next-hop node satisfy the user and ask constraint, if can, ant agent selects this node as next-hop node, preserve the QoS information and the cost information on limit between present node and this node, ant agent finishes in the present node behavior, proceeds to this next-hop node then; Otherwise, ant agent death;

Step (1-3-3-3): investigate the sequence number of ant agent, judge that whether this agent is this iteration last in sending, if not, then jump to step (1-3-4);

Step (1-3-3-4): investigate present node worldlet limit table, whether have the worldlet limit that arrives destination node, if there is no, then jump to step (1-3-4);

Step (1-3-3-5): deposit the worldlet limit in ant agent, and ant agent use worldlet limit is set, ant agent writes down this node address then, jumps to (1-3-3-1).

Step (1-3-4-2) realizes that the worldlet behavior comprises:

If the sequence number of ant agent is an even number, then select next-hop node at random; Otherwise, calculate and select next to jump, calculate the user satisfaction value computing formula of corresponding sides: $W_l={\mathrm{\&alpha;}}_{\mathrm{bw}}\&CenterDot;g_{\mathrm{bw}}^l+{\mathrm{\&alpha;}}_{\mathrm{dl}}\&CenterDot;g_{\mathrm{dl}}^l+{\mathrm{\&alpha;}}_{\mathrm{jt}}\&CenterDot;g_{\mathrm{jt}}^l+{\mathrm{\&alpha;}}_{\mathrm{ls}}\&CenterDot;g_{\mathrm{ls}}^l$ In the formula, α _Bw, α _Dl, α _JtAnd α _LsBe respectively the weight coefficient of each qos parameter, α _Bwα _Dlα _Jtα _Ls∈ [01], and α _Bw+ α _Dl+ α _Jt+ α _Ls=1, Wl is the numerical value between 0 and 1, Expression bandwidth satisfaction,

Expression time-delay satisfaction,

Expression delay jitter satisfaction,

Represent the error rate satisfaction, calculate the route success rate of i node: $I_i=\frac{{\mathrm{rcs}}_i}{{\mathrm{rct}}_i},$ Rcs is successfully the routing times counter in the formula, and rct is that the total routing times counter of node is.When node was used in route, rct just increased 1, and after the success of this time route, and it is 1 ageing for guaranteeing that rcs just increases, and rcs, rct value are every some cycles Δ t _RcWith regard to zero clearing, calculate the progress of thinking of i node and this route destination address: 32 bit address are scanned relatively binary digit from left to right by turn,,, write down identical figure place M if difference just finishes if identical just the continuation scanned _Ik, phase recency then $V_i=\frac{M_{\mathrm{ik}}}{32}:$ Calculate next bar and select probability: pro _j=P _Ij* W _j* I _j* V _jP _IjBe the pheromone concentration of corresponding Nj in the present node pheromones table, W _jUser satisfaction value, the I of expression node j _jRoute success rate, the V of expression node j _JjThe phase recency of expression node j is according to pro _j, select corresponding pro _jThe maximum next-hop node of value is jumped as next and is selected.

Step (2) examination path evaluation of estimate J _Path, find the path that meets user's request to comprise:

Step (2-1): at the source node Δ that waits for a period of time _tAfter, examination path evaluation of estimate J _Path

Step (2-2): if J _Path＜∞, then algorithm successfully finds the path that meets user's request, at source node current surviving path is added the worldlet table, goes to step K;

Step (2-3): if J _PathWhether=∞ then judges IN less than TIN, if, then go to step (1-3), otherwise routing failure.

The present invention has designed a kind of method for routing, this method is carried out algorithm design based on a kind of distributed routing algorithm-bee colony network algorithm, can effectively ask data are carried out route and forwarding in routing procedure by this method for routing based on QoS, ant colony algorithm can be designed as hop-by-hop pathfinding in network, the present invention has designed next jumping selection strategy of this algorithm, and the QoS route is supported in expansion.In order to reduce route time, improve the route success rate, this paper has introduced the worldlet behavior in algorithm.Introduced the coarse link parameter of fuzzy mathematics knowledge description simultaneously, introduced and to receive that assorted (Nash) is balanced, the most excellent microeconomics method of Pareto (Pareto) is carried out the link policy selection.The present invention carries out emulation to the self organizing network model and the self-organizing QoS routing method of design on the NS2 platform, the route method is realized, NS2 is meant Network Simulator version2, be a kind of at disclosed, the free software simulation platform of the source code of network technology, the researcher uses it can carry out the exploitation of network technology easily, has become the widely used a kind of network analog software of academia at present.By performance evaluation is carried out in emulation and realization, draw self organizing network model and have an enormous advantage than legacy network model tool, the method for routing of this design has good performance and practicality.

Description of drawings

Fig. 1 is the self-organizing QoS routing method step block diagram that the present invention is based on ant colony algorithm;

Fig. 2 is the self-organizing QoS routing method particular flow sheet that the present invention is based on ant colony algorithm;

Fig. 3 is the self-organized network topology schematic diagram;

Fig. 4 is a worldlet limit schematic diagram;

Fig. 5 is an implementation platform schematic diagram of the present invention;

Fig. 6 is a CERNET CERNET topology of the present invention.

Embodiment

A kind of self-organizing QoS routing method of the present invention based on ant colony algorithm, ant colony algorithm is embedded into Quagga, and (Quagga is a route software of increasing income, on Quagga, can move OSPF, Routing Protocols such as BGP) in the source code in the routing program of increasing income, use 24 PCs as the prototype router, wherein 4 as application server; 24 Intel (INTEL) network interface card, 24 Pu Ruier (TP_LINK) network interface card, 26 netting twines.

The configuration of prototype router software is as follows: LinuxFC7 operating system; Use ospf6d and two finger daemons of bgpd in the GNU C expansion Quagga software router under the Linux, GNU C writes the interface module of docking with algorithm under the use Liunx, realizes the self adaptation unicast routing protocol in the self-organizing network; GNU C writes finger daemon under the use Liunx, and expansion realizes the self-adaptive multicast Routing Protocol based on IPv6.Network topology structure as shown in Figure 5,20 prototype routers are disposed with reference to the CERNET2 network configuration, 4 application servers and prototype router are direct-connected.The present invention has realized the needed agreement support of binding algorithm, the obtaining of neighbor information, network state information obtain the forwarding of route agent.

Route simulating system is based upon on CERNET (CERNET) topology, and as shown in Figure 6, the parameter setting of topology has certain objectivity substantially with reference to practical situation.

In network model, a given route requests, the method for routing that moves this paper is respectively looked for the road and is added routing table entry, and then sends the packet of a period of time according to the bandwidth value of setting in the route requests, obtains following the tracks of (trace) file.Re-use the built-in distance vector of NS2 system (DV) Routing Protocol, send the data of identical time, obtain the trace file by same source, destination node and bandwidth value.Respectively the trace file that obtains is analyzed, obtained throughput, end-to-end delay, shake and the packet loss of data transfer path, calculate the end-to-end transmission satisfaction of user in conjunction with user's request by the satisfaction formula then.The present invention as depicted in figs. 1 and 2, as follows:

Steps A: receive the data message that neighbor router sends;

Step B: whether the destination address according to data message is that unicast address judges whether type of message is unicast message;

It is characterized in that: according to step B is unicast message, execution in step C then, otherwise execution in step G;

Step C: the QoS method for routing that enters the clean culture modeling;

Step D: the initialization router register, send forward direction bee colony agent pathfinding, call ant colony algorithm, wherein agent is responsible for collecting record current network state information;

Step e: examination path evaluation of estimate J _Path, find the path that meets user's request;

Step F: forward step K to;

Step G: enter the QoS method for routing of multicast modeling, given multicast request: $R(v_s,v_d,{\mathrm{\&Delta;}}_{\mathrm{bw}}^d,{\mathrm{\&Delta;}}_{\mathrm{dl}}^d,{\mathrm{\&Delta;}}_{\mathrm{jt}}^d,{\mathrm{\&Delta;}}_{\mathrm{ls}}^d,p^d),$ Be its structure multicast tree, wherein a v _sRepresent source node, v _dRepresent destination node, Represent between the bandwidth demand confining region,

Represent between the delay requirement confining region,

Represent between delay jitter demand confining region,

Represent between the error rate desired region p ^dRepresentative of consumer is willing to pay and is used the upper limit;

Step H: the initialization router register makes up multicast tree;

Step I: multicast tree cost allocation;

After forming multicast tree, because selected limit selects for use the user shared, so also naturally by selecting for use the user to share jointly, the principle of sharing is rate: the part that the high more then user of expense that the user takies required pair of this paths alone shares in multicast tree is paid is big more

The setting source node to the set of each group membership's the required paying in path is:

W _p={ pay ₁, pay ₂..., pay _N-1, pay _nPay in the formula _nBe the required paying in each group membership's path, n=1,2 ..., N, N belongs to natural number, then i group membership v _iRequired multicast tree expense ratio of sharing is:

${\mathrm{per}}_d=\frac{{\mathrm{pay}}_i}{\underset{k\&Element;\{\mathrm{1,2},...\}}{\mathrm{\&Sigma;}}{\mathrm{pay}}_k}$ I ∈ n in the formula, k ∈ n;

Step J: calculate multicast tree evaluation of estimate J _T, find out the present feasible multicast tree;

Step K: data message forwarding is arrived next hop router according to the present feasible multicast tree that calculates.

Calling ant colony algorithm among the step D comprises:

Step D1: the initialization router register, maximum iteration time TIN is set, iterations IN=0 is set, set each iteration and send honeybee agent quantity BN, the honeybee agent quantity bn=0 that has sent, path evaluation of estimate J _Path=∞, feasible path are empty, and the initialization network is divided into several territory with network, and each zone is exactly a region of search, and stipulating has a representation node in each region of search, and the node of ip address minimum is elected as representation node in the territory.If this node failure, then the next node with bigger ip address of election is as representation node, and every node that arrives in given jumping figure from representation node all belongs to same region of search;

Step D2: judge at source node whether destination node belongs to the search band of source node, if, then set to send the short distance honeybee, if, then set and send longly apart from honeybee, if do not know, then set and long and shortly respectively account for half ratio apart from honeybee and send;

Step D3:IN ← IN+1, source node send BN honeybee agent with interval delta, and agent of every transmission is with regard to bn ← bn+1, each honeybee agent of initialization, wherein: honeybee agent sequence number seq=bn;

Step D4: forward direction agent record path, the node that lives through is preserved, write down relevant network state information simultaneously, forward direction honeybee agent is after source node is issued, and the node of each process in network is act of execution by the following step;

Step D41: judge whether current honeybee agent jumping figure transfinites, if, then honeybee agent death;

Step D42: honeybee agent is that destination node is explored the way with territory, destination node place representation node earlier if source node and destination node not in same territory, are then explored the way; After arriving this territory representation node, be that destination node is explored the way with the actual purpose node again;

Step D43: if the direct next-hop node of present node has this route destination node, then investigate with this node between the limit whether satisfy the user and ask constraint, as satisfied, then honeybee agent jumps to destination node, forwards step D46 to;

Step D44: realize the worldlet limit at present node;

Step D45: carry out next at present node and jump selection;

Step D451:, route total degree counter is increased by 1 at present node;

Step D452: realize the worldlet behavior;

Step D453: if next jumping is selected successfully to return, then honeybee agent proceeds to this node;

Step D454: select failure, honeybee agent death if next is jumped;

Step D46: forward direction honeybee agent task is finished;

Step D5: when forward direction agent arrived destination node, one of triggering for generating was inherited the back to agent of forward direction agent relevant information, and the back oppositely arrives source node to the information of agent along original route from destination node, upgrades the information in the routing table simultaneously;

Step D51: in destination node, according to forward direction honeybee agent information, according to formula $J_P=\frac{{\mathrm{\&alpha;}}_{\mathrm{up}}}{{\mathrm{UU}}_P}+\frac{{\mathrm{\&alpha;}}_{\mathrm{np}}}{{\mathrm{UN}}_P},$ Calculating path evaluation of estimate J _p, and give the back of generation to agent, and wherein, α _Up, α _NpBe inclination weights to the user, 0＜α _Up, α _Np＜1, α _Up+ α _Up=1, give the back to honeybee agent with the path that forward direction honeybee agent finds;

Step D52: apart from honeybee,, upgrade the IFZ routing table for long, after flying over territory, destination node place representation node, also will upgrade the IFR routing table oppositely not flying to before the representation node in territory, destination node place;

Step D53: the back by the routing information of preserving, is oppositely jumped to upstream node to agent, upgrades IFZ or IFR table then; At present node, honeybee increases by 1 with route number of success counter, if honeybee agent has introduced the worldlet limit at present node, then upgrades worldlet limit table;

Step D54: judge whether the back arrives source node to honeybee agent, if no show then jumps to step D53;

Step D55: the back is compared to evaluation of estimate and the current path evaluation of estimate that honeybee agent carries the path,, then new route is saved as the present feasible path if agent carries the evaluation of estimate in path less than the current path evaluation of estimate;

Step D56: finish to the agent task back, calls ant colony algorithm and finish.

Step D44 realizes that at present node the worldlet limit comprises:

Step D441: judge whether current agent is using the worldlet limit, if then directly take out the worldlet limit as next-hop node; Otherwise, jump to step D443;

Step D442: judge that can this next-hop node satisfy the user and ask constraint, if can, honeybee agent selects this node as next-hop node, preserve the QoS information and the cost information on limit between present node and this node, honeybee agent finishes in the present node behavior, proceeds to this next-hop node then; Otherwise, honeybee agent death;

Step D443: investigate the sequence number of honeybee agent, judge this agent whether for this iteration send in last, if not, then jump to step D45;

Step D444: investigate present node worldlet limit table, whether have the worldlet limit that arrives destination node, if there is no, then jump to step D45;

Step D445: deposit the worldlet limit in honeybee agent, and honeybee agent use worldlet limit is set, honeybee agent writes down this node address then, jumps to D441.

Step D452 realizes that the worldlet behavior comprises:

If the sequence number of honeybee agent is an even number, then select next-hop node at random; Otherwise, calculate and select next to jump, calculate the user satisfaction value computing formula of corresponding sides: $W_l={\mathrm{\&alpha;}}_{\mathrm{bw}}\&CenterDot;g_{\mathrm{bw}}^l+{\mathrm{\&alpha;}}_{\mathrm{dl}}\&CenterDot;g_{\mathrm{dl}}^l+{\mathrm{\&alpha;}}_{\mathrm{jt}}\&CenterDot;g_{\mathrm{jt}}^l+{\mathrm{\&alpha;}}_{\mathrm{ls}}\&CenterDot;g_{\mathrm{ls}}^l$ In the formula, α _Bw, α _Dl, α _JtAnd α _LsBe respectively the weight coefficient of each qos parameter, α _Bwα _Dlα _Jtα _Ls∈ [01], and α _Bw+ α _Dl+ α _Jt+ α _Ls=1, W _iBe the numerical value between 0 and 1,

Expression bandwidth satisfaction,

Expression time-delay satisfaction,

Expression delay jitter satisfaction,

Represent the error rate satisfaction, calculate the route success rate of i node: $I_i=\frac{{\mathrm{rcs}}_i}{{\mathrm{rct}}_i},$ Rcs is successfully the routing times counter in the formula, and rct is that the total routing times counter of node is, when node is used in route, rct just increases 1, and after the success of this time route, rcs just increases 1, ageing for guaranteeing, rcs, rct value are every some cycles Δ t _RcWith regard to zero clearing; Calculate the progress of thinking of i node and this route destination address: 32 bit address are scanned relatively binary digit from left to right by turn,,, write down identical figure place M if difference just finishes if identical just the continuation scanned _k, phase recency then $V_i=\frac{M_{\mathrm{ik}}}{32}:$ M _kRepresent two 32 ip address to do binary system relatively, the number that figure place is identical, span are calculated next bar and are selected probability: pro between 0-32 _j=P _Ij* W _j* I _j* V _j, wherein: P _IjBe corresponding N in the present node pheromones table _jPheromone concentration, W _jUser satisfaction value, the I of expression node j _jRoute success rate, the V of expression node j _jThe phase recency of expression node j is according to pro _j, select corresponding pro _jThe maximum next-hop node of value is jumped as next and is selected.

Step e examination path evaluation of estimate J _Path, find the path that meets user's request to comprise:

Step e 1: at the source node Δ that waits for a period of time _tAfter, examination path evaluation of estimate J _Path

Step e 2: if J _Path＜∞, then algorithm successfully finds the path that meets user's request, at source node current surviving path is added the worldlet table, goes to step K;

Step e 3: if J _PathWhether=∞ then judges IN less than TIN, if then go to step D ₃, otherwise routing failure.

Step H initialization router register makes up multicast tree and comprises:

Step H1: the initialization router register is provided with iterations CN, iterations cn=0, multicast tree evaluation of estimate J _Tree=∞, the present feasible multicast tree is for empty;

Step H2: all group memberships by the descending arrangement of bandwidth request, are set pending group membership and gather W={v ₁, v ₂..., v _N-1, v _n, v _nBe pending group membership, n=1,2 ..., N, N belongs to natural number, processed group member set A=Φ, feasible path set P=Φ, wherein Φ is null set;

Step H3: get v _d∈ W utilizes the singlecast router algorithm based on the ant network, by v _aTo v _dSeek one and satisfy member v _dThe feasible path P of request constraint _dIf, do not find, then this time make up the multicast tree failure, jump to step J3;

Step H4: with node v _dFrom the W deletion, and add A;

Step H5: check feasible path P _dOn whether contain the group membership who is arranged in W, if do not have, then jump to step H7;

Step H6: if P _dOn contain the group membership who is arranged in W, judge that can these group memberships' request constraint at P _dBe met, the node that can be satisfied is deleted from W, and adds A;

Step H7: with feasible path P _dAdd set P;

Step H8:, then jump to step H1 if set W is not empty;

Step H9: the path among the P in the feasible path is combined into a multicast tree, and in the process of Cheng Shu, each group membership's request constraint all will be satisfied, and ring can not occur, otherwise this makes up the multicast tree failure, jumps to step J3.

Step J calculates multicast tree evaluation of estimate J _T, find out the present feasible multicast tree and comprise:

Step J1: calculate multicast tree evaluation of estimate J according to formula _T

Step J2: if J _T＜J _Tree, J _TreeBe the evaluation of estimate of present feasible multicast tree, then replace present feasible multicast tree, J with the multicast tree that calculates _Tree=J _T

Step J3:cn ← cn+1 is if cn＜CN jumps to step H;

Step J4: judge J _TreeWhether＜∞ sets up, if set up, and then algorithm success, otherwise algorithm failure.

Use the result of the inventive method as shown in table 2,

Load	Lighter	Generally	Heavier	Very heavy
					Use the bee colony method	0.664269	0.65776	0.581287	0.478431
NS2-DV	0.631488	0.510569	0.49874	0.455068

Table 2 transfer of data satisfaction

Table 2 has been added up the satisfaction result under the heterogeneous networks load respectively, the transmission satisfaction of two kinds of method for routing was very nearly the same when load was very light, along with increasing the weight of of load, this paper method for routing begins obviously to be better than the DV method of service range vector, and the transmission satisfaction value of two kinds of method for routing belongs to close again when load is very heavy.

Claims (7)

1, a kind of self-organizing QoS routing method based on ant colony algorithm, as follows:

Steps A: receive the data message that neighbor router sends;

Step B: whether the destination address according to data message is that unicast address judges whether type of message is unicast message;

It is characterized in that: according to step B is unicast message, execution in step C then, otherwise execution in step G;

Step C: the QoS method for routing that enters the clean culture modeling;

Step D: the initialization router register, send forward direction bee colony agent pathfinding, call ant colony algorithm, wherein agent is responsible for collecting record current network state information;

Step e: examination path evaluation of estimate J _Path, find the path that meets user's request;

Step F: forward step K to;

Step G: enter the QoS method for routing of multicast modeling, given multicast request: $R(v_s,v_d,{\mathrm{\&Delta;}}_{\mathrm{bw}}^d,{\mathrm{\&Delta;}}_{\mathrm{dl}}^d,{\mathrm{\&Delta;}}_{\mathrm{jt}}^d,{\mathrm{\&Delta;}}_{\mathrm{ls}}^d,p^d),$ Be its structure multicast tree, wherein a v _sRepresent source node, v _dRepresent destination node,

Represent between the bandwidth demand confining region,

Represent between the delay requirement confining region, Represent between delay jitter demand confining region, Represent between the error rate desired region p ^dRepresentative of consumer is willing to pay and is used the upper limit;

Step H: the initialization router register makes up multicast tree;

Step I: multicast tree cost allocation;

After forming multicast tree, because selected limit selects for use the user shared, so also naturally by selecting for use the user to share jointly, the principle of sharing is rate: the part that the high more then user of expense that the user takies required pair of this paths alone shares in multicast tree is paid is big more

The setting source node to the set of each group membership's the required paying in path is:

W _p={ pay ₁, pay ₂..., pay _N-1, pay _nPay in the formula _nBe the required paying in each group membership's path, n=1,2 ..., N, N belongs to natural number, then i group membership v _iRequired multicast tree expense ratio of sharing is:

${\mathrm{per}}_d=\frac{{\mathrm{pay}}_i}{\underset{k\&Element;\{\mathrm{1,2},...\}}{\mathrm{\&Sigma;}}{\mathrm{pay}}_k}$ I ∈ n in the formula, k ∈ n;

Step J: calculate multicast tree evaluation of estimate J _T, find out the present feasible multicast tree;

Step K: data message forwarding is arrived next hop router according to the present feasible multicast tree that calculates.

2, the described a kind of self-organizing QoS routing method based on ant colony algorithm of claim 1 is characterized in that calling ant colony algorithm among the described step D comprises:

Step D1: the initialization router register, maximum iteration time TIN is set, iterations IN=0 is set, set each iteration and send honeybee agent quantity BN, the honeybee agent quantity bn=0 that has sent, path evaluation of estimate J _Path=∞, feasible path is empty, the initialization network, network is divided into several territory, and each zone is exactly a region of search, and stipulating has a representation node in each region of search, the node of ip address minimum is elected as representation node in the territory, if this node failure, then the next node with bigger ip address of election is as representation node, and every node that arrives in given jumping figure from representation node all belongs to same region of search;

Step D2: judge at source node whether destination node belongs to the search band of source node, if, then set to send the short distance honeybee, if, then set and send longly apart from honeybee, if do not know, then set and long and shortly respectively account for half ratio apart from honeybee and send;

Step D3:IN ← IN+1, source node is with interval delta _tSend BN honeybee agent, agent of every transmission is with regard to bn ← bn+1, each honeybee agent of initialization, wherein: honeybee agent sequence number seq=bn;

Step D4: forward direction agent record path, the node that lives through is preserved, write down relevant network state information simultaneously, forward direction honeybee agent is after source node is issued, and the node of each process in network is act of execution by the following step;

Step D41: judge whether current honeybee agent jumping figure transfinites, if, then honeybee agent death;

Step D42: honeybee agent is that destination node is explored the way with territory, destination node place representation node earlier if source node and destination node not in same territory, are then explored the way; After arriving this territory representation node, be that destination node is explored the way with the actual purpose node again;

Step D43: if the direct next-hop node of present node has this route destination node, then investigate with this node between the limit whether satisfy the user and ask constraint, as satisfied, then honeybee agent jumps to destination node, forwards step D46 to;

Step D44: realize the worldlet limit at present node;

Step D45: carry out next at present node and jump selection;

Step D451:, route total degree counter is increased by 1 at present node;

Step D452: realize the worldlet behavior;

Step D453: if next jumping is selected successfully to return, then honeybee agent proceeds to this node;

Step D454: select failure, honeybee agent death if next is jumped;

Step D46: forward direction honeybee agent task is finished;

Step D5: when forward direction agent arrived destination node, one of triggering for generating was inherited the back to agent of forward direction agent relevant information, and the back oppositely arrives source node to the information of agent along original route from destination node, upgrades the information in the routing table simultaneously;

Step D51: in destination node, according to forward direction honeybee agent information, according to formula $J_P=\frac{{\mathrm{\&alpha;}}_{\mathrm{up}}}{{\mathrm{UU}}_P}+\frac{{\mathrm{\&alpha;}}_{\mathrm{np}}}{{\mathrm{UN}}_P},$ Calculating path evaluation of estimate J _P, and give the back of generation to agent, and wherein, α _Up, α _NpBe inclination weights to the user, 0＜α _Up, α _Np＜1, α _Up+ α _Np=1, give the back to honeybee agent with the path that forward direction honeybee agent finds;

Step D52: apart from honeybee,, upgrade the IFZ routing table for long, after flying over territory, destination node place representation node, also will upgrade the IFR routing table oppositely not flying to before the representation node in territory, destination node place;

Step D53: the back by the routing information of preserving, is oppositely jumped to upstream node to agent, upgrades IFZ or IFR table then; At present node, honeybee increases by 1 with route number of success counter, if honeybee agent has introduced the worldlet limit at present node, then upgrades worldlet limit table;

Step D54: judge whether the back arrives source node to honeybee agent, if no show then jumps to step D53;

Step D55: the back is compared to evaluation of estimate and the current path evaluation of estimate that honeybee agent carries the path,, then new route is saved as the present feasible path if agent carries the evaluation of estimate in path less than the current path evaluation of estimate;

Step D56: finish to the agent task back, calls ant colony algorithm and finish.

3, the described a kind of self-organizing QoS routing method based on ant colony algorithm of claim 2 is characterized in that described step D44 realizes that at present node the worldlet limit comprises:

Step D441: judge whether current agent is using the worldlet limit, if then directly take out the worldlet limit as next-hop node; Otherwise, jump to step D443;

Step D442: judge that can this next-hop node satisfy the user and ask constraint, if can, honeybee agent selects this node as next-hop node, preserve the QoS information and the cost information on limit between present node and this node, honeybee agent finishes in the present node behavior, proceeds to this next-hop node then; Otherwise, honeybee agent death;

Step D443: investigate the sequence number of honeybee agent, judge that whether this agent is this iteration last in sending, if not, then jump to step D45;

Step D444: investigate present node worldlet limit table, whether have the worldlet limit that arrives destination node, if there is no, then jump to step D45;

Step D445: deposit the worldlet limit in honeybee agent, and honeybee agent use worldlet limit is set, honeybee agent writes down this node address then, jumps to D441.

4, the described a kind of self-organizing QoS routing method based on ant colony algorithm of claim 2 is characterized in that described step D452 realizes that the worldlet behavior comprises:

If the sequence number of honeybee agent is an even number, then select next-hop node at random; Otherwise, calculate and select next to jump, calculate the user satisfaction value computing formula of corresponding sides: $W_l={\mathrm{\&alpha;}}_{\mathrm{bw}}\&CenterDot;g_{\mathrm{bw}}^l+{\mathrm{\&alpha;}}_{\mathrm{dl}}\&CenterDot;g_{\mathrm{dl}}^l+{\mathrm{\&alpha;}}_{\mathrm{jt}}\&CenterDot;g_{\mathrm{jt}}^l+{\mathrm{\&alpha;}}_{\mathrm{ls}}\&CenterDot;g_{\mathrm{ls}}^l$ In the formula, α _Bw, α _Dl, α _JtAnd α _LsBe respectively the weight coefficient of each qos parameter, α _Bwα _Dlα _Jtα _Ls∈ [01], and α _Bw+ α _Dl+ α _Jt+ α _Ls=1, w _lBe the numerical value between 0 and 1, Expression bandwidth satisfaction,

Expression time-delay satisfaction,

Expression delay jitter satisfaction, Represent the error rate satisfaction, calculate the route success rate of i node: $I_i=\frac{{\mathrm{rcs}}_i}{{\mathrm{rct}}_i},$ Rcs is successfully the routing times counter in the formula, and rct is that the total routing times counter of node is, when node is used in route, rct just increases 1, and after the success of this time route, rcs just increases 1, ageing for guaranteeing, rcs, rct value are every some cycles Δ t _RcWith regard to zero clearing; Calculate the progress of thinking of i node and this route destination address: 32 bit address are scanned relatively binary digit from left to right by turn,,, write down identical figure place M if difference just finishes if identical just the continuation scanned _Ik, phase recency then $V_i=\frac{M_{\mathrm{ik}}}{32}:$ M _IkRepresent two 32 ip address to do binary system relatively, the number that figure place is identical, span are calculated next bar and are selected probability: pro between 0-32 _j=P _Ij* W _j* I _j* V _j, wherein: P _IjBe corresponding N in the present node pheromones table _jPheromone concentration, W _jUser satisfaction value, the I of expression node j _jRoute success rate, the V of expression node j _jThe phase recency of expression node j is according to pro _j, select corresponding pro _jThe maximum next-hop node of value is jumped as next and is selected.

5, the described a kind of self-organizing QoS routing method based on ant colony algorithm of claim 1 is characterized in that described step e examination path evaluation of estimate J _Path, find the path that meets user's request to comprise:

Step e 1: at the source node Δ that waits for a period of time _tAfter, examination path evaluation of estimate J _Path

Step e 2: if J _Path＜∞, then algorithm successfully finds the path that meets user's request, at source node current surviving path is added the worldlet table, goes to step K;

Step e 3: if J _PathWhether=∞ then judges IN less than TIN, if, then go to step D3, otherwise routing failure.

6, the described a kind of self-organizing QoS routing method based on ant colony algorithm of claim 1 is characterized in that described step H initialization router register, makes up multicast tree and comprises:

Step H1: the initialization router register is provided with iterations CN, iterations cn=0, multicast tree evaluation of estimate J _Tree=∞, the present feasible multicast tree is for empty;

Step H2: all group memberships by the descending arrangement of bandwidth request, are set pending group membership and gather W={v ₁, v ₂..., v _N-1, v _n, v _nBe pending group membership, n=1,2 ..., N, N belongs to natural number, processed group member set A=Φ, feasible path set P=Φ, wherein Φ is null set;

Step H3: get v _d∈ W utilizes the singlecast router algorithm based on the ant network, by v _sTo v _dSeek one and satisfy member v _dThe feasible path p of request constraint _dIf, do not find, then this time make up the multicast tree failure, jump to step J3;

Step H4: with node v _dFrom the W deletion, and add A;

Step H5: check feasible path p _dOn whether contain the group membership who is arranged in W, if do not have, then jump to step H7;

Step H6: if p _dOn contain the group membership who is arranged in W, judge that can these group memberships' request constraint at p _dBe met, the node that can be satisfied is deleted from W, and adds A;

Step H7: with feasible path p _dAdd set P;

Step H8:, then jump to step H1 if set W is not empty;

Step H9: the path among the P in the feasible path is combined into a multicast tree, and in the process of Cheng Shu, each group membership's request constraint all will be satisfied, and ring can not occur, otherwise; This makes up the multicast tree failure, jumps to step J3.

7, the described a kind of self-organizing QoS routing method based on ant colony algorithm of claim 1 is characterized in that described step J calculates multicast tree evaluation of estimate J _T, find out the present feasible multicast tree and comprise:

Step J1: calculate multicast tree evaluation of estimate J according to formula _T

Step J2: if J _T＜J _Tree, J _TreeBe the evaluation of estimate of present feasible multicast tree, then replace present feasible multicast tree, J with the multicast tree that calculates _Tree=J _T

Step J3:cn ← cn+1 is if cn＜CN jumps to step H;

Step J4: judge J _TreeWhether＜∞ sets up, if set up, and then algorithm success, otherwise algorithm failure.

Images