CN104994508A - Cognitive radio mesh network resource distribution and routing method - Google Patents

Abstract

The invention discloses a cognitive radio mesh network resource distribution and routing method, comprising: introducing a buffer memory mechanism into a cognitive radio mesh network to generate two special network structures, i.e. a multi-source single-purpose network and a single-source multi-purpose network; in order to solve the problem of dynamic routing and resource distribution of the two structures, constructing a four-dimensional conflict graph of channels, radio interfaces, power and link resources in a controller; finding partial maximal independent sets covering all elements from the conflict graph; and employing an optimization method to solve a regulation scheme best meeting user requirements.

Description

A kind of cognition wireless mesh Resource Allocation in Networks and method for routing

Technical field

The invention belongs to cognitive radio networks field, particularly in cognition wireless mesh network to the dynamic assignment of channel, wave point, transmitting power, link four kinds of resources and method for routing.

Background technology

Along with the development of wireless communication technology, the Wireless Telecom Equipments such as smart mobile phone, panel computer, Smart Home increase sharply, the requirement of people to message transmission rate improves constantly, the data rate of mobile communication rises to Mb level from Kb level, but even so also cannot meet the growing bandwidth demand of people.Meanwhile, new radio communication service continues to bring out, and as WLAN (wireless local area network), vehicle network etc., and usable spectrum almost distributes totally, and therefore, cognitive radio technology has become the key technology of wireless network evolution.

In addition, the proportion that content of multimedia accounts for Internet flow is increasing, certainly will cause in network the transmission being flooded with a large amount of content of multimedia, and to cognitive radio networks, and the dynamic change of frequency spectrum is also not suitable for the multi-medium data of repetition transmission of large capacity.Research finds (see document: K.Wang, Z.Chen and H.Liu, Push-Based Wireless Converged Networks for Massive Multimedia Content Delivery [J], Wireless Communications, IEEE Transactions on, 2014, 13 (5): 2894-2905), the request frequency of content of multimedia presents Zipf's law, namely small part content is by frequent requests, some content is but seldom requested, according to this rule, frequent requested content caching will be reduced the demand of bandwidth greatly to cognitive radio networks node.At present someone caching mechanism (J.Zhao of beginning one's study in cognitive radio networks, W.Gao, Y.Wang, G.Cao, Delay-constrained caching in cognitive radio networks [C], INFOCOM, 2014), due to the existence of buffer memory, the data that can produce request have backup and the same data two kinds of special circumstances of multiple node request at multiple node, which forms the network configuration that the many objects in single source and multi-source monocular two kinds are special, route under these two kinds of structures is a new challenge, traditional routing mode is only suitable for static state or change is not network environment too frequently, dynamically Resources allocation is a urgent problem to meet consumers' demand.

Different from legacy network, the frequency spectrum of cognition wireless mesh network is dynamic change, the available channel of each cognition wireless mesh node, available radio interface, transmitting power and routed path are also time dependent, thus in traditional mesh network only to channel build conflict figuremethod and be not suitable for cognition wireless mesh network, and existing solution is all tree-like routing mode, namely only has from source node to destination node and fixes one or two (another backup path) paths.

Summary of the invention

[goal of the invention]: the present invention in order to solve in a dynamic way in cognitive radio networks, introduce caching mechanism after the Resourse Distribute that produces and routing issue.

[technical scheme]: the object of the invention is to be reached by following measure:

The present invention adopts centralized control mode, the controller that in network, existence one is centralized, be responsible for the environmental information of collecting each cognition wireless mesh node in cognition wireless mesh network, environmental information here comprises available channel, available radio interface, adjustable transmitting power and neighbor node.By Common Control Channel transmission information between controller and cognition wireless mesh node, this control mode realizes by the framework of wireless SDN (software defined network).

Emphasis of the present invention considers the scene having content of multimedia buffer memory, the mechanism of buffer memory takes simple random cache, which this content of multimedia of node request i.e., it is just by its buffer memory, in the middle of path, the node of process is only responsible for forwarding, do not carry out buffer memory, there is a mapping table safeguarding each content of multimedia position in the controller, receive at every turn one group of user ask after (all user request all can pass to controller by Common Control Channel), first content map table can be searched for each request, be redirected source and destination address, some request may have multiple source node, the content of identical destination node that what some request may be asked is, controller is needed to make different process.

The characteristic of binding cache, the invention provides the resource allocation algorithm under two kinds of network configurations of the many objects in single source and multi-source monocular.This algorithm realizes in the controller, comprises three relatively independentmodule: four-dimensional conflict figurebuild module, greatly independentcollection generation module and resource distribution module.

1, four-dimensional conflict figurebuild module.This module utilizes the full mesh topology obtained in the controller figure, the wave point had each cognition wireless mesh node, available channel, conflict relationship between power grade and link circuit resource are used figurestructure store in the controller so that other modules are called, it should be noted that this conflicts figurereal-time servicing, when wireless environment changes, conflict figurealso change, concrete structure flow process is as follows thereupon:

Step 1: according to network topology in controller figure, right in figureevery bar link builds the CPLR four-tuple of channel, power, link and wave point, is expressed as (c, p, (i, j), (s, t)), wherein, cognition wireless mesh node i and j form link (i, j), c represents the wherein available channel on link (i, j), p represents the emitting power grade that node i is selected, s, t represent the available radio interface at node i and node j place respectively, and the set of all four-tuple is designated as T={T ₁, T ₂..., T _k;

Step 2: conflict relationship is judged to the four-tuple that step 1 produces, if any two tuples meet one of following two each and every one conditions, then thinks that these two tuples can not be dispatched simultaneously, a limit will be produced between them:

A) two tuples use identical channel, and the link of one of them is in another interference territory,

B) two tuples use identical wave point;

Step 3: according to the predicting relation of step 2, sets up four-dimensional conflict figure G(T, L).

2, very big independentcollection generation module.Undirected at one in figurelook for one greatly independentcollection is proved to be np complete problem, and the present invention adopts and finds out a part greatly independentthe method of collection, reduce time complexity, concrete steps are as follows:

Step 1: calculate shortest path, utilize breadth-first search algorithm at full mesh topology in figurefind out source and destination node in all user's request set between shortest path, notice because breadth-first search is the weight of not considering limit, shortest path means minimum hop count, may there are many shortest paths between a pair node, and the present invention is designated as different shortest paths.Conflict in figureevery bar limit counting through its shortest path number, often increase by one, counting adds 1, and all can there be a count value on final every bar limit, is designated as E={E ₁, E ₂..., E _n, wherein n is full mesh topology in figurethe number of link;

Step 2: by tuple descending, the present invention considers that shortest path, channel capacity and power level are to (the conflict of a tuple in figuresummit) dispatching priority order impact, provide an index value ID to measure the priority of tuple, the computational methods of such as i-th tuple are:

${\mathrm{ID}}_i=\mathrm{\α}\·\frac{E_i}{E_{\max }}+\mathrm{\β}\·\frac{b_{\mathrm{xy}}^c}{b_{\max }}+(1-\mathrm{\α}-\mathrm{\β})\·\frac{P_i}{P_{\max }}$

Wherein, α, β > 0,0 < alpha+beta < 1 is the size of two instruction each several part proportions, Ei, p _irepresent the channel capacity of c channel and the value of emitting power grade on i tuple link count value, (x, y) link respectively, E _max, b _max, P _maxrepresent the maximum in all tuple link count values, channel capacity and power grade respectively.After the index value of each tuple all calculates, by its descending from big to small;

Step 3: find out the very big of optimal scheduling needs independentcollection, from the tuple that ID value is maximum, finds out and it independenttuple-set, then find out successively by ID value and comprise the very big of this tuple independentcollection, is designated as I={I ₁, I ₂..., I _m, the condition of end is found very big independentcollection contains all tuples;

Step 4: inspection gathers disturbed condition, in four-dimension conflict in figuredisturbed condition just for link relation between any two, this be also agreement interference model can not cogent place, the interference of real network is that the value of detection signal-to-noise ratio is on a physical layer determined, in order to make the more realistic environment of algorithm, needs greatly each to what obtain independentinterference conflict is gathered in collection inspection, judgement according to being exactly whether signal interference ratio SINR is greater than threshold value ρ, if be less than ρ, need this node from greatly independentconcentrate and reject, SINR is defined as follows:

${\mathrm{SINR}}_j=\frac{G_{i,j}{P}_i}{{\mathrm{\&Sigma;}}_{l=1,l\&NotEqual;j}^m{G}_{i,j}{P}_l+n_j}$

Wherein have m-1 node (except j itself) on the same channel simultaneously with power (P1 ..., P1 ... Pm) send data, only have node i to be the sender of node j, all the other are interference, G _{i, j}represent the channel gain of link (i, j), n _jrepresent the additive white Gaussian noise at node j place;

Step 5: circulation step 3 and step 4, until all tuples all join, certain is very big independentcollection.

Further, a part is calculated in step 3 greatly independentcollection, obtains as follows:

Step 2.1: initialization covers collection F=T, independentthe set of collection t is the tuple chosen according to ID descending order, each in S < t > independentcollection all comprises tuple t;

Step 2.2: choose a tuple q in the sequence of tuple descending, judge whether it has non-neighboring contact, namely in conflict in figuredo not have direct limit to be connected, if not, q is put into S < t > and gathers, and q is left out from covering collection F=T/q, otherwise next step;

Step 2.3: each non-neighboring contact of traversal q, if can add in S < t >, i.e. certain of q and S < t > independentenergy collecting synthesis is larger independentcollection, so adds this by q independentcollection, otherwise next step;

Step 2.4: by new for q and t composition independent{ q, t} put into S < t > to collection, and q, t are left out from covering collection F=T/{q, t}, perform next step;

Step 2.5: judge whether F is empty, if be not empty execution step 2.2, otherwise performs next step;

Step 2.6: judge the S={S < t1 > that obtains, S < t2 > ..., S < tm > in all independentwhether collection is very big independentcollection, if not, increase tuple and make it reach very big, otherwise terminate.

3, resource distribution module.By reasonably dispatching greatly independentit is very big that collection generation module produces independentcollection, maximizes the demand meeting user, and takies band as much as possible less, definition f _r(i, j), for request r is through the stream of link (i, j), concrete steps are as follows:

Step 1: according to the user's request received in controller, search content map table, its source and destination address is redirected, if multi-source monocular or the structure of single source monocular then perform step 2, if single source many objects structure then performs step 3;

Step 2: the structure of multi-source monocular, the different piece of request msg is passed to requestor by multiple source node respectively under such configuration, present invention employs structure make voidintend the method (if single source monocular, virtual source node is exactly actual source node) of source node, unrestricted to the channel capacity of each source node from virtual source node, obtain following Formal Representation accordingly:

$\begin{array}{cc}\mathrm{Maximize}& \underset{k=1}{\overset{\left|R\right|}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_r\end{array}$

$\begin{array}{cc}S.t.:& {\mathrm{\&Sigma;}}_{i=\mathrm{VS}\left(r\right),j=S\left(r\right)}{f}_r(i,j)={\mathrm{\&lambda;}}_k{b}_r,\&ForAll;r\&Element;R\end{array}---\left(1\right)$

${\mathrm{\&Sigma;}}_{(i,j)\&Element;L}{f}_r(i,j)={\mathrm{\&Sigma;}}_{(j,q)\&Element;L}{f}_r(j,q),\&ForAll;r\&Element;\mathrm{R\; and\; j}\&NotEqual;\mathrm{VS}\left(r\right),D\left(r\right)---\left(2\right)$

${\mathrm{\&Sigma;}}_{(i,j)\&Element;L}{f}_r(i,j)\&le;b_{\mathrm{ij}},\&ForAll;r\&Element;R---\left(3\right)$

$f_r(i,j)\&GreaterEqual;0,\&ForAll;r\&Element;\mathrm{R\; and}(i,j)\&Element;L---\left(4\right)$

${\mathrm{\&Sigma;}}_{j=\mathrm{VS}\left(r\right)}{f}_r(i,j)=0,\&ForAll;r\&Element;\mathrm{R\; and}(i,j)\&Element;L---\left(5\right)$

${\mathrm{\&Sigma;}}_{i=D\left(r\right)}{f}_r(i,j)=0,\&ForAll;r\&Element;\mathrm{R\; and}(i,j)\&Element;L---\left(6\right)$

$0\&le;{\mathrm{\&lambda;}}_k\&le;1,\&ForAll;k\&Element;(1,|R\left|\right)---\left(7\right)$

${\mathrm{\&Sigma;}}_{m=1}^M{\mathrm{\&alpha;}}_m\&le;1---\left(8\right)$

$b_{\mathrm{ij}}={\mathrm{\&Sigma;}}_{m=1}^M{\mathrm{\&alpha;}}_m{w}_{\mathrm{ij}}^c\left(I_m\right)---\left(9\right)$

Wherein, the flow sum retraining the true source node S (r) that (1) shows from dummy node VS (r) to r is bandwidth demand and a certain percentage λ _kproduct, percentage here represents that, when network can not provide enough bandwidth, each actual bandwidth obtained is the percentage of bandwidth on demand, and target function is exactly that the percentage sum of all requests is maximized, and namely at utmost meets all users.Constraint (2) shows to ask stream becoming a mandarin of intermediate node except virtual source and destination node of r process to equal out stream, and (1) and (2) can ensure that the bandwidth sum of destination node equals virtual source node bandwidth out.Constraint (3) and (9) shows that the stream sum on any one section of link can not exceed channel capacity, due to greatly independentcollection is timesharing scheduling, greatly each independentcollecting the time scale assigned to is α _m, total M is greatly individual independentcollection participates in the scheduling of request r, and thus ergodic capacity is greatly each independentconcentrate the summation of the product of channel capacity and corresponding time scale.Constraint (4) (5) (6) (7) (8) shows that the value flowed is non-negative, entering from virtual source node the stream gone out with object node is all 0, just can obtain concrete Resourse Distribute and routing plan according to the above solution optimized;

Step 3: single source many objects structure, this structure is exactly the multicast structure in legacy network, and different from the mode of conventional construction multicast tree, the present invention adopts Dynamic Scheduling Strategy, and distributes bandwidth with the form of network flow, so the route topological produced is figurestructure, but not multicast tree, in order to reduce the waste of bandwidth, need to find out the shared link of multiple stream, flowing through of such as a certain chain route two request identical content, so wherein partial content can share, need not repeat completely to transmit same content, bandwidth just can save, and accordingly, the present invention devises following algorithm to reduce unnecessary allocated bandwidth:

Step 3.1: calculate the bandwidth sum Resourse Distribute from source node to respective destination node respectively with the optimized algorithm in multi-source monocular;

Step 3.2: find out all shared links, to distribute the very big person of bandwidth for reference, by content ratio determination allocated bandwidth.Such as there are the first and second the third three requests through (i, j) link, bandwidth demand is respectively 9Mbp, 8Mbps and 20Mbps, and at (i, j) link is assigned with 3Mbp respectively, 4Mbps and 5Mbps, namely account for 1/3 of bandwidth demand, 1/2 and 1/4, be with reference to (because the third bandwidth of distributing on this link is maximum) with third, second still needs the bandwidth of distributing to be (1/2-1/4) * 8Mbps=2Mbps, due to the third 1/3 < 1/2, so without the need to distributing bandwidth;

Step 3.3: the bandwidth allocation scheme obtaining QoS routing.

According to above allocative decision, the distribution of bandwidth can be greatly reduced, and route topological is dynamic assignment figurestructure instead of multicast tree, improve the utilance of network, maximizes the demand meeting user simultaneously.

[beneficial effect]: the invention has the beneficial effects as follows: what (1) legacy network will optimize multiple resources formation is a np hard problem, and solution often complexity is very high, and the present invention will conflict figurestructure and resource optimization be divided into independentthe module safeguarded, executed in parallel, spatially decomposes a complicated problem, is more suitable for running in real system; (2) the present invention considers to introduce caching mechanism in cognition wireless mesh network, more adapts to the environment of dynamic change, decreases user's request time; (3) Resource Allocation Formula that optimization method obtains is the scheme of timesharing scheduling, to different very big independentcollection is assigned with the timeslice of different length, improves the utilance of resource.

Accompanying drawing explanation

fig. 1for wireless mesh node resource distribution citing in actual scene;

fig. 2for the four-dimension conflict built for actual scene figure;

fig. 3for the network configuration of multi-source monocular;

fig. 4for single source many objects network configuration.

Embodiment

Below in conjunction with accompanying drawingwith instantiation, concrete introduction is done to the present invention.

After network is set up, four-dimensional conflict in controller figurebuild module just to start working, as Fig. 3shown in, three nodes had in network, node A has an available channel c ₁, power-adjustable is L ₁, L ₂two grades, wave point only has a h ₁; Node B has two available channel c ₁, c ₂, power-adjustable is L ₁, have two wave point h ₁, h ₂; Node C has two available channel c ₁, c ₂, power-adjustable is L ₁, also only have a wave point h ₁. in figurehave three links (A, C), (B, A), (B, C), wherein (A, C) dotted line represents to only have the transmitting power of node A to be higher value L ₂shi Caineng communicates with node C.According to these information, four-dimensional topology in controller, can be built figure:

Step 1: according to network topology in controller figure, right in figureevery bar link builds the CPLR four-tuple of channel, power, link and wave point, according to fig. 37 CPLR tuples can be built, respectively (c ₁, L ₁, (B, A), (h ₁, h ₁)), (c ₁, L ₁, (B, C), (h ₁, h ₁)), (c ₁, L ₂, (A, C), (h ₁, h ₁)), (c ₁, L ₁, (B, C), (h ₂, h ₁)), (c ₁, L ₁, (B, A), (h ₂, h ₁)), (c ₂, L ₁, (B, C), (h ₁, h ₁)), (c ₂, L ₁, (B, C), (h ₂, h ₁)), be designated as T respectively from left to right ₁, T ₂..., T ₇;

Step 2: conflict relationship is judged to the four-tuple that step 1 produces, if any two tuples meet one of following two each and every one conditions, then thinks that these two tuples can not be dispatched simultaneously, a limit will be produced between them:

A) two tuples use identical channel, and the link of one of them is in another interference territory,

B) two tuples use identical wave point;

According to above two conditions, T can be found out ₁and T ₇, T ₅and T ₆it is not conflict.

Step 3: according to the predicting relation of step 2, sets up four-dimensional conflict figure G(T, L), as Fig. 4shown in, except T ₁and T ₇, T ₅and T ₆, be connected all between two between all the other nodes.

Obtain four-dimensional conflict figureafter, next will find out this conflict in figurevery big independentcollection, calculation procedure is as follows:

Step 1: calculate shortest path, utilize breadth-first search algorithm at full mesh topology in figurefind out source and destination node in all user's request set between shortest path, if controller receives request B → A and A → C, because topology in case is very little, shortest path is exactly (B, A) and (A, C), thus the count value of every bar link is E _bA=1, E _aC=1, E _bC=0;

Step 2: by tuple descending, suppose fig. 3in L1=5dBm, L2=10dBm, channel capacity proportionality coefficient α=0.5, β=0.3, can obtain E according to step 1 ₁=1, E ₂=0, E ₃=1, E ₄=0, E ₅=1, E ₆=0, E ₇=0, the index value that can calculate each tuple in case is thus ID ₁=0.73, ID ₂=0.25, ID ₃=0.84, ID ₄=0.25, ID ₅=0.73, ID ₆=0.23, ID ₇=0.23.So the result of sequence is T ₃→ T ₁→ T ₅→ T ₂→ T ₄→ T ₆→ T ₇, here tuple equal for ID value is successively sorted by first group #;

Step 3: find out the very big of optimal scheduling needs independentcollection, from the tuple that ID value is maximum, finds out and it independenttuple-set, then find out successively by ID value and comprise the very big of this tuple independentcollection, the condition of end is found very big independentcollection contains all tuples, and detailed step is as follows:

Step 2.1: initialization covers collection F={T ₁, T ₂..., T ₇, independentthe set of collection t is the tuple chosen according to ID descending order, each in S < t > independentcollection all comprises tuple t;

Step 2.2: choose a tuple T in the sequence of tuple descending ₃, it does not have non-neighboring contact, namely in conflict in figures < T ₃>={ { T ₃, F={T ₁, T ₂, T ₄, T ₅, T ₆, T ₇, get next tuple T ₁, it has non-neighboring contact T ₇, perform next step;

Step 2.3: traversal T ₁each non-neighboring contact because now S < T ₁> is empty, so next step;

Step 2.4: by T ₁and T ₇form new independentcollection { T ₁, T ₇put into S < T ₁>={ { T ₁, T ₇, F={T ₂, T ₄, T ₅, T ₆;

Step 2.5: judge F not as empty, perform step 2.2, choose tuple T ₅, there is non-neighboring contact T in it ₆, by { T ₅, T ₆put into S < T ₅>={ { T ₅, T ₆, F={T ₂, T ₄, F is not still 0, continues to get tuple T ₂, it does not have non-neighboring contact, puts into S < T ₂>={ { T ₂, F={T ₄, continue to get tuple T ₄, there is no non-neighboring contact yet, put into S < T ₄>={ { T ₄, F is empty, next step;

Step 2.6: the S < T obtained ₃>={ { T ₃, S < T ₁>={ { T ₁, T ₇, S < T ₅>={ { T ₅, T ₆, S < T ₂>={ { T ₂be all very big independentcollection, terminates;

Step 4: disturbed condition is gathered in inspection, through inspection, appeals maximum independentconcentrate the tuple not having cochannel, thus necessarily satisfy condition;

Step 5: circulation step 3 and step 4, until all tuples all join, certain is very big independentcollection.

Very big from what obtain independentcollection can be found out, only has T ₁and T ₇, T ₅and T ₆can dispatch, its physical significance is Node B h simultaneously ₁communicate with node A, use h simultaneously ₂communicate with node C, another kind of situation is, Node B h ₂communicate with node A, use h simultaneously ₁communicate with node C.

Due to emphasis of the present invention it is considered that buffer memory impact under multi-source monocular with the many objects in single source two kinds of route patterns, controller receives one group of user's request within a dispatching cycle, search distribution of content mapping table, their source and destination address is redirected, according to source and destination address situation determination resource allocation methods (if be clean culture, directly with the optimization method in multi-source monocular).

As shown in Figure 3, for the structure of multi-source monocular, build a virtual source node in logic, the down hop of virtual source node is each real source node, and these link channel capacity are not limited, obtain the model of similar clean culture accordingly, then try to achieve Resourse Distribute solution according to following optimization method.

$\begin{array}{cc}\mathrm{Maximize}& \underset{k=1}{\overset{\left|R\right|}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_r\end{array}$

$\begin{array}{cc}S.t.:& {\mathrm{\&Sigma;}}_{i=\mathrm{VS}\left(r\right),j=S\left(r\right)}{f}_r(i,j)={\mathrm{\&lambda;}}_k{b}_r,\&ForAll;r\&Element;R\end{array}---\left(1\right)$

${\mathrm{\&Sigma;}}_{(i,j)\&Element;L}{f}_r(i,j)={\mathrm{\&Sigma;}}_{(j,q)\&Element;L}{f}_r(j,q),\&ForAll;r\&Element;\mathrm{R\; and\; j}\&NotEqual;\mathrm{VS}\left(r\right),D\left(r\right)---\left(2\right)$

${\mathrm{\&Sigma;}}_{(i,j)\&Element;L}{f}_r(i,j)\&le;b_{\mathrm{ij}},\&ForAll;r\&Element;R---\left(3\right)$

$f_r(i,j)\&GreaterEqual;0,\&ForAll;r\&Element;\mathrm{R\; and}(i,j)\&Element;L---\left(4\right)$

${\mathrm{\&Sigma;}}_{j=\mathrm{VS}\left(r\right)}{f}_r(i,j)=0,\&ForAll;r\&Element;\mathrm{R\; and}(i,j)\&Element;L---\left(5\right)$

${\mathrm{\&Sigma;}}_{i=D\left(r\right)}{f}_r(i,j)=0,\&ForAll;r\&Element;\mathrm{R\; and}(i,j)\&Element;L---\left(6\right)$

$0\&le;{\mathrm{\&lambda;}}_k\&le;1,\&ForAll;k\&Element;(1,|R\left|\right)---\left(7\right)$

${\mathrm{\&Sigma;}}_{m=1}^M{\mathrm{\&alpha;}}_m\&le;1---\left(8\right)$

$b_{\mathrm{ij}}={\mathrm{\&Sigma;}}_{m=1}^M{\mathrm{\&alpha;}}_m{w}_{\mathrm{ij}}^c\left(I_m\right)---\left(9\right)$

Be illustrated in figure 4 single source many objects structure, in figured1, D2, D3 asks identical content to S, bandwidth demand is respectively 9Mbp, 8Mbps and 20Mbps, the first step calculates S to D1 with the optimization method in multi-source monocular respectively, D2, the path of D3, wherein (A, B) be three's process jointly, be assigned with 3Mbp respectively, 4Mbps and 5Mbps, namely account for 1/3 of bandwidth demand, 1/2 and 1/4, the bandwidth that D3 distributes on this link is maximum, therefore its speed must ensure, and D2, the content of D3 has had D1 transmission in advance to B node, therefore D2 still needs the bandwidth of distributing to be (1/2-1/4) * 8Mbps=2Mbps, and due to D3 content all by speed faster D2 help it to transmit (1/3 < 1/2), so without the need to distributing bandwidth.According to such computational methods, whole shared link is found out, and redistributes, massive band width can be reduced and distribute.

Claims (6)

1. a cognition wireless mesh Resource Allocation in Networks and method for routing, it is characterized in that, create multi-source monocular introduce caching mechanism in cognition wireless mesh network after with the many objects in single source two kinds of special constructions, provide three relatively independent modules by the mode of centerized fusion: four-dimensional conflict graph builds module, maximal independent set generation module and resource distribution module, maximize and meet user's request.

2. cognition wireless mesh Resource Allocation in Networks according to claim 1 and method for routing, is characterized in that, four-dimensional conflict graph construction step is:

Step 1: according to network topological diagram in controller, bar link every in figure is built to the CPLR four-tuple of channel, power, link and wave point, be expressed as (c, p, (i, j), (s, t)), wherein, cognition wireless mesh node i and j form link (i, j), c represents at link (i, j) the wherein available channel on, p represents the emitting power grade that node i is selected, s, t represents the available radio interface at node i and node j place respectively, and the set of all four-tuple is designated as T={T ₁, T ₂..., T _k;

Step 2: conflict relationship is judged to the four-tuple that step 1 produces, if any two tuples meet one of following two each and every one conditions, then think that these two tuples can not be dispatched simultaneously, a limit will be produced: a) two tuples use identical channel between them, and the link of one of them is in another interference territory, and b) two tuples use identical wave point;

Step 3: according to the predicting relation of step 2, sets up four-dimensional conflict graph G (T, L).

3. cognition wireless mesh Resource Allocation in Networks according to claim 1 and method for routing, is characterized in that, the step of maximal independent set generation module is:

Step 1: calculate shortest path, utilize breadth-first search algorithm find out in full mesh topology figure source and destination node in all user's request set between shortest path, notice because breadth-first search is the weight of not considering limit, shortest path means minimum hop count, may there are many shortest paths between a pair node, the present invention is designated as different shortest paths.Every bar limit counting in conflict graph, through its shortest path number, often increases by one, and counting adds 1, and all can there be a count value on final every bar limit, is designated as E={E ₁, E ₂..., E _n, wherein n is the number of full mesh topology figure link;

Step 2: by tuple descending, the present invention considers the impact on a tuple (summit in conflict graph) dispatching priority order of shortest path, channel capacity and power level, provide an index value ID to measure the priority of tuple, the computational methods of such as i-th tuple are:

${\mathrm{ID}}_i=\mathrm{\&alpha;}\&CenterDot;\frac{E_i}{E_{\max }}+\mathrm{\&beta;}\&CenterDot;\frac{b_{\mathrm{xy}}^c}{b_{\max }}+(1-\mathrm{\&alpha;}-\mathrm{\&beta;})\&CenterDot;\frac{P_i}{P_{\max }}$

Wherein, α, β > 0,0 < alpha+beta < 1 is the size of two instruction each several part proportions, E _i, p _irepresent the channel capacity of c channel and the value of emitting power grade on i tuple link count value, (x, y) link respectively, E _max, b _max, P _maxrepresent the maximum in all tuple link count values, channel capacity and power grade respectively.After the index value of each tuple all calculates, by its descending from big to small;

Step 3: find out the maximal independent set that optimal scheduling needs, from the tuple that ID value is maximum, finds out and its independently tuple-set, then finds out the maximal independent set comprising this tuple by ID value successively, be designated as I={I ₁, I ₂..., I _m, the condition of end is that the maximal independent set found contains all tuples;

Step 4: disturbed condition is gathered in inspection, disturbed condition in four-dimensional conflict graph is just for link relation between any two, this be also agreement interference model can not cogent place, the interference of real network is that the value of detection signal-to-noise ratio is on a physical layer determined, in order to make the more realistic environment of algorithm, need to gather interference conflict to each maximal independent set inspection obtained, the foundation judged is exactly whether signal interference ratio SINR is greater than threshold value ρ, if be less than ρ, need this node to reject from maximal independent set, SINR is defined as follows:

${\mathrm{SINR}}_j=\frac{G_{i,j}{P}_i}{{\mathrm{\&Sigma;}}_{l=1,l\&NotEqual;j}^m{G}_{i,j}{P}_l+n_j}$

Wherein have m-1 node (except j itself) on the same channel simultaneously with power (P1 ..., P1 ... Pm) send data, only have node i to be the sender of node j, all the other are interference, G _{i, j}represent the channel gain of link (i, j), n _jrepresent the additive white Gaussian noise at node j place;

Step 5: circulation step 3 and step 4, until all tuples all join certain maximal independent set.

4. maximal independent set generation method according to claim 3, is characterized in that, described step 3 computational methods are:

Step 2.1: initialization covers collection F=T, the set of independent sets t is the tuple chosen according to ID descending order, and in S<t>, each independent sets comprises tuple t;

Step 2.2: choose a tuple q in the sequence of tuple descending, judge whether it has non-neighboring contact, namely in conflict graph, do not have direct limit to be connected, if, q is not put into S<t> set, and q is left out from covering collection F=T/q, otherwise next step;

Step 2.3: each non-neighboring contact of traversal q, if can add in S<t>, namely the independent sets that certain independent energy collecting synthesis of q and S<t> is larger, so adds this independent sets by q, otherwise next step;

Step 2.4: { q, t} put into S<t>, and q, t are left out from covering collection F=T/{q, t}, perform next step q and t to be formed new independent sets;

Step 2.5: judge whether F is empty, if be not empty execution step 2.2, otherwise performs next step;

Step 2.6: judge the S={S<t1> obtained, S<t2>, ..., in S<tm>}, whether all independent sets are maximal independent set, if not, increasing tuple makes it reach very big, otherwise terminates.

5. cognition wireless mesh Resource Allocation in Networks according to claim 1 and method for routing, is characterized in that, in the structure of multi-source monocular, bandwidth sum resource allocation optimization method is:

Step 1: according to the user's request received in controller, search content map table, its source and destination address is redirected, if multi-source monocular or the structure of single source monocular then perform step 2, if single source many objects structure then performs step 3;

Step 2: the structure of multi-source monocular, the different piece of request msg is passed to requestor by multiple source node respectively under such configuration, present invention employs the method for constructing virtual source node (if single source monocular, virtual source node is exactly actual source node), unrestricted to the channel capacity of each source node from virtual source node, obtain following Formal Representation accordingly:

$\begin{array}{cc}\mathrm{Maximize}& \underset{k=1}{\overset{\left|R\right|}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_r\end{array}$

$S.t.:{\mathrm{\&Sigma;}}_{i=\mathrm{VS}\left(r\right),j=S\left(r\right)}{f}_r(i,j)={\mathrm{\&lambda;}}_k{b}_r,\&ForAll;r\&Element;R---\left(1\right)$

${\mathrm{\&Sigma;}}_{(i,j)\&Element;L}{f}_r(i,j)={\mathrm{\&Sigma;}}_{(i,q)\&Element;L}{f}_r(j,q),\&ForAll;r\&Element;\mathrm{Randj}\&NotEqual;\mathrm{VS}\left(r\right),D\left(r\right)---\left(2\right)$

${\mathrm{\&Sigma;}}_{(i,j)\&Element;L}{f}_r(i,j)\&le;b_{\mathrm{ij}},\&ForAll;r\&Element;R---\left(3\right)$

$f_r(i,j)\&GreaterEqual;0,\&ForAll;r\&Element;\mathrm{Rand}(i,j)\&Element;L---\left(4\right)$

${\mathrm{\&Sigma;}}_{j=\mathrm{VS}\left(r\right)}{f}_r(i,j)=0,\&ForAll;r\&Element;\mathrm{Rand}(i,j)\&Element;L---\left(5\right)$

${\mathrm{\&Sigma;}}_{i=D\left(r\right)}{f}_r(i,j)=0,\&ForAll;r\&Element;\mathrm{Rand}(i,j)\&Element;L---\left(6\right)$

$0\&le;{\mathrm{\&lambda;}}_k\&le;1,\&ForAll;k\&Element;(1,|R\left|\right)---\left(7\right)$

${\mathrm{\&Sigma;}}_{m=1}^M{\mathrm{\&alpha;}}_m\&le;1---\left(8\right)$

$b_{\mathrm{ij}}={\mathrm{\&Sigma;}}_{m=1}^M{\mathrm{\&alpha;}}_m{w}_{\mathrm{ij}}^c\left(I_m\right)---\left(9\right)$

Wherein, the flow sum retraining the true source node S (r) that (1) shows from dummy node VS (r) to r is bandwidth demand and a certain percentage λ _kproduct, percentage here represents that, when network can not provide enough bandwidth, each actual bandwidth obtained is the percentage of bandwidth on demand, and target function is exactly that the percentage sum of all requests is maximized, and namely at utmost meets all users.Constraint (2) shows to ask stream becoming a mandarin of intermediate node except virtual source and destination node of r process to equal out stream, and (1) and (2) can ensure that the bandwidth sum of destination node equals virtual source node bandwidth out.Constraint (3) and (9) shows that the stream sum on any one section of link can not exceed channel capacity, and because maximal independent set is timesharing scheduling, the time scale that each maximal independent set is assigned to is α _m, total M maximal independent set participates in the scheduling of request r, and thus ergodic capacity is the summation of the product of channel capacity and corresponding time scale in each maximal independent set.Constraint (4) (5) (6) (7) (8) shows that the value flowed is non-negative, entering from virtual source node the stream gone out with object node is all 0, just can obtain concrete Resourse Distribute and routing plan according to the above solution optimized;

Step 3: single source many objects infrastructure resource distributes and routing plan.

6. single source according to claim 5 many objects infrastructure resource distributes and method for routing, and it is characterized in that, described step 3 method is:

Step 3.1: calculate the bandwidth sum Resourse Distribute from source node to respective destination node respectively with the optimized algorithm in multi-source monocular;

Step 3.2: find out all shared links, to distribute bandwidth the maximum for reference, by content ratio determination allocated bandwidth.

Step 3.3: the bandwidth allocation scheme obtaining QoS routing.