CN104301996A - Wireless sensor network positioning method - Google Patents

Abstract

The invention relates to a wireless sensor network positioning method and belongs to the technical field of a wireless sensor network. The method is characterized by to begin with, utilizing an artificial fish swarm location algorithm to estimate the position of each beacon node and calculating location errors; then, processing measurement distance between the beacon nodes and obtained by calculation and corresponding location errors to construct a location expert decision table; next, utilizing a rough set to serve as a front-end system to carry out reduction on expert knowledge to obtain an attribute reduction set; and finally, broadcasting the location information of the beacon nodes in the obtained attribute reduction set to unknown nodes, and the unknown nodes determines estimation positions of the unknown nodes by utilizing the artificial fish swarm location algorithm according to the received locating information of the beacon nodes, and thus the location of the wireless sensor network is realized. The method integrates the rough set and the artificial fish swarm location algorithm to obtain estimated node coordinates of the wireless sensor, and the estimation is accurate; convergence speed is fast; and the method is low in computation complexity and good in location performance.

Description

A kind of wireless sensor network locating method

Technical field

The present invention relates to a kind of wireless sensor network locating method, belong to wireless sensor network technology field.

Background technology

Wireless sensor network (WSN, wireless sensor network) be a kind of brand-new information acquisition platform, complicated monitoring on a large scale and tracking task can be realized in agriculture field, node locating technique is one of critical support technology of agriculture wireless sensor network, and the node locating realizing high efficient and reliable is significant to aspects such as event observation, target following and raising router efficiencies.The location algorithm of current proposition is divided into substantially based on range finding and the algorithm without the need to finding range.Without the need to the location algorithm of range finding according to the information realization node locating such as network connectivty, but precision is lower.And based on the location algorithm of range finding by the distance between measured node or angle information computing node position, conventional distance-finding method has RSSI(Received signal strength indicator), TOA(Time of arrival), AOA(Angle of arrival), TDOA(Time difference of arrival) etc.Wherein, the ranging technology based on RSSI directly utilizes radio transmitting and receiving chip measure signal intensity, without the need to installing extra means additional, cost and energy consumption lower, be easy to realize, become wireless sensor network location main method.

In the agriculture application scenario of majority, by the impact of node energy, power consumption, cost, only have beaconing nodes to obtain the positional information of self by GPS navigation system, other unknown node must be positioned by beaconing nodes.Artificial fish-swarm algorithm (artificial fish swarm algorithm, AFSA) be the modern heuristic random searching algorithm of one based on the commune optimizing of animal behavior, have initial value and Selecting parameter is insensitive, strong robustness, simple, the easy advantage such as realization, this algorithm has been applied to wireless sensor network RSSI and has located at present.The basis that beaconing nodes is located as wireless sensor network, by the impact of the factors such as multipath effect, self-characteristic, geometry distribution, wherein only have the beaconing nodes of some keys more responsive to positioning result, complementary information can be provided, contribute to improving Position location accuracy, the beaconing nodes of redundancy then can increase position error, and the beaconing nodes how choose reasonable participates in locating is a major issue needing solution badly.

Summary of the invention

The object of this invention is to provide a kind of wireless sensor network locating method, to solve the problem that the beaconing nodes producing redundancy in current localization method causes position error large.

The present invention is for solving the problems of the technologies described above and providing a kind of wireless sensor network locating method, and this network locating method comprises the following steps:

1) select a beaconing nodes as aggregation node, calculate the measuring distance between each beaconing nodes and be uploaded to aggregation node;

2) artificial fish-swarm location algorithm is utilized to calculate position and the position error of each beaconing nodes;

3) measuring distance between beaconing nodes and corresponding position error are processed, build location expert decision-making table;

4) utilize rough set as front-end system, yojan is carried out to the content in expert decision-making table, obtains Attribute Reduction Set;

5) locating information of beaconing nodes yojan concentrated broadcasts to unknown node;

6) unknown node concentrates the locating information of beaconing nodes to calculate the measuring distance between unknown node and respective beacon node by artificial fish-swarm algorithm according to receiving yojan, thus determine the estimated position of each unknown node, namely obtain the position of each node in this wireless sensor network.

Distance in described step 1) between each beaconing nodes is calculated by RSSI value.

Described step 2) in adopt artificial fish-swarm localization method in target function be:

$y=1/\underset{j=1}{\overset{n-1}{\mathrm{\Σ}}}\mathrm{abs}(\sqrt{{(x_{i1}-x_j)}^2+{(x_{i2}-y_j)}^2+{(x_{i3}-z_j)}^2}-d_j)$

Wherein y is target function value, represents the food concentration of Artificial Fish current state, x _ibe the state of i-th Artificial Fish, dimension is 3, represents the three-dimensional space position residing for beaconing nodes to be positioned; N is beaconing nodes sum, (x _i1, x _i2, x _i3) be the estimated coordinates of beaconing nodes to be positioned, i.e. Artificial Fish state x _i, (x _j, y _j, z _j) be the actual coordinate of beaconing nodes, d _jfor the measuring distance between beaconing nodes to be positioned and other beaconing nodes.

Location expert decision-making table in described step 3) to build as conditional attribute and decision attribute form by measuring distance between the node obtained in beaconing nodes position fixing process and corresponding position error being adopted K-means clustering method to carry out sliding-model control, expert knowledge representation after discretization is S=(U, A, V, f), in formula, S is wireless sensor network location expertise, U={x ₁, x ₂..., x _nbe domain, corresponding beaconing nodes anchored object collection; A=C ∪ D is community set, c={c _k, k=1,2 ..., m} is conditional attribute collection, and corresponding each beaconing nodes is to other beaconing nodes measuring distance property sets; D={d} is decision kind set, corresponding beaconing nodes position error property value; V is the set of all attribute codomains; F is information function, namely determines the value of each object under each attribute in U.

Described K-means clustering algorithm estimates using Euclidean distance as diversity, asks corresponding a certain initial cluster center vector optimal classification, makes clustering criteria function E value minimum, wherein c _jrepresent the class bunch divided, x _irepresent bunch C _iin data point, c _irepresent bunch C _iaverage, the classification number that k represents bunch.

Carrying out yojan to the content in expert decision-making table in described step 4) to realize by adopting Skowron differential matrix and Attributions selection, comprising the following steps:

A. differential matrix M is asked _{n × n}, list M _{n × n}=(c _ij) _{n × n}lower triangular matrix, wherein i, j=1,2 ..., n;

B. the relatively core CORE of decision table is calculated _d(C), B=CORE is made _d(C);

C. to any c _ij, i, j=1,2 ..., n, if then

D. to any c _ij, i, j=1,2 ..., n, if had then forward step F to, otherwise forward step e to;

E. current matrix M is added up _{n × n}in the number of times that occurs of each attribute, choosing the maximum element of occurrence number is a _m, make B=B ∪ { a _m, forward step C to;

F. export B and be required yojan, its mathematical sense is the minimum form of decision attribute to conditional attribute set dependencies.

The invention has the beneficial effects as follows: first the present invention adopts artificial fish-swarm location algorithm to estimate each beaconing nodes position, and calculation of position errors; Then the measuring distance calculated between each beaconing nodes and corresponding position error are processed, build location expert decision-making table; Then utilize rough set to carry out yojan as front-end system to expertise, obtain Attribute Reduction Set; Finally the locating information of beaconing nodes is concentrated to broadcast to unknown node the yojan obtained, unknown node adopts the estimated position of artificial fish-swarm location algorithm determination unknown node according to the beaconing nodes locating information received, thus realizes the location of this wireless sensor network.Rough set is merged in the present invention and artificial fish-swarm location algorithm obtains the estimation of each node coordinate of wireless senser accurately, fast convergence rate, and the method computation complexity is low, and positioning performance is good.

Accompanying drawing explanation

Fig. 1 is the schematic flow sheet of wireless sensor network locating method of the present invention;

Fig. 2 is beaconing nodes deployment model schematic diagram in the embodiment of the present invention;

Fig. 3 is that localization method of the present invention compares schematic diagram with the position error of existing localization method.

Embodiment

Below in conjunction with accompanying drawing, the specific embodiment of the present invention is further described.

Wireless sensor node S={S _i| i=1,2 ..., m} random placement is in three-dimensional rectangular parallelepiped space region, and each node isomorphism, and have identical computing capability, beaconing nodes deployment model as shown in Figure 1.All nodes are divided into beaconing nodes and unknown node by function.Front n node S ₁(x ₁, y ₁), S ₂(x ₂, y ₂) ..., S _n(x _n, y _n) self-position can be obtained in advance, as beaconing nodes by the external equipments such as GPS or the actual arrangement known; Node S _i(x _i, y _i) (n<i≤m) Location-Unknown and do not have special hardware device can obtain self information, as unknown node in a network itself.Network node in the present invention takes following hypothesis:

1) unknown node can arbitrarily movement in region;

2) the radio signal propagation mode of beaconing nodes is desirable spheroid;

3) all the sensors node time stringent synchronization, and can direct communication.

Rough set is one of powerful of current uniform data acess, and from Polish scholar Pawlak since nineteen eighty-two proposes this theory, it has become the powerful describing imperfect, inaccuracy and Noise Data.The constitutive relations of rough set theory between the uncertain knowledge of process, elimination redundant information, discovery data attribute has outstanding advantage, the priori of its independent of model, provide reduction of condition attributes and the value reduction method of complete set, thus the minimum prediction rule collection of descriptive system normal model can be found, provide new approach for completing location feature Attributions selection and improving positioning precision.The present invention is by artificial fish-swarm algorithm and rough set theory, the position of selected distance error minimize is as estimated coordinates, have positioning precision high, anti-interference strong, computation complexity is low, not easily affected by environment, adapt to the fusion rough set of agricultural environment and the wireless sensor network locating method of artificial fish-swarm algorithm, the specific implementation process of the method is as follows:

1. gather the distance between each beaconing nodes

Adopt radio signal propagation path loss in the Lognormal shadowing model description greenhouse simplified, its expression formula is:

${\left[P_r\left(d\right)\right]}_{\mathrm{dBm}}={\left[P_r\left(d_0\right)\right]}_{\mathrm{dBm}}-10\mathrm{\&beta;}{\log }_{10}\frac{d}{d_0}---\left(1\right)$

Wherein d ₀for near-earth reference distance, the unit distance that to be m, d be between receiving terminal and transmitting terminal, unit is m, P _r(d ₀) be apart from being d ₀time the signal strength signal intensity that receives, unit is dBm; P _rd () is the signal strength signal intensity of distance for receiving during d, unit is dBm; β is the path loss index relevant with environment such as obstacles, and scope is between 2 ~ 6.During location first Stochastic choice beaconing nodes as aggregation node, each beaconing nodes sends the packet comprising self-position to aggregation node, secondly each beaconing nodes intercoms mutually, calculates measuring distance each other, and be uploaded to aggregation node by RSSI value.

2. localizer beacon node

Adopt an artificial fish-swarm location algorithm estimation beaconing nodes position, and calculation of position errors.Artificial fish-swarm algorithm is a kind of Stochastic search optimization algorithm of simulating fish school behavior, mainly utilizes looking for food and behavior of bunching of fish, realizes global optimizing by the local optimal searching of individuality each in the shoal of fish.

1) xi is the state of i-th Artificial Fish, and dimension is 3, represents the three-dimensional space position residing for beaconing nodes to be positioned; d _ij=|| x _i-x _j|| represent the distance between Artificial Fish individuality; Visual represents the perceived distance of Artificial Fish; δ represents the crowding factor of Artificial Fish, and try number represents the number of times of repeated attempt in foraging behavior.

2) y is target function value, represents the food concentration of Artificial Fish current state

$y=1/\underset{j=1}{\overset{n-1}{\mathrm{\&Sigma;}}}\mathrm{abs}(\sqrt{{(x_{i1}-x_j)}^2+{(x_{i2}-y_j)}^2+{(x_{i3}-z_j)}^2}-d_j)---\left(2\right)$

In formula, n is beaconing nodes sum, is the estimated coordinates of beaconing nodes to be positioned, i.e. Artificial Fish state x _i, (x _j, y _j, z _j) be the actual coordinate of beaconing nodes, d _jfor the measuring distance between beaconing nodes to be positioned and other beaconing nodes.Because RSSI range finding exists error, orientation problem essence makes error minimize, then target function value is larger, and the positioning result obtained is more excellent.

Beaconing nodes location summary based on artificial fish-swarm algorithm has typical behavior in 4: look for food, bunch, knock into the back and random behavior.

1) foraging behavior

The current state arranging Artificial Fish is x _i, at its (d within sweep of the eye _ij≤ visual) Stochastic choice state x _jif, y _j>y _i, then further forward to the party, otherwise, then reselect state, judge whether to meet forward travel state, repeatedly try number, if still do not meet progress bar part, then move at random and move a step, namely

Wherein rand (N (x _i, visual)) represent the state that random selecting one is new in the visual field of xi.

2) to bunch behavior

If Artificial Fish current state is, explore its number of partners nf within the vision and center thereof, when, and show that there is more food at partner center and not too crowded, then the direction towards partner takes a step forward, otherwise performs foraging behavior.

3) to knock into the back behavior

If Artificial Fish current state is, exploring in its partner within the vision is maximum partner, if showing that the state of partner has higher is substrate concentration, and not too crowded around, then the direction towards partner takes a step forward, otherwise performs foraging behavior.

4) random behavior

The behavior refers in Stochastic choice state within sweep of the eye, then thinks that this direction is moved, belongs to a default behavior of foraging behavior.

Ask for the character of extreme value according to orientation problem, adopt heuristic to perform behaviors such as bunching and knock into the back, then evaluate the value after action, select the maximum wherein to carry out actual execution, default behavior is foraging behavior.A bulletin board is set up, in order to record the food concentration of optimum Artificial Fish individual state and this Artificial Fish position in artificial fish-swarm location algorithm.Every bar Artificial Fish in action once after, just the food concentration of self current state and bulletin board are compared, if be better than bulletin board, then replace bulletin board state with oneself state.

3. locate expert decision-making table

After measuring distance between the node that obtains in beaconing nodes position fixing process and corresponding position error sliding-model control, location expert decision-making table is built respectively as conditional attribute and decision attribute, under the prerequisite of not drop-out, yojan is carried out to expert decision-making table, obtain the minimum form of decision attribute to conditional attribute set associativity, namely positioning precision is had to the set of appreciable impact beaconing nodes.

Discretization for Continuous Attribute

Rough set theory can only process discrete type attribute, and therefore before attribute reduction process, need to carry out discrete processes to locator data, the present invention adopts K-means clustering method to Discretization for Continuous Attribute.Concrete grammar carries out cluster analysis one by one by the conditional attribute of decision table, to the cluster result under each attribute by ascending sort, using corresponding cluster classification as its centrifugal pump, wherein each beaconing nodes is zero to the measuring distance of self, without the need to participating in cluster, directly compose with minimum class label.K-means clustering algorithm estimates using Euclidean distance as diversity, corresponding a certain initial cluster center vector optimal classification, makes clustering criteria function E value minimum.Wherein c _jrepresent the class bunch divided, x _irepresent bunch C _iin data point, c _irepresent bunch C _iaverage, the classification number that k represents bunch.The algorithm flow of K-means clustering algorithm:

Input: object X={x to be sorted ₁, x ₂..., x _n, clusters number k

Export: k class bunch C _j, j=1,2 ..., k

Step1 specifies k Ge Cu center { m at random ₁, m ₂..., m _k;

Step2 is for each data point x _i, find from its nearest bunch center, and be assigned to such;

Step3 recalculates Ge Cu center $m_i=\frac{1}{N}\underset{j=1}{\overset{N_i}{\mathrm{\&Sigma;}}}{x}_{\mathrm{ij}},i=\mathrm{1,2},...,k;$

Step4 calculates clustering criteria function

If Step5 E value restrains, then return { m ₁, m ₂..., m _k, algorithm stops; Otherwise turn Step2.

Build expert decision-making table

Expert knowledge representation after discretization is, in formula, S is agriculture wireless sensor network location expertise, is domain, corresponding beaconing nodes anchored object collection; For community set, be conditional attribute collection, corresponding each beaconing nodes is to other beaconing nodes measuring distance property set; Decision kind set, corresponding beaconing nodes position error property value, V is the set of all attribute codomains; F is information function, namely determines the value of each object under each attribute in U.

4. Attribute Reduction Set

Attribute reduction is keeping, under the condition that expert knowledge library classification capacity is constant, deleting wherein inessential or incoherent attribute, and the Algorithm for Reduction applied based on Skowron differential matrix and Attributions selection realizes the yojan of decision table.Detailed process is as follows:

1) differential matrix M is asked _{n × n}, the lower triangular matrix M listed _{n × n}=(c _ij) _{n × n}, wherein i, j=1,2 ..., n;

2) the relatively core CORE of decision table is calculated _d(C), B=CORE is made _d(C);

3) to any c _ij, i, j=1,2 ..., n, if then

4) to any c _ij, i, j=1,2 ..., n, if had then forward step 6) to, otherwise forward step 5) to;

5) current matrix M is added up _{n × n}in the number of times that occurs of each attribute, choosing the maximum element of occurrence number is a _m, make B=B ∪ { a _m, forward step 3) to;

6) export B and be required yojan, its mathematical sense is the minimum form of decision attribute to conditional attribute set dependencies.

5. positioning properties complete for yojan is broadcast to unknown node, unknown node is according to the measuring distance received between beaconing nodes broadcast message calculating unknown node and respective beacon node, its computational process adopts formula (1) to realize, and be uploaded to aggregation node, the estimated position of using artificial shoal of fish location algorithm determination unknown node again.

6. to export in artificial fish-swarm location algorithm colony's optimal location as the estimated coordinates of mobile node mobile node the whole network broadcasts this location end signal, and aggregation node receives rear steering step 1.

Simulating scenes is set as follows: 1) testing 3D region is 100m × 50m × 20m; 2) node total number amount is 40, and wherein 20 is beaconing nodes; 3) the transmit signal power P of beaconing nodes _tfor 30dBm, reference distance d ₀for 20m, transmitter antenna gain (dBi) G _t, receiving antenna gain G _rbe 1dBi, path loss index β is 2; 4) Artificial Fish number R=50, crowding factor delta=0.618, repeated attempt number of times try_number=50, bunch classification number k=4; 5) positioning result be emulation under identical parameters 100 times obtain the mean value of result; 6) for simulating the impact that actual environment is found range on RSSI, according to the corresponding received signal strength of Node distribution position calculation, increase zero-mean gaussian stochastic variable λ on this basis as environmental interference, then this received signal strength is obtained measuring distance as RSSI value.

The locating effect of location algorithm more in this paper and artificial fish-swarm algorithm under identical simulated conditions, get Gaussian random variable λ=N (0,25), positioning result as shown in Figure 3, can find out when the larger situation of error appears in artificial fish-swarm location algorithm, adopt the localization method merging rough set and artificial fish-swarm algorithm effectively can improve positioning precision, error compensation Be very effective.The average localization error of two kinds of location algorithms is respectively 1.6052m, 2.0688m, show to utilize rough set as front-end system, knowledge of orientation expression of space dimension can be simplified, obtain the main beaconing nodes information affecting positioning result, effectively improve positioning precision.

Claims (6)

1. a wireless sensor network locating method, is characterized in that, this network locating method comprises the following steps:

1) select a beaconing nodes as aggregation node, calculate the measuring distance between each beaconing nodes and be uploaded to aggregation node;

2) artificial fish-swarm location algorithm is utilized to calculate position and the position error of each beaconing nodes;

3) measuring distance between beaconing nodes and corresponding position error are processed, build location expert decision-making table;

4) utilize rough set as front-end system, yojan is carried out to the content in expert decision-making table, obtains Attribute Reduction Set;

5) locating information of beaconing nodes yojan concentrated broadcasts to unknown node;

6) unknown node concentrates the locating information of beaconing nodes to calculate the measuring distance between unknown node and respective beacon node by artificial fish-swarm algorithm according to receiving yojan, thus determine the estimated position of each unknown node, namely obtain the position of each node in this wireless sensor network.

2. wireless sensor network locating method according to claim 1, is characterized in that, the distance in described step 1) between each beaconing nodes is calculated by RSSI value.

3. wireless sensor network locating method according to claim 2, is characterized in that, described step 2) in adopt artificial fish-swarm localization method in target function be:

$y=1/\underset{j=1}{\overset{n-1}{\mathrm{\&Sigma;}}}\mathrm{abs}(\sqrt{{(x_{i1}-x_j)}^2+{(x_{i2}-y_j)}^2+{(x_{i3}-z_j)}^2}-d_j)$

Wherein y is target function value, represents the food concentration of Artificial Fish current state, x _ibe the state of i-th Artificial Fish, dimension is 3, represents the three-dimensional space position residing for beaconing nodes to be positioned; N is beaconing nodes sum, (x _i1, x _i2, x _i3) be the estimated coordinates of beaconing nodes to be positioned, i.e. Artificial Fish state x _i, (x _j, y _j, z _j) be the actual coordinate of beaconing nodes, d _jfor the measuring distance between beaconing nodes to be positioned and other beaconing nodes.

4. wireless sensor network locating method according to claim 3, it is characterized in that, location expert decision-making table in described step 3) to build as conditional attribute and decision attribute form by measuring distance between the node obtained in beaconing nodes position fixing process and corresponding position error being adopted K-means clustering method to carry out sliding-model control, expert knowledge representation after discretization is S=(U, A, V, f), in formula, S is wireless sensor network location expertise, U={x ₁, x ₂..., x _nbe domain, corresponding beaconing nodes anchored object collection; A=C ∪ D is community set, c={c _k, k=1,2 ..., m} is conditional attribute collection, and corresponding each beaconing nodes is to other beaconing nodes measuring distance property sets; D={d} is decision kind set, corresponding beaconing nodes position error property value; V is the set of all attribute codomains; F is information function, namely determines the value of each object under each attribute in U.

5. wireless sensor network locating method according to claim 4, it is characterized in that, described K-means clustering algorithm estimates using Euclidean distance as diversity, asks corresponding a certain initial cluster center vector optimal classification, make clustering criteria function E value minimum, wherein c _jrepresent the class bunch divided, x _irepresent bunch C _iin data point, c _irepresent bunch C _iaverage, the classification number that k represents bunch.

6. wireless sensor network locating method according to claim 5, is characterized in that, carries out yojan and realizes by adopting Skowron differential matrix and Attributions selection, comprise the following steps in described step 4) to the content in expert decision-making table:

A. differential matrix M is asked _{n × n}, list M _{n × n}=(c _ij) _{n × n}lower triangular matrix, wherein i, j=1,2 ..., n;

B. the relatively core CORE of decision table is calculated _d(C), B=CORE is made _d(C);

C. to any c _ij, i, j=1,2 ..., n, if then

D. to any c _ij, i, j=1,2 ..., n, if had then forward step F to, otherwise forward step e to;

E. current matrix M is added up _{n × n}in the number of times that occurs of each attribute, choosing the maximum element of occurrence number is a _m, make B=B ∪ { a _m, forward step C to;

F. export B and be required yojan, its mathematical sense is the minimum form of decision attribute to conditional attribute set dependencies.