CN105491587B - Distributed Kalman common recognition method for tracking moving target based on pairs of gossip algorithms - Google Patents

Distributed Kalman common recognition method for tracking moving target based on pairs of gossip algorithms Download PDF

Info

Publication number
CN105491587B
CN105491587B CN201511009138.9A CN201511009138A CN105491587B CN 105491587 B CN105491587 B CN 105491587B CN 201511009138 A CN201511009138 A CN 201511009138A CN 105491587 B CN105491587 B CN 105491587B
Authority
CN
China
Prior art keywords
node
information
kth moment
formula
pairs
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201511009138.9A
Other languages
Chinese (zh)
Other versions
CN105491587A (en
Inventor
马康健
吴少川
魏宇明
潘斯琦
王昕阳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Harbin Institute of Technology
Original Assignee
Harbin Institute of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Harbin Institute of Technology filed Critical Harbin Institute of Technology
Priority to CN201511009138.9A priority Critical patent/CN105491587B/en
Publication of CN105491587A publication Critical patent/CN105491587A/en
Application granted granted Critical
Publication of CN105491587B publication Critical patent/CN105491587B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W84/00—Network topologies
    • H04W84/18—Self-organising networks, e.g. ad-hoc networks or sensor networks

Abstract

Distributed Kalman common recognition method for tracking moving target based on pairs of gossip algorithms, is related in wireless sensor network to the distributed tracking technology of mobile target.The present invention is to solve the problems, such as that the tracking system robustness of existing distributed Kalman's common recognition filtering technique is low, data storage capacity is big, systematic tracking accuracy is low low with common recognition precision.The present invention introduces pairs of gossip algorithms in distributed Kalman knows together filtering algorithm and solves the above problems.Tracking of the present invention suitable for wireless sensor network to mobile target.

Description

Distributed Kalman common recognition method for tracking moving target based on pairs of gossip algorithms
Technical field
The present invention relates to the distributed tracking technologies to moving target in wireless sensor network.
Background technology
1, distributed average common recognition
In the network that one has N number of sensor node, arbitrary node i ∈ { 1,2 ..., N } in gossip algorithms After t iteration, all deposits there are one state value, be expressed as xi(t), { 0,1 ... } t ∈.The initial state value of arbitrary node i indicates For xi(0), then averagely common recognition problem can be expressed as：
The end-state value of all nodes all reaches consistent, and is the average value of all node initial state values.
2, pairs of gossip algorithms
The basic skills of pairs of gossip algorithms is that some node in network is randomly enabled to select some neighbor node Being averaged for state value is carried out after carrying out information exchange, it, can be so that all sections of the whole network after being averaged by continuous, limited number of time data The state value of point reaches average common recognition.This is widely used in the cooperation of multimachine device, measurement and control area.Fig. 1 gives in pairs The signal of gossip algorithms.
It is assumed that each node is there are one the state value x of oneself, any time t, the random node i in network is waken up, Its state value is xi(t), and its a neighbor node j, state value x is selectedj(t).Node i, j is respectively the shape of oneself State value is sent to other side, and it is that data are average, and the state value of other nodes is constant in network respectively to carry out state value update.This when Gossip iteration has carried out in gap, and each node state value update is as shown in formula (2) in network.
3, distributed Kalman's common recognition filtering algorithm
Gradually the Kalman filtering algorithm of recurrence optimal estimation can effectively track mobile target.On this basis It is improved, information sharing is realized by being exchanged with each other the respectively predicted value to target location between sensor node, by melting The predicted value for closing different nodes reaches all the sensors being distributed as a result, being achieved that and utilizing to moving target position consistent Estimation Formula Kalman common recognition filtering algorithm effectively tracks the mobile object in sensor network.
Specifically, it is assumed that object moves shown in match state equation such as formula (3).
X (k+1)=A (k) x (k)+B (k) w (k) (3)
In formula：X (k) is state of the mobile object at the kth moment；A (k) be state-transition matrix, indicate object from kth when The state transfer relationship for being moved to+1 moment of kth is carved, is a time-varying matrix；B (k) matrixes in order to control, can be added controlled quentity controlled variable As controller, influences of the state of a control noise w (k) to mobile object transfering state；W (k) is state-noise, is to meet 0 It is worth the random noise of Gaussian Profile, noise statistics amount is：
E[w(k)w(l)T]=Qkδkl,δkl=1 (k ≠ l), δkl=0 (k=l)
QkFor kth moment mobile object state-noise covariance matrix.
Observation zs of the kth moment arbitrary node i to mobile objecti(k) as shown in formula (4).
zi(k)=Hi(k)x(k)+vi(k) (4)
In formula：zi(k) it is observations of the kth moment arbitrary node i to mobile object；Hi(k) it is kth moment arbitrary node i It is a time-varying matrix to the observing matrix of mobile object；X (k) is state of the mobile object at the kth moment；vi(k) it is kth Moment arbitrary node i is to the observation noise of mobile object, noise statistics amount：
E[vi(k)vj(l)T]=Ri(k)δklδij,δkl=1 (k ≠ l), δkl=0 (k=l)
Ri(k) it is the observation noise covariance matrix of kth moment node i.
Any sensor node i has the observation z to mobile target x (k) at the arbitrary kth momenti(k), observation noise Covariance matrix Ri(k), predicted valueFilter estimated valuePredicting covariance matrix Pi(k) such as formula (5) institute Show, filtering evaluated error covariance matrix Mi(k), as shown in formula (6)：
On this basis, during distributed Kalman knows together filtering algorithm, each node will also generate information vector ui(k), Information matrix Ui(k)。
Given Pi(k),The information of exchange is in communication with each other with sensorIts InJi(k)=Ni(k) ∪ { i }, Ni(k) it is the sensor node set that can be in communication with each other with node i at the k moment That is its neighbor node set, Ji(k) set being made of kth moment sensor i and its neighbor node.Distributed Kalman is total It is as follows to know filtering algorithm：
Information transmission mode of any sensor node i at the kth moment, all the sensors complete filtering estimation and are known as a wheel Terminate；
1, sensor node i obtains the observation z at kth momenti(k), observation covariance matrix Ri(k) and node i is to kth The predicted value of moment object space
2, the information vector u of calculate node ii(k) and information matrix Ui(k)；
ui(k)=Hi(k)TRi(k)-1zi(k), Ui(k)=Hi(k)TRi(k)-1Hi(k)
3, by informationIt is broadcast to neighbor node；
4, the information that neighbor node transmits is received from neighbor node
5, fused data generates information vector yi(k) and information matrix Si(k)；
6, Kalman's common recognition state estimation is calculated.
Wherein e is the smaller constant with mobile object traveling time step-length same order.
7, predicted value and prediction covariance matrix update：
Pi(k+1)=A (k) Mi(k)A(k)T+B(k)Qi(k)B(k)T
It is known together and is filtered by distributed Kalman, each node is while observing alone in wireless sensor network, phase Interchangeable information so that each node obtains the information of neighbor node, while improving to mobile object tracking accuracy, network Middle all the sensors gradually reach unanimity to the tracking estimated value of mobile object, complete the task of complete distributed object tracking.
4, the defect of existing distributed Kalman's common recognition filtering technique：
Defect mainly has at following 3 points existing for existing distribution Kalman common recognition filtering technique：
1), tracking system robustness is low.In each round filtering, each node has and only once by the letter of oneself Cease mi(k) neighbor node is passed to, this needs each sensor node to whether being transmitted across data and being marked and remember. Once there is node to carry out repeating to send by the information of oneself, algorithm performance will deteriorate, it cannot be guaranteed that algorithm stability.
2), data storage capacity is big.In each round filtering, each node is receiving information of neighbor nodes and is merging letter While ceasing vector y and information matrix S, the information vector u for retaining oneself and information matrix U is needed, to wait for a certain moment will The information of oneself is sent to neighbor node, thus increases storage burden to each sensor node, and store the data phase Once data have loss or deviation that algorithm will be caused unstable between, and the filtering estimated value of tracking performance variation and all nodes is not Common recognition state can be reached.
3), systematic tracking accuracy is low, and common recognition precision is low.In each round filtering, each sensor has perception model It encloses and communication range, when mobile object moves within the sensing range of sensor, which can just have the sight at the moment Measured value can just receive the information of the sensor when other sensors node is located in the sensor node communication radius.I.e. Distributed Kalman's common recognition filtering algorithm shown in above-mentioned steps is limited by sensor node sensing range and communication range, In the case that sensing range is constant, expanding communication range, the tracking accuracy and common recognition precision of the technology have larger improvement, but In practical application expand communication range energy expenditure be it is bigger, if only by the swapping data of neighbor node just reality The purpose for now expanding communication range will be one and more preferably select.
Invention content
The present invention is in order to which the tracking system robustness for solving existing distributed Kalman's common recognition filtering technique is low, data Amount of storage is big, the low problem low with common recognition precision of systematic tracking accuracy, to provide a kind of point based on pairs of gossip algorithms Cloth Kalman common recognition method for tracking moving target.
Distributed Kalman common recognition method for tracking moving target based on pairs of gossip algorithms, it is characterized in that：Wireless In sensor network, this method is realized by following steps：
Step 1: node i obtains observation z of the kth moment to mobile targeti(k), observation covariance matrix Ri(k) it and saves Predicted values of the point i to kth moment object spaceI is positive integer；K is positive number；
Step 2: according to formula：
ui(k)=Hi(k)TRi(k)-1zi(k)
Calculate the information vector u of kth moment node ii(k)；
In formula：Hi(k) it is observing matrix of the kth moment node i to mobile target, is a time-varying matrix；
According to formula：
Ui(k)=Hi(k)TRi(k)-1Hi(k)
Calculate the information matrix U of kth moment node ii(k)；
Step 3: a pair of of adjacent node (i, j) arbitrarily in selection network, is exchanged with each other each self-information：
uj(k) be kth moment node j information vector；Uj(k) be kth moment node j information matrix；It is The predicted value of k moment nodes j；
And merged into row information according to pairs of gossip algorithms, complete primary gossip iteration in pairs；
Step 4: repeating step 3, gossip iteration in pairs is carried out repeatedly, until owning in wireless sensor network The information m (k) of node reaches average common recognition, that is, passes through gossip iteration, the information m at the kth moment of arbitrary node ii(k) more It is newlyIt is shown below：
Step 5: node i is in information miOn the basis of ' (k), according to formula：
Generate information vector yi(k)；
In formula：N is the total number of sensor node in network；ut(k) be kth moment node t information vector；
According to formula：
Generate information matrix Si(k)；
Step 6: according to formula：
Calculate Kalman's common recognition state estimation；
In formula：Tr () is Matrix Calculating trace operator；γi(k) the common recognition coefficient of kth moment node i is indicated；Indicate warp Cross forecast updating values of the gossip iteration posterior nodal point i to kth moment object space；Node i before expression gossip iteration To the predicted value of kth moment object space；ε is the constant with mobile object traveling time step-length same order；It is the kth moment The filtering estimated value of node i；Pi(k) be kth moment node i predicting covariance matrix；Mi(k) it is kth moment node i Filtering evaluated error covariance matrix；
Step 7: the predicted value and prediction covariance matrix to sensor node i are updated；Work as wireless sensor network In all node is complete Kalmans and know together after state estimations, complete a distributed Kalman of the wheel based on pairs of gossip algorithms Common recognition movable object tracking.
In step 3, primary gossip iteration in pairs is according to formula：
It realizes；
In formula：T indicates the moment.
In step 4, the information m (k) of all nodes reaches average common recognition in wireless sensor network, i.e.,：
The information update of node i is：
Wherein：
In step 7, it is according to formula that predicted value and prediction covariance matrix to sensor node i, which are updated,：
Pi(k+1)=A (k) Mi(k)A(k)T+B(k)Qi(k)B(k)T
It realizes；
In formula：A (k) is state-transition matrix, indicates that object is moved to the state at+1 moment of kth from the kth moment and shifts pass System, is a time-varying matrix；B (k) is control matrix；Qi(k) it is the state-noise covariance matrix of kth moment node i.
The advantageous effect that the present invention obtains：
1) tracking system robustness, is improved.The random node that wakes up carries out information exchange, and need not remember some node is It is no to be once waken up.Movable object tracking is carried out using the distributed kalman filter algorithm based on pairs of gossip algorithms, in net All nodes calculate respective information vector u in network, after U, constantly wake up any pair of adjacent node at random and are handed over into row information It changes, arbitrary node can be repeated wake-up, need not remember whether this node was once waken up, and not influence final average common recognition As a result.
2), reduce data storage capacity.Each sensor receive after the information that neighbor node is sent just with the letter of oneself Breath is merged, oneself information vector u and information matrix U need not be individually stored.Using based on pairs of gossip algorithms Distributed kalman filter algorithm carries out movable object tracking, and in gossip iterative process, each sensor node need not be protected Information vector u when itself iteration being stayed to start and information matrix U, receive every time the information of neighbor node just with local information into Row fusion reduces the burden for individually preserving information vector u and information matrix U.
3) tracking accuracy and common recognition precision of tracking system, are improved.Utilize the distribution based on pairs of gossip algorithms Kalman filtering algorithm carries out movable object tracking, and by the gossip iteration of limited number of time, each sensor finally obtains network The average value of middle all the sensors information just obtains the information of all the sensors in network multiplied by with sensor sum in network And value, broken algorithm and limited by sensor finite communication so that each sensor improves the tracking accuracy of target, receives Hold back precision raising.
Description of the drawings
Fig. 1 is the schematic diagram of pairs of gossip algorithms in background technology；
Fig. 2 is network topology structure of wireless sensor schematic diagram；
Fig. 3 is the mobile route schematic diagram of target；
Fig. 4 is each sensing station tracking mode schematic diagram under distributed Kalman's common recognition filtering algorithm；
Fig. 5 is each sensor under the distributed Kalman common recognition method for tracking moving target based on pairs of gossip algorithms Position tracking status diagram；
Specific implementation mode
Specific implementation mode one, the distributed Kalman common recognition method for tracking moving target based on pairs of gossip algorithms, It is characterized in that：In wireless sensor network, this method is realized by following steps：
Step 1: node i obtains observation z of the kth moment to mobile targeti(k), observation covariance matrix Ri(k) it and saves Predicted values of the point i to kth moment object spaceI is positive integer；K is positive number；
Step 2: according to formula：
ui(k)=Hi(k)TRi(k)-1zi(k)
Calculate the information vector u of kth moment node ii(k)；
In formula：Hi(k) it is observing matrix of the kth moment node i to mobile target, is a time-varying matrix；
According to formula：
Ui(k)=Hi(k)TRi(k)-1Hi(k)
Calculate the information matrix U of kth moment node ii(k)；
Step 3: a pair of of adjacent node (i, j) arbitrarily in selection network, is exchanged with each other each self-information：
uj(k) be kth moment node j information vector；Uj(k) be kth moment node j information matrix；It is The predicted value of k moment nodes j；
And merged into row information according to pairs of gossip algorithms, complete primary gossip iteration in pairs；
Step 4: repeating step 3, gossip iteration in pairs is carried out repeatedly, until owning in wireless sensor network The information m (k) of node reaches average common recognition, that is, passes through gossip iteration, the information m at the kth moment of arbitrary node ii(k) more It is newlyIt is shown below：
Step 5: node i is in information miOn the basis of ' (k), according to formula：
Generate information vector yi(k)；
In formula：N is the total number of sensor node in network；ut(k) be kth moment node t information vector；
According to formula：
Generate information matrix Si(k)；
Step 6: according to formula：
Calculate Kalman's common recognition state estimation；
In formula：Tr () is Matrix Calculating trace operator；γi(k) the common recognition coefficient of kth moment node i is indicated；Indicate warp Cross forecast updating values of the gossip iteration posterior nodal point i to kth moment object space；Node i before expression gossip iteration To the predicted value of kth moment object space；ε is the constant with mobile object traveling time step-length same order；It is the kth moment The filtering estimated value of node i；Pi(k) be kth moment node i predicting covariance matrix；Mi(k) it is kth moment node i Filtering evaluated error covariance matrix；
Step 7: the predicted value and prediction covariance matrix to sensor node i are updated；Work as wireless sensor network In all node is complete Kalmans and know together after state estimations, complete a distributed Kalman of the wheel based on pairs of gossip algorithms Common recognition movable object tracking.
In step 3, primary gossip iteration in pairs is according to formula：
It realizes；
In formula：T indicates the moment.
In step 4, the information m (k) of all nodes reaches average common recognition in wireless sensor network, i.e.,：
The information update of node i is：
Wherein：
In step 7, it is according to formula that predicted value and prediction covariance matrix to sensor node i, which are updated,：
Pi(k+1)=A (k) Mi(k)A(k)T+B(k)Qi(k)B(k)T
It realizes；
In formula：A (k) is state-transition matrix, indicates that object is moved to the state at+1 moment of kth from the kth moment and shifts pass System, is a time-varying matrix；B (k) is control matrix；Qi(k) it is the state-noise covariance matrix of kth moment node i.
Below with the effect of the specific l-G simulation test verification present invention：
100 wireless sensor nodes are uniformly placed in 100 × 100 square metres of square planar region, between node Minimum spacing be 10 meters, the perception radius of node is 15 meters, and communication radius is 32 meters, and sensor can measure in sensing range Mobile object position, measured value by mean value be 0, variance be 9 meters white Gaussian noise interfere.
Mobile object does nonlinear motion in 100 × 100 square metres of square planar region, i.e., is done in plane domain Class linear motion has small disturbance on the basis of linear motion, once reaching boundary, goes to, and returns to plane domain.Herein The initial position co-ordinates of object are (0,0), and the initial velocity magnitude of transverse and longitudinal coordinate is 7 meter per seconds and 10 meter per seconds.It is rectangular contacting When zone boundary, velocity magnitude and direction are gradually changed, so that object remains at movement in square region.Object level side 5 meter per seconds and 8 meter per seconds are respectively may be about to the average speed with vertical direction.
Every 0.04 second wireless sensor network observation moving object position, and distributed Kalman's common recognition filtering is carried out respectively It knows together and filters with the distributed Kalman based on pairs of gossip, the traveling time step-length of mobile object is 0.04 second at this time. The specific implementation step of two methods is as follows.
(1), distributed Kalman's common recognition filter tracking technology
Step 1：All nodes carry out perception measurement to the moving object position in region and obtain this in every 0.04 second network Moment is to the measured value of moving object position and oneself measurement vector sum calculation matrix to moving object position is calculated.
Step 2：In this 0.04 second, arbitrary node sends the information of oneself to neighbor node, including measures vector, surveys The predicted value of moment matrix and last moment to this moment object space.
Step 3：In this 0.04 second, arbitrary node receives the information that all neighbor nodes are sent, later with oneself Information is merged to obtain fuse information vector sum information matrix.
Step 4：In this 0.04 second, arbitrary node calculates Kalman's common recognition state estimation.
Step 5：In this 0.04 second, arbitrary node is filtered update, obtains the predicted value of subsequent time object space With corresponding prediction covariance matrix.
Step 6：Every 0.04 second repetition step 1 is gradually completed to step 5 to wireless sensor network moving target position The tracking set.
(2), the distributed Kalman common recognition filter tracking technology based on pairs of gossip algorithms
Step 1：All nodes carry out perception measurement to the moving object position in region and obtain this in every 0.04 second network Moment is to the measured value of moving object position and oneself measurement vector sum calculation matrix to moving object position is calculated.
Step 2：In this 0.04 second, arbitrary a pair of of adjacent node of selection sends mutually respective information, including measure to Amount, calculation matrix and last moment to the predicted value of this moment object space, carry out data fusion according to pairs of gossip later I.e. each sensor node is averaged the information received and the information of oneself to obtain oneself new information.
Step 3：In this 0.04 second, step 2 is repeated 4000 times, 0.01 millisecond of each used time.
Step 4：Arbitrary node by the information of oneself final updated be multiplied by sensor sum in network obtain fuse information to Amount and information matrix.
Step 5：In this 0.04 second, arbitrary node calculates Kalman's common recognition state estimation.
Step 6：In this 0.04 second, arbitrary node is filtered update, obtains the predicted value of subsequent time object space With corresponding prediction covariance matrix.
Step 7：Every 0.04 second repetition step 1 is gradually completed to step 5 to wireless sensor network moving target position The tracking set.
It observes mobile object 40 seconds tracking results to be compared as follows, application distribution formula Kalman common recognition filtering algorithm carries out The result of movable object tracking is indicated with KCF, using the distributed Kalman common recognition filtering algorithm based on pairs of gossip algorithms The result for carrying out movable object tracking is indicated with PBKCF.
Fig. 2 is wireless sensor network topology figure.Circle indicates sensor node in figure, and the line between node indicates node Between have side, can be in communication with each other.
Fig. 3 is the mobile route figure of object.Object does nonlinear motion, this motion model is known as particle model in box.
Fig. 4 and Fig. 5 is each sensing station tracking mode figure.Curve 41 is the motion track of object, and curve 42 is not With sensor node to the filtering estimation curve of mobile object.
By network to the instantaneous tracking mode figure of mobile object and each sensing station tracking mode figure it can be found that answering The tracking to object, tracking accuracy and each biography are realized with the distributed kalman filter algorithm based on pairs of gossip algorithms The common recognition precision that sensor node filters object space estimated value is above application distribution formula Kalman common recognition filtering algorithm realization To the tracking situation of object.

Claims (4)

  1. The method for tracking moving target 1. the distributed Kalman based on pairs of gossip algorithms knows together, it is characterized in that：Wirelessly passing In sensor network, this method is realized by following steps：
    Step 1: node i obtains observation z of the kth moment to mobile targeti(k), observation covariance matrix Ri(k) and node i To the predicted value of kth moment object spaceI is positive integer；K is positive number；
    Step 2: according to formula：
    ui(k)=Hi(k)T Ri(k)-1zi(k)
    Calculate the information vector u of kth moment node ii(k)；
    In formula：Hi(k) it is observing matrix of the kth moment node i to mobile target, is a time-varying matrix；
    According to formula：
    Ui(k)=Hi(k)T Ri(k)-1Hi(k)
    Calculate the information matrix U of kth moment node ii(k)；
    Step 3: a pair of of adjacent node (i, j) arbitrarily in selection network, is exchanged with each other each self-information：
    uj(k) be kth moment node j information vector；Uj(k) be kth moment node j information matrix；When being kth Carve the predicted value of node j；
    And merged into row information according to pairs of gossip algorithms, complete primary gossip iteration in pairs；
    Step 4: repeating step 3, gossip iteration in pairs is carried out repeatedly, until all nodes in wireless sensor network Information m (k) reach average common recognition, that is, pass through gossip iteration, the information m at the kth moment of arbitrary node ii(k) it is updated toIt is shown below：
    Step 5: node i is in information m 'i(k) on the basis of, according to formula：
    Generate information vector yi(k)；
    In formula：N is the total number of sensor node in network；ut(k) be kth moment node t information vector；
    According to formula：
    Generate information matrix Si(k)；
    Step 6: according to formula：
    Calculate Kalman's common recognition state estimation；
    In formula：Tr () is Matrix Calculating trace operator；γi(k) the common recognition coefficient of kth moment node i is indicated；It indicates to pass through Forecast updating values of the gossip iteration posterior nodal point i to kth moment object space；Node i before expression gossip iteration To the predicted value of kth moment object space；ε is the constant with mobile object traveling time step-length same order；It is the kth moment The filtering estimated value of node i；Pi(k) be kth moment node i predicting covariance matrix；Mi(k) it is kth moment node i Filtering evaluated error covariance matrix；
    Step 7: the predicted value and prediction covariance matrix to sensor node i are updated；When institute in wireless sensor network After thering is node i to complete Kalman's common recognition state estimation, distributed Kalman common recognition of the wheel based on pairs of gossip algorithms is completed Movable object tracking.
  2. The movable object tracking side 2. the distributed Kalman according to claim 1 based on pairs of gossip algorithms knows together Method, it is characterised in that in step 3, primary gossip iteration in pairs is according to formula：
    It realizes；
    In formula：T indicates the moment.
  3. The movable object tracking side 3. the distributed Kalman according to claim 1 based on pairs of gossip algorithms knows together Method, it is characterised in that in step 4, the information m (k) of all nodes reaches average common recognition in wireless sensor network, i.e.,：
    The information update of node i is：
    Wherein：
  4. The movable object tracking side 4. the distributed Kalman according to claim 1 based on pairs of gossip algorithms knows together Method, it is characterised in that in step 7, it is according to public affairs that predicted value and prediction covariance matrix to sensor node i, which are updated, Formula：
    Pi(k+1)=A (k) Mi(k)A(k)T+B(k)Qi(k)B(k)T
    It realizes；
    In formula：A (k) is state-transition matrix, indicates that object is moved to the state transfer relationship at+1 moment of kth from the kth moment, is One time-varying matrix；B (k) is control matrix；Qi(k) it is the state-noise covariance matrix of kth moment node i.
CN201511009138.9A 2015-12-28 2015-12-28 Distributed Kalman common recognition method for tracking moving target based on pairs of gossip algorithms Active CN105491587B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201511009138.9A CN105491587B (en) 2015-12-28 2015-12-28 Distributed Kalman common recognition method for tracking moving target based on pairs of gossip algorithms

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201511009138.9A CN105491587B (en) 2015-12-28 2015-12-28 Distributed Kalman common recognition method for tracking moving target based on pairs of gossip algorithms

Publications (2)

Publication Number Publication Date
CN105491587A CN105491587A (en) 2016-04-13
CN105491587B true CN105491587B (en) 2018-11-02

Family

ID=55678220

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201511009138.9A Active CN105491587B (en) 2015-12-28 2015-12-28 Distributed Kalman common recognition method for tracking moving target based on pairs of gossip algorithms

Country Status (1)

Country Link
CN (1) CN105491587B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106054171B (en) * 2016-05-27 2021-07-13 中国电子科技集团公司信息科学研究院 Multi-radar node self-adaptive selection tracking method based on information entropy
CN107767405B (en) * 2017-09-29 2020-01-03 华中科技大学 Nuclear correlation filtering target tracking method fusing convolutional neural network

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103024759A (en) * 2013-01-18 2013-04-03 哈尔滨工业大学 Wireless communication method based on quantization consensus distributed Gossip algorithm
CN103648164A (en) * 2013-12-19 2014-03-19 哈尔滨工业大学 Time difference of arrival and Gossip algorithm based wireless sensor network distributed positioning method
CN103648083A (en) * 2013-12-27 2014-03-19 哈尔滨工业大学 Distributed average-consensus broadcast Gossip wireless communication method

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103024759A (en) * 2013-01-18 2013-04-03 哈尔滨工业大学 Wireless communication method based on quantization consensus distributed Gossip algorithm
CN103648164A (en) * 2013-12-19 2014-03-19 哈尔滨工业大学 Time difference of arrival and Gossip algorithm based wireless sensor network distributed positioning method
CN103648083A (en) * 2013-12-27 2014-03-19 哈尔滨工业大学 Distributed average-consensus broadcast Gossip wireless communication method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
《移动自组织网络一种自适应gossip机制的路由算法》;袁永琼等;《遥测遥控》;20100531;第31卷(第3期);第1-6页 *
Distributed Averaging in Sensor Networks Based on Broadcast Gossip Algorithms;Mauro Franceschelli等;《IEEE SENSORS JOURNAL》;20110331;第11卷(第3期);第808-817页 *

Also Published As

Publication number Publication date
CN105491587A (en) 2016-04-13

Similar Documents

Publication Publication Date Title
CN103648108B (en) Sensor network distributed consistency object state estimation method
CN105263113B (en) A kind of WiFi location fingerprints map constructing method and its system based on crowdsourcing
CN105491587B (en) Distributed Kalman common recognition method for tracking moving target based on pairs of gossip algorithms
CN108051999A (en) Accelerator beam path control method and system based on deeply study
CN107084714A (en) A kind of multi-robot Cooperation object localization method based on RoboCup3D
CN106570563A (en) Deformation prediction method and apparatus based on Kalman filtering and BP neural network
CN104331630B (en) A kind of multi tate observes the state estimation and data fusion method of data
Barcelo et al. Dynamic OD matrix estimation exploiting bluetooth data in urban networks
CN107390517A (en) Robust adaptive non-singular terminal sliding-mode control for train ATO systems
CN101216546B (en) Wireless sensor network target positioning location estimation method
CN101635941B (en) Target tracking method based on master/slaver mobile agent in wireless sensor network
CN105722030B (en) A kind of DTN nodes position predicting method
CN110278571A (en) It is a kind of based on simple forecast-correction link distributed signal tracking
CN105898691B (en) Wireless sensor network target tracking method based on particlized sum-product algorithm
CN108347694B (en) Node positioning method and system based on boundary conditions
CN106019253A (en) Box particle CPHD based multi-expansion-target tracking method
CN103889053B (en) Automatic establishing method of self-growing-type fingerprint
CN105491588B (en) Distributed Kalman common recognition method for tracking moving target based on unbiased broadcast gossip algorithm
CN113342029B (en) Maximum sensor data acquisition path planning method and system based on unmanned aerial vehicle cluster
CN104808486B (en) The forecast Control Algorithm and device of piezoelectric ceramic actuator based on fuzzy T-S model
Hegyi et al. Parallelized particle filtering for freeway traffic state tracking
CN104714520A (en) Fault estimation method under sensor network environment and based on Green space theory
CN103940430B (en) Multisensor configuration and multi-object tracking method are moved in a kind of state constraint
CN103313384A (en) Wireless sensor network target tracking method based on informational consistency
Zhou et al. Dynamic origin-destination demand flow estimation utilizing heterogeneous data sources under congested traffic conditions

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant