CN109246637B - Distributed sensor network collaborative registration method and system - Google Patents
Distributed sensor network collaborative registration method and system Download PDFInfo
- Publication number
- CN109246637B CN109246637B CN201810950034.5A CN201810950034A CN109246637B CN 109246637 B CN109246637 B CN 109246637B CN 201810950034 A CN201810950034 A CN 201810950034A CN 109246637 B CN109246637 B CN 109246637B
- Authority
- CN
- China
- Prior art keywords
- sensor
- kalman filtering
- module
- node
- representing
- 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
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/30—Services specially adapted for particular environments, situations or purposes
- H04W4/38—Services specially adapted for particular environments, situations or purposes for collecting sensor information
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/02—Arrangements for optimising operational condition
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Navigation (AREA)
Abstract
The invention provides a distributed sensor network collaborative registration method and a distributed sensor network collaborative registration system, wherein initial sensor registration error parameters and target parameters are formed at each sensor node; starting EM iterative computation, and carrying out forward Kalman filtering on the target parameters; performing reverse Kalman filtering on the target parameters subjected to forward Kalman filtering; smoothing the target state estimation by each sensor node by using the results of the forward Kalman filtering step and the reverse Kalman filtering step; each sensor node uses the smoothed target state estimation obtained in the target state estimation smoothing step to solve the respective sensor registration error estimation value; if the EM iteration is not finished, returning to the forward Kalman filtering step; and if the EM iteration is finished, outputting respective sensor registration error estimated values by each sensor node. The registration process of the invention does not need a central node and a full connection structure between nodes, and the method is simple, effective and easy to implement.
Description
Technical Field
The invention relates to the technical field of communication, in particular to a distributed sensor network collaborative registration method and system.
Background
The distributed sensor network is a sensor network without a central node and full connection between nodes. In a distributed sensor network, each node can only communicate with a portion of the nodes within its communication range. The characteristic of the distributed sensor network endows the distributed sensor network with a flexible and changeable topological structure, strong environment adaptability and robustness to node failure in a system, and has wide application value and application prospect. However, while this feature of a distributed sensor network provides the aforementioned benefits and advantages, it also provides significant challenges to the information processing of the sensor network. A completely distributed information processing method is required for a structure without a central node and without full connection. In order to solve this problem, many scholars represented by r.olfati-Saber, g.battistelli, and the like have proposed a distributed sensor network information processing method based on a consistency (consensus) policy. According to the method, local communication between adjacent nodes in the distributed sensor network is utilized, consistency iteration is carried out between the adjacent nodes in the sensor network, and all nodes in the distributed sensor network can obtain globally consistent estimated values.
The core significance of the sensor network lies in that the external environment is sensed by the multi-source sensor, the system obtains higher precision compared with a single sensor through the fusion of multi-source information, and the performance of the whole sensor network system is improved. However, in practical system applications, the presence of sensor registration errors can significantly degrade the performance of the fusion and even lead to fusion failures. For this reason, the registration of the sensors is an essential and important link in the application of sensor networks. In centralized systems, the registration problem of sensors has been studied intensively, and many improved algorithms including Least squares (Least squares) method, Maximum likelihood (Maximum likelihood) method, and the like have been proposed. However, the foregoing methods all require information processing to be performed centrally at the central node. This requirement limits the application of these methods in distributed sensor networks.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a distributed sensor network collaborative registration method and system.
The invention provides a distributed sensor network collaborative registration method, which comprises the following steps:
an initialization step: forming initial sensor registration error parameters according to prior information and forming initial target parameters according to a given initial target state and an error covariance matrix at each sensor node;
a forward Kalman filtering step: starting EM iterative computation, and carrying out forward Kalman filtering on the target parameters;
and (3) an inverse Kalman filtering step: performing reverse Kalman filtering on the target parameters subjected to forward Kalman filtering;
and a target state estimation smoothing step: smoothing the target state estimation by each sensor node by using the results of the forward Kalman filtering step and the reverse Kalman filtering step;
solving the sensor registration error estimation value: each sensor node uses the smoothed target state estimation obtained in the target state estimation smoothing step to solve the respective sensor registration error estimation value;
a judging step: if the EM iteration is not finished, returning to the forward Kalman filtering step; and if the EM iteration is finished, outputting respective sensor registration error estimated values by each sensor node.
Preferably, the forward kalman filtering in the forward kalman filtering step includes:
forward initial step filtering: performing forward consistency Kalman filtering by using the initial sensor registration error parameter and the initial target parameter in the initialization step as initial values;
forward non-initial step filtering: and performing forward consistency Kalman filtering by using the sensor registration error estimated value obtained in the sensor registration error estimated value solving step and the initial target parameter in the initialization step as initial values.
Preferably, the inverse kalman filtering in the inverse kalman filtering step includes:
and (3) reverse initial step filtering: performing inverse consistency Kalman filtering by using the target parameters obtained in the forward Kalman filtering step after the forward Kalman filtering and the initial sensor registration error parameters in the initialization step as initial values;
reverse non-initial step filtering: and performing inverse consistency Kalman filtering by using the target parameters obtained in the forward Kalman filtering step after the forward Kalman filtering and the sensor registration error parameter estimation value obtained in the sensor registration error estimation value solving step as initial values.
Preferably, the EM iteration adopts M-step EM iteration, the forward kalman filtering adopts N-step forward kalman filtering, and the inverse kalman filtering adopts N-step inverse kalman filtering;
wherein M, N is a positive integer.
The invention provides a distributed sensor network collaborative registration system, which comprises modules:
an initialization module: forming initial sensor registration error parameters according to prior information and forming initial target parameters according to a given initial target state and an error covariance matrix at each sensor node;
a forward Kalman filtering module: starting EM iterative computation, and carrying out forward Kalman filtering on the target parameters;
an inverse Kalman filtering module: performing reverse Kalman filtering on the target parameters subjected to forward Kalman filtering;
a target state estimation smoothing module: each sensor node smoothes the target state estimation by using the results of the forward Kalman filtering module and the reverse Kalman filtering module;
a sensor registration error estimation value solving module: each sensor node uses the smoothed target state estimation obtained by the target state estimation smoothing module to solve the respective sensor registration error estimation value;
a judging module: if the EM iteration is not finished, returning to the forward Kalman filtering module; and if the EM iteration is finished, outputting respective sensor registration error estimated values by each sensor node.
Preferably, the forward kalman filtering in the forward kalman filtering module includes:
forward initial step filtering: performing forward consistency Kalman filtering by using an initial sensor registration error parameter and an initial target parameter in an initialization module as initial values;
forward non-initial step filtering: and performing forward consistency Kalman filtering by using the sensor registration error estimation value obtained by the sensor registration error estimation value solving module and an initial target parameter in the initialization module as initial values.
Preferably, the inverse kalman filtering in the inverse kalman filtering module includes:
and (3) reverse initial step filtering: performing inverse consistency Kalman filtering by using the target parameters obtained after the forward Kalman filtering in the forward Kalman filtering module and the initial sensor registration error parameters in the initialization module as initial values;
reverse non-initial step filtering: and performing inverse consistency Kalman filtering by using the target parameters obtained after the forward Kalman filtering in the forward Kalman filtering module and the sensor registration error parameter estimation value obtained by the sensor registration error estimation value solving module as initial values.
Preferably, the EM iteration adopts M-step EM iteration, the forward kalman filtering adopts N-step forward kalman filtering, and the inverse kalman filtering adopts N-step inverse kalman filtering;
wherein M, N is a positive integer.
Compared with the prior art, the invention has the following beneficial effects:
the invention can register the sensor network nodes in the redundant information of the distributed sensor network multi-sensor. The registration process does not need a central node and a full connection structure between nodes, each node is subjected to iterative calculation through local communication between adjacent nodes, the method is simple, effective and easy to implement, is particularly suitable for the application of a distributed sensor network without the central node, and can be widely applied to various fields such as robots, intelligent traffic, air traffic control, aerospace, aviation, navigation and the like.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
FIG. 1 is a flow chart of the operation of the present invention;
FIG. 2 is a schematic diagram of a topology of a sensor network and a motion trajectory of a target according to an embodiment of the present invention;
FIG. 3 is a plot of angular registration error estimate as a function of EM iterations for an embodiment of the present invention;
fig. 4 is a plot of range registration error estimate versus number of EM iterations for an embodiment of the present invention.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the present invention.
As shown in fig. 1, a distributed sensor network collaborative registration method provided by the present invention includes:
step S1: forming initial sensor registration error parameters according to prior information and forming initial target parameters according to a given initial target state and an error covariance matrix at each sensor node;
step S2: and starting M steps of EM iterative calculation. Each sensor node performs N-step forward consistency Kalman filtering on target parameters according to the measurement of N time points acquired by the sensor node and the information of adjacent nodes, wherein the N-step forward consistency Kalman filtering comprises forward initial step filtering and forward non-initial step filtering, and M, N is a positive integer; wherein:
forward initial step filtering: performing forward consistency Kalman filtering by using the initial sensor registration error parameter and the initial target parameter in the step S1 as initial values;
forward non-initial step filtering: and performing forward consistency Kalman filtering by using the sensor registration error estimated value obtained in the step S5 and the initial target parameter in the step S1 as initial values.
Step S3: each sensor node performs N-step reverse consistency Kalman filtering on target parameters subjected to N-step forward Kalman filtering according to the measurement of N time points acquired by the sensor node and the information of adjacent nodes, wherein the N-step reverse consistency Kalman filtering comprises reverse initial time filtering and reverse non-initial time filtering; wherein:
and (3) reverse initial step filtering: performing inverse consistency Kalman filtering by using the target parameters obtained in the S2 after the forward Kalman filtering and the initial sensor registration error parameters in the S1 as initial values;
reverse non-initial step filtering: and performing inverse consistency Kalman filtering by using the target parameters obtained in the step S2 after the forward Kalman filtering and the sensor registration error parameter estimation value obtained in the step S5 as initial values.
Step S4: each sensor node smoothes the target state estimate using the results of S2 and S3.
Step S5: each sensor node solves for a respective sensor registration error estimate using the smoothed target state estimate obtained at S4.
Step S6: if the EM iteration is not finished, returning to S2; and if the EM iteration is finished, outputting respective sensor registration error estimated values by each sensor node. (if M is M, then returning to S2; if M is M, then each sensor node outputs its own sensor registration error estimation value)
In step S1, each sensor node is given an initial target stateAnd corresponding error covariance matrixAnd respective a priori registration errors etai(0)WhereinEach representing a respective one of the different sensor nodes,is a set of all sensor nodes.
In step S2, N recursive forward kalman filter calculations (k ═ 1, 2.., N) are performed, each recursive calculation including:
step S2.1, each sensor node i utilizes the target state at the moment k-1Sum error covariance matrixEstimating the target state to be predicted in one step, and respectively estimating the target state at the moment of k and the predicted value of the error covariance matrixAndfrom the obtained predicted values, a predicted information matrix is calculatedAnd information vectorThe calculation formula is as follows:
step S2.2, measuring each sensor node i according to k timeCalculating new information vectorsAnd new information matrixThe calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,sensor measurement equation, partial differential, representing time kSubscript ofIs expressed as partial differentialThe value of (A) is selected,representing the measured noise variance matrix at time k for sensor i,representing the measurement value of the sensor node i at time k, m representing the number of current EM iterations, ηi(m-1)Representing the registration error estimate for the sensor node i of the previous EM iteration.
Step S2.3, each sensor node i pair information matrix obtained in step S2.1 and step S2.2Information vectorNew information matrixAnd new information vectorCarrying out L-step consistency iteration; computing unit for each step of consistency iteration of each sensor node iThe formula is as follows:
in the formula: all superscripts i represent sensor nodes i,represents a set formed by all nodes capable of directly communicating with the node i, including the node i, j represents all nodes capable of directly communicating with the node i, including the node i, L represents the step number of the current consistency stack, and L is 1,2 πi,jSatisfy pi for consistency weighti,jIs not less than 0 and
s2.4, measuring and updating each sensor node i;
the updated calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,representing the number of sensor nodes in the sensor network.
Step S2.5, estimating and extracting the target state at the current moment k, wherein the calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,a forward filtered estimate representing the target state at time k,the estimated error variance corresponding to the forward filtering estimated value of the target at the moment k;
step S2.6, when k < N, k ═ k +1 and return to performing step S2.1; when k is equal to N, all are output Andand
in step S3, N recursive inverse kalman filter calculations (k ═ N, N-1.., 1) are performed, each recursive calculation including:
s3.1, each sensor node i carries out one-step backward prediction on the target state by utilizing the target state and covariance estimation at the moment k to respectively obtain predicted values of the state and covariance at the moment k-1Andfrom the obtained predicted values, a predicted information matrix is calculatedAnd information vectorThe calculation formula is as follows:
step S3.2, measuring each sensor node i according to k timeCalculating new information vectorsAnd new information matrixThe calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,sensor measurement equation, partial differential, representing time kSubscript ofIs expressed as partial differentialThe value of (A) is selected,representing the measured noise variance matrix at time k for sensor i,representing the measurement value of a sensor node i at the moment k, representing that all superscripts b represent that the variable is a variable in the inverse Kalman filtering, representing the number of current EM iterations, etai(m-1)Representing the registration error estimate for the sensor node i of the previous EM iteration.
Step S3.3, each sensor node i pair information matrix obtained in step S3.1 and step S3.2Information vectorNew information matrixAnd new information vectorCarrying out L-step consistency iteration; the calculation formula of each step of consistency iteration of each sensor node i is as follows:
in the formula: all superscripts i represent sensor nodes i,represents a set formed by all nodes capable of directly communicating with the node i, including the node i, j represents all nodes capable of directly communicating with the node i, including the node i, L represents the step number of the current consistency stack, and L is 1,2 πi,jSatisfy pi for consistency weighti,jIs not less than 0 and
s3.4, measuring and updating each sensor node i;
the updated calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,representing the number of sensor nodes in the sensor network.
Step S3.5, estimating and extracting the target state at the current moment k, wherein the calculation formula is as follows:
Step S4 includes:
the sensor nodes i smooth the state estimation of the target by using the structures of the step S3 and the step S4 to obtain the smoothed state estimationAnd corresponding covariance matrixThe calculation formula is as follows:
step S5 includes:
each sensor node i respectively calculates the registration error estimation value eta of the sensor node ii(m)The calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,sensor measurement equation, partial differential, representing time kSubscript η ofi=ηi(m-1)Is expressed as partial differential at etai(m-1)Value of ηi(m-1)Representing the registration error estimate for the sensor node i of the previous EM iteration.
Step S6 includes: when the EM iteration is not ended, returning to execute step S2; when the EM iteration is finished, each sensor node i outputs etai(M)As an estimate of the registration error.
The technical solution of the present embodiment is further described in detail with reference to the accompanying drawings.
Step one, initializing each node of the sensor to form initial parameters.
At each sensor nodeA priori registration error η for a given sensori(0)And initial target state estimationAnd estimate error varianceWherein the content of the first and second substances,is a set of all sensor nodes.
And step two, at each step M (M is 1, 2.. multidot.M) of the EM iteration, calculating N-step forward consistency Kalman filtering by using local information and local information acquired by local iterative communication at each sensor node i. And step three, at each step M (M is 1, 2.. multidot.M) of the EM iteration, calculating N-step inverse consistency Kalman filtering by using local information and local information acquired by local iterative communication at each sensor node i.
Step four, at each step M (M is 1, 2.. multidot.m) of the EM iteration, each sensor node i smoothes the target state estimation by using the results of the step two and the step three.
Step five, at each step M (M is 1,2,.. multidot.M) of the EM iteration, each sensor node i utilizes the smoothed target state estimation obtained in the step four to solve the respective registration error estimation value
Step six, after M-step EM iteration is finished, each sensor node i outputs an estimation value eta of self-registration errori(M)。
Consider the tracking problem of a two-dimensional plane. Consider a distributed sensor network consisting of 36 sensor nodes, each sensor being randomly deployed in a 5000m planar space, each sensor being able to communicate only with its neighboring nodes within its communication range. The sensor measures the azimuth angle and the relative distance to the target, and the measurement equation is
Wherein the content of the first and second substances,in order to measure the noise, the noise is measured,is the position coordinates of the sensor. The registration error of each sensor is constant and is initially distributed by GaussianWherein is randomly generated, wherein0=[0.5°,5m]T,Pη=diag([(3°)2,(3.3m)2]) The state of the target at the initial time is
x0=[1700m,18m/s,4200m,-12m/s]T
P0=diag([102m2,3.22m2/s2,102m2,3.22m2/s2]T)
The target motion adopts a constant velocity model (CV). The topology of the sensor network and the motion trajectory of the target are shown in fig. 2.
After the initial values and the simulation parameters are given, the specific steps are as follows:
step S1: EM iteration
For 1.. times, M (For each M1.. times, M, the following operations are performed)
Step S2.1: n-step forward consistency Kalman filtering
For k 1.., N (For each k 1.., N, the following operations are performed)
And (3) prediction: each sensor node i independently predicts the target state by adopting the prediction step of Kalman filtering, and calculates the predicted target stateSum estimation error variance matrixAnd calculating corresponding information matrixAnd information vector
consistency iteration:
for 1.. times, L (For each L ═ 1.. times, L, the following operations are performed)
End (End)
Updating:
end (End)
Step S2.2: n-step inverse consistency Kalman filtering
For k 1.., N (For each k 1.., N, the following operations are performed)
And (3) backward prediction: each sensor node i independently performs target state reverse prediction by adopting prediction steps of Kalman filtering, and calculates the predicted target stateSum estimation error variance matrixAnd calculating corresponding information matrixAnd information vector
consistency iteration:
for 1.. times, L (For each L ═ 1.. times, L, the following operations are performed)
End (End)
Updating:
end (End)
Step S2.3: smoothing target state estimates
Step S2.4: solving registration error estimates for each sensor i
End (End)
Step S2: output η of each sensor node ii(M)As an estimate of the registration error.
The present embodiment uses Matlab language to test the proposed algorithm at different number L of consistent iterations and compare with the centralized sensor EM registration algorithm. Fig. 3 and 4 show the variation of sensor angular and range registration error estimates as a function of the number of EM iterations, respectively.
As can be seen from fig. 3 and 4, the proposed method can perform registration on sensor nodes in a distributed sensor network under the condition of distributed non-fusion center, and the registration convergence speed gradually approaches to a centralized EM registration method along with the increase of the number of consistency iterations.
The distributed sensor network collaborative registration method provided by the embodiment is a distributed sensor network registration algorithm. In particular, the present invention relates to a distributed registration method based on a consistency (consensus) method and an Expectation Maximization (EM) method. The method embeds the consistent iteration of the consistent algorithm into the calculation process of the EM iteration, so that the conditional expectation of a log-likelihood function can be calculated in a fully distributed mode, and the estimated value of the sensor registration error is solved by maximizing the conditional expectation. Simulation results show that the embodiment can effectively register each sensor node in the distributed sensor network. The present embodiments may be applied to sensor registration scenarios for various types of distributed sensor networks.
On the basis of the distributed sensor network collaborative registration method, the invention also provides a distributed sensor network collaborative registration system, which comprises the following steps:
an initialization module: forming initial sensor registration error parameters according to prior information and forming initial target parameters according to a given initial target state and an error covariance matrix at each sensor node;
a forward Kalman filtering module: starting EM iterative computation, and carrying out forward Kalman filtering on the target parameters;
an inverse Kalman filtering module: carrying out reverse Kalman filtering on the target parameters subjected to the N-step forward Kalman filtering;
a target state estimation smoothing module: each sensor node smoothes the target state estimation by using the results of the forward Kalman filtering module and the reverse Kalman filtering module;
a sensor registration error estimation value solving module: each sensor node uses the smoothed target state estimation obtained by the target state estimation smoothing module to solve the respective sensor registration error estimation value;
a judging module: if the EM iteration is not finished, returning to the forward Kalman filtering module; and if the EM iteration is finished, outputting respective sensor registration error estimated values by each sensor node.
The forward kalman filtering in the forward kalman filtering module includes:
forward initial step filtering: performing forward consistency Kalman filtering by using an initial sensor registration error parameter and an initial target parameter in an initialization module as initial values;
forward non-initial step filtering: and performing forward consistency Kalman filtering by using the sensor registration error estimation value obtained by the sensor registration error estimation value solving module and an initial target parameter in the initialization module as initial values.
The inverse kalman filtering in the inverse kalman filtering module includes:
and (3) reverse initial step filtering: performing inverse consistency Kalman filtering by using the target parameters obtained after the forward Kalman filtering in the forward Kalman filtering module and the initial sensor registration error parameters in the initialization module as initial values;
reverse non-initial step filtering: and performing inverse consistency Kalman filtering by using the target parameters obtained after the forward Kalman filtering in the forward Kalman filtering module and the sensor registration error parameter estimation value obtained by the sensor registration error estimation value solving module as initial values.
Those skilled in the art will appreciate that, in addition to implementing the system and its various devices, modules, units provided by the present invention as pure computer readable program code, the system and its various devices, modules, units provided by the present invention can be fully implemented by logically programming method steps in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Therefore, the system and various devices, modules and units thereof provided by the invention can be regarded as a hardware component, and the devices, modules and units included in the system for realizing various functions can also be regarded as structures in the hardware component; means, modules, units for performing the various functions may also be regarded as structures within both software modules and hardware components for performing the method.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.
Claims (4)
1. A distributed sensor network collaborative registration method is characterized by comprising the following steps:
an initialization step: forming initial sensor registration error parameters according to prior information and forming initial target parameters according to a given initial target state and an error covariance matrix at each sensor node;
a forward Kalman filtering step: starting EM iterative computation, and carrying out forward Kalman filtering on the target parameters;
and (3) an inverse Kalman filtering step: performing reverse Kalman filtering on the target parameters subjected to forward Kalman filtering;
and a target state estimation smoothing step: smoothing the target state estimation by each sensor node by using the results of the forward Kalman filtering step and the reverse Kalman filtering step;
solving the sensor registration error estimation value: each sensor node uses the smoothed target state estimation obtained in the target state estimation smoothing step to solve the respective sensor registration error estimation value;
a judging step: if the EM iteration is not finished, returning to the forward Kalman filtering step; if the EM iteration is finished, each sensor node outputs a respective sensor registration error estimation value;
the forward kalman filtering in the forward kalman filtering step includes:
forward initial step filtering: performing forward consistency Kalman filtering by using the initial sensor registration error parameter and the initial target parameter in the initialization step as initial values;
forward non-initial step filtering: performing forward consistency Kalman filtering by using the sensor registration error estimated value obtained in the sensor registration error estimated value solving step and the initial target parameter in the initialization step as initial values;
the inverse kalman filtering in the inverse kalman filtering step includes:
and (3) reverse initial step filtering: performing inverse consistency Kalman filtering by using the target parameters obtained in the forward Kalman filtering step after the forward Kalman filtering and the initial sensor registration error parameters in the initialization step as initial values;
reverse non-initial step filtering: performing inverse consistency Kalman filtering by using the target parameters obtained in the forward Kalman filtering step after the forward Kalman filtering and the sensor registration error parameter estimation value obtained in the sensor registration error estimation value solving step as initial values;
in the initialization step, initial target states are respectively given to all the sensor nodesAnd corresponding error covarianceMatrix ofAnd respective a priori registration errors etai(0)WhereinEach representing a respective one of the different sensor nodes,a set formed by all sensor nodes;
in the forward kalman filtering step, N recursive forward kalman filtering calculations (k ═ 1, 2.., N) are performed, each recursive calculation step including:
step S2.1, each sensor node i utilizes the target state at the moment k-1Sum error covariance matrixEstimating the target state to be predicted in one step, and respectively estimating the target state at the moment of k and the predicted value of the error covariance matrixAndfrom the obtained predicted values, a predicted information matrix is calculatedAnd information vectorThe calculation formula is as follows:
step S2.2, measuring each sensor node i according to k timeCalculating new information vectorsAnd new information matrixThe calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,sensor measurement equation, partial differential, representing time kSubscript ofIs expressed as partial differentialThe value of (A) is selected,representing the measured noise variance matrix at time k for sensor i,representing the measurement value of the sensor node i at time k, m representing the number of current EM iterations, ηi(m-1)Representing a registration error estimated value of a previous EM iteration sensor node i;
step S2.3, each sensor node i pair information matrix obtained in step S2.1 and step S2.2Information vectorNew information matrixAnd new information vectorCarrying out L-step consistency iteration; the calculation formula of each step of consistency iteration of each sensor node i is as follows:
in the formula: all superscripts i represent sensor nodes i,represents a set formed by all nodes capable of directly communicating with the node i, including the node i, j represents all nodes capable of directly communicating with the node i, including the node i, L represents the step number of the current consistency stack, and L is 1,2 πi,jSatisfy pi for consistency weighti,jIs not less than 0 and
s2.4, measuring and updating each sensor node i;
the updated calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,representing the number of sensor nodes in the sensor network;
step S2.5, estimating and extracting the target state at the current moment k, wherein the calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,a forward filtered estimate representing the target state at time k,the estimated error variance corresponding to the forward filtering estimated value of the target at the moment k;
step S2.6, when k < N, k ═ k +1 and return to performing step S2.1; when k is equal to N, all are output Andand
the inverse kalman filtering step performs N-step recursive inverse kalman filtering calculations (k ═ N, N-1.., 1), each step comprising:
s3.1, each sensor node i carries out one-step backward prediction on the target state by utilizing the target state and covariance estimation at the moment k to respectively obtain predicted values of the state and covariance at the moment k-1Andfrom the obtained predicted values, a predicted information matrix is calculatedAnd information vectorThe calculation formula is as follows:
step S3.2, measuring each sensor node i according to k timeCalculating new information vectorsAnd new information matrixThe calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,sensor measurement equation, partial differential, representing time kSubscript ofIs expressed as partial differentialThe value of (A) is selected,representing the measured noise variance matrix at time k for sensor i,representing the measurement value of a sensor node i at the moment k, representing that all superscripts b represent that the variable is a variable in the inverse Kalman filtering, representing the number of current EM iterations, etai(m-1)Representing a registration error estimated value of a previous EM iteration sensor node i;
step S3.3, each sensor node i pair information matrix obtained in step S3.1 and step S3.2Information vectorNew information matrixAnd new information vectorCarrying out L-step consistency iteration; the calculation formula of each step of consistency iteration of each sensor node i is as follows:
in the formula: all superscripts i represent sensor nodes i,represents a set formed by all nodes capable of directly communicating with the node i, including the node i, j represents all nodes capable of directly communicating with the node i, including the node i, L represents the step number of the current consistency stack, and L is 1,2 πi,jSatisfy pi for consistency weighti,jIs not less than 0 and
s3.4, measuring and updating each sensor node i;
the updated calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,representing the number of sensor nodes in the sensor network;
step S3.5, estimating and extracting the target state at the current moment k, wherein the calculation formula is as follows:
The target state estimate smoothing step includes:
the sensor nodes i smooth the state estimation of the target by using the structures of the step S3 and the step S4 to obtain the smoothed state estimationAnd corresponding covariance matrixThe calculation formula is as follows:
the sensor registration error estimation value solving step comprises the following steps:
each sensor node i respectively calculates the registration error estimation value eta of the sensor node ii(m)The calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,sensor measurement equation, partial differential, representing time kSubscript η ofi=ηi(m-1)Is expressed as partial differential at etai(m-1)Value of ηi(m-1)Representing a registration error estimated value of a previous EM iteration sensor node i;
the judging step comprises: when the EM iteration is not finished, returning to the step of executing the forward Kalman filtering; when the EM iteration is finished, each sensor node i outputs etai(M)As an estimate of the registration error.
2. The method for collaborative registration of a distributed sensor network according to claim 1, wherein the EM iteration employs an M-step EM iteration, the forward kalman filtering employs an N-step forward kalman filtering, and the inverse kalman filtering employs an N-step inverse kalman filtering;
wherein M, N is a positive integer.
3. A distributed sensor network collaborative registration system is characterized by comprising modules:
an initialization module: forming initial sensor registration error parameters according to prior information and forming initial target parameters according to a given initial target state and an error covariance matrix at each sensor node;
a forward Kalman filtering module: starting EM iterative computation, and carrying out forward Kalman filtering on the target parameters;
an inverse Kalman filtering module: performing reverse Kalman filtering on the target parameters subjected to forward Kalman filtering;
a target state estimation smoothing module: each sensor node smoothes the target state estimation by using the results of the forward Kalman filtering module and the reverse Kalman filtering module;
a sensor registration error estimation value solving module: each sensor node uses the smoothed target state estimation obtained by the target state estimation smoothing module to solve the respective sensor registration error estimation value;
a judging module: if the EM iteration is not finished, returning to the forward Kalman filtering module; if the EM iteration is finished, each sensor node outputs a respective sensor registration error estimation value;
the forward Kalman filtering in the forward Kalman filtering module comprises:
forward initial step filtering: performing forward consistency Kalman filtering by using an initial sensor registration error parameter and an initial target parameter in an initialization module as initial values;
forward non-initial step filtering: performing forward consistency Kalman filtering by using the sensor registration error estimation value obtained by the sensor registration error estimation value solving module and an initial target parameter in the initialization module as initial values;
the inverse Kalman filtering in the inverse Kalman filtering module comprises:
and (3) reverse initial step filtering: performing inverse consistency Kalman filtering by using the target parameters obtained after the forward Kalman filtering in the forward Kalman filtering module and the initial sensor registration error parameters in the initialization module as initial values;
reverse non-initial step filtering: performing inverse consistency Kalman filtering by using the target parameters obtained after the forward Kalman filtering in the forward Kalman filtering module and the sensor registration error parameter estimation value obtained by the sensor registration error estimation value solving module as initial values;
in the initialization module, initial target states are respectively given to all the sensor nodesAnd corresponding error covariance matrixAnd respective a priori registration errors etai(0)WhereinEach representing a respective one of the different sensor nodes,a set formed by all sensor nodes;
the forward kalman filtering module performs N-step recursive forward kalman filtering calculations (k ═ 1, 2.., N), each step of recursive calculation including:
module S2.1, each sensor node i uses the target state at the moment k-1Sum error covariance matrixEstimating the target state to be predicted in one step, and respectively estimating the target state at the moment of k and the predicted value of the error covariance matrixAndfrom the obtained predicted values, a predicted information matrix is calculatedAnd information vectorThe calculation formula is as follows:
module S2.2, measurement of each sensor node i according to the time kCalculating new information vectorsAnd new information matrixThe calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,sensor measurement equation, partial differential, representing time kSubscript ofIs expressed as partial differentialThe value of (A) is selected,representing the measured noise variance matrix at time k for sensor i,representing the measurement value of the sensor node i at time k, m representing the number of current EM iterations, ηi(m-1)Representing a registration error estimated value of a previous EM iteration sensor node i;
module S2.3, information matrix obtained by each sensor node i for module S2.1 and module S2.2Information vectorNew information matrixAnd new information vectorCarrying out L-step consistency iteration; the calculation formula of each step of consistency iteration of each sensor node i is as follows:
in the formula: all superscripts i represent sensor nodes i,represents a set formed by all nodes capable of directly communicating with the node i, including the node i, j represents all nodes capable of directly communicating with the node i, including the node i, L represents the step number of the current consistency stack, and L is 1,2 πi,jSatisfy pi for consistency weighti,jIs not less than 0 and
a module S2.4, each sensor node i carries out measurement updating;
the updated calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,representing the number of sensor nodes in the sensor network;
module S2.5, the state of the k target at the current time is estimated and extracted, and the calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,a forward filtered estimate representing the target state at time k,the estimated error variance corresponding to the forward filtering estimated value of the target at the moment k;
a module S2.6, when k is less than N, k equals k +1 and returns to execute the module S2.1; when k is equal to N, all are output Andand
the inverse kalman filter module performs N-step recursive inverse kalman filter calculations (k ═ N, N-1.., 1), each step of the recursive calculations including:
and a module S3.1, each sensor node i carries out one-step backward prediction on the target state by utilizing the target state and covariance estimation at the moment k to respectively obtain predicted values of the state and covariance at the moment k-1Andfrom the obtained predicted values, a predicted information matrix is calculatedAnd information vectorThe calculation formula is as follows:
module S3.2, measurement of each sensor node i according to the time kCalculating new information vectorsAnd new information matrixThe calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,sensor measurement equation, partial differential, representing time kSubscript ofIs expressed as partial differentialThe value of (A) is selected,representing the measured noise variance matrix at time k for sensor i,representing the measurement value of a sensor node i at the moment k, representing that all superscripts b represent that the variable is a variable in the inverse Kalman filtering, representing the number of current EM iterations, etai(m-1)Representing a registration error estimated value of a previous EM iteration sensor node i;
module S3.3, information matrix obtained by each sensor node i for module S3.1 and module S3.2Information vectorNew information matrixAnd new information vectorCarrying out L-step consistency iteration; the calculation formula of each step of consistency iteration of each sensor node i is as follows:
in the formula: all superscripts i represent sensor nodes i,represents a set formed by all nodes capable of directly communicating with the node i, including the node i, j represents all nodes capable of directly communicating with the node i, including the node i, L represents the step number of the current consistency stack, and L is 1,2 πi,jSatisfy pi for consistency weighti,jIs not less than 0 and
a module S3.4, each sensor node i carries out measurement updating;
the updated calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,representing the number of sensor nodes in the sensor network;
and a module S3.5, estimating and extracting the state of the k target at the current moment, wherein the calculation formula is as follows:
The target state estimate smoothing module includes:
each sensor node i utilizes the structures of the module S3 and the module S4 to smooth the state estimation of the target to obtain the state estimation after smoothingAnd corresponding covariance matrixThe calculation formula is as follows:
the sensor registration error estimation value solving module comprises:
each sensor node i respectively calculates the registration error estimation value eta of the sensor node ii(m)The calculation formula is as follows:
in the formula: all superscripts i represent sensor nodes i,sensor measurement equation, partial differential, representing time kSubscript η ofi=ηi(m-1)Is expressed as partial differential at etai(m-1)Value of ηi(m-1)Representing a registration error estimated value of a previous EM iteration sensor node i;
the judging module comprises: when the EM iteration is not finished, returning to execute the forward Kalman filtering module; when the EM iteration is finished, each sensor node i outputs etai(M)As an estimate of the registration error.
4. The distributed sensor network co-registration system of claim 3, wherein the EM iteration employs an M-step EM iteration, the forward Kalman filtering employs an N-step forward Kalman filtering, and the inverse Kalman filtering employs an N-step inverse Kalman filtering;
wherein M, N is a positive integer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810950034.5A CN109246637B (en) | 2018-08-20 | 2018-08-20 | Distributed sensor network collaborative registration method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810950034.5A CN109246637B (en) | 2018-08-20 | 2018-08-20 | Distributed sensor network collaborative registration method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109246637A CN109246637A (en) | 2019-01-18 |
CN109246637B true CN109246637B (en) | 2020-11-06 |
Family
ID=65071146
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810950034.5A Active CN109246637B (en) | 2018-08-20 | 2018-08-20 | Distributed sensor network collaborative registration method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109246637B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101482607A (en) * | 2009-02-19 | 2009-07-15 | 武汉理工大学 | Target tracking method and device used for wireless movable sensor network |
CN105676181A (en) * | 2016-01-15 | 2016-06-15 | 浙江大学 | Underwater moving target extended Kalman filtering tracking method based on distributed sensor energy ratios |
CN106646356A (en) * | 2016-11-23 | 2017-05-10 | 西安电子科技大学 | Nonlinear system state estimation method based on Kalman filtering positioning |
CN106685427A (en) * | 2016-12-15 | 2017-05-17 | 华南理工大学 | Sparse signal reconstruction method based on information consistency |
-
2018
- 2018-08-20 CN CN201810950034.5A patent/CN109246637B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101482607A (en) * | 2009-02-19 | 2009-07-15 | 武汉理工大学 | Target tracking method and device used for wireless movable sensor network |
CN105676181A (en) * | 2016-01-15 | 2016-06-15 | 浙江大学 | Underwater moving target extended Kalman filtering tracking method based on distributed sensor energy ratios |
CN106646356A (en) * | 2016-11-23 | 2017-05-10 | 西安电子科技大学 | Nonlinear system state estimation method based on Kalman filtering positioning |
CN106685427A (en) * | 2016-12-15 | 2017-05-17 | 华南理工大学 | Sparse signal reconstruction method based on information consistency |
Non-Patent Citations (4)
Title |
---|
A Consensus Nonlinear Filter With Measurement Uncertainty in Distributed Sensor Networks;Kai Shen;《IEEE SIGNAL PROCESSING LETTERS》;20170913;正文第2-4节 * |
Consensus and EM based Sensor Registration in Distributed Sensor Networks;Kai Shen;《2018 21st International Conference on Information Fusion》;20180713;正文第1-4节,表I * |
Distributed Variational Fiin Wireless Sensor Networks in Wireless Sensor Networksltering for Simultaneous Distributed Variational Filtering for Simultaneous;Jing Teng;《IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY》;20120630;全文 * |
Simultaneous target tracking and sensor location refinement in distributed sensor networks;Kai Shen;《ELSEVIER Signal Processing》;20180720;正文第2-4节 * |
Also Published As
Publication number | Publication date |
---|---|
CN109246637A (en) | 2019-01-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108896047B (en) | Distributed sensor network collaborative fusion and sensor position correction method | |
Han et al. | An improved IMM algorithm based on STSRCKF for maneuvering target tracking | |
Chen et al. | Distributed cubature information filtering based on weighted average consensus | |
Huang et al. | On the complexity and consistency of UKF-based SLAM | |
CN111178385A (en) | Target tracking method for robust online multi-sensor fusion | |
Petitti et al. | Consensus-based distributed estimation for target tracking in heterogeneous sensor networks | |
CN110289989A (en) | A kind of distributed state estimation method based on volume Kalman filtering algorithm | |
CN104777469A (en) | Radar node selection method based on measurement error covariance matrix norm | |
CN111798494A (en) | Maneuvering target robust tracking method under generalized correlation entropy criterion | |
CN115930949A (en) | Multi-sensor distributed cooperative detection method and system and electronic equipment | |
CN109509207B (en) | Method for seamless tracking of point target and extended target | |
CN113709662B (en) | Autonomous three-dimensional inversion positioning method based on ultra-wideband | |
CN114139109A (en) | Target tracking method, system, equipment, medium and data processing terminal | |
CN109246637B (en) | Distributed sensor network collaborative registration method and system | |
CN116819975A (en) | Multi-target geometric center estimation method based on pure angle observation | |
Yousuf | Robust output-feedback formation control design for nonholonomic mobile robot (nmrs) | |
CN107966697B (en) | Moving target tracking method based on progressive unscented Kalman | |
Wang et al. | Estimating the position and orientation of a mobile robot using neural network framework based on combined square‐root cubature Kalman filter and simultaneous localization and mapping | |
CN107886058B (en) | Noise-related two-stage volume Kalman filtering estimation method and system | |
CN115718426A (en) | Event-triggered STF fault detection method for satellite attitude control system | |
CN116908777A (en) | Multi-robot random networking collaborative navigation method based on explicit communication with tag Bernoulli | |
CN112333236B (en) | Fault-tolerant cooperative positioning method based on two-layer filtering in three-dimensional dynamic cluster network | |
Messaoudi et al. | Comparison of interactive multiple model particle filter and interactive multiple model unscented particle filter for tracking multiple manoeuvring targets in sensors array | |
CN111426322B (en) | Adaptive target tracking filtering method and system for simultaneously estimating state and input | |
CN109474892B (en) | Strong robust sensor network target tracking method based on information form |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |