A distributed adaptive target tracking method with communication protection
By employing a state decomposition-based consensus algorithm and adaptive weights in the target tracking system, communication protection of target tracking results is achieved, improving the convergence speed of information interaction and the accuracy of target tracking. This solves the problems of communication protection being susceptible to interference and complex information processing in existing technologies.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING INST OF CONTROL ENG
- Filing Date
- 2024-11-05
- Publication Date
- 2026-06-23
AI Technical Summary
In the target tracking process, existing technologies suffer from communication protection mechanisms that are susceptible to interference and complex information processing, which affects the convergence accuracy of distributed filtering algorithms.
A consensus algorithm based on state decomposition is adopted. By observing some sub-states of the real state of the interaction between nodes, combined with adaptive consensus weights and the steepest gradient descent method, the communication protection and accuracy maintenance of the target tracking results are achieved.
Without affecting tracking accuracy, it improves the convergence speed of information interaction and the collaborative accuracy of target tracking, and protects the observation information from being eavesdropped.
Smart Images
Figure CN119544763B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of target tracking technology, and in particular to a distributed adaptive target tracking method with communication protection. Background Technology
[0002] Target tracking technology is an important field of scientific research today, playing a vital role and having wide applications in military, industrial, and daily life, and is also a hot topic of research for scholars. For example, in the military, when carrying out missions to intercept enemy aircraft, it is necessary to locate, navigate, and engage the target. If the target's motion status information cannot be effectively obtained, subsequent interception is impossible; the enemy often uses devices with eavesdropping capabilities to obtain data such as communication information between our observation nodes, thus causing our interception mission to fail. Therefore, it is necessary to protect the information exchanged between observation nodes to prevent our effective observation data from being obtained by the enemy.
[0003] Noise injection and cryptography are two of the most commonly studied communication protection mechanisms, but they are easily affected by the complex noise of the battlefield and the information processing is quite complex for the performance of airborne computers. Therefore, a more easily implemented and robust algorithm is needed. Summary of the Invention
[0004] The technical problem solved by this invention is to overcome the shortcomings of the prior art and provide a distributed adaptive target tracking method with communication protection, which can realize communication protection of target tracking results without affecting the convergence accuracy of the consensus algorithm in distributed filtering.
[0005] The technical solution of this invention is:
[0006] A distributed adaptive target tracking method with communication protection includes:
[0007] 1) Establish a discrete linear model of the distributed target tracking system;
[0008] 2) Using the discrete linear model of the distributed target tracking system described in step 1), obtain the measurement vector of the observed target obtained by each observation node i;
[0009] 3) For each observation node i, the prediction results of the observation value at time k+1, the prediction results of the information vector, and the prediction results of the information matrix are obtained respectively;
[0010] 4) For each observation node i, using the measurement vector of the observation target obtained by each observation node i in step 2), and based on the prediction results of the observation value at time k+1, the prediction results of the information vector, and the prediction results of the information matrix obtained in step 3), calculate the correction results of the information vector and the correction results of the information matrix at time k+1.
[0011] 5) Based on the information vector correction results and information matrix correction results obtained in step 4), determine the fusion result;
[0012] 6) Repeat step 5) L times to obtain the L iterative fusion results corresponding to time k+1. Based on the L iterative fusion results, and based on the information vector correction results and information matrix correction results obtained in step 4) at time k+1, determine the tracking result corresponding to observation node i at time k+1.
[0013] Preferably, the discrete linear model of the distributed target tracking system in step 1) is as follows:
[0014] x k =A k-1 x k-1 +ω k-1
[0015]
[0016] Where i = 1, 2, ..., N; N is a positive integer;
[0017] For the kth sampling time, x k To observe the target state vector, Let ω be the measurement vector of the observation node i. k For the process noise of the observed target, For the measurement noise at observation node i, Let be the observation matrix of observation node i;
[0018] A k-1 Let be the state transition matrix at the (k-1)th sampling time.
[0019] Preferably, the method for obtaining the prediction result of the observation at time k+1 in step 3) is as follows:
[0020]
[0021] in, This is the correction step information at time k. Predict the state value of the target at time k+1; The prediction results of the observations at time k+1 are corrected.
[0022] Preferably, the method for obtaining the information vector prediction result at time k+1 in step 3) is as follows:
[0023]
[0024] in, The information matrix of the observation target at time k+1 is used as the prediction result of the information vector at time k+1; Ω k This is the information matrix of the correction step for the observed target at time k.
[0025] Preferably, the method for obtaining the prediction result of the information matrix at time k+1 in step 3) is as follows:
[0026]
[0027] in, The information vector of the observation target at time k+1 is used as the prediction result of the information matrix at time k+1.
[0028] Preferably, the method for calculating the information vector correction result at time k+1 in step 4) is as follows:
[0029]
[0030] in, The information vector of the observation target at time k+1 is corrected, and it is used as the information vector correction result at time k+1.
[0031] i k+1 The contribution of the information vector of the observed node i at time k+1; For the k-th sampling time, the inverse matrix of the noise covariance matrix is measured.
[0032] Preferably, the method for calculating the corrected information matrix at time k+1 in step 4) is as follows:
[0033]
[0034] in, The information matrix of the observed target at the correction steps at time k+1 is used as the correction result of the information matrix at time k+1; I k+1 It contributes to the information matrix of the observation node i at time k+1.
[0035] Preferably, the method for determining the fusion result in step 5) is as follows:
[0036] Initialize the consensus algorithm for communication protection with initial values:
[0037]
[0038]
[0039] Where rand(1) is any random number with a value range of (0, 1);
[0040] The information vector and information matrix are fused according to the consensus algorithm:
[0041]
[0042] in,
[0043] Let α represent the sub-state of the information vector of observation node i at time k+1 and during the l-th iteration;
[0044] This represents the β sub-state of the information vector at time k+1 during the l-th iteration;
[0045] Let α represent the sub-state of the information matrix at time k+1 and during the l-th iteration;
[0046] This represents the β sub-state of the information matrix at time k+1 during the l-th iteration;
[0047] l = 0, 1, ..., L-1; L is the total number of iterations at each time step, with L iterations at each time step, and the value of L ranges from 20 to 50.
[0048] ω ij (l) represents the value of the i-th row and j-th column of the consensus weight matrix of the distributed target tracking system at the l-th iteration; the consensus weight matrix of the distributed target tracking system is an N-row N-column matrix, j = 1, 2, ..., N; that is, the element ω in the consensus weight matrix of the distributed target tracking system. ij (l) represents the consistency correlation weight between observation node i and observation node j in the lth iteration;
[0049] a i,αβ Let a be the correlation coefficient of the substates of observed node i. i,αβ ∈(0,1];
[0050] Update the consistency weight matrix coefficients e i (l):
[0051]
[0052] Preferably, step 6) determines the tracking result corresponding to observation node i at time k+1. The method is as follows:
[0053]
[0054] The advantages of this invention compared to the prior art are:
[0055] This invention employs a state decomposition-based consensus algorithm, which can protect the communication of target tracking results without affecting tracking accuracy, ensuring that target information is not eavesdropped. Furthermore, this invention designs an adaptive consensus weight; the consensus weight calculated using the steepest gradient descent method can effectively improve the convergence speed of information interaction in cooperative target tracking methods, achieving higher cooperative target tracking accuracy with fewer communication cycles. Attached Figure Description
[0056] Figure 1 A simulation diagram illustrating the effectiveness of communication protection;
[0057] Figure 2 This is the tracking error curve for target tracking in this invention;
[0058] Figure 3 The present invention aims to track the error covariance curves of each sensor node;
[0059] Figure 4 This is a flowchart of the method of the present invention. Detailed Implementation
[0060] This invention provides a distributed adaptive target tracking method with communication protection. By allowing different observation nodes to interact only with a subset of sub-states in the real state, the method can ensure target tracking accuracy while protecting the communication between sensors. Simulation results verify that this method improves the consistency iteration convergence speed without sacrificing tracking accuracy and can achieve higher cooperative target tracking accuracy with fewer communication cycles.
[0061] To better describe the present invention, the invention will be explained in detail below with reference to schematic diagrams and examples. For example... Figure 4 As shown, the distributed adaptive target tracking method with communication protection of the present invention includes the following steps:
[0062] Step 1: Establish a discrete linear model (target motion model and sensor observation model) for the distributed target tracking system. The distributed target tracking system includes N observation nodes, and each observation node is equipped with a sensor, such as microwave radar or lidar.
[0063] Consider the following discrete linear model of a distributed target tracking system consisting of N observation nodes:
[0064] x k =A k-1 x k-1 +ω k-1
[0065]
[0066] Where i = 1, 2, ..., N; N is a positive integer;
[0067] For the kth sampling time, x k To observe the target state vector, Let ω be the measurement vector of observation node i (i.e., the sensor's output information). k For the process noise of the observed target, The measurement noise at observation node i is represented by Gaussian white noise, where both process noise and measurement noise are considered as such, and their corresponding covariance matrices are Q. k and R k Q k Let R be the covariance matrix of the process noise. k The covariance matrix of the measurement noise; Let be the observation matrix of observation node i.
[0068] A k-1 The state transition matrix is given at the (k-1)th sampling time.
[0069] Define matrix
[0070] W k For the k-th sampling time, the process noise covariance matrix Q k The inverse matrix;
[0071] For the k-th sampling time, the measurement noise covariance matrix R k The inverse matrix.
[0072] Step 2: Obtain the measurement vector of the observed target for each observation node i. Sample and store;
[0073] Step 3: For each observation node i, execute the prediction step of the distributed filtering algorithm to obtain the prediction results of the observation value at time k+1, the prediction results of the information vector, and the prediction results of the information matrix.
[0074] State prediction:
[0075]
[0076] Predicted Observations:
[0077]
[0078] Information vector prediction:
[0079]
[0080] Information matrix prediction:
[0081]
[0082] in, This is the correction step information at time k. Predict the state value of the target at time k+1;
[0083] The prediction results of the observations at time k+1 are corrected.
[0084] Ω k =W k The information matrix of the target at the correction step at time k is the value of the target. The information matrix of the observation target at time k+1 is used as the prediction result of the information vector at time k+1.
[0085] The information vector of the observation target at time k+1 is used as the prediction result of the information matrix at time k+1.
[0086] Step 4: For each observation node i, execute the correction step of the distributed filtering algorithm. Using the measurement vector of the observed target obtained by each observation node i in Step 2, and based on the prediction results of the observation value at time k+1, the prediction results of the information vector, and the prediction results of the information matrix obtained in Step 3, calculate the correction results of the information vector and the correction results of the information matrix at time k+1.
[0087] 41) Information contribution vector:
[0088]
[0089] 42) Information vector correction:
[0090]
[0091] 43) Information Contribution Matrix:
[0092]
[0093] 44) Information matrix correction:
[0094]
[0095] Among them, i k+1 The contribution of the information vector of the observation node i at time k+1; I k+1 Contribute to the information matrix of observation node i at time k+1;
[0096] The information vector of the observation target at time k+1 is corrected, and it is used as the information vector correction result at time k+1.
[0097] The information matrix of the observation target at time k+1 is used as the correction result of the information matrix at time k+1.
[0098] Step 5: Based on the information vector correction results and information matrix correction results obtained in Step 4, determine the fusion result; the fusion result in Step 5 includes:
[0099] At time k+1, the observation node i... l The α sub-state of the information vector at the next iteration
[0100] At time k+1, the observation node i... l The β substate of the information vector at the next iteration
[0101] At time k+1, the observation node i... l The α-substate of the information matrix at the next iteration
[0102] At time k+1, the observation node i... l The β substate of the information matrix at the next iteration
[0103] The method for determining the fusion result in step 5 is as follows:
[0104] 51) Initialize the consensus algorithm for communication protection with initial values:
[0105]
[0106] Rand(1) is an arbitrary random number with a value range of (0, 1).
[0107] 52) The information vector and information matrix are fused according to the consensus algorithm:
[0108]
[0109]
[0110] in,
[0111] Let α represent the sub-state of the information vector of observation node i at time k+1 and during the l-th iteration;
[0112] This represents the β sub-state of the information vector at time k+1 during the l-th iteration;
[0113] Let α represent the sub-state of the information matrix at time k+1 and during the l-th iteration;
[0114] This represents the β sub-state of the information matrix at time k+1 during the l-th iteration;
[0115] Where l = 0, 1, ..., L-1; L is the total number of iterations at each time step, with L iterations at each time step, and the value of L ranges from 20 to 50.
[0116] ω ij (l) represents the value of the i-th row and j-th column of the consensus weight matrix of the distributed target tracking system at the l-th iteration; the consensus weight matrix of the distributed target tracking system is an N-row N-column matrix, j = 1, 2, ..., N; that is, the element ω in the consensus weight matrix of the distributed target tracking system. ij (l) represents the consistency correlation weight between observation node i and observation node j in the lth iteration;
[0117] a i,αβ Let a be the correlation coefficient of the substates of observed node i. i,αβ ∈(0,1];a i,αβ It is a constant related to the observation node and is independent of the number of iterations and the sampling time.
[0118] 53) After each consensus iteration, update the coefficients e of the consensus weight matrix. i (l):
[0119]
[0120] Step 6: Repeat Step 5 L times to obtain the L iterative fusion results corresponding to time k+1. Based on the L iterative fusion results, and based on the information vector correction results and information matrix correction results obtained in Step 4, determine the tracking result corresponding to observation node i at time k+1.
[0121] 61) Consensus algorithm iteration results:
[0122]
[0123] 62) Obtain the distributed target tracking results at time k+1:
[0124]
[0125] Where L is the total number of iterations at each time step; each time step iterates L times, and the value of L ranges from 20 to 50.
[0126] Example
[0127] This invention proposes a distributed adaptive target tracking filtering algorithm with communication protection, which can protect the communication between sensors of each observation node while ensuring the target tracking accuracy, and complete the target tracking task.
[0128] To demonstrate the effectiveness of this invention, the proposed distributed adaptive target tracking filtering algorithm with communication protection was simulated and verified. The simulation results for the communication protection effect are as follows: Figure 1 As shown, the observation states of all observation nodes converge to the expected average state. Within a certain number of iterations, the observation node states tend to be consistent. Furthermore, the sub-states of information exchange during the algorithm process are always far away from the actual observation states of each observation node, thus protecting the communication security of observation information in the consensus algorithm. Figure 1 The green curve in the diagram represents the α sub-state of the information matrix from step 5 above.
[0129] For the verification of target tracking accuracy, the estimation results are as follows: Figure 2 The figure shows a comparison of tracking errors for the observations of the observation node, the volumetric information filtering of the local single node, and the distributed consensus volumetric information filtering with communication protection. Figure 2 The DCIF curve in the image is the tracking result output in step 6 above.
[0130] Figure 3 The simulation results show a comparison of the RMSE of different filtering methods. The results demonstrate that the adaptive distributed target tracking filtering method with communication protection has higher tracking accuracy and convergence speed, illustrating the feasibility and effectiveness of the proposed method. Figure 3 The black curve in the image represents the tracking result output from step 6, as described above.
[0131] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Any person skilled in the art can make possible variations and modifications to the technical solutions of the present invention using the disclosed methods and techniques without departing from the spirit and scope of the invention. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention, without departing from the content of the technical solutions of the present invention, shall fall within the protection scope of the present invention. Where there is no conflict, the embodiments of this application and the technical features thereof can be combined with each other.
[0132] The contents not described in detail in this specification are common knowledge to those skilled in the art.
Claims
1. A distributed adaptive target tracking method with communication protection, characterized in that, include: 1) Establish a discrete linear model of the distributed target tracking system; 2) Obtain the measurement vector of the observed target using the discrete linear model of the distributed target tracking system described in step 1) the obtained measurement vector of the observed target; 3) Each observation node , respectively obtained Prediction results of time-time observations, prediction results of information vectors, and prediction results of information matrices; 4) For each observation node Using each observation node obtained in step 2) The obtained measurement vector of the observed target, and based on the data obtained in step 3). The prediction results of time-time observations, information vectors, and information matrices are used to calculate... The results of time-information vector correction and information matrix correction; 5) Based on the results obtained in step 4) The fusion result is determined by using the corrected time-information vector and the corrected information matrix. 6) Repeat step 5) L times to obtain the L-fold iterative fusion result corresponding to time k+1. Based on the L-fold iterative fusion result and the result obtained in step 4), The time-information vector correction results and information matrix correction results are used to determine... Observation nodes at all times Corresponding tracking results ; The method for determining the fusion result in step 5) is as follows: Initialize the consensus algorithm for communication protection with initial values: , , , , Where rand(1) is any random number with a value range of (0, 1); The information vector and information matrix are fused according to the consensus algorithm: in, Indicates the observation node exist At that moment, the The information vector at the next iteration Substate; Indicates the observation node exist At that moment, the The information vector at the next iteration Substate; Indicates the observation node exist At that moment, the The information matrix at the next iteration Substate; Indicates the observation node exist At that moment, the The information matrix at the next iteration Substate; L is the total number of iterations at each time step, with L iterations at each time step. The value of L ranges from 20 to 50. For the first In the nth iteration, the consensus weight matrix of the distributed target tracking system... Line number The values of the columns; the consistency weight matrix of the distributed target tracking system is an N x N matrix. That is, the elements in the consistency weight matrix of the distributed target tracking system. Indicates the observation node and observation nodes In the Consistency-related weights in the next iteration; For observation nodes The correlation coefficient of the substates, ; Update consistency weight matrix coefficients : 。 2. The distributed adaptive target tracking method with communication protection according to claim 1, characterized in that, The discrete linear model of the distributed target tracking system in step 1) is specifically as follows: in, ; It is a positive integer; Corresponding to the Each sampling time, To observe the target state vector, For observation nodes The measurement vector, For the process noise of the observed target, For observation nodes Measurement noise, For observation nodes The observation matrix; For the first The state transition matrix at each sampling time.
3. The distributed adaptive target tracking method with communication protection according to claim 2, characterized in that, The acquisition in step 3) The method for predicting the results of time-time observations is as follows: in, For the first Correction step information at each moment, For the observation target in Predict the state value at each step; In order to be in Predicted results of observations at the time-correction step.
4. The distributed adaptive target tracking method with communication protection according to claim 3, characterized in that, The acquisition in step 3) The method for predicting time-information vector results is as follows: in, For the observation target in The information matrix of the time-prediction step and as a reference Information vector prediction results at time step; For the observation target in the first The information matrix of the correction step at each time step.
5. A distributed adaptive target tracking method with communication protection according to claim 3 or 4, characterized in that, Step 3) Obtaining The method for predicting the results of the time-information matrix is as follows: in, For the observation target in The information vector of the time-prediction step is used as... The prediction result of the information matrix at each time point.
6. The distributed adaptive target tracking method with communication protection according to claim 5, characterized in that, Step 4) Calculation The method for correcting the time-information vector is as follows: in, For the observation target in The information vector of the time-correction step, and used as... The result of the information vector correction at any given time; for Observation nodes at all times Information vector contribution; For the first At each sampling time, the inverse of the noise covariance matrix is measured.
7. A distributed adaptive target tracking method with communication protection according to claim 6, characterized in that, The calculation in step 4) The method for correcting the time-information matrix is as follows: in, For the observation target in The information matrix of the correction step at each time step, and used as Correction results of the information matrix at each time step; for Observation nodes at all times Information matrix contribution.
8. A distributed adaptive target tracking method with communication protection according to claim 7, characterized in that, Step 6) determines Observation nodes at all times Corresponding tracking results The method is as follows: ， ， 。