Multi-base radar receiving station distribution and networking architecture optimization method for target tracking

Abstract

The invention discloses a target tracking-oriented multi-base radar receiving station distribution and networking architecture optimization method. The method comprises the following steps of: firstly, acquiring a hierarchical distributed Cramer-Rao lower bound considering bilateral packet loss; secondly, combining the radar node resource overhead model and the communication resource overhead model to obtain a multi-base radar resource overhead model; and then according to the hierarchical distributed Cramer-Rao lower bound, establishing a target tracking precision model with resource variables. And finally, based on the multi-base radar resource overhead model and the target tracking precision model, a multi-base radar receiving station allocation and networking architecture optimization model is established and solved, and according to a solving result, multi-base radar receiving station allocation and networking architecture optimization is completed. According to the method, the radar network resource overhead is minimized under the condition that the preset multi-target tracking precision is ensured, and the method can be well applied to the field of target tracking and information fusion.

Description

Technical Field

[0001] The present invention belongs to the field of target tracking and information fusion, and in particular relates to a method for allocating multi-base radar receiving stations and optimizing a networking architecture. Background Art

[0002] In the multi-base radar network collaborative tracking task, radar resources are always limited. Faced with the target tracking demand, the required resources may exceed the resource availability of the radar network. Therefore, the resource limitation of the radar network must be considered, and reasonable configuration can be used to reduce resource overhead and improve target tracking performance. The resources that can be allocated by the radar network mainly include various transmitting resources, mobile resources, and receiving resources. In addition, multi-base radars also have some potential resources, such as the multi-base radar receiving station networking architecture. Different receiving station networking architectures will not only affect the target tracking performance, but also the communication overhead of the radar network. In previous studies, the impact of the potential resource of the receiving station networking architecture on the radar network resource overhead and target tracking accuracy has been ignored. Summary of the invention

[0003] In view of the deficiencies of existing research, the present invention proposes a method for optimizing the allocation and networking architecture of multi-base radar receiving stations. The hierarchical distributed Cramer-Rao lower bound (PCRLB) considering bilateral packet loss is adopted as the theoretical bound of target tracking accuracy, and the optimization objective function is established based on the node overhead and communication overhead of multi-base radar receiving stations. A multi-base radar receiving station allocation and networking architecture optimization model is constructed, and a solution method is proposed for the optimization model, which minimizes the resource overhead and communication resource overhead of multi-base radar receiving station nodes (hereinafter referred to as radar nodes) while ensuring target tracking accuracy. The following steps are included:

[0004] S1. Hierarchical distributed Cramer-Rao lower bound PCRLB considering bilateral packet loss.

[0005] In the actual communication model, the propagation signal gradually attenuates as the distance increases. Therefore, the packet loss probability of the communication information between two radar nodes is proportional to the distance between them, and the packet reception rate is inversely proportional to the distance between them. The packet loss rate γ' and the packet reception rate γ are described as:

[0006]

[0007] Among them, ψ is composed of many communication parameters, such as channel parameters, signal-to-noise ratio, modulation parameters, and coding parameters. d is the Euclidean distance between radar nodes.

[0008] Considering the hierarchical distributed fusion architecture with bilateral packet loss, when the multi-base radar network is fully observable and there is no packet loss, the Fisher information matrix J of the target state at time k is k|k , that is, the inverse of PCRLB can be expressed as:

[0009]

[0010] Where M is the total number of local fusion centers, N is the total number of radar nodes under each local fusion center, m is the number of local fusion centers, n is the number of radar nodes under each local fusion center, and J is the number of local fusion centers. k|k-1 is the Fisher information matrix of the prior target state at time k, F k is the target state transfer matrix, Q k-1 Process noise matrix, R mn,k is the measurement noise matrix of node mn at time k, H mn,k is the Jacobian matrix of the measurement model of node mn at time k relative to the target state.

[0011] Fisher Information Matrix J k|k Related to the local measurement at the radar node, it depends on the packet arrival rate γ m ,ξ m , γ mn ,ξ mn Among them, γ m is the arrival rate of data packets transmitted from the fusion center to the cluster head node m, ξ m is the arrival rate of data packets transmitted from cluster head node m to the fusion center, γ mn is the arrival rate of data packets transmitted from cluster head node m to ordinary node mn, ξ mn is the arrival rate of data packets transmitted from ordinary node mn to cluster head node m. Therefore, considering the bilateral packet loss, the Fisher information matrix J' k|k It can be expressed as:

[0012]

[0013] Then the hierarchical distributed PCRLB considering bilateral packet loss can be expressed as:

[0014]

[0015] S2. Combining the radar node resource overhead model and the communication resource overhead model, a multi-base radar resource overhead model is obtained.

[0016] Considering that the transmission distance of radar nodes is limited and the communication resource consumption of radar nodes increases sharply with the increase of transmission distance, in order to ensure the effective transmission of information between radar nodes and reduce the consumption of communication resources, a hierarchical distributed fusion architecture is adopted for radar nodes that track targets. All assigned nodes are networked and their networking architecture is optimized. The core idea is to divide these nodes into several clusters, select a cluster head node from each cluster to act as a local fusion center, and then the local fusion center transmits the information of each cluster to the fusion center for fusion processing.

[0017] Radar node resource overhead model:

[0018] Define the radar node selection matrix U for tracking the target at time k k It is expressed as:

[0019]

[0020] Where N is the total number of radar nodes, and the node selection variable Indicates whether the nth radar node at time k is used to track the target. If the nth radar node is used to track the target, then If the nth radar node is not used to track the target, then

[0021] Then the radar node resource consumption Θ at time k k It is expressed as:

[0022]

[0023] Among them, 1 N is an N-dimensional column vector whose elements are all 1, and (·) represents matrix multiplication operation.

[0024] Communication resource overhead model:

[0025] Since the communication overhead is usually positively correlated with the square of the distance between radar nodes, the communication overhead matrix D between each radar node is k Build as:

[0026]

[0027] Among them, the matrix D k is a symmetric matrix, d i,j represents the distance between radar node i and radar node j, i≠j and i,j∈{1,2,…,N}, d i,FC represents the distance between radar node i and the fusion center (FC).

[0028] The nodes assigned to perform the target tracking task at time k will form a network, and its cluster head selection matrix S k and the clustering matrix G k Respectively expressed as:

[0029]

[0030] in, A 0-1 binary variable representing cluster head selection, which is used to represent whether radar node i is selected as the cluster head at time k (1 means selected as the cluster head, 0 means not selected as the cluster head); A 0-1 binary variable representing cluster member selection, used to represent whether radar node j is assigned to cluster i at time k (1 indicates that radar node j is assigned to cluster i, 0 indicates that radar node j is not assigned to cluster i).

[0031] So the total communication overhead of the radar network at time k is Φ k It is expressed as:

[0032]

[0033] Among them, (·) represents the matrix multiplication operation, Denotes the Hadamard product.

[0034] Overall resource cost model:

[0035] Combining the radar node resource overhead model and the communication resource overhead model, the overall multi-base radar resource overhead model is expressed as the sum of the two, namely:

[0036]

[0037] S3. According to the hierarchical distributed Cramer-Rao lower bound (PCRLB), a target tracking accuracy model with resource variables is established.

[0038] Combined with the hierarchical distributed PCRLB considering bilateral packet loss in S1, the hierarchical distributed PCRLB considering bilateral packet loss with resource variables can be written as:

[0039]

[0040] Among them, ξ i,k is the arrival rate of data packets transmitted from radar node i to FC at time k, is the arrival rate of the data packet transmitted by FC to radar node i at time k, γ ij,k is the arrival rate of data packets transmitted from radar node i to radar node j, γ ji,k is the arrival rate of data packets transmitted from radar node j to radar node i; R j,k is the measurement noise matrix of node j at time k, H j,k is the Jacobian matrix of the measurement model of node j at time k relative to the target state.

[0041] The elements on the diagonal represent the lower bound of the estimated variance of the target state. In order to evaluate the overall target tracking performance, the trace of PCRLB is selected as the target tracking accuracy model with resource variables, which is expressed as:

[0042]

[0043] S4. Based on the multi-base radar resource overhead model and target tracking accuracy model, a multi-base radar receiving station allocation and networking architecture optimization model is established.

[0044] Based on the multi-base radar resource cost model and target tracking accuracy model, a multi-base radar receiving station allocation and networking structure optimization model can be established. The optimization criterion is to minimize the radar network resource cost while ensuring the preset target tracking accuracy. The optimization variable is the radar node selection matrix U k , cluster head selection matrix S k and the clustering matrix G k . Assume that the preset target tracking accuracy model needs to satisfy F(U k ,S k ,G k )≤T max Among them, T max is the target tracking accuracy index to be achieved, the optimization model is expressed as:

[0045]

[0046] S5. Solve the multi-base radar receiving station allocation and networking architecture optimization model, and complete the multi-base radar receiving station allocation and networking architecture optimization based on the solution results.

[0047] According to the multi-base radar receiving station allocation and networking architecture optimization model, the optimization variable is the radar node selection matrix U k , cluster head selection matrix S k and the clustering matrix G k , we only need to solve U k , S k and G k However, G k Included in U k and S k information, so we only need to solve G k The variable to be solved is a discrete random variable in a finite space (a matrix consisting of 0-1 binary variables).

[0048] Due to the binary constraints of the optimization variables, this problem is a nonlinear and non-convex problem. The computational complexity of the exhaustive search algorithm is too large. Therefore, a method based on genetic algorithm and greedy strategy is designed to solve the problem. The steps are as follows:

[0049] Step 1: Take out G k The main diagonal elements of form the cluster head vector

[0050] Step 2: Set the maximum number of iterations and randomly generate several cluster head vectors G' k .

[0051] Step 3: Start iteration, each generated G' k Indicates that the cluster head node has been determined. A greedy strategy is used to assign the remaining nodes to different clusters (each node assignment minimizes the communication overhead until F(U k ,S k ,G k )≤T max or node exhaustion), the optimal

[0052] Step 4: When the current number of iterations is less than the maximum number of iterations, a new G' is generated through selection, crossover, and mutation operations. k , return to Step 3 to continue execution and update When the current number of iterations reaches the maximum number of iterations, the output

[0053] Step 5: Pass get and

[0054] Technical Effects of the Invention

[0055] Compared with the existing research, the present invention has the following excellent effects:

[0056] 1. When establishing the target tracking accuracy model, the PCRLB used in the present invention comprehensively considers the packet loss problem in the actual communication process, and derives a hierarchical distributed PCRLB considering bilateral packet loss, which is more accurate than the PCRLB under ideal conditions;

[0057] 2. The present invention takes into account the impact of networking architecture on target tracking performance and radar network communication resource overhead, and based on this, establishes a multi-base radar receiving station allocation and networking architecture optimization model, and proposes a solution method based on genetic algorithm and greedy strategy to minimize the radar network resource overhead while ensuring the preset multi-target tracking accuracy.

[0058] 3. Compared with other optimization methods, the method proposed in the present invention is closer to the actual situation and makes full use of the potential resource of the networking architecture. While ensuring the target tracking performance of the radar network, it can minimize the resource overhead of the radar network and can be well applied to the field of target tracking and information fusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 Diagram of hierarchical distributed fusion architecture considering bilateral packet loss;

[0060] Figure 2 This is a schematic diagram of the radar network architecture;

[0061] Figure 3 is the simulation scene graph;

[0062] Figure 4 This is a comparison chart of the two methods of PCRLB in Experiment 1;

[0063] Figure 5 This is a comparison chart of resource consumption of the two methods in Experiment 1;

[0064] Figure 6 This is a comparison chart of the two methods of PCRLB in Experiment 2;

[0065] Figure 7 This is a comparison chart of resource consumption of the two methods in Experiment 2. DETAILED DESCRIPTION

[0066] The present invention will be further explained below in conjunction with simulation experiments and accompanying drawings;

[0067] The present invention proposes a method for allocating multi-base radar receiving stations and optimizing networking architecture for target tracking, comprising the following steps:

[0068] S1. Obtain the hierarchical distributed Cramer-Rao lower bound PCRLB considering bilateral packet loss. The hierarchical distributed fusion architecture considering bilateral packet loss is shown in the figure below: Figure 1 As shown in FIG, a hierarchical distributed fusion architecture usually includes a fusion center (FC), several local fusion centers (cluster heads, CH), and each CH includes several common radar nodes.

[0069] S2. Combining the radar node resource overhead model and the communication resource overhead model, we get the multi-base radar resource overhead model. The schematic diagram of the radar network architecture is as follows: Figure 2 shown.

[0070] S3. According to the hierarchical distributed Cramer-Rao lower bound (PCRLB), a target tracking accuracy model with resource variables is established.

[0071] S4. Based on the multi-base radar resource overhead model and target tracking accuracy model, a multi-base radar receiving station allocation and networking architecture optimization model is established.

[0072] S5. Solve the multi-base radar receiving station allocation and networking architecture optimization model, and complete the multi-base radar receiving station allocation and networking architecture optimization based on the solution results.

[0073] In order to verify the effectiveness and superiority of the method for optimizing the multi-base radar receiving station allocation and networking architecture for target tracking proposed in the present invention, this embodiment conducts a simulation experiment for optimizing the multi-base radar receiving station allocation and networking architecture. Two sets of simulation experiments are designed:

[0074] 1. The conventional method of optimizing the allocation of multi-base radar receiving stations (the selected radar receiving stations are not networked, that is, each radar receiving station is a cluster head and is directly connected to the FC) is compared with the optimization method proposed in the present invention.

[0075] 2. Use the same radar receiving station as the optimization method proposed in the present invention, but do not optimize the network architecture, and compare the results with the optimization method proposed in the present invention.

[0076] The scenarios and parameter settings of the two simulation experiments designed in this embodiment are as follows:

[0077] The simulation scene is shown in the figure Figure 3 As shown in the figure, the radar network in the scene is a single-transmitter multi-receiver radar network, with 1 transmitting station and 12 receiving stations. Each receiving station has the ability to become a local fusion center (cluster head), and a fusion center is set. The scene contains a target, which moves in a uniform straight line. Set the target tracking accuracy threshold T max =20m, the specific simulation scenario parameters are shown in Table 1:

[0078] Table 1 Simulation scenario parameters

[0079]

[0080]

[0081] The ATSC signal is used as the radar transmission signal model, and the reference angle measurement standard deviation in the simulation experiment is set to σ θ0 =3°, set the signal constant R0 in the radar signal-to-noise ratio to 80000, and the false alarm rate P FA It is set to 0.01. The radar signal parameters in the simulation experiment are shown in Table 2.

[0082] Table 2 Radar signal parameters

[0083]

[0084] The experimental results of simulation experiment 1 are as follows Figure 4 and Figure 5 shown. Figure 4 This is a comparison of the change curves of PCRLB of the two methods during the tracking process. It can be concluded from the figure that both methods can ensure that the PCRLB of the target reaches the preset target tracking accuracy threshold. Figure 5The following is a comparison of the change curves of radar network resource consumption during the tracking process of the two methods. The radar network resource consumption is not counted in the first frame and is set to 0. In the initial frames of the simulation cycle, even if all radar receiving stations are used up, the preset target tracking accuracy cannot be achieved, so the resource consumption reaches the maximum; however, the method proposed in this article takes into account the impact of the network architecture on resource consumption, so the optimized resource overhead is smaller. It can be concluded from the figure that the method for optimizing the allocation and networking architecture of multi-base radar receiving stations proposed in the present invention can reduce the resource consumption of the radar network while ensuring the preset target tracking accuracy.

[0085] The experimental results of simulation experiment 2 are as follows Figure 6 and Figure 7 shown. Figure 6 The following is a comparison of the change curves of PCRLB of the two methods during the tracking process. It can be concluded from the figure that the method proposed in the present invention can ensure that the PCRLB of the target reaches the preset target tracking accuracy threshold, but when the radar receiving station with the same optimization method as the optimization method proposed in the present invention is used without optimizing the network architecture, the preset target tracking accuracy cannot be achieved. This is because the influence of the network architecture on the target tracking accuracy is ignored, and the preset target tracking accuracy cannot be achieved without optimizing the network architecture. Figure 7 The following is a comparison of the change curves of radar network resource consumption during the tracking process of the two methods. The radar network resource consumption is not counted in the first frame and is set to 0. It can be concluded from the figure that the method for optimizing the allocation and networking architecture of multi-base radar receiving stations proposed in the present invention takes into account the influence of the networking architecture on resource consumption, optimizes the networking architecture, and can reduce the resource consumption of the radar network while ensuring the preset target tracking accuracy.

[0086] The above two groups of simulation experiments verified and analyzed the impact of the networking architecture on target tracking accuracy and radar network resource overhead. Compared with other optimization methods, the method proposed in the present invention is closer to the actual situation and makes full use of the potential resource of the networking architecture. While ensuring the target tracking performance of the radar network, it can minimize the resource overhead of the radar network. The multi-base radar receiving station allocation and networking architecture optimization method proposed in the present invention can be well applied to the field of target tracking and information fusion.

Claims

1. A method for allocating multi-base radar receiving stations and optimizing network architecture for target tracking, characterized in that: The following steps are involved: S1, obtain the hierarchical distributed Cramer-Rao lower bound PCRLB considering bilateral packet loss; S2. Combining the radar node resource overhead model and the communication resource overhead model, a multi-base radar resource overhead model is obtained; S3. According to the hierarchical distributed Cramer-Rao lower bound PCRLB, a target tracking accuracy model with resource variables is established; S4. Based on the multi-base radar resource overhead model and target tracking accuracy model, a multi-base radar receiving station allocation and networking architecture optimization model is established; S5. Solve the multi-base radar receiving station allocation and networking architecture optimization model, and complete the multi-base radar receiving station allocation and networking architecture optimization based on the solution results.

2. The method for allocating and optimizing the network architecture of multi-base radar receiving stations for target tracking according to claim 1, characterized in that: The specific implementation process of step S1 is as follows: The packet loss rate γ' and the packet reception rate γ are described as: Among them, ψ consists of communication parameters, d is the Euclidean distance between radar nodes; Considering the hierarchical distributed fusion architecture with bilateral packet loss, when the multi-base radar network is fully observable and there is no packet loss, the Fisher information matrix J of the target state at time k is k|k , that is, the inverse of PCRLB is: Where M is the total number of local fusion centers, N is the total number of radar nodes under each local fusion center, m is the number of local fusion centers, n is the number of radar nodes under each local fusion center, and J is the number of local fusion centers. k|k-1 is the Fisher information matrix of the prior target state at time k, F k is the target state transfer matrix, Q k-1 Process noise matrix, R mn,k is the measurement noise matrix of node mn at time k, H mn,k is the Jacobian matrix of the measurement model of node mn at time k relative to the target state; Fisher Information Matrix J k|k Depends on the packet arrival rate γ m , m , γ mn and mn , where γ m is the arrival rate of data packets transmitted from the fusion center to the cluster head node m, ξ m is the arrival rate of data packets transmitted from cluster head node m to the fusion center, γ mn is the arrival rate of data packets transmitted from cluster head node m to ordinary node mn, ξ mn is the arrival rate of data packets transmitted from ordinary node mn to cluster head node m, so the Fisher information matrix J' considering bilateral packet loss k|k It is expressed as: The hierarchical distributed PCRLB considering bilateral packet loss is expressed as:

3. The method for allocating and optimizing the network architecture of multi-base radar receiving stations for target tracking according to claim 2, characterized in that: The specific implementation process of step S2 is as follows: A hierarchical distributed fusion architecture is used for radar nodes that track targets. All assigned nodes are networked and their networking architecture is optimized. Radar nodes are divided into several clusters. Each cluster selects a cluster head node to serve as a local fusion center. The local fusion center then transmits the information of each cluster to the fusion center for fusion processing. Define the radar node selection matrix U for tracking the target at time k k It is expressed as: Where N is the total number of radar nodes, and the node selection variable Indicates whether the nth radar node at time k is used to track the target; if the nth radar node is used to track the target, then If the nth radar node is not used to track the target, then Then the radar node resource consumption Θ at time k k It is expressed as: Among them, 1 N is an N-dimensional column vector whose elements are all 1, (·) represents matrix multiplication operation; The communication cost matrix D between each radar node k Build as: Among them, the matrix D k is a symmetric matrix, d i,j represents the distance between radar node i and radar node j, i≠j and i,j∈{1,2,…,N}, d i,FC represents the distance between radar node i and fusion center FC; The nodes assigned to perform the target tracking task at time k will form a network, and its cluster head selection matrix S k and the clustering matrix G k Respectively expressed as: in, A 0-1 binary variable representing cluster head selection, representing whether radar node i is selected as the cluster head at time k, 1 means it is selected as the cluster head, and 0 means it is not selected as the cluster head; A 0-1 binary variable representing cluster member selection, representing whether radar node j is assigned to cluster i at time k, 1 means radar node j is assigned to cluster i, 0 means radar node j is not assigned to cluster i; So the total communication overhead of the radar network at time k is Φ k It is expressed as: Among them, (·) represents the matrix multiplication operation, represents the Hadamard product; Combining the radar node resource overhead model and the communication resource overhead model, the overall multi-base radar resource overhead model is expressed as the sum of the two, namely:

4. The method for allocating and optimizing the network architecture of multi-base radar receiving stations for target tracking according to claim 3 is characterized in that: The specific process of establishing the target tracking accuracy model with resource variables is as follows: Combined with the hierarchical distributed PCRLB considering bilateral packet loss, the hierarchical distributed PCRLB considering bilateral packet loss with resource variables is written as: Among them, ξ i,k is the arrival rate of data packets transmitted from radar node i to FC at time k, is the arrival rate of the data packet transmitted by FC to radar node i at time k, γ ij,k is the arrival rate of data packets transmitted from radar node i to radar node j, γ ji,k is the arrival rate of data packets transmitted from radar node j to radar node i; R j,k is the measurement noise matrix of node j at time k, H j,k is the Jacobian matrix of the measurement model of node j at time k relative to the target state; The elements on the diagonal represent the lower bound of the estimated variance of the target state. The trace of PCRLB is chosen as the target tracking accuracy model with resource variables, which is expressed as:

5. The method for allocating and optimizing the network architecture of multi-base radar receiving stations for target tracking according to claim 4, characterized in that: The establishment of a multi-base radar receiving station allocation and networking architecture optimization model is specifically as follows: Based on the multi-base radar resource overhead model and target tracking accuracy model, a multi-base radar receiving station allocation and networking structure optimization model is established. The optimization criterion is to minimize the radar network resource overhead while ensuring the preset target tracking accuracy. The optimization variable is the radar node selection matrix U k , cluster head selection matrix S k and the clustering matrix G k ; The preset target tracking accuracy model satisfies F(U k ,S k ,G k )≤T max , where T max To achieve the target tracking accuracy index, the optimization model is expressed as:

6. The method for allocating and optimizing the network architecture of multi-base radar receiving stations for target tracking according to claim 5, characterized in that: The specific process of solving the multi-base radar receiving station allocation and networking architecture optimization model is as follows: According to the multi-base radar receiving station allocation and networking architecture optimization model, the optimization variable is the radar node selection matrix U k , cluster head selection matrix S k and the clustering matrix G k , G k Included in U k and S k So we need to solve G k That is, the variable to be solved is a discrete random variable in a finite space, that is, a matrix composed of 0-1 binary variables; A method based on genetic algorithm and greedy strategy is designed to solve the problem. The specific process is as follows: S5.1: Take G k The main diagonal elements of constitute the cluster head vector S5.2: Set the maximum number of iterations and randomly generate several cluster head vectors G' k ; S5.3: Start iteration, each generated G' k Indicates that the cluster head node has been determined; a greedy strategy is used to assign the remaining nodes to different clusters, and each node assignment minimizes the communication overhead until F(U k ,S k ,G k )≤T max Or node exhaustion, get the optimal S5.4: When the current number of iterations is less than the maximum number of iterations, a new G' is generated through selection, crossover, and mutation operations. k , return to S5.3 to continue execution and update When the current number of iterations reaches the maximum number of iterations, the output S5.5: Pass get and

Images