Node selection method for wide-area sparse array based on improved binary particle swarm optimization
By adopting an improved binary particle swarm optimization algorithm based on pheromone in a wide-area random thinning array, the problem of node selection under fixed nodes is solved, better sidelobe optimization effect and resource savings are achieved, and radar signal processing is improved performance.
Patent Information
- Application Number
- CN202311374802.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-10-23
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2043-10-23
AI Technical Summary
In the selection of wide-area random thin array nodes with fixed number of nodes, it is difficult to meet the constraints of resource optimization and maximizing the main side lobe ratio, especially the problem of not meeting the constraints during the iteration process.
The improved binary particle swarm optimization algorithm based on pheromone is adopted, and the solution is adjusted based on the node pheromone after particle position update is used to ensure that the number of nodes is met, and a balance is achieved between global optimization and local area exploration is achieved, thereby optimizing the sidelobe effect.
In the 20*20m2 scenario, 200 nodes are randomly distributed, and the selection number is set to 100. Compared with the traditional binary particle swarm method, the sidelobe optimization result has the advantage of 1dB, achieving better resource saving and detection capabilities.
Smart Images

Figure CN117406191B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of radar signal processing, and in particular to a wide-area sparse array node selection method based on improved binary particle swarm optimization. Background Art
[0002] The wide-area random sparse array improves the radar detection power without increasing the radar aperture of a single station through the wide-area random distribution of distributed sub-nodes. At the same time, because the transmitting and receiving nodes are arranged in a decentralized manner, the transmitting and receiving angles are expanded, and they have strong anti-interference and anti-stealth capabilities. However, the wide-area random sparse array also faces the problem of resource optimization. It is crucial to select the optimal node under the constraints of the number of nodes and the maximum main-sidelobe ratio.
[0003] Currently, the node selection methods for scenarios such as wide-area random sparse arrays and sensor networks mainly include swarm intelligence algorithm-based and machine learning algorithm-based.
[0004] The Chinese invention patent with application number 202310018283.1 discloses a fast node selection method based on transfer reinforcement learning. First, the Monte Carlo tree is trained, and the result parameters are saved for real-time call, which improves the optimization selection speed and can realize the fast selection of non-fixed number of nodes.
[0005] The paper "Improved Node Selection Algorithm for Binary Particle Swarm Optimization" improves the traditional binary optimization selection method, adds a search strategy with a guidance factor in the particle velocity and position update link, and proposes a fitness function with a penalty factor, which has better optimization effect than the traditional binary particle swarm method.
[0006] Existing methods are all aimed at selecting nodes with a non-fixed number, and face the problem of not satisfying constraints during the iteration process when selecting nodes with a fixed number. Summary of the invention
[0007] The purpose of the present invention is to provide a wide-area sparse array node selection method based on improved binary particle swarm optimization in view of the problems existing in the prior art. This method improves the traditional binary particle swarm algorithm and proposes a binary particle swarm optimization algorithm based on pheromone. Under the condition of a fixed number of nodes, when the particle position is updated, the solution of each particle is adjusted based on the node pheromone to ensure that the limited number of nodes is met, and at the same time, the particles are moved closer to the optimal solution, achieving a balance between global optimization and local area exploration, and having a better sidelobe optimization effect.
[0008] In order to achieve the above object, the technical solution adopted by the present invention is as follows:
[0009] A method for selecting nodes of a wide - area sparse array based on improved binary particle swarm optimization is used to select P nodes from N already - arranged nodes to form a wide - area sparse array, where P < N, so as to save resources while meeting the detection ability. The method includes the following steps:
[0010] Step 1: Construct a particle swarm and randomly generate the initial position and velocity of each particle. The specific method is as follows:
[0011] For each particle, construct a zero - vector with a length of N, and the N positions of the vector correspond one - to - one with the N already - arranged nodes. Then, randomly permute the integers from 1 to N, and take the first P numbers as the selected node labels. Set the 0s at the corresponding positions in the zero - vector to 1 as the initial position X0 of the particle. In addition, randomly generate a vector with a length of N as the initial velocity V0 of the particle.
[0012] Step 2: Calculate the fitness function value corresponding to each particle in the particle swarm, that is, the highest sidelobe value, and find the individual extreme value P i k and the global extreme value
[0013] Step 3: Initialize the pheromone of each node. Sequentially retrieve the selected nodes in each particle, and use the fitness function value corresponding to the particle as the pheromone value of the selected node. If a node is selected by multiple particles, use the largest fitness function value among them as the pheromone value of the node.
[0014] Step 4: Update the particle velocity and position. Calculate the velocity value of each particle according to the following velocity update formula:
[0015]
[0016] In the formula, ω, c1, and c2 respectively represent the influence weights of the particle velocity by the current velocity, individual extreme value, and global extreme value. c1 and c2 are constants, ω is a dynamic inertia weight, r1 and r2 are random numbers distributed in the interval [0, 1], V i k , X i k respectively represent the particle velocity, particle position, and individual optimal position of the i - th particle in the k - th iteration, represents the global optimal position in the k - th iteration;
[0017] Then, update the individual position according to the following binary particle swarm position calculation formula:
[0018]
[0019] Step 5, determine whether the number of selected nodes after the particle position is updated meets the constraint number P. If the number of selected nodes is less than the constraint number P, select the node with the largest pheromone value from the unselected nodes to make up the number of selected nodes. If the number of selected nodes is greater than the constraint number P, delete the node with the lowest pheromone value from the selected nodes.
[0020] Step 6, calculate the fitness function value of each particle, and update the individual extreme value and the group extreme value;
[0021] Step 7, update the pheromone value of each node;
[0022] Step 8, determine whether the set number of iterations has been reached, if the iterations are completed, output the final selection result, otherwise return to step 4 to continue execution.
[0023] Furthermore, in step 2, the fitness function value corresponding to each particle in the particle swarm is calculated as follows:
[0024] Step 201, calculate when the target is at the rectangular coordinate pos goal =[0,0,r] T The array steering vector at this time, r is the distance from the target to the origin of the coordinate system, and the coordinates of the pth selected node are marked as pos node_p =[x p ,y p ,z p ] T , p=1,2,...P, calculate the distance between the selected node and the target:
[0025] d goal_node_p =||pos goal -pos node_p ||
[0026] The array steering vector is obtained as:
[0027] w=[exp(2πfd goal_node_1 / c),exp(2πfd goal_node_2 / c),...,exp(2πfd goal_node_P / c)]
[0028] Where f is the frequency, c is the speed of light, and exp represents the exponential function of the natural constant e;
[0029] Step 202, divide the search space into L search grids, and calculate the array flow pattern corresponding to each search grid, specifically in the following manner:
[0030] The polar coordinates of the lth search grid are represented by pos wg =[a l ,θ l ,r]T , convert it into rectangular coordinates as pos=[x l ,y l ,z l ] T , the distance between the node and the grid is d wg_node_p =||pos-pos node_p ||, the guiding vector of the lth grid is denoted as a(α l ,θ l )=[exp(2πfd l_node_1 / c),exp(2πfd l_node_2 / c),...,exp(2πfd l_node_P / c)], so the array flow pattern is A = [a(α1,θ1),a(α2,θ2),...,a(a L ,θ L )];
[0031] Step 203, multiply the array steering vector by the array flow pattern conjugate to obtain the array direction pattern:
[0032] E=w H A
[0033] Convert the directivity pattern E into power form:
[0034] E dB =20lg(abs(E). / max(abs(E)))
[0035] Where, E represents the amplitude data of the array pattern, E dB represents the array pattern power data calculated after amplitude normalization;
[0036] Get the array power spectrum;
[0037] Step 204, setting the value in the main lobe region to zero, finding the maximum value of the power spectrum after setting to zero, and obtaining the highest side lobe value.
[0038] Furthermore, in step 4, ω is the decreasing inertia weight, which decreases gradually as the number of iterations increases. The ω value of the kth iteration is:
[0039]
[0040] Where T is the total number of iterations.
[0041] Compared with the background technology, the present invention has the following beneficial effects:
[0042] 1. Compared with the existing traditional binary particle swarm node selection method, the present invention adds a pheromone selection method, which has a better optimization effect under the constraint of a fixed number of nodes.
[0043] 2. In 20*20m 2 200 nodes are randomly distributed in the scene, the selection number is set to 100, and the Monte Carlo experiment is performed. Compared with the traditional binary particle swarm method, the sidelobe optimization result of this method has a 1dB advantage. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 It is a flowchart of wide area random sparse array node selection in an embodiment of the present invention.
[0045] Figure 2 It is a diagram of the sidelobe optimization result based on the improved binary particle swarm in an embodiment of the present invention.
[0046] Figure 3 It is the array directional diagram after node selection in the embodiment of the present invention.
[0047] Figure 4 It is a node location distribution diagram after node selection in an embodiment of the present invention. DETAILED DESCRIPTION
[0048] The specific implementation modes of the present invention are described in detail below with reference to the accompanying drawings.
[0049] Wide-area sparse array node selection method based on improved binary particle swarm optimization, such as Figure 1 As shown, the following steps are included:
[0050] Step 1, population initialization: Initialize each particle in the population, randomly generate the initial position and speed of each particle, first construct a zero vector of length N, then randomly arrange the integers from 1 to N, take the first P numbers as the selected node numbers, set the 0 in the corresponding position of the zero vector to 1, as the initial position X0 of the particle, and randomly generate a vector of length N as the initial speed V0 of the particle, completing the particle group initialization. N represents the total number of nodes, and P represents the limited number of nodes to be selected;
[0051] Step 2: Calculate the fitness function corresponding to each particle in the population, that is, the highest sidelobe value, and find the individual extreme value P i k and group extreme value
[0052] First, calculate when the target is at the rectangular coordinate pos goal =[0,0,r] T The array guide vector at that time. The coordinates of the selected node in the particle position, that is, the p=1, 2, ..., Pth node with a value of 1, are marked as pos node_p =[x p ,y p ,z p ]T , calculate the distance between the selected node and the target as d goal_node_p =||pos goal -pos node_p ||, the array steering vector is: w = [exp(2πfd goal_node_1 / c),exp(2πfd goal_node_2 / c),...,exp(2πfd goal_node_P / c)].
[0053] Then, the search space is gridded and the array flow pattern corresponding to each search grid is calculated. The polar coordinates of the lth grid point in a total of L grids can be expressed as pos wg =[α l ,θ l ,r] T , convert it into rectangular coordinates as pos=[x l ,y l ,z l ] T , the distance between the node and the grid point is d wg_node_p =||pos-pos node_p ||, the guiding vector of the lth grid is denoted as a(α l ,θ l )=[exp(2πfd l_node_1 / c),exp(2πfd l_node_2 / c),...,exp(2πfd l_node_P / c)], the array flow pattern can be obtained as A = [a(α1,θ1),a(a2,θ2),...,a(a L ,θ L )].
[0054] Next, the array steering vector is multiplied by the array flow pattern conjugate to obtain the array pattern E = w H A. Convert the directional pattern E into power form E dB =20lg(abs(E). / max(abs(E))), and the array power spectrum is obtained.
[0055] Finally, after setting the value in the main lobe area to zero, find the maximum value of the power spectrum after setting to zero and obtain the maximum side lobe value.
[0056] Step 3, initialize the pheromone of each node: retrieve the selected node in each particle at this time in turn, and use the fitness function value corresponding to the particle at this time as the pheromone value of the selected node. If a node is selected by multiple particles, the largest fitness function value is retained as the pheromone of the node;
[0057] Step 4, particle speed and position update: follow the speed update formula Calculate the velocity value of each particle, and then calculate the position of the binary particle group according to the formula Update individual locations;
[0058] Among them, ω, c1, and c2 represent the weights of the particle velocity affected by the current velocity, individual extreme value, and population extreme value, respectively. c1 and c2 are constants, and ω is usually the dynamic inertia weight.
[0059] In this method, ω is the decreasing inertia weight, which decreases gradually with the increase of the number of iterations. According to the formula Determined by the initial inertia weight, the total number of iterations, and the current number of iterations.
[0060] Step 5, particle value constraint. Determine whether the number of selected nodes after the particle position is updated meets the fixed node number constraint. If the number of selected nodes is less than the constraint number, select the node with the largest pheromone value from the unselected nodes to make up the number of selected nodes. If the number of selected nodes is greater than the constraint number, delete some nodes with the lowest pheromone value from the selected nodes.
[0061] Step 6: Update the particle fitness value, individual extreme value and population extreme value.
[0062] Step 7: Update the node pheromone value.
[0063] Step 8: Determine whether the iteration is complete. If the iteration is complete, output the optimization result; otherwise, return to step 4.
[0064] So far, the node selection method for wide-area random sparse array with fixed number of nodes based on improved binary particle swarm algorithm has been completed.
[0065] Figure 2 The maximum sidelobe power of the optimal particle in each generation of the group during the iteration process is given, which is optimized from the initial -13.2dB to 15.2dB, with an optimization effect of 2dB. Figure 3 is the position of the 100 nodes selected after optimization, Figure 3 Draw a directional graph of the 100 nodes in Figure 4 , it can be seen from the figure that the maximum sidelobe level is about -15.2dB.
[0066] In summary, compared with the existing traditional binary particle swarm node selection method, the present invention adds a pheromone selection strategy, strikes a balance between global optimization and local area exploration, has a better optimization effect under the constraint of a fixed number of nodes, and solves the problem of wide-area random sparse array node selection under the constraint of a fixed number of nodes in the field of radar signal processing.
Claims
1. A wide-area sparse array node selection method based on improved binary particle swarm optimization, characterized in that: Used to select P nodes from N arranged nodes to form a wide-area sparse array, where P < N, so as to save resources on the premise of meeting the detection ability, including the following steps: Step 1, construct a particle swarm, and randomly generate the initial position and velocity of each particle; the specific method is: For each particle, construct a zero vector with a length of N, and the N positions of the vector correspond one by one to the N arranged nodes; then randomly permute the integers from 1 to N, take the first P numbers as the selected node labels, and set the 0 at the corresponding position in the zero vector to 1 as the initial position X0 of the particle; in addition, randomly generate a vector with a length of N as the initial velocity V0 of the particle; Step 2: Calculate the fitness function value corresponding to each particle in the particle swarm, that is, the highest sidelobe value, and find the individual extreme value P i k and group extreme value Among them, the fitness function value corresponding to each particle in the particle swarm is calculated in the following way: Step 201, calculate when the target is at the rectangular coordinate pos goal =[0,0,r] T The array steering vector at this time, r is the distance from the target to the origin of the coordinate system, and the coordinates of the pth selected node are marked as pos node_p =[x p ,y p ,z p ] T , p=1,2,...P, calculate the distance between the selected node and the target: d goal_node_p =||pos goal -pos node_p || The array steering vector is obtained as: w=[exp(2πfd goal_node_1 / c),exp(2πfd goal_node_2 / c),...,exp(2πfd goal_node_P / c)] where f is the frequency, c is the speed of light, and exp represents the exponential function of the natural constant e; Step 202, divide the search airspace into L search grids, and calculate the array manifold corresponding to each search grid. The specific method is: The polar coordinates of the lth search grid are represented by pos wg =[α l ,θ l ,r] T , convert it into rectangular coordinates as pos=[x l ,y l ,z l ] T , the distance between the node and the grid is d wg_node_p =||pos-pos node_p ||, the guiding vector of the lth grid is denoted as a(α l ,θ l )=[exp(2πfd l_node_1 / c),exp(2πfd l_node_2 / c),...,exp(2πfd l_node_P / c)], so the array flow pattern is A = [a(α1,θ1),a(α2,θ2),...,a(α L ,θ L )]; Step 203, multiply the array steering vector conjugate by the array manifold to obtain the array pattern: E=w H A Convert the pattern E to a power form: AND dB =20lg(abs(E). / max(abs(E))) Where, E represents the amplitude data of the array pattern, E dB represents the array pattern power data calculated after amplitude normalization; Obtain the array power spectrum; Step 204, set the values in the main lobe region to zero, find the maximum value of the power spectrum after setting to zero, and obtain the highest sidelobe value; Step 3, initialize the pheromone of each node: sequentially retrieve the selected nodes in each particle, and use the fitness function value corresponding to the particle as the pheromone value of the selected node. If a node is selected by multiple particles, then use the largest fitness function value among them as the pheromone value of the node; Step 4, update the particle velocity and position: calculate the velocity value of each particle according to the following velocity update formula: Where ω, c1, and c2 represent the weights of the particle velocity affected by the current velocity, individual extreme value, and population extreme value, respectively. c1 and c2 are constants. ω is the dynamic inertia weight. r1 and r2 are random numbers distributed in the interval [0,1]. V i k , They represent the particle velocity, particle position and individual optimal position of the i-th particle in the k-th iteration, represents the optimal position of the group in the kth iteration; Then, update the individual position according to the following binary particle swarm position calculation formula: Step 5, judge whether the number of selected nodes after updating the particle position meets the constraint number P. If the number of selected nodes is less than the constraint number P at this time, select the node with the largest pheromone value from the unselected nodes to make up the number of selected nodes. If the number of selected nodes is greater than the constraint number P at this time, delete the node with the lowest pheromone value from the selected nodes; Step 6, calculate the fitness function value of each particle, and update the individual extreme value and the group extreme value; Step 7, update the pheromone value of each node; Step 8, judge whether the set number of iterations is reached. If the iteration is completed, output the final selection result, otherwise return to Step 4 to continue execution.
2. The wide-area sparse array node selection method based on improved binary particle swarm optimization according to claim 1, characterized in that: In Step 4, ω is a decreasing inertia weight, which gradually decreases as the number of iterations increases. The ω value of the kth iteration is: where T is the total number of iterations.
