A new method for node position prediction based on hybrid particle swarm optimization and grey wolf optimization

Abstract

The application discloses a new node position prediction method based on hybrid particle swarm optimization and grey wolf optimization, comprising the following steps: determining a room or an area where a target wireless sensor node is located; measuring received signal strength at different positions away from the target wireless sensor node; filtering the signal strength obtained by using a Gaussian filter; fitting the obtained signal strength with distance, constructing a signal strength-distance relationship model and evaluating the model to select a model with the best fitting degree; dividing the room or the area to be positioned by using a dichotomy method, determining and narrowing a reference node selection area; limiting the position of a reference node deployment; selecting a reference node with strong and stable signal strength by using a weighted distance strategy; and positioning and optimizing the position of the target wireless sensor node by using a hybrid PSOGWO algorithm based on the obtained data. The application improves the positioning accuracy of the target node and also improves the positioning efficiency.

Description

Technical Field

[0001] This invention belongs to the field of wireless sensor node location prediction technology, and in particular relates to a new method for node location prediction based on hybrid particle swarm optimization and gray wolf optimization. Background Technology

[0002] Wireless sensor nodes are widely used in smart homes, environmental monitoring, and healthcare due to their small size, ease of packaging, low power consumption, and low cost. While these devices bring convenience to people's daily lives and work, they may also pose some security risks. For example, the presence of unauthorized wireless sensor nodes in public places such as conference rooms, offices, and hotels could potentially lead to the leakage of internal information or secrets of government agencies and enterprises, and even the exposure of personal privacy. Therefore, how to quickly and accurately locate such wireless sensor nodes has become an urgent problem to be solved.

[0003] Most wireless sensor nodes have built-in wireless communication modules, so we can locate them by acquiring their signal strength. Currently, the Received Signal Strength (RSS)-based positioning method is widely used due to its low cost, low power consumption, and low implementation complexity. This method is easy to use, but its positioning accuracy is easily affected by environmental factors such as multipath effects, non-line-of-sight, and shadow fading. Therefore, some researchers have made improvements. Kang et al. proposed an RSS self-calibration protocol to improve positioning accuracy. This protocol can adjust calibration parameters based on environmental changes over time, thereby reducing the ranging error of the RSS method. Nguyen et al. used deep learning to refine the collected raw RSS data and used machine learning to extract the inherent spatial network geometry hidden in the RSS dataset. These correlations were used to reduce the amplitude of RSS fluctuations, thus improving positioning accuracy. In addition, some other positioning methods can be extended from RSS-based methods. Zhao et al. and Tang et al. evaluated the distance between reference nodes and target nodes based on the acquired RSS, identifying the centroid of the overlapping area of ​​all reference nodes within the target node's communication radius as the target node's location, and thus proposed a centroid positioning method. Cho et al. and Luo et al. proposed a location fingerprint-based positioning method. This method requires establishing a location fingerprint database for a given indoor environment beforehand, and then matching the RSS measured by the user in real time with the location fingerprints in the database, selecting the location with the highest matching degree as the estimated location. Han et al. and Sun et al. proposed a method based on received signal strength approximation. This method uses the gradient value of RSS, measures at multiple points, compares the received signal strength values, analyzes the trend of RSS changes, and thus estimates the location of the target node.

[0004] However, existing RSS-based methods still suffer from problems such as unreliable reference node location selection and low target node positioning accuracy. Summary of the Invention

[0005] This invention addresses the problem that existing positioning methods based on received signal strength are susceptible to environmental factors and the location of reference nodes, resulting in low positioning accuracy. It proposes a novel node location prediction method based on hybrid particle swarm optimization (PSOGWO) and grey wolf optimization (PSOGWO) algorithms. This method first constructs a signal strength-distance relationship model for the indoor environment where the target node is located. Second, it uses a bisection method to narrow down the target node's location area and employs a weighted distance strategy to select reference node locations with strong and stable signal strength. Finally, it uses the intersection area of ​​the reference nodes as the target node's location area and then uses PSOGWO and PSOGWO algorithms to locate and optimize the target node's position. This method not only improves the reference node location selection strategy but also introduces a hybrid PSOGWO algorithm to achieve high-precision and high-efficiency positioning.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A novel node location prediction method based on hybrid particle swarm optimization and gray wolf optimization includes:

[0008] Step 1: Determine the room or area where the target wireless sensor node is located, and measure the received signal strength at different locations at a distance from the target wireless sensor node;

[0009] Step 2: Use a Gaussian filter to filter the signal strength obtained in Step 1;

[0010] Step 3: Fit the acquired signal strength to the distance to construct a model of the relationship between signal strength and distance;

[0011] Step 4: Evaluate the signal strength vs. distance relationship model constructed in Step 3 and select the model with the best fit;

[0012] Step 5: Use the binary search method to divide the room or area to be located, and determine and narrow down the selection area of ​​reference nodes;

[0013] Step 6: Limit the location of the reference node deployment;

[0014] Step 7: Use a weighted distance strategy to select a reference node with a strong and stable signal;

[0015] Step 8: Based on the data obtained above and the intensity-distance relationship model with the best fit selected, use the hybrid PSOGWO algorithm to locate and optimize the position of the target wireless sensor node.

[0016] Further, step 3 includes:

[0017] The signal strength and distance relationship models were constructed using Logistic, Poly4, and Log3P1 methods, respectively.

[0018] Further, step 5 includes:

[0019] After determining the room or area where the target wireless sensor node is located, the target area is divided into four regions using a binary search method. Sampling is performed at the center of each of the four regions. If signal strength is detected at multiple points, the regions are compared, and the region with the strongest signal strength is selected as the next sampling region. If no signal strength is detected, the four regions are divided into two again, and the center points of each region are sampled after the second division. This process is iterated until the signal strength of the target node is detected, and then the next sampling region is selected.

[0020] Furthermore, step 5 also includes:

[0021] The ring formed by the fluctuation range of the signal strength collected by the reference node is used as the approximate area where the target node is located.

[0022] Further, step 6 includes:

[0023] When there are 3 or more reference nodes, it is guaranteed that at least 3 reference nodes are not collinear.

[0024] Further, step 7 includes:

[0025] Record the RSS variance obtained by the reference node. Based on the recorded variance, assign different weights to different reference nodes. The larger the variance, the more unstable the RSS obtained by the reference node is, and the smaller the weight should be assigned.

[0026] Further, step 8 includes:

[0027] Step 8.1: Initialize the gray wolf population N within the target area, where each gray wolf represents a potential location of the target wireless sensor node;

[0028] Step 8.2: Calculate the fitness of individual gray wolves, and save the top 3 gray wolves with the best fitness (a, β, δ). The fitness value calculation formula is as follows:

[0029]

[0030] In the formula:

[0031]

[0032]

[0033] Among them RMS iThis represents the fitness of gray wolf i. The smaller the fitness, the closer the gray wolf is to the actual position of the target node; (x t y t ) and (x p y p ) represent the position of the reference node and the position of each gray wolf, respectively; d t，p It is the distance from the reference node to each gray wolf; d t，ap is the distance from the reference node to the target node; n is the number of reference nodes;

[0034] Step 8.3: Update the current position of the gray wolf;

[0035] Step 8.4: Update a, A and C, where A and C are cooperative coefficient vectors, A = 2a × r1, C = 2r2, a decreases linearly from 2 to 1 during the entire iteration, and r1 and r2 are random vectors in [0, 1].

[0036] Step 8.5: Calculate the fitness of all gray wolves;

[0037] Step 8.6: Update the fitness and position of α, β, and δ;

[0038] Step 8.7: Determine whether the maximum number of iterations or the set fitness threshold has been reached. If so, exit the loop and output the position of the target node; otherwise, go to step 3 to continue execution.

[0039] Furthermore, in step 8.2, if the fitness values ​​of the gray wolves are equal, the following conditions are used to exclude them:

[0040] d 1，i =d 1，j d 2，i =d 2，j , ..., d n-1，i =d n-1，j

[0041] Where d 1，i and d 1，j These represent the distances from gray wolf i and gray wolf j to the first reference node, respectively.

[0042] Furthermore, in step 8.6, the current position of the gray wolf is updated according to the following formula:

[0043]

[0044]

[0045] In the formula:

[0046] D α =|C1×L α -L|,D β=|C2×L β -L|,D δ =|C3×L δ -L|

[0047] L1=|L α -A1×D α |,L2=|L β -A2×D β |, L3 = |L δ -A3×D δ |

[0048] Where w is the inertia factor; c1, c2 and c3 are acceleration constants, representing the learning factors of the three optimal gray wolves respectively; r1, r2 and r3 are random numbers in the interval [0, 1], which can increase the randomness of the search; This represents the velocity vector of particle i in the j-th dimension during the t-th iteration; L represents the position vector of particle i in the j-th dimension during the t-th iteration; α L β L δ L1, L2, and L3 represent the position vectors of α, β, and δ in the current population, respectively; L1, L2, and L3 represent the step size and direction of the candidate gray wolves moving towards the three optimal gray wolves; L represents the position vector of the prey; D α D β D δ α, β, and δ represent the distances between the current candidate gray wolf and α, β, and δ, respectively; A1, A2, A3, C1, C2, and C3 are the cooperative coefficient vectors.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] 1. This invention selects reference node positions using a Gaussian filtering model and a weighted distance strategy, thereby reducing the impact of weak signal strength and unstable reference nodes on the positioning results and effectively improving positioning accuracy.

[0051] 2. Through multiple measurements, this invention has found that the received signal strength value measured at the same location will fluctuate within a certain range. Compared with the circle used in traditional methods, this invention proposes using a ring as the area where the target node appears, which effectively reduces the positioning area of ​​the target node, thereby improving positioning accuracy and the stability of positioning results.

[0052] 3. This is the first time that hybrid particle swarm optimization and gray wolf optimization algorithms have been applied to the field of wireless sensor node localization. This algorithm not only improves the localization accuracy of target nodes, but also enhances localization efficiency.

[0053] 4. Experimental results show that the method of the present invention can maintain high stability in positioning results, with 90% of the positioning errors being less than 0.9012m; compared with existing methods PSO, GWO and EWCL, its average positioning error is also reduced by 28.2%-49.0%. Attached Figure Description

[0054] Figure 1 This is a schematic diagram illustrating the principle framework of a novel node location prediction method based on hybrid particle swarm optimization and gray wolf optimization according to an embodiment of the present invention.

[0055] Figure 2 This is a graph showing the relationship between distance and signal strength in an embodiment of the present invention;

[0056] Figure 3 This is the RSS processing procedure in an embodiment of the present invention;

[0057] Figure 4 This is a signal strength and distance relationship model for an embodiment of the present invention;

[0058] Figure 5 The sampling area is reduced using the binary method in this embodiment of the invention;

[0059] Figure 6 This is a schematic diagram of the area where the target node is located in an embodiment of the present invention;

[0060] Figure 7 This is a schematic diagram showing collinear reference nodes in an embodiment of the present invention;

[0061] Figure 8 This is a schematic diagram showing the deployment of reference node locations in an embodiment of the present invention;

[0062] Figure 9 This is a schematic diagram of the signal strength collected in an embodiment of the present invention;

[0063] Figure 10 This is a schematic diagram illustrating the selection of reference nodes for weighted distance in an embodiment of the present invention;

[0064] Figure 11 This is a schematic diagram of target node positioning in an embodiment of the present invention;

[0065] Figure 12 This is a schematic diagram of the PSO-GWO optimized target node localization in an embodiment of the present invention;

[0066] Figure 13 This illustrates the impact of the positioning area area on positioning accuracy in an embodiment of the present invention.

[0067] Figure 14 This illustrates the impact of the number of reference nodes on positioning accuracy in an embodiment of the present invention.

[0068] Figure 15This illustrates the impact of the number of iterations on positioning accuracy in an embodiment of the present invention.

[0069] Figure 16 This represents the standard deviation of the embodiments of the present invention;

[0070] Figure 17 This represents the cumulative probability distribution in an embodiment of the present invention. Detailed Implementation

[0071] For ease of understanding, the following explanations are provided for some of the terms used in the specific embodiments of this invention:

[0072] ① Particle Swarm Optimization Algorithm

[0073] Particle Swarm Optimization (PSO) was first proposed in 1995. It's a swarm intelligence optimization algorithm inspired by the foraging behavior of birds. In PSO, each solution to the optimization problem represents a bird in the search space, which we call a "particle." All particles update their velocity and position by tracking two "extremes." One is the optimal solution found by the particle itself, called the individual extreme; the other is the optimal solution found by the entire population in the current iteration, called the global extreme. The formulas for updating the particle's velocity and position during iteration are as follows:

[0074]

[0075]

[0076] Here, w is the inertia factor, and adjusting the value of w can adjust the global and local optimization performance. c1 and c2 are acceleration constants, where c1 is the individual learning factor for each particle and c2 is the social learning factor for each particle. r1 and r2 are random numbers in the interval [0, 1], which can increase the randomness of the search. This represents the historical optimal position of particle i in the j-th dimension during the t-th iteration. This represents the historical best position of the group in the j-th dimension during the t-th iteration. This represents the velocity vector of particle i in the j-th dimension during the t-th iteration. Let represent the position vector of particle i in the j-th dimension during the t-th iteration.

[0077] ② Grey Wolf Optimization Algorithm

[0078] The Grey Wolf Optimization (GWO) algorithm is a swarm intelligence optimization search method developed by Seyedali et al., inspired by the hunting behavior of grey wolves. GWO has attracted widespread attention from scholars due to its strong convergence, few parameters, and low implementation complexity, and has been successfully applied to various fields. The GWO optimization process includes five stages: grey wolf social hierarchy division, surrounding prey, searching for prey, attacking prey, and hunting (optimization). The specific details are as follows:

[0079] Phase 1: Social Hierarchy of Gray Wolves. The fitness of each individual in the population is calculated, and the three gray wolves with the best fitness are labeled α, β, and δ, respectively. The remaining gray wolves are labeled ω. The optimization process is mainly guided by the three optimal solutions in each iteration of the population.

[0080] Phase 2: Encircling the Prey. When gray wolves search for prey, they gradually approach and surround it. The mathematical model for this behavior is as follows:

[0081] D = |C × L p (t)-L(t)|,C=2r2 (3)

[0082] L(t+1)=L p (t)-A×D,A=2a×r1 (4)

[0083] Where t is the current iteration number, A and C are the coordination coefficient vectors, and L... p Let L(t) represent the position vector of the prey, and let a(t) be the current position vector of the gray wolf. During the entire iteration, a decreases linearly from 2 to 1. r1 and r2 are random vectors in [0, 1].

[0084] Phase 3: Prey Search. In each iteration, the three gray wolves α, β, and δ possess a strong ability to identify the locations of potential prey, and update the positions of other search agents (including ω) based on their location information. The mathematical model for this behavior can be represented as follows:

[0085] D α =|C1×L α -L|,D β =|C2×L β -L|,D δ =|C3×L δ -L| (5)

[0086] L1=|L α -A1×D a |,L2=|L β -A2×D β |, L3 = |L δ -A3×D δ| (6)

[0087]

[0088] Where L α L β L δ Let α, β, and δ represent the position vectors of the current population, respectively, and L represent the position vector of the prey. α D β D δ L1, L2, and L3 represent the distances between the current candidate gray wolf and the three optimal gray wolves, respectively. L1, L2, and L3 represent the step size and direction of the candidate gray wolf moving towards the three optimal gray wolves in the wolf pack.

[0089] Phase 4: Attacking the Prey. As the value of 'a' linearly decreases from 2 to 0, the corresponding value of 'A' also changes within the range. When |A| > 1, the gray wolves disperse as much as possible across different areas and search for prey. When |A| < 1, the gray wolves will concentrate their search and begin attacking their prey.

[0090] Phase 5: Hunting. This phase involves iteratively obtaining the optimal solution. The coordination coefficient vector C is a random value during the iteration process. This coefficient helps the algorithm escape local stagnation, and it is particularly important in the later stages of the iteration.

[0091] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments:

[0092] Although PSO and GWO have been widely used in many fields, some problems still exist in the optimization process. PSO has strong search capabilities but is prone to getting trapped in local optima; GWO has strong ability to expand the solution space and is less prone to getting trapped in local optima, but the lack of information sharing among individual wolves leads to poor search capabilities. The hybrid PSOGWO algorithm can maximize the advantages of PSO and GWO while minimizing their disadvantages. Therefore, this invention proposes to apply it to target node localization. To increase information sharing among individual wolves, when GWO calculates the final position, it does not simply use the average position of the three best wolves as the target position, but instead introduces the speed formula of the PSO algorithm to improve the search capability of GWO. The hybrid aspects of this model are as follows:

[0093]

[0094]

[0095] Where c1, c2 and c3 are acceleration constants, representing the learning factors of the three optimal gray wolves, and r1, r2 and r3 are random numbers in the interval [0, 1], which can increase the randomness of the search.

[0096] The pseudocode for PSOGWO is shown in Algorithm 1.

[0097]

[0098] The implementation process of this invention, a novel node location prediction method based on hybrid particle swarm optimization and gray wolf optimization (ILP-PSOGWO), can be divided into the following three stages: construction and analysis of the signal strength and distance relationship model, selection of reference node locations, and target node localization. The principle framework of this method is as follows: Figure 1 As shown, the detailed process is described below:

[0099] Step 1: Data Acquisition. Determine the room or area where the target wireless sensor node is located, and begin collecting signal strength data at different locations at different distances from the target wireless sensor node to provide data support for building a signal attenuation model;

[0100] Step 2: Data Processing. To obtain more reliable data, this embodiment uses Gaussian filtering to filter the signal strength obtained in Step 1;

[0101] Step 3: Fit the data. Fit the acquired signal intensity to the distance to construct a model of the relationship between signal intensity and distance;

[0102] Step 4: Evaluate the model. Use model evaluation criteria to evaluate the signal strength vs. distance relationship model constructed in Step 3, and select the model with the best fit.

[0103] Step 5: Determine and narrow down the selection area for the reference node. Use a bisection method to divide the room or area to be located, further narrowing down the range of the target node;

[0104] Step 6: Deploy reference node locations. Several possible deployment scenarios for reference nodes are presented, and limitations on their deployment locations are proposed.

[0105] Step 7: Select a reference node using a weighted distance strategy. The selection of the reference node directly affects the positioning accuracy of the target node. This embodiment proposes using a weighted distance strategy to select a reference node with strong and stable signal strength.

[0106] Step 8: Localize using the PSOGWO algorithm. Based on the data obtained above and the selected strength-distance relationship model with the best fit, the PSOGWO hybrid algorithm is used to locate and optimize the position of the target node.

[0107] The steps described above will be explained in detail below.

[0108] 1. Construction and Analysis of the Relationship Model between Signal Strength and Distance

[0109] This section is mainly divided into the following three stages: data acquisition and processing, construction of signal strength and distance relationship model, and model evaluation, which will be introduced in detail below.

[0110] 1.1 Data Acquisition and Processing

[0111] The Received Signal Strength (RSS)-based positioning method locates the target node by detecting the signal strength from it. This embodiment measures the received signal strength at different locations away from the target node. For example... Figure 2 As shown, the closer to the target node, the stronger the detected signal strength, meaning that the distance is inversely proportional to the received signal strength.

[0112] Figure 2 This also indicates that the signal strength acquired by the receiver is usually affected by reflection, refraction, and noise during transmission, causing fluctuations in the acquired signal strength. Therefore, before constructing a model of the relationship between signal strength and distance, it is necessary to filter the acquired signal strength.

[0113] Common filtering methods include mean filtering, median filtering, and Gaussian filtering. Since signal strength follows or approximately follows a normal distribution, this embodiment selects Gaussian filtering to process the acquired signal strength. This method uses a Gaussian model to filter the RSS values ​​of high-probability regions as valid values, and obtains their average value as the measurement value. Assuming a reference node collects n RSS values, this can be represented as S. n ={s1, s2, ..., s n-1 s n}, where the probability P corresponding to each RSS value is P n ={p1, p2, ..., p h-1 p n The mean and variance are μ and δ, respectively. 2 Then the Gaussian filter model is

[0114]

[0115] Note:

[0116]

[0117]

[0118] By setting a minimum high-probability value greater than 0.6, the impact of low-probability events on the RSS value is reduced. The RSS value is calculated as follows:

[0119] 0.15δ+μ≤x≤3.09δ+μ (13)

[0120] Substituting μ and δ into equation (13) determines the range of RSS. To eliminate errors, the value obtained from equation (13) is used to optimize the mean using equation (14) to obtain the optimized value of RSS.

[0121]

[0122] Figure 3 The results of RSS value processing are presented. It can be seen that Gaussian filtering eliminates the RSS value deviation caused by noise superposition to a certain extent, resulting in better stability.

[0123] 3.2 Data Fitting

[0124] like Figure 4 As shown, this embodiment selects a distance within the range of 0-9m to measure the signal strength obtained by the receiving end. The black curve represents the original data, and the other curves are the fitted signal strength attenuation models. After comparing with various curve models, this embodiment selects three functions with better fitting effects, Logistic, Poly4, and Log3P1, to construct a signal strength and distance relationship model.

[0125] 3.3 Model Evaluation

[0126] To further select the model with the best fit, this embodiment evaluates the three types of models using the following three criteria:

[0127] (1) Reduced Chi-Sqr: The sum of the squared differences between the estimated and actual values ​​of y at each point is called the residual sum of squares. The smaller the residual sum of squares, the better the model fits. Figure 4 The results show that the Poly4 model has the smallest sum of squared residuals, which is 4.82058, while the Log3P1 model has the largest sum of squared residuals, which is 5.38572.

[0128] (2) R-Square(R 2 ):R 2 The closer the value is to 1, the better the fit. Among the three models, the Log3P1 model has the highest R-value. 2 The value was 0.95437, indicating the best fit.

[0129] (3) Adjusted R 2 Adjusted R 2 With R 2 The greater the difference between them, the worse the fit. From Figure 4 As can be seen from the data, the adjusted R values ​​of the three models Logistic, Poly4, and Log3P1 are... 2 With R 2 The differences between the values ​​were 0.00508, 0.00629, and 0.00352, respectively. The adjusted R values ​​for the three models...2 With R 2 If the difference between the two sides is small, the fitting effect is quite good.

[0130] In summary, this embodiment will select the Poly4 curve to construct the signal strength-distance relationship model. The Poly4 fitting curve equation is as follows:

[0131] y = 0.01889x 4 -0.50415x 3 +4.49659x 2 -17.72742x-35.48047 (15)

[0132] 2. Strategy for Selecting Reference Node Locations

[0133] The reference node location selection strategy mainly includes three stages: determining and narrowing the reference node selection area, the reference node location deployment strategy, and the weighted distance strategy for selecting reference nodes. The following is a detailed introduction to these three stages.

[0134] 2.1 Determine and narrow down the selection area of ​​reference nodes

[0135] After determining that the target node is in a specific room or area, this embodiment uses a binary search method to divide the target area into two parts. For example... Figure 5 As shown in (a), sampling is performed at the center of the four divided regions. If signal strength is detected at multiple points, the regions are compared, and the region with the strongest signal strength (i.e., the blue region in the figure) is selected as the next sampling region; if no signal strength is detected, the sampling is performed as follows: Figure 5 As shown in (b), the four regions are further divided into two, and sampling is performed at the green points. This process is repeated until the signal strength of the target node is detected, and then the next sampling region (i.e., the pink region in the figure) is selected.

[0136] During the measurement process, we found that signal strength is highly susceptible to environmental interference, leading to significant instability in signal strength measurements at fixed locations. To reduce the impact of variability on positioning accuracy, this embodiment proposes using an error interval to construct the approximate area where the target node is located, such as... Figure 6 As shown, assuming the signal strength measured at the reference node is in the range of [-53, -45], this is converted to a distance [d2, d1]. The circle formed by this distance range represents the actual region where the target node is located. Therefore, in this embodiment, the signal strength value collected by a reference node during sampling will no longer be represented by a circle (i.e.,...). Figure 6 Instead of using a circle with radius d, the approximate area of ​​the target node is defined by an annulus formed by the range of signal strength fluctuations at that location.

[0137] 2.2 Reference Node Location Deployment Strategy

[0138] The binary search method can quickly narrow down the approximate area where the target node is located. Once this area is determined, this embodiment will select as few reference nodes as possible to locate the target node. For example... Figure 7 As shown, when the three reference nodes are collinear, there are three possible scenarios: The first is when there is no intersecting region, such as... Figure 7 As shown in (a), the second type is the existence of a large intersecting region, such as Figure 7 As shown in (b), the third type has multiple intersecting regions, such as Figure 7 As shown in (c), the above three situations will increase the difficulty of locating the true location of the target node.

[0139] To avoid the above situation, this embodiment proposes to limit the position of the reference nodes. When the number of reference nodes is equal to or greater than 3, we will ensure that at least 3 reference nodes are not collinear. For example... Figure 8 As shown, firstly, the first reference node A and the second reference node B are selected. To avoid the three reference nodes being collinear, the selection of the third reference node can be one of three cases: The first case is as follows... Figure 8 As shown in (a), the third reference node C or C′ lies on the perpendicular line of line segment AB; the second case is as follows. Figure 8 As shown in (b), where C or C′ lies on the perpendicular line to the extension of AB; the third case is as follows. Figure 8 As shown in (c), where C or C′ lies on the perpendicular line to the reverse extension of AB. This ensures that the annulus formed by the three reference nodes has a single, smaller intersection area.

[0140] 2.3 Weighted Distance Strategy for Selecting Reference Nodes

[0141] The signal strength and stability of the target node vary depending on the location of the reference node. For example... Figure 9 As shown, this embodiment selects four reference nodes D1, D2, D3, and D4. The mean RSS values ​​of these four reference nodes are -30.26, -50.24, -65.05, and -80.54, respectively, and the variances are 1.952, 7.822, 6.868, and 3.028, respectively. It is evident that although the signal strength at reference node D2 is relatively strong, its signal stability is the worst. If the signal strength value is used as the sole criterion for selecting reference nodes, it will affect the positioning accuracy.

[0142] Therefore, to select reference nodes with strong and stable signal strength, this embodiment proposes recording the RSS variance of the reference nodes to improve the reliability of reference node selection. Different weights are assigned to different reference nodes based on the recorded variance. The signal strength collected N times consecutively at the i-th reference node can be represented as follows:

[0143]

[0144] According to the obtained RSS i We can calculate the variance of the reference node as σ. i The larger the variance, the more unstable the RSS received by the reference node, and therefore the smaller the weight should be assigned. The weight w assigned to the i-th reference node... i The calculation is as follows:

[0145]

[0146] Where t represents the number of reference nodes, and d represents the weighted distance between the reference nodes and the target node. ij The following can be calculated:

[0147] d ij =d i ×w i (18)

[0148] Where j represents the target node, d i This represents the distance between the i-th reference node and the target node.

[0149] like Figure 10 As shown in (a), taking the deployment of three reference nodes as an example, S is the first reference node to be deployed. In order to deploy the next reference node position that is closest to the target node, this embodiment selects a position in each of the four directions above, below, left, and right of position S and compares the signal strength of these four positions. The position with the strongest signal strength (i.e., S2) is selected as the candidate position for the second reference node position. Figure 10 In (b), P1, P2, and P3 represent three reference nodes selected within the distance range S[R1, 2R1]. According to the weighted distance strategy, the weighted distance at P1 is the shortest, so this position is selected as the second reference node position. Figure 10 (c) describes the selection process for the third reference node position. According to the reference node position deployment strategy, this embodiment selects three reference node positions, Q1, Q2, and Q3, on the perpendicular bisectors of positions S and P1, respectively. The distances between these three reference node positions and the first two reference node positions satisfy the range [R1, 2R1]. Based on the weighted distance strategy, Q2 will be selected as the third reference node position. Finally, the three reference node positions S, P1, and Q2 are selected sequentially.

[0150] 3 Target Node Localization

[0151] like Figure 11 As shown, simply using the centroid of the intersection of the three rings (i.e., the green hexagon) as the true position of the target node will lead to reduced positioning accuracy. Therefore, this embodiment proposes using a hybrid PSOGWO to optimize the position of the target node. The main implementation steps of the positioning process are as follows:

[0152] Step 1: As Figure 12 As shown in (a), a gray wolf population N is initialized within the target area, where each gray wolf represents a potential location of the target node;

[0153] Step 2: Calculate the fitness of individual gray wolves, and save the top 3 gray wolves with the best fitness, namely α, β, and δ. The fitness value is calculated using the following formula:

[0154]

[0155] Note:

[0156]

[0157]

[0158] Among them RMS i This represents the fitness of gray wolf i. The smaller the fitness, the closer the gray wolf is to the actual position of the target node. (x) t y t ) and (x p y p ) represent the position of the reference node and the position of each gray wolf, respectively, d t，p d is the distance from the reference node to each gray wolf. t，ap This represents the distance from the reference node to the target node, where n is the number of reference nodes. When sorting the fitness of individual gray wolves, situations may arise where fitness values ​​are equal, such as... Figure 12 As shown in (b), the distances from gray wolf i and gray wolf j to the three reference nodes are equal, i.e., d1 + d2 + d3 = D1 + D2 + D3. At this time, the fitness values ​​of the two gray wolves are also equal, indicating that the gray wolves' positions overlap. However, in reality, the position of gray wolf i is not equal to the position of gray wolf j, which will affect the target node's localization. Therefore, if the fitness values ​​of the gray wolves are equal, this embodiment proposes to use the following condition for exclusion:

[0159] d 1，i =d 1，j d 2，i =d 2，j , ..., d n-1，i =d n-1，j (20)

[0160] Where d 1，i and d 1，j These represent the distances from gray wolf i and gray wolf j to the first reference node, respectively.

[0161] Step 3: Update the current position of the gray wolf;

[0162] Step 4: Update a, A, and C;

[0163] Step 5: Calculate the fitness of all gray wolves;

[0164] Step 6: Update the fitness and position of α, β, and δ;

[0165] Step 7: Determine whether the maximum number of iterations or the set fitness threshold has been reached. If so, exit the loop and output the position of the target node; otherwise, go to step 3 to continue execution.

[0166] Time complexity affects the execution efficiency of an algorithm. The time complexity of ILP-PSOGWO mainly consists of two aspects: firstly, using binary search to narrow down the selection area of ​​reference nodes; and secondly, using a hybrid PSOGWO algorithm to optimize the target node position. The optimal time complexity of binary search for the target node is O(1), and the worst-case time complexity is O(log₂n). The time complexity of the hybrid PSOGWO algorithm is O(t×n) + O(t×n×d). Therefore, the time complexities of the two methods are O(log₂n) and O(n×d), respectively. 3 The time complexity of ILP-PSOGWO is O(n^2). 3 ).

[0167] 4 Experiments

[0168] To verify the effectiveness and feasibility of the proposed method, this invention acquired a dataset in a real-world environment, conducted multiple experiments, and compared the experimental results with existing methods PSO, GWO, and EWCL.

[0169] 4.1 Experimental Setup

[0170] In real-world scenarios, the propagation environment of wireless communication systems is typically categorized into line-of-sight (LOS) and not-line-of-sight (NLOS) environments. In a LOS environment, the wireless signal can propagate in a straight line between the transmitter and receiver without obstruction. If the conditions are not met, the signal strength will attenuate, which is the NLOS environment. To verify the impact of the experimental environment on ILP-PSOGWO, this embodiment conducts experiments in both LOS and NLOS environments. The experiments were performed on a Windows 10 professional edition platform using an Intel(R) Core(TM) i7-11700@2.50GHz, 32GB RAM, and Python version 3.7.10. Detailed parameter settings during the experiments are shown in Table 1.

[0171] Table 1 Experimental Setup

[0172]

[0173] 4.2 Performance Evaluation

[0174] The experiments tested the impact of four factors—location area, number of reference nodes, and number of iterations—on the positioning accuracy of the proposed method under both LOS and NLOS environments. Positioning accuracy is a key performance indicator for a positioning method and an important criterion for evaluating the quality of a positioning model. Positioning accuracy is typically measured using the Euclidean distance between the estimated and true positions, i.e., the positioning error. A smaller positioning error indicates higher positioning accuracy. i It can be calculated as follows:

[0175]

[0176] Where (x) true y true ), (x est y est ) represent the actual position and estimated position of target node i, respectively.

[0177] 4.2.1 The Influence of Positioning Area Area on Positioning Accuracy

[0178] When the number of reference nodes is 5, this embodiment tests the impact of different positioning area areas on the positioning accuracy of the four methods in indoor LOS and NLOS environments.

[0179] Figure 13 (a) illustrates the impact of the location area on positioning accuracy in a LOS environment. As the location area increases, the positioning errors of the EWCL and ILP-PSOGWO methods increase accordingly. The positioning errors obtained by the PSO and GWO methods exhibit an unstable trend because they search for target nodes across the entire location area, with inconsistent search direction and speed, leading to unstable positioning results. This embodiment reduces the search space for target nodes before using the PSOGWO algorithm, thus obtaining stable and relatively accurate positioning results. Figure 13 (b) shows the impact of the positioning area on positioning accuracy in the NLOS environment. In the NLOS environment, the positioning error of all four methods increases. Compared with the EWCL and ILP-PSOGWO methods, the PSO and GWO methods are more sensitive to the environment and have larger positioning errors.

[0180] 4.2.2 Impact of Target Node Location on Positioning Accuracy

[0181] During the experiment, when the positioning area was 17×10 and the number of reference nodes was 5, this embodiment selected 6 different target node locations for testing. The experimental results are shown in Table 2. Compared with NLOS, the positioning errors obtained by the four methods were lower in the LOS environment. This is because in the NLOS environment, the signal strength is easily affected by the environment. As can be seen from the table, the positioning errors obtained by the PSO (

[31] Gul F, Rahiman W, Alhady SS, et al. Meta-heuristic approach for solving multi-objective path planning for autonomous guided robot using PSO-GWO optimization algorithm with evolutionary programming[J]. Journal of Ambient Intelligence and Humanized Computing, 2021, 12(7): 7873-7890.) and GWO (

[38] El-Hasnony IM, Barakat SI, Elhoseny M, et al. Improved feature selection model for big data analytics[J]. IEEE Access, 2020, 8: 66989-67004.) methods differ greatly in the two environments, indicating that these two methods are easily affected by the environment. However, the positioning errors measured by EWCL (

[27] Alhammadi A, Hashim F, Fadlee M, et al.An adaptive localization system using particle swarm optimization in a circular distribution form[J].Jurnal Teknologi,2016:105-110.) and ILP-PSOGWO methods were relatively similar in the two environments, indicating that these two methods have good stability.

[0182] Table 2. Impact of Target Node Location on Positioning Accuracy

[0183]

[0184]

[0185] 4.2.3 Impact of the Number of Reference Nodes on Positioning Accuracy

[0186] The number of reference nodes is one of the factors affecting the positioning accuracy of the target node. In order to observe the impact of the number of reference nodes on the positioning accuracy, this embodiment selects 3, 5, 7, 9, 11 and 13 reference nodes respectively under the condition that the positioning area is 17×10.

[0187] Figure 14 The impact of the number of reference nodes on the positioning accuracy of the three methods is presented in both LOS and NLOS environments. Figure 14 (a) shows that as the number of reference nodes increases, the positioning error of the three methods decreases accordingly, and the positioning error remains within 1.5m. Figure 14 (b) shows that, compared to the GWO and ILP-PSOGWO methods, the change in the number of reference nodes has a greater impact on the localization error of the PSO method. This is because, although the PSO method has a strong ability to develop new spaces, the lack of communication between particles and its tendency to get trapped in local optima lead to unstable localization results. Although the GWO method has a weaker ability to develop new spaces, resulting in lower localization accuracy, the information sharing among individual gray wolves allows it to maintain higher stability in localization results. The ILP-PSOGWO method significantly reduces the area where the target node exists, thus maintaining high localization accuracy even when the number of reference nodes changes.

[0188] 4.2.4 Impact of Iteration Number on Positioning Accuracy

[0189] Figure 15 Figure (a) shows the relationship between the number of iterations and the localization error in the LOS environment. As the number of iterations increases, the localization errors of the three methods decrease significantly. When the number of iterations exceeds 15, the localization errors obtained by the three methods remain basically unchanged: the localization errors of the PSO and GWO methods remain around 1m, while the localization error of the ILP-PSOGWO method is around 0.4m. It can also be seen from the figure that the ILP-PSOGWO method reaches the optimal solution faster, i.e., has stronger convergence ability. This is because the hybrid PSOGWO algorithm used in this embodiment combines the advantages of both PSO and GWO algorithms, giving the method strong search and convergence capabilities. Figure 15 (b) presents the experimental results under the NLOS environment. Environmental changes significantly increased the localization errors of the three methods, with a greater impact on the localization errors of the PSO and GWO methods, and a smaller impact on the localization error of the ILP-PSOGWO method. When the number of iterations increased to 35, the localization errors of the PSO, GWO, and ILP-PSOGWO methods decreased significantly to 1.9073m, 1.8351m, and 0.9112m, respectively.

[0190] 4.2.5 Stability

[0191] To verify the stability of the ILP-PSOGWO method, this embodiment evaluates it from the following two aspects: standard deviation and cumulative probability distribution.

[0192] (1) Standard deviation

[0193] Standard deviation reflects the dispersion of a set of data, thus reflecting the stability of that data set. std The calculation formula is as follows:

[0194]

[0195] Where n represents the number of experimental measurements, and error j The positioning error representing the j-th measurement can be calculated using formula (21), where error is the positional error. avg This represents the average positioning error of the data set after the j-th measurement.

[0196] When the number of sample points is 7 and the location area is 17×10, this embodiment conducted 5, 10, 15, 20, 25, 30, 35, 40, 45, and 50 experiments in both LOS and NLOS environments. The experimental results are as follows. Figure 16 As shown. Under LOS conditions, the standard deviations of the PSO and GWO methods are below 0.7, and the standard deviations fluctuate significantly with the number of experiments. The EWCL and ILP-PSOGWO methods exhibit good stability, with their standard deviations consistently remaining below 0.4. Under NLOS conditions, the stability of all four methods decreases significantly. Compared to the EWCL and ILP-PSOGWO methods, the PSO and GWO methods are less stable, with the PSO method exhibiting the worst stability.

[0197] (2) Cumulative probability distribution

[0198] To further evaluate the stability of the ILP-PSOGWO method, this embodiment conducted 100 tests at a single target node location in a LOS environment. The experimental results are as follows: Figure 17 As shown.

[0199] Figure 17 The cumulative frequency distribution and cumulative frequency distribution of 100 positioning errors obtained by four methods are shown respectively. The mean positioning errors obtained by the PSO, GWO, EWCL and ILP-PSOGWO methods are 1.0848m, 1.15672m, 0.54455m and 0.47539m respectively. Figure 17This also indicates that the positioning errors obtained by the PSO and GWO methods are mostly distributed between 0.6m and 1.8m, and the frequency of occurrence is less than 10 times, indicating that these two methods have poor stability. In contrast, the positioning errors of the EWCL and ILP-PSOGWO methods are mostly distributed between 0.3m and 0.7m, and the frequency of occurrence is as high as 20 times, indicating that they have better stability. Figure 17 (d) reflects that the ILP-PSOGWO method has a positioning error of less than 0.9012m in 90% of cases and less than 0.4837m in 50% of cases, which are better than the comparison method.

[0200] In summary, to quickly and accurately map the locations of wireless sensor nodes to the indoor environment and reduce the impact of non-line-of-sight and improper reference node location selection on positioning accuracy, this invention proposes a novel node location prediction method (ILP-PSOGWO) based on hybrid particle swarm optimization and gray wolf optimization. This method uses a weighted distance strategy in the reference node location selection stage and a hybrid PSOGWO algorithm in the positioning stage. Compared with several existing methods, a series of experimental results show that the ILP-PSOGWO method can achieve higher positioning accuracy and convergence speed. Furthermore, experiments conducted in LOS and NLOS environments demonstrate that ILP-PSOGWO maintains high stability in positioning results.

[0201] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A new method for node position prediction based on hybrid particle swarm optimization and grey wolf optimization, characterized in that, The method comprises the following steps: Step 1: determining a room or area where the target wireless sensor node is located, and measuring received signal strength at different positions away from the target wireless sensor node; Step 2: filtering the signal strength obtained in step 1 by using Gaussian filtering; Step 3: fitting the obtained signal strength with distance to construct a signal strength-distance relationship model; Step 4: evaluating the signal strength-distance relationship model constructed in step 3 to select the model with the best fitting degree; Step 5: dividing the room or area to be positioned by using dichotomy to determine and narrow down the reference node selection area; Step 6: limiting the position of the reference node deployment; Step 7: selecting the reference node with strong and stable signal strength by using a weighted distance strategy; Step 8: positioning and optimizing the position of the target wireless sensor node by using a hybrid PSOGWO algorithm based on the data obtained above and the selected signal strength-distance relationship model with the best fitting degree; The step 8 comprises: Step 8.1: initializing a grey wolf population N in the target area, and each grey wolf represents a potential position of the target wireless sensor node; Step 8.2: calculating the fitness of the grey wolf individuals, and saving the top three grey wolves α, β and δ with the best fitness, and the fitness value calculation formula is as follows: In the formula: wherein represents the fitness of gray wolf i, the smaller the fitness, the closer the actual position of the gray wolf to the target node; ) and ( ) represent the positions of the reference nodes and each gray wolf, respectively; is the distance from the reference node to each gray wolf; is the distance from the reference node to the target node; n is the number of reference nodes; Step 8.3: updating the position of the current grey wolf; Step 8.4: Update a, A and C, where A and C are the vectors of the correlation coefficients, , , a is linearly reduced from 2 to 1 throughout the iteration process, and r1 and r2 are random vectors in [0, 1]; Step 8.5: calculating the fitness of all the grey wolves; Step 8.6: updating the fitness and position of α, β and δ; Step 8.7: judging whether the maximum iteration number is reached or the set fitness threshold is reached, if yes, exiting the loop and outputting the position of the target node; otherwise, going to step 3 to continue execution; In the step 8.6, the position of the current grey wolf is updated according to the following formula: In the formula: , , , , where, is the inertia factor; , and are acceleration constants, representing the learning factors of the three best grey wolves, respectively; , and are random numbers in the interval [0, 1], which can increase the randomness of the search; denotes the velocity vector of particle i in the jth dimension in the tth iteration; denotes the position vector of particle i in the jth dimension in the tth iteration; , , denote the position vectors of a, b, d in the current population, respectively; , and denote the step size and direction of the candidate grey wolf in the wolf pack towards the best three grey wolves, respectively; denotes the position vector of the prey; , , denote the distances between the current candidate grey wolf and a, b, d, respectively; , , , , , is the cooperation coefficient vector.

2. The method of claim 1, wherein the method is characterized by: The step 3 comprises: Logistic, Poly4 and Log3P1 are respectively used to construct the signal strength-distance relationship model.

3. The method of claim 1, wherein the method is characterized by: The step 5 comprises: After the room or area where the target wireless sensor node is located is determined, dichotomy is used to divide the target area, and the center of each of the four divided areas is sampled; if signal strength is detected at multiple points, comparison is performed to select the area with the maximum signal strength as the sampling area in the next step; if no signal strength is detected, the four areas are divided again, and the center points of the areas after the second division are sampled, and iteration is performed in sequence until the signal strength of the target node is detected, and the sampling area in the next step is selected.

4. The method of claim 1, wherein the method is characterized by: The step 5 further comprises: The circular ring formed by the signal strength floating range collected by the reference node is used as the approximate area where the target node is located.

5. The method of claim 1, wherein the method is characterized by: The step 6 comprises: When the number of reference nodes is more than three, it is ensured that at least three reference nodes are not collinear.

6. The method of claim 1, wherein the method is characterized by: The step 7 comprises: The RSS variance obtained by the reference node is recorded, different weights are assigned to different reference nodes according to the recorded variance, and the greater the variance, the less stable the RSS obtained by the reference node, and the smaller the weight to be assigned.

7. The method of claim 1, wherein the method is characterized by: In the step 8.2, if the fitness values of the grey wolves are equal, the following conditions are used for exclusion: wherein denotes the distance of the gray wolves i and j to the first reference node.

Images