WWB positioning method based on wild dog optimization algorithm
By introducing a wild dog optimization algorithm with the target tracking mechanism and an adaptive elimination probability mechanism in UWB positioning, the problem of degradation of UWB positioning accuracy in complex indoor environments is solved, and higher positioning accuracy and stability are achieved.
Patent Information
- Application Number
- CN202411163643.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-23
- Publication Date
- 2024-11-26
AI Technical Summary
In complex indoor environments, especially under non-line-of-sight conditions, the accuracy of UWB positioning technology has significantly decreased and it is impossible to provide effective positioning services, resulting in large positioning errors and low accuracy.
UWB positioning method based on wild dog optimization algorithm is adopted to improve wild dog optimization by constructing distance measurement between base station and unknown nodes, combining target tracking mechanism, domain solution optimization strategy, non-inferior solution conversion strategy and adaptive elimination probability mechanism The algorithm's global optimization ability and positioning accuracy.
Under NLOS error interference, positioning accuracy and stability are significantly improved, positioning errors are reduced, and positioning accuracy and search capabilities of the wild dog optimization algorithm are improved.
Smart Images

Figure CN119031471A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of indoor positioning technology, and in particular to a UWB positioning method based on a wild dog optimization algorithm. Background Art
[0002] In recent years, according to the differences in the demand for location services in different scenarios, people have proposed various positioning systems and positioning optimization algorithms for different scenarios to locate relevant targets in real time. Although the global positioning system has high positioning accuracy and good real-time performance, GPS positioning technology will be affected by obstructions in indoor terrains such as tunnels, mines, and basements, resulting in poor positioning accuracy and even failure to receive positioning signals. Therefore, GPS positioning is not suitable for indoor environments. To achieve accurate indoor positioning, other technical solutions are needed.
[0003] At present, UWB technology has achieved relatively mature research results and application practices in the field of indoor positioning, and its positioning accuracy is also continuously improving, which is reflected in open indoor environments, such as classrooms, gymnasiums, waiting rooms, etc. However, in complex indoor environments, especially under non-line-of-sight (NLOS) conditions, the positioning accuracy of UWB technology decreases significantly, and even cannot provide positioning services. Summary of the invention
[0004] In view of the above-mentioned shortcomings of the prior art, the present invention provides a UWB positioning method based on a wild dog optimization algorithm.
[0005] To achieve the above effects, the technical solution of the present invention is as follows:
[0006] The present invention provides a UWB positioning method based on a wild dog optimization algorithm, comprising the following steps:
[0007] Step 1: Build base stations B1, B2, and B3, use the TDOA algorithm to measure the propagation time of the electromagnetic wave emitted from the unknown node A to base stations B1-B3, and calculate the distance between the unknown node A and base stations B1-B3 based on the propagation speed of the electromagnetic wave;
[0008] Step 2: Take the distance measurement value from unknown node A to base stations B1-B3 as the radius, draw a circle with the corresponding base station as the center, and use the coordinates of the intersection of the circles as the first coordinates of unknown node A;
[0009] Step 3: Build the wild dog optimization algorithm:
[0010] Assume that the number of initial wild dog population is N. The position of each individual in the wild dog population will change continuously during the search for prey. The position set of each individual in the wild dog population is X = (x1, x2, ..., x n); In the wild dog optimization algorithm, each individual will perform encirclement, pursuit, scavenging and survival behaviors. Each individual must go through encirclement, pursuit and scavenging behaviors to survive, and the final surviving individual is the target location. The final target location is solved to obtain the second coordinate of the unknown node A;
[0011] Step 4: Set up the target tracking mechanism: Initialize the number of wild dog populations and the problem dimension, and randomly generate wild dog population samples:
[0012] Through the search capability of the target tracking mechanism, all solvable targets in the search space are decomposed and the updated statistics of all solvable positions are completed;
[0013] Step 5: Introduce the adaptive elimination probability mechanism into the wild dog optimization algorithm to obtain the AEM-wild dog optimization algorithm; apply the target tracking mechanism and the AEM-wild dog optimization algorithm as the target tracking wild dog optimization algorithm in the UWB positioning method.
[0014] Furthermore, the distance between the unknown node A and the base stations B1 to B3 is calculated in step 1 and is expressed as:
[0015]
[0016] Wherein, d1~d3 represent the distances between the unknown node A and the base stations B1~B3 respectively, (x1, y1), (x2, y2), (x3, y3) represent the coordinate positions of the base stations B1~B3 respectively, and (x0, y0) represents the coordinates of the unknown node A.
[0017] Further, step 2 includes:
[0018] The equation group (1) can be rearranged to obtain:
[0019] 2(x i x0+y i y0)=x i 2 +y i 2 +x0 2 +y0 2 +d i 2 ,i=1,2,3 (2)
[0020] Where i represents the i-th base station; the solution of equation (2) is the position of the unknown node A;
[0021] Under the influence of NLOS error, the circles where base stations B1 to B3 are located will not intersect at one point. The least square method is used to estimate the position of the unknown node A. By squaring both sides of the equation group (2), the following equation group is obtained:
[0022]
[0023] Let x 2 +y 2 =C, then formula (3) is expressed in matrix as:
[0024] MX=N (4)
[0025] in,
[0026] According to formula (4), the error vector θ is defined as:
[0027] θ=MX-N (5)
[0028] Let the square of the error vector θ be Q, which is expressed as:
[0029] Q=θ 2 =(MX-N) T (MX-N) (6)
[0030] The derivatives of both sides of equation (6) are:
[0031]
[0032] When the derivative of equation (7) is set to 0, the extreme value is obtained:
[0033] X=(M T M) -1 M T N (8)
[0034] The position of the unknown node A is determined at the intersection of two hyperbolas with B1, B2 and B1, B3 as foci. The first coordinate of the unknown node A (d 21 , d 31 ):
[0035]
[0036] Furthermore, in the wild dog optimization algorithm described in step 3, each individual will perform the following behaviors: encirclement, pursuit, scavenging, and survival, including: (1) encirclement behavior
[0037] In the wild dog optimization algorithm, a certain number of initialized wild dog populations is set, and the initialized wild dog population is placed within the set range to search and update the location of the suspected target. The specific expression is as follows:
[0038]
[0039]
[0040] Among them, k represents the kth individual, represents the kth individual at time d; represents the i-th individual at time d; represents the current best individual at time d;
[0041] na represents the number of wild dog individuals surrounding the prey in the wild dog population, which is a random integer in [2,N / 2]; is the distance between na wild dogs and the target object; β is a uniform random number between [-2,2], and we get The updated new position is expected to be -2X best ;X best represents the best individual at present;
[0042] (2) Pursuit behavior
[0043] The pursuit behavior is to guide the wild dog optimization algorithm to find the relatively optimal solution by simulating the pursuit strategy. After the individual surrounds the suspected target, it searches near the local optimal individual position. The formula is as follows:
[0044]
[0045] Among them, β1 is a uniform random number between [-2,2], β2 is a uniform random number between [-1,1], is a random individual in a wild dog population;
[0046] (3) Degradation behavior
[0047] Through the hunted target, search again near the relatively better individuals; the formula is as follows:
[0048]
[0049] The coefficient β2 in formula (13) is a random number in [-2, 2], and the coefficient b is a random number in [-1, 1].
[0050] (4) Survival
[0051] After completing the scavenging behavior, wild dogs will still face the problem of survival. The survival probability of each wild dog is calculated based on its survival function F(X). Wild dog individuals with low survival rates will repeat the encirclement, pursuit, and scavenging behaviors and search for the target location again to increase the final survival probability of the wild dogs.
[0052] Furthermore, the survival probability of the individual is calculated by formula (14), that is, the final survival probability value. If the individual survival probability is less than 30%, the individual re-estimates the new position using formula (15). After the relevant positions are updated, the survival probability is calculated again. Two different individuals in the group are randomly selected. and Ensure that the survival probability of each wild dog reaches more than 70%;
[0053]
[0054]
[0055] In the formula, and represents two different individuals in a group; F(X), F(X i ) represent the survival function of the individual; ε represents the fitness function in solving the survival probability.
[0056] Furthermore, in step 4, all solvable targets in the search space are decomposed to complete the updated statistics of all solvable positions, including:
[0057] The statistical process of all solvable positions is as follows:
[0058]
[0059] Where N is the number of wild dogs, X is n (t) represents all the solvable positions within the initial range, and the domain solution optimization strategy and non-inferior solution conversion strategy are selected and used according to all the solvable positions;
[0060] a) Domain solution optimization strategy:
[0061] The mathematical model of the neighborhood solution optimization strategy is as follows:
[0062]
[0063] Among them, D′1, D′2, and D′3 represent the three individuals closest to the relatively optimal solution in the current range;
[0064] Therefore, the position update of the wild dog optimization algorithm based on the neighborhood optimization strategy in the search phase is adjusted to:
[0065]
[0066]
[0067] Among them, λ1, λ2, λ3 represent control parameters, and X(t+1) represents the latest solution position obtained through the three best solutions in the neighborhood;
[0068] b) Non-inferior solution conversion strategy
[0069] Get five solutions B~F within the range of the wild dog optimization algorithm, perform an independent search near the five solutions B~F, find a non-inferior solution with a small error in the fitness value of each individual, convert through the non-inferior solution, search near the non-inferior solution to find a better solution, when the wild dog optimization algorithm performs a certain number of iterations, obtain five new best non-inferior solutions B1~F1 near the initial five solutions that are closer to the current optimal point A*;
[0070] The best non-inferior solution B1~F1 is closer to the current best point A*. The best non-inferior solution and the current best point A* are grouped together to conduct a new search. The best solution among the best non-inferior solutions B1~F1 and the current best point A* is compared by equations (4)-(11) to obtain:
[0071] Xb i (t)-X c (t)<α[Xb(t)-X c (t)] (20)
[0072] α=log 0.5 (-t / L) (21)
[0073] Among them, α is the update factor, Xb i (t) is the position of the i-th non-inferior solution obtained by the wild dog optimization algorithm after t iterations, Xb(t) is the position of the initial non-inferior solution, and X c (t) is the current optimal solution position; the non-inferior solution after t iterations is compared with the current optimal point A*, and the optimal solution position is obtained by converting the best non-inferior solution B1~F1 and the current optimal point A*:
[0074]
[0075] Among them, X i (t+1) is the position of the ith optimal solution after t iterations of the wild dog optimization algorithm; k(t) is the standard normal distribution conversion coefficient at the t iteration; L is the maximum number of iterations;
[0076] Therefore, by introducing the non-inferior solution conversion strategy, the conversion between the non-inferior solution and the current best point A* is found and the best point in the new non-inferior solution replaces the current best point A*.
[0077] Furthermore, the adaptive elimination probability mechanism in step 5 provides an individual information retention search strategy and a group information sharing search strategy, and adaptively adjusts the value in response to the search demand in different situations, wherein:
[0078] (1) Individual information retention search strategy:
[0079] The higher the individual information retention rate, the higher the corresponding search and mining capability value of the individual. Conversely, the lower the individual information retention rate, the lower the corresponding search and mining level of the individual. The individual is an individual in the wild dog population, and the individual information retention rate is defined as:
[0080]
[0081]
[0082] in, represents the individual information retention rate of the individual value n in the wild dog population after the mth iteration, It represents the search ability value of an individual after the mth iteration. The search ability value of each individual is different. When the search ability reaches the lowest and highest value, it is replaced by low and high respectively. represents the amount of information that the individual has collected after the mth iteration;
[0083] Keep the information added by the individual after each iteration. The individuals in the wild dog population are represented by x n , the new search position obtained by retaining individual information is shown in formula (25);
[0084]
[0085] in, represents the individual information retention rate of the individual value n in the wild dog population after the mkth iteration;
[0086] The update method of the individual information preservation search strategy based on the adaptive elimination rate is shown in formula (26):
[0087]
[0088] Among them, L is the maximum number of iterations, x g,high Indicates the highest value of information retention in the current individual, r1 and r2 are selected as random numbers in (0,1) respectively;
[0089] (2) Group information sharing search strategy:
[0090] The closest individuals form a group, and the groups share information with each other. The individual position is updated according to the total amount of group information. The new search position obtained by group information sharing is shown in formula (27):
[0091]
[0092] Among them, S * Represents the new location obtained by group information sharing search update after m iterations, and Represents the information retention rate and individual quantity of individuals after M iterations;
[0093] The update method of the group information sharing search strategy based on the adaptive elimination rate is shown in formula (28):
[0094]
[0095] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:
[0096] In order to solve the problem of large positioning error and low positioning accuracy caused by NLOS error interference during positioning, the present invention proposes a UWB positioning method based on wild dog optimization algorithm. On the basis of the traditional wild dog optimization algorithm, a target tracking mechanism is introduced in the search stage, and the global optimization ability of the wild dog optimization algorithm is improved by using two strategies: domain solution search and non-inferior solution conversion, and error interference is reduced; an adaptive elimination probability mechanism is introduced, and the wild dog optimization algorithm's ability to mine dimensional information is improved by using individual information retention strategy and group information sharing strategy, and the optimal search space is adaptively formed, further improving the accuracy and stability of the wild dog optimization algorithm's position estimation. BRIEF DESCRIPTION OF THE DRAWINGS
[0097] Figure 1 This is a schematic diagram of TOA positioning based on the present invention;
[0098] Figure 2 This is a schematic diagram of TDOA positioning based on the present invention;
[0099] Figure 3 It is a schematic diagram of the neighborhood solution optimization process of the present invention;
[0100] Figure 4 It is a schematic diagram of generating a non-inferior solution of the present invention;
[0101] Figure 5 It is a schematic diagram of the non-inferior solution conversion process of the present invention;
[0102] Figure 6 It is a schematic diagram of the use of the target tracking mechanism strategy of the present invention;
[0103] Figure 7 is a schematic diagram of the individual information retention model of the present invention;
[0104] Figure 8 is a schematic diagram of a group information sharing model of the present invention;
[0105] Fig. 9 It is a schematic diagram of the simulation environment of the present invention;
[0106] Fig.10 RMSE distribution diagram of the present invention;
[0107] Fig.11 Schematic diagram of CDF distribution of the present invention. DETAILED DESCRIPTION
[0108] The following will describe the embodiments of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art will easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention is also implemented or applied through other different specific embodiments. The details in this specification are also modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention, not for limiting the scope of protection of the present invention.
[0109] It should be noted that the illustrations provided in the following embodiments are only used to schematically illustrate the basic concept of the present invention. Therefore, the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component are subject to arbitrary changes, and the component layout type can also be more complex.
[0110] Before introducing the embodiments of the present invention, the relevant terms involved in the embodiments of the present invention are explained as follows:
[0111] Ultra Wide Band (UWB) technology is a wireless carrier communication technology that does not use a sinusoidal carrier but instead uses nanosecond-level non-sinusoidal narrow pulses to transmit data.
[0112] Target Tracking Dingo Optimization (TTDO) is a ranging method based on TDOA.
[0113] TDOA: Time Difference of Arrival, a UWB positioning method based on the time of arrival (TOA) of the received signal.
[0114] Example
[0115] See also Figure 1 In order to reduce the influence of positioning error caused by NLOS error interference in the positioning process and improve positioning accuracy in a complex indoor environment, a target tracking wild dog optimization UWB positioning method with improved adaptive elimination rate is proposed. This embodiment proposes a UWB positioning method based on wild dog optimization algorithm. It includes the following steps:
[0116] Step 1: Build base stations B1, B2, and B3, use the TDOA algorithm to measure the propagation time of the electromagnetic wave emitted from the unknown node A to base stations B1-B3, and calculate the distance between the unknown node A and base stations B1-B3 based on the propagation speed of the electromagnetic wave;
[0117] Step 2: Take the distance measurement value from unknown node A to base stations B1-B3 as the radius, draw a circle with the corresponding base station as the center, and under ideal conditions, use the coordinates of the intersection of these circles as the first coordinates of unknown node A;
[0118] Step 3: Build the wild dog optimization algorithm:
[0119] Assume that the number of initial wild dog population is N. The position of each individual in the wild dog population will change continuously during the search for prey. The position set of each individual in the wild dog population is X = (x1, x2, ..., x n ), calculate the final survival function F(X) to judge the performance of the UWB positioning method; in the wild dog optimization algorithm, each individual will perform encirclement, pursuit, decomposition and survival behaviors. Each individual must go through encirclement, pursuit, decomposition to survive, and the final surviving individual is the target location. The final target location is solved to obtain the second coordinate of the unknown node A;
[0120] Step 4: Set up the target tracking mechanism: Initialize the number of wild dog populations and the problem dimension, and randomly generate wild dog population samples:
[0121] That is, the number of wild dog populations is set to N, the problem dimension of the wild dog optimization algorithm is set to 2, the generation positions of wild dog individuals are randomly initialized to obtain fitness values, and all solvable targets in the search space are decomposed through the search capability of the target tracking mechanism, completing the first update statistics of all solvable positions;
[0122] Step 5: Introduce the adaptive elimination probability mechanism into the wild dog optimization algorithm to obtain the AEM-wild dog optimization algorithm. The adaptive elimination probability mechanism provides individual information retention search strategy and group information sharing search strategy. Facing the search demand degree in different situations, the value is adjusted adaptively.
[0123] Through the above four steps, the target tracking mechanism and the AEM-wild dog optimization algorithm are applied as the target tracking wild dog optimization algorithm in the UWB positioning method, thereby obtaining more accurate coordinate information of the unknown node A.
[0124] The present invention introduces the target tracking mechanism into the wild dog optimization algorithm, implements two strategies in the search stage - domain solution optimization strategy and non-inferior solution conversion strategy, conducts integrated search for all target positions that can exist within the range, improves the search capability and prevents the wild dog optimization algorithm from falling into the local optimal solution; obtains the coordinate value in the positioning environment according to the individual survival probability, obtains the coordinate value error range according to the fitness function, and evaluates the performance of the AEM-wild dog optimization algorithm.
[0125] As a preferred technical solution, in this embodiment, the distance between the unknown node A and the base stations B1 to B3 is calculated in step 1, which is expressed as:
[0126]
[0127] Wherein, d1~d3 represent the distances between the unknown node A and the base stations B1~B3 respectively, (x1, y1), (x2, y2), (x3, y3) represent the coordinate positions of the base stations B1~B3 respectively, and (x0, y0) represents the coordinates of the unknown node A.
[0128] 3. The method according to claim 2, characterized in that step 2 comprises:
[0129] The equation group (1) can be rearranged to obtain:
[0130] 2(x i x0+y i y0)=x i 2 +y i 2 +x0 2 +y0 2 +d i 2 ,i=1,2,3 (2)
[0131] Where i represents the i-th base station; the solution of equation (2) is the position of the unknown node A;
[0132] However, in actual applications, under the influence of NLOS error, the circles where base stations B1 to B3 are located will not intersect at one point. The least squares method is used to estimate the position of the unknown node A. By squaring both sides of the equation group (2), the following equation group is obtained:
[0133]
[0134] Let x 2 +y 2 =C, then formula (3) is expressed in matrix as:
[0135] MX=N (4)
[0136] in,
[0137] According to formula (4), the error vector θ is defined as:
[0138] θ=MX-N (5)
[0139] Let the square of the error vector θ be Q, which is expressed as:
[0140] Q=θ 2 =(MX-N) T (MX-N) (6)
[0141] The derivatives of both sides of equation (6) are:
[0142]
[0143] When the derivative of equation (7) is set to 0, the extreme value is obtained:
[0144] X=(M T M) -1 M T N (8)
[0145] According to mathematical principles, the position of the unknown node A is determined at the intersection of two hyperbolas with B1, B2 and B1, B3 as foci; according to equation (9) listed in the figure above, the first coordinate (d 21 , d 31 ):
[0146]
[0147] TDOA positioning principle: The positioning method based on the time difference of arrival (TDOA) of the received signal is improved based on the TOA positioning method. By using the mathematical knowledge that the difference in distance from a point on the hyperbola to the two foci is always equal, the problem of clock asynchrony between the tag and the base station is solved. However, unlike the TOA positioning method, it is not necessary to obtain the absolute arrival time during measurement, which reduces the positioning complexity and has a relatively high positioning accuracy. Therefore, this is the most commonly used positioning method in UWB positioning systems. Figure 2 As shown in the figure, the distances between base stations B1, B2, B3 and unknown node A are set to d1, d2, d3 respectively, and the distances between unknown node A and base stations B are obtained. 12 and B 13 The distance d 21 and d 31 .
[0148] Dingo Optimization (DO) has strong local search capabilities and good convergence performance. Through the four stages of the African wild dog's hunting process, it summarizes and imitates four corresponding behaviors, namely, searching, chasing, scavenging, and survival. At the same time, combined with the indicator of performance evaluation of the Dingo Optimization Algorithm - survival rate, the higher the survival rate, the higher the probability that the Dingo Optimization Algorithm will find the optimal solution. Therefore, the Dingo Optimization Algorithm is applied to the UWB positioning method, and through a series of "hunting" processes, relatively stable and high-precision position coordinates are found.
[0149] As a preferred technical solution, in this embodiment, in the wild dog optimization algorithm described in step 3, each individual will perform the behaviors of encirclement, pursuit, scavenging and survival, including:
[0150] (1) Encirclement
[0151] The definition of encirclement behavior is a strategy to simulate the behavior of biological groups. It uses multiple search individuals to conduct a range search around potential high-quality solutions to improve search efficiency and solution quality. In the wild dog optimization algorithm, a certain number of initial wild dog populations is set, and the initial wild dog population is placed within the set range to search and update the location of the suspected target. The specific expression is as follows:
[0152]
[0153]
[0154] Among them, k represents the kth individual, represents the kth individual at time d; represents the i-th individual at time d; represents the current best individual at time d;
[0155] na represents the number of wild dog individuals surrounding the prey in the wild dog population, which is a random integer in [2,N / 2]; is the distance between na wild dogs and the target object; β is a uniform random number between [-2,2], and we get The updated new position is expected to be -2X best ;
[0156] (2) Pursuit behavior
[0157] The pursuit behavior is to guide the wild dog optimization algorithm to find the relatively optimal solution by simulating the pursuit strategy. After the individual surrounds the suspected target, it searches near the local optimal individual position. The formula is as follows:
[0158]
[0159] Among them, β1 is a uniform random number between [-2,2], β2 is a uniform random number between [-1,1], is a random individual in a wild dog population;
[0160] (3) Degradation behavior
[0161] The specific implementation of the scavenging behavior in the wild dog optimization algorithm is to search again near the relatively better individuals through the hunted target; the formula is as follows:
[0162]
[0163] The coefficient β2 in formula (13) is a random number in [-2, 2], and the coefficient b is a random number in [-1, 1].
[0164] (4) Survival
[0165] After completing the scavenging behavior, wild dogs will still face the problem of survival. The survival probability of each wild dog is calculated based on its survival function F(X). Wild dog individuals with low survival rates will repeat the encirclement, pursuit, and scavenging behaviors and search for the target location again to increase the final survival probability of the wild dogs.
[0166] As a preferred technical solution, in this embodiment, the survival probability of an individual, that is, the final survival probability value, is calculated by formula (14). If the individual survival probability is less than 30%, the individual re-estimates the new position using formula (15). After the relevant position is updated, the survival probability is calculated again. Two different individuals in the group are randomly selected. and Ensure that the survival probability of each wild dog reaches more than 70%;
[0167]
[0168]
[0169] In the formula, and represents two different individuals in a group; F(X), F(X i ) respectively; ε represents the fitness function in solving the survival probability.
[0170] In the hunting process, the traditional Dingo Optimization Algorithm only relies on individuals to search and then hunt, but each individual can have more than one target. When an individual hunts multiple targets, the number of individuals that are still alive will increase, and the calculated value of the final survival rate will be falsely high, resulting in the premature end of the Dingo Optimization Algorithm iteration, and mistakenly outputting the local relative optimal solution as the global optimal solution. In response to this situation, a Target Tracking Dingo Optimization (TTDO) algorithm based on target tracking is proposed, which introduces an improved target tracking mechanism to guide each individual to focus on its own hunting target and improve the accuracy of the survival rate, so that the Dingo Optimization Algorithm can fully iterate until the optimal solution in the global search space is found and output.
[0171] The traditional target tracking mechanism is a method for dynamically monitoring and tracking the position and state of mobile targets. In the wild dog optimization algorithm, it is necessary to solve the problem of a single individual chasing multiple targets, which leads to an artificially high survival rate. An improved target tracking mechanism is proposed, which divides the whole process into two stages: search and hunting. In the search stage, a domain solution optimization strategy and a non-inferior solution conversion strategy are implemented to integrate and search all target positions that can exist within the range, improve the search capability, and avoid the wild dog optimization algorithm from falling into the local optimal solution, thereby improving the iteration speed of the wild dog optimization algorithm. The search stage refers to the initial stage in which individuals explore unknown areas or solution spaces within a certain space to discover targets. It relies on individuals to widely collect target information, identify the optimal path or choice that can exist, and make an initial estimate of potential directions. In the first half of the improved target tracking mechanism, the wild dog optimization algorithm is used to search for all relatively optimal solutions that can appear in a large range of solution spaces, providing all possible ways or means for the subsequent hunting stage to carry out single target tracking.
[0172] As a preferred technical solution, in this embodiment, step 4 decomposes all solvable targets in the search space and completes the first update statistics of all solvable positions, including:
[0173] The statistical process of all solvable positions is as follows:
[0174]
[0175] Where N is the number of wild dogs, X is n (t) represents all the solvable positions within the initial range, and the domain solution optimization strategy and non-inferior solution conversion strategy are selected and used according to all the solvable positions;
[0176] a) Domain solution optimization strategy:
[0177] In the search phase of the traditional wild dog optimization algorithm, individuals search for the target location in a small range, and the search ability is limited. They can only perform a rough search of the global situation, and the relatively optimal solution found locally can easily make the wild dog optimization algorithm stay near the local optimal solution. Figure 3 As shown in the figure, A represents the relative optimal solution position in the field, A* represents the actual location of the target, and B, C, D, E, and F represent the positions of random energy solutions within a certain range.
[0178] The unknown node A is set as the relatively optimal solution position found by the wild dog optimization algorithm in the initial search stage. The initial large-scale rough search will cause the target to appear around the unknown node A instead of around the current best point A*. The wild dog optimization algorithm will mistakenly output the local search optimal solution as the optimal solution, causing the wild dog optimization algorithm to end its iteration early. The domain solution optimization strategy is to further strengthen the local search capability and deeply search the neighborhood of the optimal point, that is, near the current best point A*, to improve the accuracy of the solution. Since the wild dog optimization algorithm is searched by individuals in the wild dog population, a range set is formed in the form of three individuals as a group to search, and the search result takes the average of the three relatively optimal solutions for output.
[0179] The mathematical model of the improved neighborhood solution optimization strategy is as follows:
[0180]
[0181] Among them, D′1, D′2, and D′3 represent the three individuals closest to the relatively optimal solution in the current range;
[0182] Therefore, the position update of the wild dog optimization algorithm based on the neighborhood optimization strategy in the search phase is adjusted to:
[0183]
[0184]
[0185] Among them, λ1, λ2, λ3 represent control parameters, and X(t+1) represents the latest solution position obtained through the three best solutions in the neighborhood;
[0186] b) Non-inferior solution conversion strategy
[0187] The neighborhood solution optimization strategy introduced in the search phase can greatly enhance the local discovery of better new solutions to replace old solutions. However, it can only perform local neighborhood depth search when three individuals are in a group. Once the global optimal solution falls into the local neighborhood, the subsequent iteration process of the wild dog optimization algorithm will continue to iterate around the relatively optimal solution in the neighborhood. Figure 3As shown in the figure, the wild dog optimization algorithm obtains five solutions B~F within the range, performs an independent search near the five solutions B~F, finds each non-inferior solution with a small fitness value error, and converts through the non-inferior solution. The subsequent search direction of the wild dog optimization algorithm will be changed to search near the non-inferior solution to find a better solution, so the non-inferior solution conversion strategy is introduced. Figure 4 As shown, after the wild dog optimization algorithm performs a certain number of iterations, five new best non-inferior solutions B1~F1 that are closer to the current optimal point A* are obtained near the initial five solutions;
[0188] The best non-inferior solution B1~F1 is closer to the current best point A*. The best non-inferior solution and the current best point A* are grouped together to conduct a new search. The best solution among the best non-inferior solutions B1~F1 and the current best point A* is compared by equations (4)-(11) to obtain:
[0189] Xb i (t)-X c (t)<α[Xb(t)-X c (t)] (20)
[0190] α=log 0.5 (-t / L) (21)
[0191] Among them, α is the update factor, Xb i (t) is the position of the i-th non-inferior solution obtained by the wild dog optimization algorithm after t iterations, Xb(t) is the position of the initial non-inferior solution, and X c (t) is the current optimal solution position; the non-inferior solution after t iterations is compared with the current optimal point A*, and the optimal solution position is obtained by converting the best non-inferior solution B1~F1 and the current optimal point A*:
[0192]
[0193] Among them, X i (t+1) is the position of the ith optimal solution after t iterations of the wild dog optimization algorithm; k(t) is the standard normal distribution conversion coefficient at the t iteration; L is the maximum number of iterations;
[0194] Therefore, by introducing the non-inferior solution conversion strategy, the conversion between the non-inferior solution and the current optimal point A* is as follows: Figure 5 As shown, find and replace the current best point A* with the best point E1 in the new non-inferior solution.
[0195] Two strategies are introduced in the search phase: the domain solution optimization strategy and the non-inferior solution conversion strategy. All possible target locations within the range are searched in an integrated manner to improve the search capability while preventing the wild dog optimization algorithm from falling into the local optimal solution. The complete strategy flow of the improved target tracking mechanism search phase is as follows: Figure 6 shown.
[0196] The wild dog optimization algorithm relies on the OLV of individuals in the wild dog population to search-hunt-search-survive. The number of surviving individuals will be continuously updated in each round of iteration. The surviving wild dogs will continue to look for new prey. In fact, the number of individuals eliminated in each stage is large, but the eliminated individuals are not completely unable to find the target location. Instead, there may be individuals who have approached the target location in the search and stayed at the target location but failed to search. Therefore, in order to improve the wild dog population when the problem dimension is increased, the search difficulty and scope are increased. The elimination rate is introduced to adjust the number of wild dog populations in the early and late stages of the wild dog optimization algorithm, and the appropriate number of wild dog populations is allocated to improve the wild dog optimization algorithm's ability to search for better target locations locally, improve the wild dog optimization algorithm's global optimization ability, speed up the convergence of the wild dog optimization algorithm, and better find the best solution when facing NLOS error interference.
[0197] Considering that the infection of NLOS errors in the actual environment can come from many aspects, which increases the breadth and depth of the wild dog population search, the implementation speed of the wild dog optimization algorithm will slow down when the wild dog optimization algorithm is set to search and hunt through individuals in the early stage. It may happen that when the iteration reaches the maximum number of times, the optimal solution position cannot be found, resulting in a decrease in the accuracy of the wild dog optimization algorithm. Therefore, an adaptive elimination probability mechanism (Adaptive Elimination Mechanism, AEM) is adopted for the dynamic elimination probability value p. Through the individual information retention search strategy and the group information sharing search strategy, the dynamic elimination probability value p is adaptively adjusted relatively according to different situations, and the optimal search strategy is selected to balance the global and local search allocation capabilities of the wild dog optimization algorithm on different problems, accelerate the convergence speed of the wild dog optimization algorithm, and further improve the positioning accuracy of the optimal target.
[0198] The information search model is used to optimize the search mode in the wild dog optimization algorithm, and the wild dog optimization algorithm search phase is internally developed and mined. In the early stage of the wild dog optimization algorithm, the information storage value in the entire wild dog population is unknown. Each individual has the ability to find the target location, but individuals cannot cooperate with each other, and the overall information degree and development capacity of the information are limited. As the information search model is adjusted and optimized, the optimized information is used to guide individuals to perform subsequent search tasks, thereby improving the mining ability of the wild dog optimization algorithm.
[0199] As a preferred technical solution, in this embodiment, the adaptive elimination probability mechanism in step 5 provides an individual information retention search strategy and a group information sharing search strategy, and adaptively adjusts the value in response to the search demand in different situations, wherein:
[0200] (1) Individual information retention search strategy:
[0201] The core of the search strategy based on individual information retention is that individuals keep their own information private in the group and do not share it with other members of the group. The individual information retention search strategy is regarded as each individual relying only on his own experience to make decisions and actions, rather than based on the knowledge or information shared by the group. The purpose of the individual information retention search strategy is to maintain the independence of the individual, reduce external interference, and increase the diversity of the exploration search space, and continuously optimize the target position by referring to the past values of its own search. The higher the individual information retention rate, the higher the corresponding search and mining ability value of the individual, and vice versa, the lower the individual information retention rate, the lower the corresponding search and mining level of the individual; the individual is an individual in the wild dog population, and the individual information retention rate is defined as:
[0202]
[0203]
[0204] in, represents the individual information retention rate of the individual value n in the wild dog population after the mth iteration, It represents the search ability value of an individual after the mth iteration. The search ability value of each individual is different. When the search ability reaches the lowest and highest value, it is replaced by low and high respectively. represents the amount of information that the individual has collected after the mth iteration;
[0205] Formula (24) can clearly show the relationship between individual information retention rate and search ability value. When the information retention rate is higher, the individual will be stronger in search ability mining; conversely, the lower the information retention rate, the weaker the search mining ability.
[0206] The information added by the individual after each iteration is retained, and the search mining ability value of the individual information retention search strategy is increased. In each iteration, the useful information left by the previous iteration is referred to, and the useful information is retained until the iteration is completed. The target position is obtained by retaining the information each time, so that the search process is distributed in breadth and the search range is wide. The individuals in the wild dog population are represented by x n , the new search position obtained by retaining individual information is shown in formula (25);
[0207]
[0208] in, represents the individual information retention rate of the individual value n in the wild dog population after the mkth iteration;
[0209] Based on the individual information retention process model Figure 7 As shown:
[0210] In each iteration of the wild dog optimization algorithm, only whether the target is searched is used as the selection of the elimination probability value p, and the useful value of the information left by the individual iteration process in the wild dog population is not considered; the key to the individual information retention strategy is to be able to retain these useful information historical values, and will not be eliminated along with the individual as the wild dog optimization algorithm iterates, and make full use of the useful retained information in the search process to speed up the wild dog optimization algorithm to search for the target location; therefore, adding an individual information retention strategy to the wild dog optimization algorithm can ensure that while the individual continuously iterates and updates the location information, the search and mining capabilities are strengthened through historical useful information, so that the entire wild dog optimization algorithm can further improve the search breadth, and improve the performance of the wild dog optimization algorithm in the three-dimensional space problem; the update method of the individual information retention search strategy based on the adaptive elimination rate is shown in formula (26):
[0211]
[0212] Among them, N is the initial number of wild dog population, L is the maximum number of iterations, x g,high Indicates the highest value of information retention in the current individual, r1 and r2 are selected as random numbers in (0,1) respectively;
[0213] (2) Group information sharing search strategy:
[0214] The core of the group information sharing search strategy is to guide individual behavior and decision-making by sharing information about the target location between groups. Through this strategy, individuals can use the collective wisdom of the group to improve their own search strategies, thereby increasing the speed and accuracy of the entire wild dog population in finding the target location during the search phase.
[0215] In this method, individuals in the wild dog population calculate the search rate based on the amount of information they have collected. The closest individuals form a group, which shares information with each other and updates the individual position based on the total amount of group information. The wild dog optimization algorithm can continuously update the new position while searching. In each iteration of the search phase, it absorbs useful information from the group, rather than simply obtaining the target position through the retained information of the individual, further reducing the probability of elimination. At the same time, it also selects and summarizes the information of all individuals in the group, and summarizes the target position with a very high repetition rate, which expands the depth range of the wild dog population search and further improves the accuracy and practicality of the wild dog optimization algorithm in obtaining the target position. The new search position obtained by sharing group information is shown in formula (27):
[0216]
[0217] Among them, S * Represents the new location obtained by group information sharing search after m iterations, and Represents the information retention rate and individual quantity of individuals after M iterations. The group information sharing process model is as follows: Figure 8 As shown:
[0218] The update method of the group information sharing search strategy based on the adaptive elimination rate is shown in formula (28):
[0219]
[0220] At this point, the respective advantages of the individual information retention strategy and the group information sharing search strategy are fully utilized. During the iteration process of the wild dog optimization algorithm, the appropriate search strategy is adaptively selected through different elimination probability values to improve the actual positioning performance of the wild dog optimization algorithm in the NLOS environment.
[0221] In summary, the present invention applies the target tracking mechanism and the AEM-wild dog optimization algorithm to UWB positioning, and obtains a TTDO ultra-wideband positioning wild dog optimization algorithm based on an adaptive elimination rate; the TTDO ultra-wideband positioning wild dog optimization algorithm realizes the positioning of the position of the tag to be tested in the positioning scene, and also autonomously reduces the influence of the NLOS error in the NLOS environment, reduces the positioning error, and obtains better positioning accuracy, which mainly includes:
[0222] Step 1: TDOA method for ranging;
[0223] The base station with known location and the tag with unknown location are set up, and the estimated coordinate information is obtained by the arrival time difference method of sending and receiving signals between two nodes. At this time, due to the influence of NLOS error, the location of the tag cannot be tracked in real time, and the positioning accuracy is poor;
[0224] Step 2: Fitness function calculation
[0225] The position of the unknown node A is estimated by using the wild dog optimization algorithm based on target tracking. Therefore, it is necessary to formulate an accurate fitness function to evaluate the accuracy of the positioning result. In this process, the measured distance between the base station and the tag node to be located and its actual distance obtained in Step 1 are used to calculate the fitness function value, which is used as a benchmark for measuring the performance of the positioning wild dog optimization algorithm.
[0226] Step 3: Initialize data
[0227] The number of wild dog populations is determined. However, since the convergence speed of the wild dog optimization algorithm is affected by the initialization of the wild dog population, if the method of randomly generating the initialization of the wild dog population is adopted, the convergence speed can be prolonged, resulting in a decrease in positioning accuracy. Therefore, by setting a fitness function as a standard, according to the fitness function, an initialization of the wild dog population data N is reversely set, and an improved initialization of the wild dog population is obtained. Starting from the local search, the position with the highest survival rate is continuously updated in real time according to the target tracking, which speeds up the convergence speed of the wild dog optimization algorithm.
[0228] Step 4: Positioning and solving
[0229] The AEM-wild dog optimization algorithm is used to find the optimal solution through the search and hunting stages. The search starts from the local area, and the latest survival rate is continuously updated in real time through the target tracking mechanism. The hunting process of prey is expanded to the global range, and the position will be continuously updated and iterated; through the adaptive elimination probability mechanism, the dynamic elimination probability value p is adaptively adjusted according to different situations through the individual information retention strategy and the group information sharing search strategy, the optimal search strategy is selected, and the global and local search allocation capabilities of the wild dog optimization algorithm on different problems are balanced, the convergence speed of the wild dog optimization algorithm is accelerated, and the positioning accuracy of the optimal target is further improved.
[0230] It should be noted that the test process and results are:
[0231] MATLAB is used as the simulation platform, and the simulation environment adopts a 40m×40m simulated indoor two-dimensional plane scene, such as Fig. 9As shown. A total of 8 base stations are set up, and the location coordinates are (5,5), (25,5), (45,5), (5,25), (45,25), (5,45), (25,45), (45,45), and the main base station is set at (5,5). The NLOS error in the environment is simulated and generated according to the exponential distribution with a delay of τ. 100 test points (brown five-pointed star icons) are randomly generated and distributed in the test area. The upper limit of the search plane coordinate is (45,45), the lower limit is (5,5), and the problem dimension dim is 2.
[0232] This test selects four wild dog optimization algorithms, namely enhanced whale optimization wild dog optimization algorithm (EWOA), adaptive differential evolution particle swarm wild dog optimization algorithm (APSO-DE), TTDO and AEM-TTDO, to test the positioning performance of intelligent wild dog optimization algorithms in NLOS environments. Among them, the focus is on setting up a comparative test between TTDO and AEM-wild dog optimization algorithms to test the positioning accuracy and stability of the wild dog optimization algorithms in the face of NLOS error interference after the introduction of target tracking mechanism and adaptive elimination probability mechanism.
[0233] Fig.10 The comparison of positioning performance in LOS and NLOS environments is shown using RMSE as the evaluation criterion. Fig.10 (a) and (b) show that from LOS environment to NLOS environment, AEM-TTDO and wild dog optimization algorithm show good positioning performance; the minimum RMSE value of AEM-wild dog optimization algorithm in LOS environment is only 0.1357m, which is 0.0427m, 0.2621m and 0.6345m higher than TTDO, APSO-DE and EWOA wild dog optimization algorithms respectively; and the minimum RMSE value in NLOS environment is 0.1836m, which is 0.0612m, 0.3124m and 0.6652m higher than TTDO, APSO-DE and EWOA wild dog optimization algorithms respectively.
[0234] In the NLOS environment, the positioning accuracy deviation of APSO-DE and EWOA wild dog optimization algorithms is relatively serious, while the AEM-wild dog optimization algorithm, which is an improvement on the wild dog optimization algorithm, is significantly more stable in terms of overall positioning error and has a smaller error fluctuation range. Therefore, whether the positioning error variance is small or large, the positioning accuracy of the AEM-wild dog optimization algorithm in the NLOS environment is higher than that of other compared wild dog optimization algorithms.
[0235] Fig.11 The comparison of positioning performance in LOS and NLOS environments is shown using CDF as the evaluation criterion. Figure 4 and Figure 5It is found that the AEM-wild dog optimization algorithm has the highest positioning accuracy in the LOS environment. When the CDF probability is 90%, the positioning errors of the four wild dog optimization algorithms are 0.2642m, 0.3225m, 0.4984m and 0.6424m respectively; while the AEM-wild dog optimization algorithm has the highest positioning accuracy in the NLOS environment. When the CDF is 90%, the positioning errors of the four wild dog optimization algorithms are 0.3554m, 0.4357m, 0.6175m and 0.8128m respectively. In the NLOS environment, the AEM-wild dog optimization algorithm has better stability than the wild dog optimization algorithm. The probability of positioning error within 0.4m is 84% and 66% respectively, and the positioning accuracy is significantly improved. Therefore, the positioning error of the AEM-wild dog optimization algorithm in the NLOS environment is smaller and more stable.
[0236] Aiming at the defects of the UWB positioning stray dog optimization algorithm in the existing indoor positioning technology, such as large positioning error and low positioning accuracy in complex environments, especially in NLOS environments, the present invention selects and designs a target tracking stray dog optimization UWB positioning stray dog optimization algorithm with improved adaptive elimination rate. On the basis of the traditional stray dog optimization algorithm, a target tracking mechanism is introduced in the search stage, and the global optimization ability of the stray dog optimization algorithm is improved by using two strategies, domain solution search and non-inferior solution conversion, to reduce error interference; an adaptive elimination probability mechanism is introduced, and the ability of the stray dog optimization algorithm to mine dimensional information is improved by using individual information retention strategy and group information sharing strategy, and the optimal search space is adaptively formed, so as to further improve the position estimation accuracy and stability of the stray dog optimization algorithm.
[0237] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the claims of the present invention.
Claims
1. A UWB positioning method based on wild dog optimization algorithm, It is characterized in that The following steps are involved: Step 1: Build Base Station B 1 , Base Station B 2 , Base Station B 3 , using the TDOA algorithm to measure the electromagnetic waves emitted from the unknown node A to the base station B 1 ~B 3 The propagation time of the unknown node A to the base station B is calculated according to the propagation speed of electromagnetic waves. 1 ~B 3 The distance between Step 2: From unknown node A to base station B 1 ~B 3 The ranging value is the radius, and a circle is drawn with the corresponding base station as the center. The coordinates of the intersection of the circles are taken as the first coordinates of the unknown node A; Step 3: Build the wild dog optimization algorithm: Assume that the number of the initial wild dog population is N. The position of each individual in the wild dog population will change continuously during the search for prey. The position set of each individual in the wild dog population is X = (x 1 ,x 2 ,...,x n ); In the wild dog optimization algorithm, each individual will perform encirclement, pursuit, scavenging and survival behaviors. Each individual must go through encirclement, pursuit and scavenging behaviors to survive, and the final surviving individual is the target location. The final target location is solved to obtain the second coordinate of the unknown node A; Step 4: Set up the target tracking mechanism: Initialize the number of wild dog populations and the problem dimension, and randomly generate wild dog population samples: Through the search capability of the target tracking mechanism, all solvable targets in the search space are decomposed and the updated statistics of all solvable positions are completed; Step 5: Introduce the adaptive elimination probability mechanism into the wild dog optimization algorithm to obtain the AEM-wild dog optimization algorithm; apply the target tracking mechanism and the AEM-wild dog optimization algorithm as the target tracking wild dog optimization algorithm in the UWB positioning method.
2. The method according to claim 1, It is characterized in that Step 1 calculates the distance from unknown node A to base station B 1 ~B 3 The distance between them is expressed as: Where, d 1 ~d 3 Respectively represent unknown node A to base station B 1 ~B 3 The distance between 1 ,y 1 )、(x 2 ,y 2 )、(x 3 ,y 3 ) represent base station B 1 ~B 3 The coordinate position, (x 0 ,y 0 ) represents the coordinates of the unknown node A.
3. The method according to claim 2, It is characterized in that Step 2 includes: The equation group (1) can be rearranged to obtain: 2(x i x 0 +y i y 0 )=x i 2 +y i 2 +x 0 2 +y 0 2 +d i 2 ,i=1,2,3 (2) Where i represents the i-th base station; the solution of equation (2) is the position of the unknown node A; Under the influence of NLOS error, base station B 1 ~B 3 The circles will not intersect at one point. The least square method is used to estimate the position of the unknown node A. Square both sides of the equation group (2) to obtain the following equation group: Let x 2 +y 2 =C, then formula (3) is expressed in matrix as: MX=N (4) in, According to formula (4), the error vector θ is defined as: θ=MX-N (5) Let the square of the error vector θ be Q, which is expressed as: Q=θ 2 =(MX-N) T (MX-N) (6) The derivatives of both sides of equation (6) are: When the derivative of equation (7) is set to 0, the extreme value is obtained: X=(M T M) -1 M T N (8) The location of unknown node A is determined by B 1 , B 2 and B 1 , B 3 The intersection of the two hyperbolas as the focus is solved to obtain the first coordinate of the unknown node A (d 21 , d 31 ):
4. The method according to claim 1, It is characterized in that In the wild dog optimization algorithm described in step 3, each individual will perform encirclement, pursuit, scavenging and survival behaviors, including: (1) Encirclement In the wild dog optimization algorithm, a certain number of initialized wild dog populations is set, and the initialized wild dog population is placed within the set range to search and update the location of the suspected target. The specific expression is as follows: Among them, k represents the kth individual, represents the kth individual at time d; represents the i-th individual at time d; represents the current best individual at time d; na represents the number of wild dog individuals surrounding the prey in the wild dog population, which is a random integer in [2,N / 2]; is the distance between na wild dogs and the target object; β is a uniform random number between [-2,2], and we get The updated new position is expected to be -2X best ;X best represents the best individual at present; (2) Pursuit behavior The pursuit behavior is to guide the wild dog optimization algorithm to find the relatively optimal solution by simulating the pursuit strategy. After the individual surrounds the suspected target, it searches near the local optimal individual position. The formula is as follows: Among them, β 1 is a uniform random number between [-2,2], β 2 is a uniform random number between [-1,1], is a random individual in a wild dog population; (3) Degradation behavior Through the hunted target, search again near the relatively better individuals; the formula is as follows: Among them, the coefficient β in formula (13) 2 The value is a random number in [-2,2], and the value of coefficient b is a random number in [-1,1]; (4) Survival After completing the scavenging behavior, wild dogs will still face the problem of survival. The survival probability of each wild dog is calculated based on its survival function F(X). Wild dog individuals with low survival rates will repeat the encirclement, pursuit, and scavenging behaviors and search for the target location again to increase the final survival probability of the wild dogs.
5. The method according to claim 4, It is characterized in that The survival probability of the individual, that is, the final survival probability value, is calculated by formula (14). If the individual survival probability is less than 30%, the individual re-estimates the new position using formula (15). After the relevant positions are updated, the survival probability is calculated again. Two different individuals in the group are randomly selected. and Ensure that the survival probability of each wild dog reaches more than 70%; In the formula, and represents two different individuals in a group; F(X), F(X i ) represent the survival function of individuals; ε represents the fitness function in solving the survival probability.
6. The method according to claim 5, It is characterized in that Step 4 decomposes all solvable targets in the search space and completes the updated statistics of all solvable positions, including: The statistical process of all solvable positions is as follows: Where N represents the wild dog population, X n (t) represents all the solvable positions within the initial range, and the domain solution optimization strategy and non-inferior solution conversion strategy are selected and used according to all the solvable positions; a) Domain solution optimization strategy: The mathematical model of the neighborhood solution optimization strategy is as follows: Among them, D 1 ′,D′ 2 , D 3 ′ represents the three individuals closest to the relatively optimal solution in the current range; Therefore, the position update of the wild dog optimization algorithm based on the neighborhood optimization strategy in the search phase is adjusted to: Among them, λ 1 ,λ 2 ,λ 3 represents the control parameter, X(t+1) represents the latest solution position obtained through the three best solutions in the neighborhood; b) Non-inferior solution conversion strategy Get five solutions B~F within the range of the wild dog optimization algorithm, perform an independent search near the five solutions B~F, find a non-inferior solution with a small error in the fitness value of each individual, convert through the non-inferior solution, search near the non-inferior solution to find a better solution, when the wild dog optimization algorithm performs a certain number of iterations, obtain five new best non-inferior solutions B near the initial five solutions that are closer to the current optimal point A* 1 ~F 1 ; The best non-inferior solution B 1 ~F 1 Closer to the current best point A*, the best non-inferior solution and the current best point A* are grouped together to conduct a new search to determine the best non-inferior solution B 1 ~F 1 Comparing with the optimal solution in the current best point A*, we can get: Xb i (t)-X c (t)<α[Xb(t)-X c (t)] (20) α=log 0.5 (-t / L) (21) Among them, α is the update factor, Xb i (t) is the position of the i-th non-inferior solution obtained by the wild dog optimization algorithm after t iterations, Xb(t) is the position of the initial non-inferior solution, and X c (t) is the current optimal solution position; the non-inferior solution after t iterations is compared with the current optimal point A*, and the best non-inferior solution B 1 ~F 1 The optimal solution position is obtained by converting the current best point A*: Among them, X i (t+1) is the position of the ith optimal solution after t iterations of the wild dog optimization algorithm; k(t) is the standard normal distribution conversion coefficient at the t iteration; L is the maximum number of iterations; Therefore, by introducing the non-inferior solution conversion strategy, the conversion between the non-inferior solution and the current best point A* is found and the best point in the new non-inferior solution replaces the current best point A*.
7. The method according to claim 6, It is characterized in that The adaptive elimination probability mechanism in step 5 provides an individual information retention search strategy and a group information sharing search strategy, and adaptively adjusts the value in response to the search demand in different situations, where: (1) Individual information retention search strategy: The higher the individual information retention rate, the higher the corresponding search and mining capability value of the individual. Conversely, the lower the individual information retention rate, the lower the corresponding search and mining level of the individual. The individual is an individual in the wild dog population, and the individual information retention rate is defined as: in, represents the individual information retention rate of the individual value n in the wild dog population after the mth iteration, It represents the search ability value of an individual after the mth iteration. The search ability value of each individual is different. When the search ability reaches the lowest and highest value, it is replaced by low and high respectively. represents the amount of information that the individual has collected after the mth iteration; Keep the information added by the individual after each iteration. The individuals in the wild dog population are represented by x n , the new search position obtained by retaining individual information is shown in formula (25); in, represents the individual information retention rate of the individual value n in the wild dog population after the mkth iteration; The update method of the individual information preservation search strategy based on the adaptive elimination rate is shown in formula (26): Among them, L is the maximum number of iterations, x g,high Indicates the highest value of information retention in the current individual, r 1 and r 2 Select random numbers in (0,1) respectively; (2) Group information sharing search strategy: The closest individuals form a group, and the groups share information with each other. The individual position is updated according to the total amount of group information. The new search position obtained by group information sharing is shown in formula (27): Among them, S * Represents the new location obtained by group information sharing search update after m iterations, and Represents the information retention rate and individual quantity of individuals after M iterations; The update method of the group information sharing search strategy based on the adaptive elimination rate is shown in formula (28):
Citation Information
Cited By
Multi-unmanned aerial vehicle cooperative power grid detection method and device
CN119624069A
