Low-power Bluetooth anchor node deployment method based on improved gradient optimization algorithm

Abstract

The present invention proposes a method for deploying low-power Bluetooth anchor nodes based on an improved gradient optimization algorithm, including: determining the number of randomly deployed positioning anchor nodes in a positioning area; initializing the IGBO algorithm based on the number of randomly deployed positioning anchor nodes; introducing a low-difference sequence in the positioning area to randomly generate a number of different target deployment schemes based on the initialized IGBO algorithm; calculating the evaluation performance index value of the target deployment scheme and sorting them; iterating and looping all the sorted anchor node deployment schemes to obtain the optimal anchor node deployment scheme. The present invention provides a mathematical basis for the reasonable deployment of anchor nodes, increases the effective area range of Bluetooth positioning, and improves the positioning accuracy of BLE. This solution provides a more accurate and efficient solution for the field of indoor positioning, and meets the continuous pursuit of modern technology for the accuracy and efficiency of positioning systems.

Description

Technical Field

[0001] The present invention belongs to the field of indoor positioning technology, and in particular relates to a low-power Bluetooth anchor node deployment method based on an improved gradient optimization algorithm. Background Art

[0002] Existing indoor positioning technologies cover a variety of methods, including wireless LAN (Wi-Fi) positioning, ultra-wideband (UWB) positioning, BLE positioning, Zigbee positioning, etc. In this technical field, BLE positioning technology has become the focus of attention due to its unique advantages. First, BLE can operate for a long time due to its low standby power consumption, high-speed connection and low peak power consumption, without frequent battery replacement or charging, ensuring the stability and reliability of the system, which is very suitable for indoor positioning devices. Secondly, compared with other methods, the deployment of BLE positioning system is simpler. It does not require large-scale renovation of buildings, so it can be more easily integrated into existing building structures. BLE hardware is relatively cheap, and this low-cost feature makes it possible to deploy BLE indoor positioning systems on a large scale with limited budgets. In addition, BLE performs well in providing sufficient indoor navigation and location awareness. Although it may not be as high as the ultra-high accuracy provided by some other positioning technologies (such as UWB), in many application scenarios, the positioning accuracy of BLE is sufficient to meet the needs. Since BLE is widely used in many mobile devices, communication between devices is more convenient, which also provides more possibilities for indoor positioning. In general, BLE positioning technology has obvious advantages in low power consumption, cost-effectiveness, accuracy and universality, and is a technology with great promotion prospects in the field of indoor positioning.

[0003] In BLE indoor pedestrian positioning, the reasonable deployment of anchor nodes is a key issue. The goal is to minimize positioning blind spots in the positioning area, while ensuring positioning accuracy to a certain extent and avoiding waste of resources caused by excessive deployment. The current common practice is to arrange anchor nodes in the positioning area in a regular form (such as triangulated networks, grid networks or regular hexagons) or randomly based on experience. However, this deployment method has challenges. Too sparse may lead to an increase in positioning blind spots, thereby affecting positioning accuracy, while too dense may cause waste of resources. To address this problem, it is urgent to propose an anchor node deployment scheme based on mathematical basis, which provides mathematical support for the reasonable deployment of anchor nodes, thereby improving the accuracy of BLE positioning. Summary of the invention

[0004] The purpose of the present invention is to solve the problem of high subjectivity of anchor node deployment scheme in BLE indoor positioning, and provide a deployment scheme of anchor nodes in BLE indoor positioning. Specifically, the deployment method provided by the technical solution of the present invention is based on an improved gradient optimization algorithm (Improved Gradient-based Optimizer, IGBO), and the specific operation is to introduce a low-difference sequence to initialize the population, and combine the gradient search rule (Gradient search rule, GSR) and the local escaping operator (Local escaping operator, LEO) in the gradient optimization algorithm to determine the optimal deployment scheme of the anchor node, help the algorithm converge to a better solution, so that the coverage range of the anchor node is optimized, the effective area range of Bluetooth positioning is increased, and the accuracy of BLE positioning is further improved.

[0005] To achieve the above object, the present invention provides a method for deploying low-power Bluetooth anchor nodes based on an improved gradient optimization algorithm, comprising:

[0006] Determine the number of randomly deployed positioning anchor nodes in the positioning area;

[0007] Initialize the IGBO algorithm based on the number of randomly deployed positioning anchor nodes;

[0008] Based on the initialized IGBO algorithm, a low-discrepancy sequence is introduced in the positioning area to randomly generate several different target deployment schemes;

[0009] Calculate the evaluation performance index values ​​of the target deployment schemes and sort them;

[0010] All sorted anchor node deployment plans are iterated and cycled to obtain the optimal anchor node deployment plan.

[0011] Optionally, determining the initial deployment number of anchor nodes in the positioning area using a regular hexagonal deployment method includes:

[0012] Initializing the positioning area and generating a grid;

[0013] Traversing four quadrants in the grid, gradually placing points in each quadrant;

[0014] The placement points outside the positioning area and in the obstacle area are deleted, and the upper limit of the number of anchor nodes is preset, and finally the initial number of anchor nodes deployed in the positioning area is determined.

[0015] Optionally, initializing the IGBO algorithm includes: initializing the IGBO algorithm parameters; the parameters include: population number np, maximum number of iterations Maxit, probability parameter pr, discrete granularity dr, perception radius R s, dimension dimension, redundant deployment penalty function parameters N1, N2, δ, n, and the observation value delta_nums required for network similarity.

[0016] Optionally, introducing a low-discrepancy sequence to randomly generate several different anchor node deployment schemes includes:

[0017] S1. Determine the dimension and cardinality of the low-discrepancy sequence in the positioning region, and obtain the van der Kopit sequence of each dimension in the low-discrepancy sequence;

[0018] S2. performing coordinate transformation on the van der Kopit sequence to obtain a sequence that conforms to the actual target positioning area;

[0019] S3. Pairing the van der Kopit sequence and the sequence that matches the actual target positioning region in order to generate a sequence of anchor node coordinates, which is an initial deployment plan for the anchor nodes;

[0020] S4. Repeat S1-S3 and introduce an obfuscation processing method to generate several different anchor node deployment schemes.

[0021] Optionally, the evaluation performance index value is:

[0022] f=k1*f1+k2*f2+k3*f3+k4*f4+k5*f5

[0023] Among them, k1, k2, k3, k4 and k5 are coefficients, f1 is the proportion of unlocatable nodes to all target nodes, f2 is the average GDOP value of locatable nodes, f3 is the penalty term, f4 is the sensor network topology evaluation, and f5 is the network uniformity.

[0024] Optionally, iterating and looping all the sorted anchor node deployment solutions includes:

[0025] Using the gradient search rule, the sorted deployment plan is updated to obtain an updated deployment plan;

[0026] Creating a preset deployment plan based on preset conditions, and determining whether to continue updating the deployment plan using the gradient search rule based on the preset deployment plan to obtain all updated deployment plans;

[0027] Calculate the evaluation performance index values ​​of all updated deployment plans to obtain the optimal deployment plan and the worst deployment plan;

[0028] Determine whether the update iteration reaches the preset maximum number of iterations. If so, output the anchor node deployment plan with the lowest evaluation performance index value, that is, the optimal anchor node deployment plan. Otherwise, continue to update the deployment plan according to the gradient search rule.

[0029] Optionally, the method for updating the sorted deployment scheme using the gradient search rule is:

[0030]

[0031] Where randn is a random number with normal distribution, GSR is the gradient search rule, Δx is the difference between the two deployment schemes, and x i is the target positioning node, x worst is the worst solution for anchor node deployment, x best is the optimal solution for the anchor node deployment scheme, and ε is a small value in the range of [0,0.1].

[0032] Optionally, the updated deployment scheme is:

[0033]

[0034] in, To update the deployment plan after iteration, r1 and r2 are two random numbers in the range of [0, 1]. and They are all different deployment schemes generated in the GSR stage, it is the number of iterations, and i is the index of a deployment location of the anchor node deployment scheme.

[0035] Optionally, the updated full deployment scheme is:

[0036]

[0037] in, is a randomly selected deployment solution x P is a random deployment scheme, u2 is a random number in the range [0,1], A deployment scheme is randomly selected.

[0038] The present invention has the following beneficial effects:

[0039] Compared with traditional technologies, the significant advantage of the present invention is that it adopts the IGBO algorithm and optimizes the deployment of anchor nodes, thereby solving the shortcomings of existing solutions and greatly improving the performance of the positioning system. Based on the original gradient optimization algorithm, this technical solution introduces a low-difference sequence to optimize the initialization process of the population, thereby enhancing the search ability of the algorithm and improving the convergence speed of the algorithm. This improvement not only enables the algorithm to reach the optimal solution faster, but also significantly increases the coverage of the anchor nodes of the Bluetooth positioning system. Through the optimized deployment scheme, the present invention achieves the dual goals of Bluetooth indoor positioning: high precision and wide coverage. The accuracy of the positioning system has been significantly improved, and the effective area has been expanded, so that the indoor positioning system can still operate stably even in complex environments. The present invention provides a more accurate and efficient solution for the field of indoor positioning, meeting the continuous pursuit of modern technology for the accuracy and efficiency of positioning systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0041] Figure 1 A schematic diagram of an anchor node coverage optimization process based on the IGBO algorithm according to an embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram of the principle of a regular hexagonal deployment solution according to an embodiment of the present invention;

[0043] Figure 3 is a k-degree coverage diagram of an embodiment of the present invention;

[0044] Figure 4 Schematic diagram of the obstacle perception model principle of an embodiment of the present invention. DETAILED DESCRIPTION

[0045] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0046] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0047] like Figure 1 The embodiment shown in the figure proposes a low-power Bluetooth anchor node deployment method based on an improved gradient optimization algorithm; including:

[0048] Determine the number of initially deployed positioning anchor nodes in the positioning area;

[0049] Specifically, in this embodiment, the number of initially deployed positioning anchor nodes is determined in the positioning area using a regular hexagonal deployment method.

[0050] According to the number of initial deployment anchor nodes, the initial IGBO algorithm is used to introduce a low-discrepancy sequence to randomly generate several different deployment schemes.

[0051] Calculate and rank the evaluation performance index values ​​of several different anchor node deployment schemes;

[0052] Specifically, in this embodiment, after sorting, the deployment schemes of the best solution and the worst solution are recorded;

[0053] The deployment plans of all the sorted plans are input into the initialized IGBO algorithm, and an iterative cycle is performed to obtain the optimal anchor node deployment plan.

[0054] Specifically, in this embodiment, the input of all the sorted anchor node deployment schemes into the initialized IGBO algorithm includes two stages, as follows:

[0055] Entering the iterative loop, in the gradient search rule (GSR) stage, the area where a better solution may be found in the anchor node deployment solution space is explored to obtain an updated deployment plan.

[0056] In the local escaping operator (LEO) stage, the algorithm's ability to escape the local solution coordinate value is improved, and it is determined whether randn is greater than pr. If so, it continues to update to obtain a new deployment plan. If not, it continues to use the updated deployment plan generated in the gradient search rule stage.

[0057] Furthermore, the low-power Bluetooth anchor node deployment method based on the improved gradient optimization algorithm proposed in this embodiment includes step 1: determining the initial deployment number of anchor nodes when the boundary size of the positioning area and internal obstacles are determined. That is, randomly deploying positioning anchor nodes in the positioning area includes:

[0058] Initializing the positioning area and generating a grid;

[0059] Traversing four quadrants in the grid, gradually placing points in each quadrant;

[0060] The placement points outside the positioning area and in the obstacle area are deleted, and the upper limit of the number of anchor nodes is preset, and the initial deployment number of anchor nodes is determined in the positioning area.

[0061] Furthermore, step 1: Determine the initial number of anchor nodes to be deployed after determining the boundary size and internal obstacles of the positioning area. First, determine the boundary size of the positioning area (lb, ub), where lb represents the lower boundary of the positioning area and ub represents the upper boundary of the positioning area; the size of the internal obstacles in the area obs = {ob1, ob2, ..., ob m}, ob i =(lo,uo), where lo represents the lower bound of the obstacle area, uo represents the upper bound of the obstacle area, and i = {1,2,…,m}; then, the number of anchor nodes nums is preliminarily determined according to the boundary of the positioning area and the size of the obstacle. The method for determining the initial number of anchor nodes generally adopts a regular hexagonal deployment scheme, as shown in the schematic diagram. Figure 2 As shown, the specific implementation steps are as follows:

[0062] (1) Initialization and mesh generation

[0063] First, define the entire drawing area, calculate the starting deployment position of the dot matrix and the step length of each step on the x-axis and y-axis. The above calculation takes into account the expected spacing between points to ensure that each point is evenly distributed.

[0064] (2) Quadrant traversal

[0065] Traverse the four quadrants and gradually place points in each quadrant. Start from the center point of the region and expand toward the border of each quadrant according to the defined direction.

[0066] (3) Boundary observation

[0067] Each newly added point is checked to see if it exceeds the defined region boundary or enters the obstacle region. If so, the point may be moved outside the obstacle boundary or deleted completely.

[0068] In addition, considering the economic cost, it is necessary to set an upper limit on the number of anchor nodes, which can be mathematically described as:

[0069] nums=n*m / 6 (1)

[0070] Then, nums anchor nodes are randomly deployed in the positioning area, denoted as AP = {ap1, ap2, ..., ap nums}.

[0071] Furthermore, step 2: initializing the parameters of the IGBO algorithm and the related parameters in the evaluation performance index.

[0072] Specifically, step 2: Initialize the parameters of the IGBO algorithm and the relevant parameters in the evaluation performance index, including the population number np, the maximum number of iterations Maxit, the probability parameter pr, the discrete granularity dr, and the perception radius R s, dimension dimension, redundant deployment penalty function parameters N1, N2, δ, n, and the observation value delta_nums required for network similarity. Among them, the population number, maximum number of iterations, and probability parameter are IGBO algorithm parameters, and the rest are related parameters in the evaluation performance index. The population number corresponds to the number of deployment schemes.

[0073] Furthermore, in this embodiment, step 3: randomly generate np different deployment schemes using a low-discrepancy sequence.

[0074] Specifically, step 3: randomly generate np different deployment schemes using low-discrepancy sequences. The low-discrepancy sequence method, also known as the quasi-Monte Carlo method, is a method that uses a deterministic quasi-random number sequence instead of a pseudo-random number sequence. It can generate a low-discrepancy distribution sequence with super-uniform properties. Low discrepancy means high uniformity, and a high-uniformity initial population can improve the exploration ability of the swarm intelligence algorithm and help the algorithm converge to a better solution.

[0075] Among them, the Halton sequence is a generalization of the VDC (Van der Corput) sequence in the multidimensional case. The VDC sequence samples the one-dimensional interval [0,1] so that one point in the sequence is as far away from the next point as possible. Suppose you want to generate N samples of the VDC sequence with base b (where b is a prime number), you can get the nth element in two simple steps:

[0076] Use formula (5) to obtain the coefficient a of the binary expression representing the integer n in base b: i .

[0077] n=a0+a1b+a2b 2 +…+a r b r (5)

[0078] a i ∈[0,b-1]

[0079] Through a i The coefficients, the nth element of the sequence can be calculated using the formula Φ(n).

[0080]

[0081] Halton sequence is a VDC sequence with different main bases on each axis. In the case of D dimensions, the first dimension in the Halton sequence is a VDC sequence based on 2, the second dimension is a VDC sequence based on 3, and so on. The Dth dimension is a VDC sequence based on the Dth prime number. Therefore, the specific steps for generating the initial deployment plan of anchor nodes are as follows:

[0082] Step 3.1: Determine the dimension and cardinality of the Halton sequence. For the positioning requirements of the two-dimensional positioning area, the dimension is set to 2. Two different prime numbers 2 and 3 are selected as cardinality. This selection is based on the characteristics of prime numbers and can generate a uniformly distributed point sequence in a multidimensional space. In this embodiment, the Halton sequence is used, which is a kind of low-discrepancy sequence.

[0083] Step 3.2: Determine the VDC sequence of each dimension. When the cardinality is 2, convert the integer n into binary according to formula (5), then reverse it and place it after the decimal point, and convert it into decimal form according to formula (6), so that the nth element of the sequence is obtained as follows. The full name of the VDC sequence is Van der Corput, which means Van der Corput sequence in Chinese.

[0084]

[0085] Similarly, when the base is 3, by converting the integer n to ternary, performing the same inversion and decimal conversion process, the corresponding sequence is as follows,

[0086]

[0087] Step 3.3: Calculate the specific deployment location of the anchor node according to the actual deployment range. Assume that the horizontal range of the target positioning area is [x min ,x max ]. The i-th quasi-random number generated by the Halton sequence is R i ∈[0,1], the horizontal coordinate of the deployment node can be expressed as:

[0088] x i =x min +(x max -x min )×R i (7)

[0089] The vertical coordinate conversion method is the same as above. According to the above formula, we can convert the basic VDC sequence into a sequence that conforms to the actual target positioning area.

[0090] Step 3.4: Generate an anchor node deployment plan. Pair the two adjusted sequences in order to generate a sequence whose elements are anchor node coordinates, which is an initial deployment plan for the anchor nodes. This is denoted as AP i .

[0091] Step 3.5: Generate deployment plans for all anchor nodes. Repeat the above steps and introduce obfuscation to generate np different deployment plans. Therefore, the initial deployment plan can be recorded as S = {AP1, AP2, …, AP np}.

[0092] Furthermore, in this embodiment, step 4: calculate the overall evaluation performance index value, sort by the evaluation performance index value, and record the deployment plans corresponding to the optimal solution and the worst solution.

[0093] Specifically, in this embodiment, the evaluation performance index of anchor node deployment can evaluate the advantages and disadvantages of the deployment scheme from different angles. The following are five evaluation indexes defined in this embodiment, including localizability rate, geometric precision factor, redundant deployment penalty item, sensor network topology evaluation, and network uniformity:

[0094] The locatable rate is the ratio of locatable measurement nodes to all measurement nodes. A locatable measurement node refers to a measurement node that can be observed by K anchor nodes, that is, a measurement node that meets K-degree coverage. The K value is set according to the actual scenario and the positioning method used. Figure 3 This is a coverage map based on four anchor nodes, where the coverage of areas S1, S2, S3, and S4 are 1, 2, 3, and 4 respectively.

[0095] Define T i It is a binary variable, indicating whether the measurement node can be located. If the measurement node x i If it can be located, the value is 1, otherwise it is 0. Add the measurement nodes that can be located to the set M T As shown below:

[0096]

[0097] M T ={x i |x i ∈M,T i =1} (9)

[0098] Where M represents the set of all measurement nodes after the positioning area is discretized. Is a collection The number of anchor nodes in is expressed as follows:

[0099]

[0100] Where N is the number of anchor nodes in the positioning area.

[0101] According to the above formula, the unlocatable rate can be obtained as follows:

[0102]

[0103] Among them, |M| and |M T | respectively represent the set of all target points M and the set of locatable nodes M T The number of nodes.

[0104] Geometric Dilution of Precision (GDOP). The quality of anchor deployment is evaluated by the geometric position of the target node and the locatable anchor node. Assuming that the measurement node x i The set of locatable anchor nodes is represented as, then GDOP is the target node x i and collection Internal anchor node b j The function of the angle between is expressed as follows:

[0105]

[0106] A js =|sin(θ j -θ s )|

[0107] Among them, θ j It is beacon b j and the target node x i The angle between Is a collection The number of beacons in the .

[0108] According to the above formula, the average GDOP value of the locatable nodes can be obtained as follows:

[0109]

[0110] Among them, |M T | represents the set of locatable nodes M T The GDOP value is obtained by solving the above formula (12).

[0111] Redundant deployment penalty term. The Boolean perception model is a commonly used anchor node perception model for anchor node deployment solutions. This perception model is an ideal perception model. Due to various interference factors in the experimental environment (such as multipath effects, non-line-of-sight propagation, etc.), the distance value obtained by RSSI ranging method has a large deviation from the actual value, further affecting the accuracy of indoor positioning. In order to solve this problem, an obstacle perception model is introduced. Its basic principle is to connect the measurement node and the anchor node. If the connection is blocked by at least one obstacle, it is considered that the anchor node is not "observed" by the measurement node, that is, the anchor node and the measurement node are non-line-of-sight (NLOS) connected, then this anchor node cannot be used as the positioning anchor node of the measurement node. Of course, if there is no obstacle between the anchor node and the measurement node, that is, it is a line-of-sight (LOS) connection, then this anchor node can be used as the positioning anchor node of the measurement node. Its principle diagram is as follows Figure 4 shown.

[0112] Considering that some measurement nodes will have a large number of corresponding locatable anchor nodes, while other measurement nodes cannot be located due to insufficient anchor nodes, in order to balance the number of locatable anchor nodes of the measurement nodes, the penalty function is designed as follows:

[0113]

[0114] Among them, N1 and N2 represent the measurement nodes x respectively. i The minimum and maximum number of anchor nodes required for positioning, ξ and ξ′ respectively penalize the unlocatable area and over-locatable area. The value of n affects the intensity of the penalty when the number of locatable anchor nodes is large. The larger n is, the weaker the penalty is.

[0115] According to the above formula, the penalty term can be obtained as follows:

[0116]

[0117] Where |M| represents the number of anchor nodes in the set M of all measured nodes, and δ is the balance coefficient of the penalty term.

[0118] Sensor network topology evaluation. The regular hexagonal topology is usually regarded as an ideal coverage model. In order to measure the similarity between the anchor node deployment topology and the ideal regular hexagonal topology, the definition of PCD value is introduced:

[0119]

[0120]

[0121] Among them, Ω is the network topology to be solved, that is, the actual deployment result, H represents a standard regular hexagonal network structure, r T is the radial distance. pc (Ω,Ω H ) value is smaller, indicating a higher similarity to the regular hexagonal structure.

[0122] Network uniformity. The average standard deviation U of the distance between nodes is a method to evaluate the uniformity of anchor nodes in a deployment scheme. This method reflects the coverage of the network by analyzing the distribution of nodes in a specific area, that is, the average standard deviation:

[0123]

[0124] Among them, U i represents the average standard deviation of the distance between anchor node i and its adjacent anchor nodes, N represents the number of anchor nodes, and K i represents the set of adjacent anchor nodes of anchor node i, D ijIndicates the distance between anchor nodes i and j, M i It represents the average distance between anchor node i and its adjacent anchor nodes.

[0125]

[0126] In summary, the final evaluation performance indicators are as follows:

[0127] f=k1*f1+k2*f2+k3*f3+k4*f4+k5*f5 (19)

[0128] Among them, k1, k2, k3, k4 and k5 are coefficients, f1 is the proportion of unlocatable nodes to all target nodes, f2 is the average GDOP value of locatable nodes, f3 is the penalty term, f4 is the sensor network topology evaluation, and f5 is the network uniformity. The smaller the six performance evaluation indicators: the unlocatable rate, the average GDOP value of locatable nodes, the penalty term, the sensor network topology evaluation and the network uniformity, the better. Therefore, the smaller the weighted sum of the evaluation performance indicator values, the better the anchor node deployment performance.

[0129] Finally, the deployment schemes corresponding to the optimal solution and the worst solution of the evaluation performance index value are recorded respectively, that is, the anchor node deployment scheme corresponding to the optimal solution is represented as AP best , the anchor node deployment scheme corresponding to the worst solution is represented by AP worst .

[0130] Further, step 5: enter the iterative loop, in the GSR stage, explore the area where a better solution may be found in the anchor node deployment solution space. In GBO, the gradient search rule (GSR) updates the optimal deployment solution in step 4 to a better deployment solution and explores the area where a better solution may be found in the anchor node deployment solution space. The definition of GSR is:

[0131]

[0132] in:

[0133]

[0134] Among them, u i+1 is a new deployment scheme generated in GBO, and randn is a random number with normal distribution. worst and x best They represent the worst deployment plan and the best deployment plan obtained in the iterative optimization process, and ε represents a small value in the range of [0,0.1]. DM represents the direction of movement, and its formula is:

[0135]

[0136] Among them, ρ2 is an adaptive parameter, which is the number of iterations. Δx refers to the difference between the "two deployment schemes" and is defined as follows:

[0137] Δx=rand×|χ| (24)

[0138] in:

[0139]

[0140]

[0141] Among them, a1, a2, a3 and a4 (a1≠a2≠a3≠a4≠i) are integers randomly selected from [1,N], and χ represents the step size.

[0142] GBO uses GSR and DM to create a new deployment plan in each iteration of the optimization process. Therefore, the new deployment plan The calculation of is as follows:

[0143]

[0144]

[0145] Among them, ρ1 and ρ2 are two adaptive proportional operators. These operators are used to balance exploration and utilization and explore areas in the anchor node deployment solution space where better solutions may be found. They are defined as follows:

[0146]

[0147] Among them, γ min and γ max are the minimum and maximum values ​​of γ, respectively, which are determined by the user. Maxit represents the maximum number of iterations.

[0148] GBO uses another deployment to augment the local search. In creating this deployment, the global optimal deployment plays an important role. In this case, the deployment can be calculated as:

[0149]

[0150] Therefore, based on the deployment plan and Deployment plan after the next iteration is defined as:

[0151]

[0152] in:

[0153]

[0154] Where r1 and r2 are two random numbers in the range [0,1].

[0155] According to formula (26), formula (28), formula (30) and formula (29), the updated deployment plan is obtained.

[0156] Furthermore, in this embodiment, step 6: in the LEO stage, the ability of the algorithm to escape from the local optimal deployment plan is improved, and a suitable deployment plan is created according to the condition rand < 0.5 Determine whether rand is greater than 0.5. If so, continue to update and obtain a new deployment plan. Otherwise, continue to use the updated deployment plan generated in step 5. The implementation process is as follows:

[0157]

[0158]

[0159] Where f1 is a uniform random number in the range [-1,1], f2 is a normally distributed random number with mean 0 and standard deviation 1, pr is the probability, and u1, u2, and u3 are three random numbers defined as follows:

[0160]

[0161] Here, rand is a random number in the range [0,1].

[0162] To specify the deployment scheme in equation (32) The following schemes are defined:

[0163]

[0164] x P =X min +rand(0,1)×(X max -X min ) (35)

[0165] where x P is a random deployment scheme. is a randomly selected deployment plan (p∈[1,2,…,N]), and u2 is a random number in the range [0,1].

[0166] Furthermore, in this embodiment, step 7: calculate the evaluation performance index value corresponding to the deployment scheme of the individual update, and determine whether to perform the update operation. If so, update the optimal solution AP best And the worst solution AP worst .

[0167] Further, in this embodiment, step 8: determine whether the loop reaches the maximum number of iterations Maxit, if so, jump out of the iteration loop and output the anchor node deployment plan with the lowest evaluation performance index value, otherwise, it=it+1, return to step 5.

[0168] This embodiment proposes an anchor node deployment scheme based on mathematical basis, which provides mathematical support for reasonable anchor node layout, enhances the utilization rate of anchor nodes, and thus improves the accuracy of BLE positioning.

[0169] This embodiment proposes an anchor node deployment scheme, which involves using a mathematical model to describe the propagation characteristics of BLE signals, taking into account factors such as walls and obstacles. An optimization algorithm can be used to determine the best anchor node deployment scheme to minimize the coverage area or maximize the positioning accuracy. Overall, this mathematically based anchor node deployment scheme is expected to provide a more scientific and effective method to overcome the problems that may be faced by traditional deployment methods and improve the performance of the BLE indoor pedestrian positioning system.

[0170] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A low-power Bluetooth anchor node deployment method based on an improved gradient optimization algorithm, characterized in that: include: Determine the number of randomly deployed positioning anchor nodes in the positioning area; Initialize the IGBO algorithm based on the number of randomly deployed positioning anchor nodes; Based on the initialized IGBO algorithm, a low-discrepancy sequence is introduced in the positioning area to randomly generate several different target deployment schemes; Calculate the evaluation performance index values ​​of the target deployment schemes and sort them; Iterate all the sorted target deployment plans to obtain the optimal anchor node deployment plan; Iterate all the sorted target deployment plans to obtain the optimal anchor node deployment plan, including: S1: Enter the iteration loop. In the GSR stage, explore the area where a better solution may be found in the anchor node deployment solution space. In IGBO, the gradient search rule updates the optimal deployment solution of all the sorted target deployment solutions to a better deployment solution, and explores the area where a better solution may be found in the anchor node deployment solution space. The definition of GSR is: in: Among them, rand is a random number in the range [0,1], GSR is the gradient search rule, and u i+1 is the new deployment scheme generated in IGBO, randn is a random number with normal distribution, Δx is the difference between the two deployment schemes, and x i is the target positioning node, x worst and x best They represent the worst deployment plan and the best deployment plan obtained in the iterative optimization process, ε represents a small value in the range of [0,0.1], and DM represents the direction of movement. The formula is: Among them, ρ2 is an adaptive scaling operator, which is the number of iterations; Δx is defined as follows: Δx=rand×|χ| (24) in: Where a1, a2, a3 and a4 (a1≠a2≠a3≠a4≠i) are integers randomly selected from [1,N], and χ represents the step size; IGBO uses GSR and DM to create a new deployment plan in each iteration of the optimization process; therefore, the new deployment plan The calculation of is as follows: Among them, ρ1 and ρ2 are two adaptive proportional operators. ρ1 and ρ2 are used to balance exploration and utilization and explore the area where a better solution may be found in the anchor node deployment solution space. They are defined as follows: Among them, γ min and γ max are the minimum and maximum values ​​of γ, respectively, which are determined by the user, and Maxit represents the maximum number of iterations; IGBO uses another deployment to augment the local search; the global optimal deployment plays an important role in creating this deployment; in this case, the deployment is calculated as: Therefore, based on the deployment plan and Deployment plan after the next iteration is defined as: in: in, To update the deployment plan after iteration, r1 and r2 are two random numbers in the range [0,1]. and They are all different deployment schemes generated in the GSR stage, it is the number of iterations, and i is the index of a deployment location of the anchor node deployment scheme; According to formula (26), formula (28), formula (30) and formula (29), an updated deployment plan is obtained; S2: In the LEO stage, improve the algorithm's ability to escape the local optimal deployment plan and create a suitable deployment plan based on the condition rand < 0.5 Determine whether rand is greater than 0.

5. If so, continue to update and obtain a new deployment plan; otherwise, continue to use the updated deployment plan generated by S1. The implementation process is as follows: Where f1 is a uniform random number in the range [-1,1], f2 is a normally distributed random number with mean 0 and standard deviation 1, pr is the probability, and u1, u2, and u3 are three random numbers defined as follows: Among them, rand is a random number in the range of [0,1]; To specify the deployment scheme in equation (32) The following schemes are defined: x P =X min +rand(X max -X min ) (35) Among them, x P is a random deployment scheme. is a randomly selected deployment plan, P∈[1,2,…,N]; S3: Calculate the evaluation performance index value corresponding to the deployment scheme of the individual update, and determine whether to perform this update operation; if so, update the optimal solution AP best And the worst solution AP worst ; S4: Determine whether the loop reaches the maximum number of iterations Maxit. If so, exit the iteration loop and output the anchor node deployment plan with the lowest evaluation performance index value. Otherwise, it=it+1, and return to S1.

2. The method for deploying low-power Bluetooth anchor nodes based on an improved gradient optimization algorithm according to claim 1, characterized in that: The number of randomly deployed positioning anchor nodes is determined in the positioning area using a regular hexagonal deployment method, including: Initializing the positioning area and generating a grid; Traversing four quadrants in the grid, gradually placing points in each quadrant; The placement points outside the positioning area and in the obstacle area are deleted, and the upper limit of the number of anchor nodes is preset, and finally the initial number of anchor nodes deployed in the positioning area is determined.

3. The method for deploying low-power Bluetooth anchor nodes based on an improved gradient optimization algorithm according to claim 1, characterized in that: Initializing the IGBO algorithm includes: initializing the IGBO algorithm parameters; the parameters include: population number np, maximum number of iterations Maxit, probability parameter pr, discrete granularity dr, perception radius R s , dimension dimension, redundant deployment penalty function parameters N1, N2, δ, n, and the observation value delta_nums required for network similarity.

4. The method for deploying low-power Bluetooth anchor nodes based on an improved gradient optimization algorithm according to claim 1, characterized in that: Introducing low-discrepancy sequences to randomly generate several different anchor node deployment schemes include: S1. Determine the dimension and cardinality of the low-discrepancy sequence in the positioning region, and obtain the van der Kopit sequence of each dimension in the low-discrepancy sequence; S2. performing coordinate transformation on the van der Kopit sequence to obtain a sequence that conforms to the actual target positioning area; S3. Pairing the van der Kopit sequence and the sequence that matches the actual target positioning region in order to generate a sequence of anchor node coordinates, which is an initial deployment plan for the anchor nodes; S4. Repeat S1-S3 and introduce an obfuscation processing method to generate several different anchor node deployment schemes.

5. The method for deploying low-power Bluetooth anchor nodes based on an improved gradient optimization algorithm according to claim 1, characterized in that: The evaluation performance index value is: <h2 style=";text-align:left;direction:ltr">f=k1<h2 style=";text-align:left;direction:ltr"> f <h2 style=";text-align:left;direction:ltr"> 1 <h2 style=";text-align:left;direction:ltr"> +k <h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> *f2+k3*f3+k4*f4+k5*f5 Among them, k1, k2, k3, k4 and k5 are coefficients, f1 is the proportion of unlocatable nodes to all target nodes, f2 is the average GDOP value of locatable nodes, f3 is the penalty term, f4 is the sensor network topology evaluation, and f5 is the network uniformity.

Images