An Indoor Wireless Localization Method for Industrial Assets Based on Ensemble Learning and RNN Neural Network

Abstract

The present invention discloses an indoor wireless positioning method for industrial assets based on ensemble learning and RNN neural network, which relates to the technical field of indoor positioning of mobile industrial asset targets. By combining WIFI positioning and UWB positioning, first, the large-scale indoor positioning area is divided into unit areas. The WIFI positioning uses an ensemble learning algorithm aggregated by various machine learning algorithm models to calculate the specific area where the moving target is located. Then, the UWB positioning RNN model within this area is used to determine the precise position or moving path of the moving target within the area. Finally, according to the backpropagation error of the accurately positioned coordinates, the weights of various machine learning algorithms are updated to correct the model, and the current model is gradually optimized through multiple iterations to achieve the best positioning model in this scenario. The present invention establishes a clearly hierarchical indoor positioning method for moving targets, provides a reliable target positioning method and process, and improves the disadvantages of the existing positioning methods, such as unclear hierarchy, limited adaptability, and inability to track the entire process.

Description

Technical Field

[0001] The present invention relates to the technical field of indoor positioning of moving targets, and in particular to an indoor wireless positioning method for industrial assets based on ensemble learning and RNN neural network. Background Art

[0002] With the rapid development of China's manufacturing industry, the processing processes of factory products are gradually becoming more diverse. From the procurement and transportation of raw materials, processing and inspection, to product assembly, packaging and warehousing, various processes will inevitably experience various transportation processes. In order to avoid economic losses caused by the loss of important industrial production materials during transportation by manufacturers, determining the precise location of industrial assets during transportation has become an urgent problem to be solved.

[0003] The positioning of industrial assets usually has higher requirements than general indoor positioning, including real-time performance, accuracy, economic benefits and stability. During the transportation of industrial assets, it is necessary to monitor their position data information in real time, which is usually achieved by deploying continuous sensors or data collectors to ensure the continuity and real-time performance of data transmission. The learning model trained with historical data is closer to the specific location of the target than general position calculation rules, and has higher accuracy. Since the transportation range of industrial assets is generally set in large-scale places such as factories, docks, workshops, etc., the existing positioning methods will result in high positioning costs and maintenance costs due to the large number of sensor deployments, resulting in low economic benefits of positioning, and are easily interfered by electromagnetic waves, spatial magnetic fields, and electric fields, resulting in unstable positioning accuracy. Summary of the Invention

[0004] In order to solve the above problems, the present invention provides an indoor wireless positioning method for industrial assets based on ensemble learning and RNN neural network, which has the advantages of saving the cost of positioning equipment, clear positioning classification, improving positioning accuracy through self-iteration, and tracking the entire process of positioning information. It proposes an effective solution to the problems such as resource waste caused by the large number of beacon points to be deployed for the transportation positioning of important production materials in the current manufacturing industry, algorithm failure, low positioning accuracy, and large error caused by a single positioning method or a single positioning algorithm in a complex indoor positioning environment.

[0005] To achieve the above object, the present invention provides an indoor wireless positioning method for industrial assets based on ensemble learning and RNN neural network, specifically including the following steps:

[0006] Step S1: Dataset construction, deploy indoor AP access points and beacon points, and collect the RSSI fingerprint library of signal strength identifiers to construct the positioning correspondence;

[0007] Step S2: WIFI positioning prediction. Based on the signal strength values measured at AP access points, real-time position calculation is performed through the ensemble learning algorithm obtained by integrating machine learning models to obtain the location area where the maximum probability of the moving target is located;

[0008] Step S3: UWB positioning prediction. Combining the specific location area determined by the WIFI positioning prediction in Step S2, the server retrieves the original data obtained by the UWB signal points in the location area, calls the RNN model to calculate the exact position where the moving target is located, and returns the position coordinate value;

[0009] Step S4: Reverse update aggregation weight. Calculate the mean square error of the result points deduced by ensemble learning based on the accurately positioned coordinate points, and correct the weight values of the weighted aggregation of the machine learning model during the WIFI positioning process;

[0010] Step S5: Return to Step S2 for iteration. According to the position change of the moving target between different areas indoors and the weights of the updated machine learning model in Step S4, calculate and output new positioning coordinates to gradually reduce the error.

[0011] Preferably, in Step S2, it specifically includes the following steps:

[0012] S21: Use the test set data measured in Step S1 to conduct positioning model accuracy tests for the same map at different times under a specific scenario to obtain a set of positioning errors, and set the error threshold to

[0013]

[0014] S22: Initialize the machine learning model and initialize the weighted fusion value Input the RSSI and position coordinate information into the machine learning model as positioning references;

[0015] S23: Use T machine learning models to calculate the real-time RSSI collected at the AP points respectively, and calculate the T coordinate points L predicted by the machine learning models corresponding to a set of signal strength values i (i = 1, 2,..., T) = (x i , y i );

[0016] S24: According to the weighted fusion value, perform heterogeneous data positioning coordinate fusion on the T predicted coordinate points to fuse them into a determined coordinate. The fusion adopts the enhanced weighted K-nearest neighbor algorithm. According to the error threshold Remove the reference points greater than the threshold, and the remaining coordinate points are fused into the WIFI area positioning coordinates in the following manner According to the specific area distribution under WIFI positioning, the positioning coordinates are located in the specific UWB area to complete the preliminary positioning:

[0017]

[0018] Preferably, in step S3, it specifically includes using the original data pair (D i , P i ) collected by the beacon points to train the RNN model to obtain the accurate position information of the moving target in the determined area;

[0019] Where D i represents the distance D i = [d i,1 , d i,2 ,..., d i,n , measured by n beacon points deployed in the area, and P i represents the actual coordinate point corresponding to the n-dimensional distance vector.

[0020] Preferably, in step S3, the UWB positioning method includes the method of mapping the local positioning of the moving target to the global positioning, specifically expressed as:

[0021] Mapping the relative coordinate points of the UWB positioning in the specific room to the absolute coordinate points of the entire map, including the following steps:

[0022] Step 1: Calculate the real-time distance vector collected by the beacon point positions through the RNN model to obtain the internal coordinate points (X i , Y i ) of the UWB positioning model;

[0023] Step 2: Combine the specific coordinates of the deployed UWB positioning beacon points and the positional relationship between the moving target and the beacon points to map the relative coordinate points to the overall positioning area range to obtain the absolute coordinates of the relative coordinate points in the global positioning area

[0024] Preferably, in step S4, the method of updating the weight includes the reverse update process based on the positioning result, specifically including the following steps:

[0025] S41: Based on the mean square error σ between the T predicted coordinate points determined by the WIFI positioning prediction in step S2 and the absolute coordinates obtained in step S3 i 2 , update the weight V i size of the weighted fusion of the machine learning model to obtain V i ' as the updated weight;

[0026]

[0027] S42: Use the updated weight V i ' Calculate the error threshold for all test points under the prediction of the existing ensemble learning model results, denoted as

[0028]

[0029] Preferably, in step S5, the specific implementation process is as follows:

[0030] S51: For the machine learning model with updated weights, re-determine the error and calculate the error threshold;

[0031] S52: Use the machine learning model with updated weights to solve the real-time RSSI data to obtain T predicted coordinate points corresponding to the machine learning model;

[0032] S53: Screen out the coordinates lower than the threshold according to the updated error threshold, fuse the coordinates lower than the threshold according to the updated weights to obtain the fused coordinate points, and enter the UWB positioning stage;

[0033] S54: Continue to execute the UWB positioning prediction process in step S3, and use the accurate coordinate values calculated by the RNN algorithm to update the weights of the machine learning model again, and continue to iterate.

[0034] Therefore, the present invention adopts the above-mentioned indoor wireless positioning method for industrial assets based on ensemble learning and RNN neural network, and has the following beneficial effects:

[0035] (1) The present invention avoids the waste of resources caused by deploying a large number of UWB positioning beacon points indoors in real scenarios with high positioning accuracy requirements. By combining low-cost WIFI positioning and high-cost UWB positioning, the indoor positioning cost is saved.

[0036] (2) The present invention updates the aggregation weights of various machine learning models in the WIFI area positioning through the UWB precise positioning information. After multiple iterations, several area positioning algorithm models most suitable for the positioning scenario are determined and given high weights, so that the predicted values of these several higher-precision models account for a higher proportion in the final result than other models, making the positioning algorithm adaptive and can be applied to different positioning scenarios through continuous update of the weights to meet different positioning requirements.

[0037] (3) The present invention combines area positioning and precise positioning within the area to gradually narrow down the possible range where the moving target may be located, reduce errors, and improve accuracy. Compared with the traditional indoor positioning method using a single positioning method and a single model, it has a more accurate prediction result.

[0038] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Description of the Drawings

[0039] Figure 1 It is a flowchart of the WIFI positioning machine learning model structure of the present invention;

[0040] Figure 2 It is a regional division diagram in the application scenario in the embodiment of the present invention;

[0041] Figure 3 It is the UWB positioning recurrent neural network (RNN) model structure of the present invention;

[0042] Figure 4 It is a program block diagram for updating various machine learning weights of WIFI according to the UWB positioning result of the present invention;

[0043] Figure 5 It is a graph showing the change of positioning error with the number of iterations drawn by using the weight update method of the present invention;

[0044] Figure 6 It is an experimental simulation result diagram of the real collected WIFI fingerprint database and UWB beacon point values of the present invention. Detailed Embodiments

[0045] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0046] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the ordinary meaning as understood by those of ordinary skill in the field to which the present invention belongs.

[0047] The terms "including" or "comprising" and the like used in the present invention mean that the elements before this word cover the elements listed after this word, and do not exclude the possibility of also covering other elements. The orientation or positional relationship indicated by the terms "inside", "outside", "above", "below", etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation to the present invention. When the absolute position of the described object changes, the relative position relationship may also change accordingly. In the present invention, unless otherwise clearly defined and limited, terms such as "attached" shall be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or integrated; it can be directly connected, or indirectly connected through an intermediate medium, and can be the internal connection of two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0048] As shown Figure 1 in the figure, an indoor wireless positioning method for industrial assets based on ensemble learning and RNN neural network specifically includes the following steps:

[0049] Step S1: Dataset construction. Deploy indoor AP access points and beacon points, and collect the RSSI fingerprint database of signal strength identifiers to construct the positioning correspondence;

[0050] Install WIFI positioning AP points in the factory building, requiring that their signals can cover the entire area. Deploy UWB beacon points in the indoor area according to the rooms or channels that need to be positioned, and divide the overall WIFI positioning area into several UWB positioning sub-areas. As Figure 2 shown in the figure, it is the plant floor plan of an actual factory and the specific division method of the positioning method.

[0051] Step S2: WIFI positioning prediction. According to the signal strength values measured by the AP access points, perform real-time position calculation through the ensemble learning algorithm obtained by fusing machine learning models to obtain the location area where the maximum moving target probability is located; Select six machine learning algorithm models, including the nearest neighbor algorithm (NN), K-nearest neighbor algorithm (KNN), weighted K-nearest neighbor method (WKNN), Bayesian probability algorithm (Prob), Gaussian kernel density algorithm (Gk), and improved algorithm of KNN method (Stg):

[0052] The nearest neighbor algorithm (NN) calculates the Euclidean distance L between the measured RSS vector and the RSS vectors of each record in the database according to the following formula i , and then selects the position coordinates of the point corresponding to the smallest distance as the result output. As shown in the following formula, where RSSI i is the RSS vector of the point to be measured, RSSI j is the RSS vector of each reference point record in the database, and j is the number of reference points:

[0053]

[0054] The main idea of the K-nearest neighbor algorithm (KNN) is: find the K (K>2) RP (database vectors) with the highest WiFi fingerprint similarity to TP (the vector obtained by positioning measurement), and then average the positions corresponding to these K RPs to finally obtain the position of TP. As shown in the following formula, where is the positioning estimation result, (x i , y i ) is the coordinate corresponding to the i-th selected fingerprint information:

[0055]

[0056] The difference between the Weighted K-Nearest Neighbor method (WKNN) and the KNN algorithm is that after selecting the K (K>2) nearest reference points, a weighting coefficient is multiplied by the coordinates of each database vector, and the weighting method is to weight by the reciprocal of the Euclidean distance. As shown in the following formula, where is the positioning estimation result, and (x i , y i ) is the coordinate corresponding to the i-th selected fingerprint information:

[0057]

[0058] The main idea of the Bayesian probability algorithm (Prob) is: through the Bayesian formula, calculate the posterior probability of the TP appearing on each RP, select the k RPs with the largest probabilities, and then average the positions corresponding to these K RPs to finally obtain the position of the TP. As shown in the following formula, where is the positioning estimation result, and (x i , y i ) is the coordinate corresponding to the i-th selected fingerprint information:

[0059]

[0060] The main idea of the Gaussian kernel density algorithm (Gk) is to calculate the probability of the relative supersaturation at each fingerprint position through a Gaussian kernel density estimator (GaussianKernelDensityEstimator, GKDE), and determine the position of the TP by averaging the positions corresponding to the highest likelihood values. As shown in the following formula, x and y are input samples, and ||x - y|| 2 represents the square of the Euclidean distance, and δ represents a hyperparameter of the Gaussian kernel function:

[0061]

[0062] The improved KNN method algorithm (Stg) is an improved method of the KNN method. It screens the RPs through the k signal access points with the strongest signals of the TP, and then applies the KNN method.

[0063] In step S2, it specifically includes the following steps:

[0064] S21: Use the test set data measured in step S1 to perform calculations on the RSSI of the test set data based on a specific scenario using the above six models respectively, compare the predicted point positions obtained with the true coordinate points, and record the median error of each positioning method as Take the average of the six median errors To obtain the threshold

[0065]

[0066] S22: Initialize the machine learning model and initialize the weighted fusion value Input the RSSI and location coordinate information into the machine learning model as the positioning reference;

[0067] S23: Use T machine learning models to calculate the real-time RSSI collected at the AP points respectively, and calculate the T coordinate points L predicted by the machine learning models corresponding to a set of signal strength values i (i = 1, 2,..., T) = (x i , y i );

[0068] S24: According to the weighted fusion value, perform heterogeneous data positioning coordinate fusion on the T predicted coordinate points, fuse them into a determined coordinate, and the fusion adopts the enhanced weighted K-nearest neighbor algorithm. According to the error threshold Remove the reference points greater than the threshold, and the remaining coordinate points are fused into the WIFI area positioning coordinate in the following way According to the specific area distribution under WIFI positioning, locate the positioning coordinate to the specific area of UWB to complete the preliminary positioning:

[0069]

[0070] In step S3, it specifically includes using the original data pair (D i , P i ) collected by the beacon points to train the RNN model to obtain the accurate position information of the moving target in the determined area;

[0071] where D i represents the distance D between the beacon point and the moving target D i = [d i,1 , d i,2 ,..., d i,n , which is measured by n beacon points deployed in the area, and P i represents the actual coordinate point corresponding to the n-dimensional distance vector.

[0072] Step S3: UWB positioning prediction. Combining the specific position area determined by the WIFI positioning prediction in step S2, the server retrieves the original data obtained by the UWB beacon points in the position area, calls the RNN model to calculate the accurate position where the moving target is located, and returns the position coordinate value; In step S3, the UWB positioning method includes the method of mapping the local positioning of the moving target to the global positioning, using the pre-measured data collected by the beacon points and the actual coordinate positions to train the positioning model, recording the network structure and node parameters, and its basic structure is as Figure 3As shown in the figure, assume that n UWB beacon points are deployed indoors. Then the input of the model is an n-dimensional vector, representing the distance between each beacon point and the moving target. The input data passes through a recurrent neural network layer (RNN), a linear layer (Linear), a batch normalization layer (BatchNormalization), a rectified linear unit layer (Relu), and a linear layer (Linear) in sequence to obtain the output result.

[0073] The input of the RNN layer is an n-dimensional vector, which contains 256 hidden layers; the output enters the first linear layer, and a bias is added to convert it into a 512-dimensional output; the batch normalization layer is used to normalize the 512 feature dimensions to satisfy the distribution law with a mean of 0 and a variance of 1; then the normalized features are input into the rectified linear unit layer for classification, and the result passes through the second linear layer and is output in two-dimensional data, where the two-dimensional data represents the predicted horizontal and vertical coordinate values.

[0074] Specifically, it is expressed as:

[0075] Mapping the relative coordinate points of UWB positioning in a specific room to the absolute coordinate points of the entire map includes the following steps:

[0076] Step 1: Use the RNN model to calculate the real-time distance vector collected at the beacon points to obtain the internal coordinate points (X i , Y i ) in the internal coordinate system calculated by the UWB positioning model;

[0077] Step 2: Combine the specific coordinates of the deployed UWB positioning beacon points and the positional relationship between the moving target and the beacon points, and map the relative coordinate points to the overall positioning area range to obtain the absolute coordinates of the relative coordinate points in the global positioning area.

[0078] Step S4: Update the aggregation weights in the reverse direction. Calculate the mean square error between the result points deduced by ensemble learning based on the accurately positioned coordinate points, and correct the weight values of the weighted aggregation in the machine learning model during the WIFI positioning process;

[0079] In step S4, the way to update the weights includes a reverse update process based on the positioning results, which specifically includes the following steps:

[0080] S41: As Figure 4 shown, use the Euclidean distance calculation formula to calculate the mean square error σ between the T predicted coordinate points determined by the WIFI positioning prediction in step S2 and the absolute coordinates i 2 obtained in step S3, and update the weight value V i of the weighted fusion of the machine learning model to obtain V i' is the updated weight, and the median of the prediction error is updated with the updated ensemble learning method with weights;

[0081]

[0082] S42: Use the updated weight V i ' Calculate the error threshold for all test points under the prediction of the existing ensemble learning model result, denoted as

[0083]

[0084] Step S5: Return to Step S2 for iteration. According to the position change of the moving target between indoor areas and the weights of the updated machine learning model in Step S4, solve and output new positioning coordinates, gradually reducing the error.

[0085] As Figure 5 shown, in Step S5, the specific implementation process is as follows:

[0086] S51: For the machine learning model with updated weights, re-determine the error and calculate the error threshold;

[0087] S52: Use the machine learning model with updated weights to solve the real-time RSSI data and obtain T predicted coordinate points corresponding to the machine learning model;

[0088] S53: Screen out the coordinates lower than the threshold according to the updated error threshold, fuse the coordinates lower than the threshold according to the updated weights to obtain the fused coordinate points, and enter the UWB positioning stage;

[0089] S54: Continue to execute the UWB positioning prediction process in Step S3, and use the accurate coordinate values calculated by the RNN algorithm to update the weights of the machine learning model again and continue to iterate.

[0090] Experimental Simulation

[0091] An experimental simulation was carried out. The experimental data was based on a real collected WIFI fingerprint database and UWB beacon position values. As Figure 6As shown, the abscissa represents different training weeks, and the predicted data for each week can be regarded as one iteration. The ordinate represents the difference between the predicted coordinate points and the true coordinate points. The error changes of all six types of algorithms in various machine learning are shown. In the current positioning environment, the solution errors of the probability-based algorithm Prob, the improved K-nearest neighbor algorithm Stg, and the nearest neighbor algorithm NN are significantly greater than those of the K-nearest neighbor algorithm KNN, the weighted K-nearest neighbor algorithm WKNN, and the Gaussian kernel density algorithm GK method. After improving the weighted fusion value of the small error algorithm, it can be observed that the "Gathered" curve (fused coordinates) gradually converges to the small error algorithm. Compared with using a single algorithm for positioning, the regional positioning method with updated weights reduces the error by 20.12% to 37.58%. Under stable environmental conditions, the error can be maintained within 2.5 meters to 3.5 meters, meeting the indoor positioning error standard.

[0092] Therefore, the present invention adopts the above-mentioned indoor wireless positioning method for industrial assets based on ensemble learning and RNN neural network, combining WIFI positioning and UWB positioning. First, the large-scale indoor positioning area is divided into unit areas. The WIFI positioning uses an ensemble learning algorithm aggregated by various machine learning algorithm models to calculate the specific area where the moving target is located. Then, the UWB positioning RNN model within this area is used to determine the precise position or moving path of the moving target within the area. Finally, according to the backpropagation error of the accurately positioned coordinates, the weights of various machine learning are updated to correct the model. After multiple iterations, the current model is gradually optimized to achieve the best positioning model in this scenario, which has the advantages of saving the cost of positioning devices, clear positioning classification, improving positioning accuracy through self-iteration, and tracking all-process positioning information. It provides an effective solution to the problems such as resource waste caused by deploying a large number of beacon points for the transportation positioning of important production materials in the current manufacturing industry, algorithm failure, low positioning accuracy, and large error caused by a single positioning method or a single positioning algorithm in a complex indoor positioning environment.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements do not make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An indoor wireless positioning method for industrial assets based on ensemble learning and RNN neural network, characterized in that: Specifically, it includes the following steps: Step S1: Dataset construction, deploy indoor AP access points and beacon points, collect the signal strength identifier RSSI fingerprint database to construct the positioning correspondence; Step S2: WIFI positioning prediction, based on the signal strength values measured at the AP access points, perform real-time position calculation through the ensemble learning algorithm obtained by fusing machine learning models, and obtain the location area where the maximum moving target probability is located; In step S2, it specifically includes the following steps: S21: Use the test set data measured in step S1 to perform accuracy tests on the positioning models at different times of the same map under a specific scenario, obtaining a set of positioning errors, and setting the error threshold as S22: Initialize the machine learning model and initialize the weighted fusion value T represents the type of the machine learning model. The RSSI and location coordinate information are passed into the machine learning model as positioning references; S23: Use T machine learning models to respectively calculate the real-time RSSI collected at the AP points, and calculate the coordinate points L predicted by the T machine learning models corresponding to a set of signal strength values i (i = 1, 2,..., T) = (x i , y i ); S24: According to the weighted fusion value, perform heterogeneous data location coordinate fusion on the T predicted coordinate points to fuse them into a determined coordinate. The fusion uses the enhanced weighted K-nearest neighbor algorithm, according to the error threshold Remove the reference points greater than the threshold, and the remaining coordinate points are fused into the WIFI area location coordinates in the following way According to the specific area distribution under WIFI positioning, locate the positioning coordinates to the specific area of UWB to complete the preliminary positioning: Step S3: UWB positioning prediction, combining the specific location area determined by the WIFI positioning prediction in step S2, the server retrieves the original data obtained from the UWB beacon points within the location area, calls the RNN model to calculate the precise location of the moving target, and returns the location coordinate value; Step S4: Reverse update the aggregation weight, calculate the mean square error of the result points deduced by the ensemble learning based on the accurately positioned coordinate points, and correct the weight values of the weighted aggregation of the machine learning model during the WIFI positioning process; In step S4, the weight update method includes a reverse update process based on the positioning result, specifically including the following steps: S41: Based on the mean square error σ between the T predicted coordinate points determined by the WIFI positioning prediction in step S2 and the absolute coordinates obtained in step S3 i 2 , update the weight V of the weighted fusion of the machine learning model i size to obtain V i ' as the updated weight; S42: Use the updated weight V i 'Calculate the error threshold for all test points under the prediction of the existing ensemble learning model results, denoted as Step S5: Return to step S2 for iteration, according to the position change of the moving target between different indoor areas and the weight values of the updated machine learning model in step S4, calculate and output new positioning coordinates, and gradually reduce the error.

2. The industrial asset indoor wireless positioning method based on ensemble learning and RNN neural network according to claim 1, characterized in that: In step S3, specifically, the original data pair (D i , P i ) collected using the beacon points is used to train the RNN model to obtain the accurate position information of the moving target in the determined area; Among them, D i represents the distance D between the fiducial point and the moving target i = [d i,1 , d i,2 ,..., d i,n , which is measured by n fiducial points deployed in the area, and P i represents the actual coordinate point corresponding to the n-dimensional distance vector.

3. The industrial asset indoor wireless positioning method based on ensemble learning and RNN neural network according to claim 2, wherein: The UWB positioning method includes the method of mapping the local positioning of the moving target to the global positioning, specifically expressed as: Map the relative coordinate points of the UWB positioning in a specific room to the absolute coordinate points of the entire map, including the following steps: Step 1: Calculate the real-time distance vector collected at the beacon points through the RNN model to obtain the coordinates of the internal coordinate system within the area calculated by the UWB positioning model (X i , Y i ); Step 2: Combine the specific coordinates of the deployed UWB positioning beacon points and the positional relationship between the moving target and the beacon points, map the relative coordinate points to the overall positioning area range, and obtain the absolute coordinates of the relative coordinate points in the global positioning area 4. The indoor wireless positioning method for industrial assets based on ensemble learning and RNN neural network according to claim 3, characterized in that: In step S5, the specific implementation process is as follows: S51: For the machine learning model with updated weights, re-determine the error and calculate the error threshold; S52: Use the machine learning model with updated weights to calculate the real-time RSSI data, and obtain T predicted coordinate points corresponding to the machine learning model; S53: Screen out the coordinates lower than the threshold according to the updated error threshold, perform K-nearest neighbor fusion on the coordinates lower than the threshold according to the updated weights to obtain the fused coordinate points, and enter the UWB positioning stage; S54: Continue to execute the UWB positioning prediction process in step S3, use the precise coordinate values calculated by the RNN algorithm to update the weights of the machine learning model again, and continue to iterate.

Images