Bluetooth mesh fingerprint compensation positioning method and device for dynamic environment

Abstract

The application discloses a Bluetooth mesh network fingerprint compensation positioning method and device for a dynamic environment, relates to the technical field of positioning data processing, and comprises the following steps: S1, periodically collecting cooperative positioning sensing data, and performing pretreatment on the collected data; S2, determining whether interference abnormity occurs to the Bluetooth signal strength, and compensating the Bluetooth signal strength in combination with the corresponding cooperative positioning sensing data; S3, evaluating the signal stability of each node, constructing a fingerprint feature self-supervision model, and outputting fingerprint feature samples; and S4, establishing a Bluetooth Mesh positioning coordinate system, obtaining a target positioning coordinate in combination with the corresponding cooperative positioning sensing data of real-time fingerprint feature samples, evaluating the cooperative consistency of target positioning, and determining whether to perform coordinate correction based on the result. The application solves the problem that in the prior art, signal fluctuation caused by a dynamic environment cannot be corrected, and thus fingerprint samples are distorted and target positioning coordinates are deviated.

Description

Technical Field

[0001] This invention relates to the field of positioning data processing technology, specifically to a Bluetooth mesh fingerprint compensation positioning method and apparatus for dynamic environments. Background Technology

[0002] With the continuous development of positioning technology, especially in fields such as drones and virtual reality, improving positioning accuracy has become crucial for enhancing system performance. However, traditional positioning methods often face various challenges, such as fusion errors from multiple sensors, signal transmission delays, and interference from environmental factors. These factors mean that the accuracy of traditional positioning technologies cannot meet the demands of high-precision applications in complex dynamic environments. To address this issue, researchers have proposed various error compensation and optimization methods, but most rely on single sensors or general models, lacking precise adjustment mechanisms for specific systems, making it difficult to effectively compensate for system errors.

[0003] For example, the invention with announcement number CN115618749B provides an error compensation method for real-time positioning of large unmanned aerial vehicles (UAVs), including the following steps: S1, acquiring an error dataset for real-time UAV positioning; S2, constructing an error model based on the error dataset; S3, applying the error model; S4, updating the error model. This invention fully considers the working mode, working state, and data transmission delay of each sensor, accurately describes the system error of real-time positioning of large UAVs, and avoids the error problem of matching traditional general models with UAV systems by using machine learning combined with the acquisition and training of specific UAV samples, thus significantly improving the real-time positioning accuracy of UAVs.

[0004] For example, the invention with publication number CN120469587A relates to the field of virtual reality positioning and discloses a VR large-space positioning and interaction system based on multimodal perception. The system includes: deploying multimodal sensors to acquire sensing data, extracting and preprocessing visual features, matching and establishing a sparse point cloud map and performing scale calibration, and constructing an environment model; pre-judging UWB signal paths based on the environment model, optimizing and compensating for errors in the NLOS state, updating observations through degradation detection and fusing them to obtain predicted state changes; constructing an interactive perception network to track hands and the whole body; using a layered haptic system to achieve haptic feedback, generating haptic effects using vibration frequency mapping; using beamforming technology to improve transmission efficiency, rendering virtual scenes using an edge cloud server cluster, and pre-rendering scenes to offset network and rendering latency; and designing an online calibration mechanism to correct system errors through visual loop closure detection, dynamic correction of UWB beacons, and IMU drift compensation.

[0005] However, despite the solutions offered by the aforementioned technologies, certain limitations remain. Traditional error compensation methods often rely on universal models when processing multi-sensor data, making them unable to effectively adapt to the characteristics of different systems. Especially in dynamic environments, mismatches may occur between the error models of different sensors, leading to unsatisfactory compensation results. Furthermore, these methods typically cannot cope with complex changes occurring in real-time environments, such as sudden changes in ambient light intensity or signal attenuation, and they also struggle to accurately fuse signals and data from different sensors, thus affecting the real-time positioning accuracy of the system.

[0006] Therefore, in order to address the above problems, there is an urgent need for a Bluetooth mesh fingerprint compensation positioning method and device for dynamic environments. Summary of the Invention

[0007] Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a Bluetooth mesh fingerprint compensation positioning method and device for dynamic environments, which solves the problem that existing virtual positioning technologies cannot correct signal fluctuations caused by dynamic environments, thus causing fingerprint sample distortion and target positioning coordinate drift.

[0009] Technical solution

[0010] To achieve the above objectives, the present invention provides the following technical solution: a Bluetooth mesh fingerprint compensation positioning method for dynamic environments, comprising: S1, periodically collecting collaborative positioning sensing data, performing time synchronization, noise suppression, abrupt change smoothing, anomaly detection, and missing data imputation on the collected data to obtain preprocessed collaborative positioning sensing data; S2, extracting the preprocessed ambient light intensity and Bluetooth signal strength, determining whether the Bluetooth signal strength has interference anomalies, and compensating the Bluetooth signal strength in conjunction with the corresponding collaborative positioning sensing data; S3, extracting real-time collaborative positioning sensing data according to node number, evaluating the signal stability of each node, constructing a fingerprint feature self-supervised model based on the evaluation results, outputting fingerprint feature samples, and constructing a fingerprint feature index matrix; S4, establishing a Bluetooth Mesh positioning coordinate system based on the fingerprint feature index matrix, obtaining the target positioning coordinates by combining the collaborative positioning sensing data corresponding to the real-time fingerprint feature samples, evaluating the collaborative consistency of target positioning, and determining whether to perform coordinate correction based on the results to achieve adaptive positioning optimization.

[0011] Furthermore, the collaborative positioning and sensing data is periodically collected. The collected data undergoes time synchronization, noise suppression, abrupt change smoothing, anomaly detection, and missing data imputation. The specific steps to obtain preprocessed collaborative positioning and sensing data are as follows: A fixed sliding time window is set as one sampling period. Collaborative positioning and sensing data generated by the Bluetooth Mesh nodes during the positioning and coordination process is collected in real time. This data includes node number, Bluetooth signal strength, antenna phase difference, node arrival time difference, ambient light intensity, ambient temperature, ambient humidity, node voltage, IMU acceleration, IMU angular velocity, signal-to-noise ratio, and packet loss rate. For the collected cooperative positioning and sensing data, time synchronization and timing unification are performed using a clock drift estimation algorithm based on Kalman filtering, and noise suppression and abrupt change smoothing are performed using an adaptive median filtering algorithm. Abnormal observation samples in the cooperative positioning and sensing data are identified and removed using a local outlier factor algorithm to eliminate outliers caused by node failures and signal reflections. Missing time segments in the cooperative positioning and sensing data are interpolated using a K-nearest neighbor-based interpolation algorithm to restore continuous feature sequences. Numerical normalization processing of the cooperative positioning and sensing data is performed using a minimum-maximum normalization algorithm to unify the numerical scale and eliminate dimensional differences.

[0012] Further, the specific steps for extracting the preprocessed ambient light intensity and Bluetooth signal strength, and determining whether the Bluetooth signal strength exhibits interference anomalies, are as follows: Extract the preprocessed ambient light intensity and Bluetooth signal strength, and construct the ambient light intensity sequence and Bluetooth signal strength sequence in timestamp order; perform sliding window differential operation on the ambient light intensity sequence to calculate the light intensity change rate of adjacent sampling periods. When the absolute value of the light intensity change rate is greater than the light intensity change threshold, the corresponding time interval is marked as the illumination change interval; within the illumination change interval, calculate the variance and change rate of the Bluetooth signal strength. When the signal change rate exceeds the signal change threshold and the signal variance exceeds the signal fluctuation threshold, it is determined that the Bluetooth signal strength exhibits interference anomalies.

[0013] Furthermore, the specific steps for compensating Bluetooth signal strength using corresponding cooperative positioning and sensing data are as follows: Extract the corresponding cooperative positioning and sensing data within the illumination change interval; calculate the absolute values ​​of the differences between ambient light intensity and ambient light intensity in the previous sampling period, ambient temperature and ambient temperature in the previous sampling period, and ambient humidity in the previous sampling period; add the three absolute values ​​of difference, add one, and then take the natural logarithm to obtain the composite environmental disturbance term; divide the absolute value of the node arrival time difference by the absolute value of the node arrival time difference plus one to obtain the signal propagation delay factor; multiply the composite environmental disturbance term by the signal propagation delay factor to obtain the signal propagation delay factor. The negative of the product is the exponential function of the exponent with respect to the natural constant e, yielding the illumination attenuation correction term. The Bluetooth signal strength, illumination attenuation correction term, signal propagation delay factor, and cosine of the antenna phase difference are multiplied sequentially to obtain the Bluetooth signal compensation value. The Bluetooth signal compensation values ​​of each Bluetooth Mesh node are synchronized and aligned according to timestamps to construct a Bluetooth signal compensation value sequence. A moving average filter is applied to the Bluetooth signal compensation value sequence, and the signal change rate is recalculated. When the signal change rate is lower than the signal change threshold for N consecutive sampling periods, the corresponding time interval is marked as the compensation stable interval, and the original Bluetooth signal strength is replaced with the Bluetooth signal compensation value.

[0014] Further, the specific steps for extracting real-time collaborative positioning and sensing data by node number and evaluating the signal stability of each node are as follows: Extract the Bluetooth signal strength, time difference of arrival (TDOA), and antenna phase difference of each node by node number; perform inter-node differential operations on the Bluetooth signal strength of multiple nodes at the same timestamp to calculate the signal strength difference and TDOA between adjacent nodes; combine the signal strength difference and TDOA between adjacent nodes into a signal spatial feature sequence by node number; perform least-squares fitting using the signal spatial feature sequence based on the spatial topological relationship between nodes to calculate the relative distance between nodes; and then extract the real-time collaborative positioning and sensing data of each node. The absolute value of the node Bluetooth signal compensation value is taken, then incremented by one and its natural logarithm is taken to obtain the amplitude quantization term. The signal amplitude logarithm quantization term, the cosine of the antenna phase difference, and the constant are added to the signal-to-noise ratio and multiplied sequentially to obtain the signal spatial consistency term. The constant, the square of the node arrival time difference, the square of the difference between the relative distance between nodes and the relative distance between nodes in the previous sampling period, the IMU acceleration magnitude, the IMU angular velocity magnitude, and the received packet loss rate are added sequentially to obtain the signal propagation disturbance term. The signal spatial consistency term is divided by the signal propagation disturbance term to obtain the stable intermediate value. The stable intermediate value is multiplied by the natural logarithm of the node voltage plus one to obtain the node signal stability evaluation value.

[0015] Furthermore, the specific steps for constructing a fingerprint feature self-supervised model based on the evaluation results, outputting fingerprint feature samples, and building a fingerprint feature index matrix are as follows: Historical fingerprint feature samples are extracted from the fingerprint feature database and fused with the node signal stability evaluation values ​​calculated from the corresponding collaborative positioning and sensing data. A self-supervised variational autoencoder algorithm is then used to construct the fingerprint feature self-supervised model. The signal distribution variation patterns during different sampling periods are learned through feature reconstruction. The change in relative distance between nodes is used as a self-supervised reference term to adaptively adjust the feature consistency index output by the model, completing model training. Real-time sampling period collaborative positioning and sensing data and node signal stability evaluation values ​​are input into the fingerprint feature self-supervised model to output fingerprint feature samples. The collaborative positioning and sensing data, relative distance between nodes, node signal stability evaluation values, and fingerprint feature samples are indexed and sorted according to node number and sampling timestamp to establish a fingerprint feature index matrix.

[0016] Furthermore, the specific steps for establishing a Bluetooth Mesh positioning coordinate system based on the fingerprint feature index matrix and obtaining the target positioning coordinates by combining the collaborative positioning sensing data corresponding to the real-time fingerprint feature samples are as follows: Extract the node signal stability evaluation value of each node based on the fingerprint feature index matrix, and sort them from largest to smallest; define the node with the highest node signal stability evaluation value as the origin, the second highest node as the positive X-axis, and the third highest node as the node in the XY plane, thus establishing the Bluetooth Mesh positioning coordinate system; input the historical fingerprint feature samples of each node into the collaborative positioning calculation engine to obtain the corresponding node spatial coordinates. The Euclidean distance similarity algorithm is used to calculate the feature distance between the real-time fingerprint feature sample and each fingerprint feature sample in the fingerprint feature index matrix. The top M matching samples are selected as candidate positioning sets according to the feature distance from low to high, and the corresponding node number and node spatial coordinates are recorded. The weighted average calculation is performed on the node spatial coordinates in the candidate positioning reference set. The reciprocal of the node signal stability evaluation value and the variance of the node arrival time difference are used as the joint weight to obtain the weighted positioning coordinates, forming the initial positioning coordinates. Then, the IMU acceleration and angular velocity data are used to perform integral compensation on the node displacement to correct the coordinate deviation caused by the motion and obtain the target positioning coordinates.

[0017] Furthermore, the specific steps for evaluating the cooperative consistency of target positioning are as follows: Summing the absolute value of the difference between the relative distance between nodes and the relative distance between nodes in the previous sampling period, the node arrival time difference, the IMU acceleration magnitude, and the IMU angular velocity magnitude yields the motion characteristic change term; taking the node signal stability evaluation value as an exponential function with respect to the natural constant e yields the signal stability amplification term; multiplying the system motion characteristic change term by the signal stability amplification term yields the cooperative change basis term; adding one to the absolute value of the difference between the Bluetooth signal strength and the Bluetooth signal compensation value yields the signal difference constraint term; taking the negative of the node voltage as an exponent with respect to the natural constant e, and adding one again yields the voltage suppression adjustment term; subtracting the received packet loss rate from the constant one and adding a minimum term yields the communication reliability correction term; dividing the cooperative change basis term by the product of the signal difference constraint term, the voltage suppression adjustment term, and the communication reliability correction term yields the positioning cooperative evaluation value.

[0018] Furthermore, based on the results, the specific steps for determining whether to perform coordinate correction and achieving adaptive optimization of positioning are as follows: Real-time comparison of positioning coordination evaluation value and coordination threshold. When the positioning coordination evaluation value is greater than the coordination threshold, the current target positioning coordinates are retained. When the positioning coordination evaluation value is less than or equal to the coordination threshold, the Mesh network self-calibration mechanism is triggered, the neighboring node with the highest node signal stability evaluation value is selected as the reference node, and local weighted adjustment is performed based on the relative distance between nodes and the coordinates of the reference node to correct the current target positioning coordinates.

[0019] The second aspect of this invention provides a Bluetooth mesh fingerprint compensation positioning device for dynamic environments, comprising: a data acquisition and preprocessing module, an environment compensation and correction module, a feature extraction and positioning calculation module, and a positioning optimization and self-calibration module. The data acquisition and preprocessing module periodically acquires collaborative positioning sensing data, performs time synchronization, noise suppression, abrupt change smoothing, anomaly detection, and missing data imputation on the acquired data to obtain preprocessed collaborative positioning sensing data. The environment compensation and correction module extracts the preprocessed ambient light intensity and Bluetooth signal strength, determines whether the Bluetooth signal strength experiences interference anomalies, and compensates the Bluetooth signal strength based on the corresponding collaborative positioning sensing data. The feature extraction and positioning calculation module extracts real-time collaborative positioning sensing data according to node numbers, evaluates the signal stability of each node, constructs a fingerprint feature self-supervised model based on the evaluation results, outputs fingerprint feature samples, and constructs a fingerprint feature index matrix. The positioning optimization and self-calibration module establishes a Bluetooth Mesh positioning coordinate system based on the fingerprint feature index matrix, obtains the target positioning coordinates by combining the collaborative positioning sensing data corresponding to the real-time fingerprint feature samples, evaluates the collaborative consistency of target positioning, and determines whether to perform coordinate correction based on the results, thereby achieving adaptive positioning optimization.

[0020] Beneficial effects

[0021] The present invention has the following beneficial effects:

[0022] (1) Bluetooth mesh fingerprint compensation positioning method and device for dynamic environment, through the exponential suppression structure of composite environmental disturbance term and signal propagation delay factor, and the linkage antenna phase difference to compensate Bluetooth signal strength online, generate Bluetooth signal compensation value, realize the instantaneous suppression of high light and rapid environmental fluctuations, and significantly reduce link interruption and fingerprint distortion caused by SNR nonlinear attenuation.

[0023] (2) Bluetooth mesh fingerprint compensation positioning method and device for dynamic environment: By using the time difference of arrival of nodes, the relative distance between nodes, the magnitude of IMU acceleration, the magnitude of IMU angular velocity, the signal-to-noise ratio, the packet loss rate and the node voltage to construct a signal stability evaluation value, the fingerprint samples are dynamically screened and weighted, and highly stable samples are retained first, thereby improving the effectiveness of fingerprint database and online matching from the source of collection.

[0024] (3) Bluetooth mesh fingerprint compensation positioning method and device for dynamic environment: By introducing a fingerprint feature self-supervised model, the relative distance change between nodes is used as a self-supervised reference term, and adaptive correction is performed in combination with signal stability evaluation value. In scenarios where there is no manual coordinate labeling or the labeling is scarce, the signal distribution in the dynamic environment is continuously learned, thereby enhancing the generalization ability across time periods and scenarios.

[0025] (4) Bluetooth mesh fingerprint compensation positioning method and device for dynamic environment: The positioning coordination evaluation value is used to uniformly quantify motion disturbance, signal difference, voltage suppression and communication reliability as the basis for determining whether to perform coordinate correction; when the positioning coordination evaluation value is lower than the coordination threshold, the relative distance between the adjacent most stable node and the node is automatically selected for local weighted adjustment, and coordinate correction is completed autonomously to avoid erroneous correction and oscillation. Attached Figure Description

[0026] Figure 1 Flowchart of a Bluetooth mesh fingerprint compensation positioning method for dynamic environments;

[0027] Figure 2 A structural diagram of a Bluetooth mesh fingerprint compensation positioning device for dynamic environments;

[0028] Figure 3 A flowchart for Bluetooth Mesh signal compensation in dynamic environments;

[0029] Figure 4 A determination map for correcting target positioning coordinates based on positioning coordination evaluation values. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Please see Figures 1-4 This invention provides a technical solution: a Bluetooth mesh fingerprint compensation positioning method for dynamic environments, comprising: S1, periodically collecting collaborative positioning sensing data, performing time synchronization, noise suppression, abrupt change smoothing, anomaly detection, and missing data imputation on the collected data to obtain preprocessed collaborative positioning sensing data; S2, extracting the preprocessed ambient light intensity and Bluetooth signal strength, determining whether the Bluetooth signal strength has interference anomalies, and compensating the Bluetooth signal strength in conjunction with the corresponding collaborative positioning sensing data; S3, extracting real-time collaborative positioning sensing data according to node number, evaluating the signal stability of each node, constructing a fingerprint feature self-supervised model based on the evaluation results, outputting fingerprint feature samples, and constructing a fingerprint feature index matrix; S4, establishing a Bluetooth Mesh positioning coordinate system based on the fingerprint feature index matrix, obtaining the target positioning coordinates in conjunction with the collaborative positioning sensing data corresponding to the real-time fingerprint feature samples, evaluating the collaborative consistency of target positioning, and determining whether to perform coordinate correction based on the results to achieve adaptive positioning optimization.

[0032] Specifically, the process involves periodically collecting collaborative positioning and sensing data, performing time synchronization, noise suppression, abrupt change smoothing, anomaly detection, and missing data imputation on the collected data, and obtaining preprocessed collaborative positioning and sensing data. The specific steps are as follows: A fixed sliding time window is set as one sampling period, using millisecond-level timestamps and Coordinated Universal Time (UTC); collaborative positioning and sensing data generated by Bluetooth Mesh nodes during the positioning and coordination process is collected in real time. This data includes node number, Bluetooth signal strength, antenna phase difference, node arrival time difference, ambient light intensity, ambient temperature, ambient humidity, node voltage, IMU acceleration, IMU angular velocity, signal-to-noise ratio, and packet loss rate. The node number is a unique identifier recorded in the uplink message; the Bluetooth signal strength is measured by RF baseband and measured in decibels and milliwatts; the antenna... Phase difference is obtained from carrier phase difference by sampling at the same frequency and measured in radians; node arrival time difference is obtained from the difference between the timestamp of the first valid message recorded by the local high-stability clock and the reference timestamp and measured in seconds; ambient light intensity is sampled by a light sensor and measured in lux; ambient temperature is sampled by a temperature sensor and measured in degrees Celsius; ambient humidity is sampled by a humidity sensor and measured in relative humidity percentage; node voltage is collected by the power monitoring channel and measured in volts; IMU acceleration is obtained by calculating the magnitude of the three-axis accelerometer of the inertial measurement unit (IMU); IMU angular velocity is obtained by calculating the magnitude of the three-axis gyroscope of the IMU; signal-to-noise ratio is calculated by the ratio of signal power to noise power within a fixed bandwidth and a fixed integration time; received packet loss rate is calculated by the percentage difference between the expected number of frames and the actual number of received frames within a sliding time window. For the collected cooperative positioning and sensing data, a clock drift estimation algorithm based on Kalman filtering is used for time synchronization and timing unification. The state includes clock offset and drift rate, and the arrival time difference of observed nodes is taken. Process noise and observation noise are adaptively updated based on mean square residuals within the sliding time window. Adaptive median filtering is used for noise suppression and abrupt smoothing. The window length is adaptively selected based on the median absolute deviation, and the time boundary uses a reflection spread strategy to maintain sampling continuity. A local outlier factor algorithm is used for anomaly detection and removal. The outlier threshold is taken as the local reachability density quantile within the sliding time window. After removal, consistency verification is performed on the same timestamp to avoid the generation of timing discontinuities. A K-nearest neighbor-based interpolation algorithm is used to interpolate missing time segments. The closer the time distance, the greater the weight. Interpolation is only performed within the sliding time window boundary and the node number and timestamp are kept in a one-to-one correspondence. The cooperative positioning and sensing data is numerically normalized using a min-max normalization algorithm to unify the numerical scale and eliminate dimensional differences.

[0033] In this implementation scheme, through a preprocessing process, the collaborative positioning and sensing data achieves consistent alignment on a millisecond-level timeline, with node numbers and timestamps maintaining a one-to-one correspondence, effectively reducing the cumulative impact of time-series drift on subsequent calculations. Bluetooth signal strength, antenna phase difference, node arrival time difference, ambient light intensity, ambient temperature, ambient humidity, node voltage, IMU acceleration, IMU angular velocity, signal-to-noise ratio, and received packet loss rate exhibit greater robustness under perturbation scenarios, with a significant decrease in the anomaly trigger rate. Missing time-series segments are reversibly completed, and the continuity of cross-cycle features is improved, enabling direct comparison between historical and real-time samples under a unified time reference. Numerical scales are unified to a comparable space, allowing heterogeneous physical quantities to directly participate in joint modeling, reducing the manual overhead of pre-feature engineering. Overall data quality and traceability are enhanced, providing highly reliable input for environmental compensation modeling, fingerprint feature learning, and collaborative positioning inversion, resulting in comprehensive benefits such as faster positioning convergence, more stable online judgment, and more timely error correction.

[0034] Specifically, the steps for extracting the preprocessed ambient light intensity and Bluetooth signal strength, and determining whether the Bluetooth signal strength shows any interference anomalies, are as follows: Figure 3 As shown, the preprocessed ambient light intensity and Bluetooth signal strength are first extracted, and an ambient light intensity sequence and a Bluetooth signal strength sequence are constructed in timestamp order to ensure that the time sequence is consistent with the sampling interval and to avoid misjudgment due to out-of-order conditions. A sliding window differential operation is performed on the ambient light intensity sequence to calculate the rate of change of light intensity between adjacent sampling periods. This is used to capture rapid light fluctuations and distinguish between steady-state and abrupt changes. When the absolute value of the rate of change of light intensity is greater than the light intensity change threshold, the corresponding time interval is marked as a light intensity abrupt change interval. The light intensity change threshold is determined based on historical steady-state interval statistics to reflect the on-site noise level. The physical impact of light abrupt changes on the Bluetooth signal link is considered: rapid changes in strong light can affect the signal-to-noise ratio of the receiving link. The decrease in light intensity, the switching of RF front-end gain state, and the change in ambient reflective surface brightness lead to multipath structure adjustments, which in turn result in increased fluctuations and gradients in Bluetooth signal strength statistics within a short period of time. This relationship is sequential in the time dimension and co-occurs in the same region in the spatial dimension. Within the interval of sudden changes in light intensity, the Bluetooth signal strength sequence is extracted to calculate the variance and rate of change of the Bluetooth signal strength. The variance reflects the intensity of energy fluctuations within this interval, and the rate of change reflects the magnitude of the instantaneous gradient. When the rate of change of the signal exceeds the signal change threshold and the signal variance exceeds the signal fluctuation threshold, it is determined that the Bluetooth signal strength has an interference anomaly, and the start and end timestamps of the abnormal interval and the node numbers involved are recorded simultaneously for reference in subsequent compensation stages.

[0035] In this implementation scheme, under the joint constraints of ambient light intensity change rate and Bluetooth signal strength statistics, it can clearly distinguish between steady-state scenes and scenes with sudden changes in illumination, significantly improving the balance between the specificity and sensitivity of interference identification. By using dual criteria of change rate threshold and variance threshold, it reduces misjudgments caused by a single threshold, lowering the probability of false alarms and false misses. It accurately locates the abnormal interval to the start and end timestamps and node numbers, limiting the effective boundary of subsequent Bluetooth signal strength compensation and avoiding overcompensation for unaffected segments. It completes the anomaly screening before the interference triggering stage, making the data entering subsequent steps purer, reducing the disturbance of unstable samples to the calculation process, and improving the stability and reproducibility of online judgment results.

[0036] Specifically, the steps for compensating Bluetooth signal strength using corresponding cooperative positioning and sensing data are as follows: Extract the corresponding cooperative positioning and sensing data within the illumination change interval; calculate the absolute values ​​of the differences between ambient light intensity and ambient light intensity in the previous sampling period, ambient temperature and ambient temperature in the previous sampling period, and ambient humidity in the previous sampling period; add the three absolute values ​​of difference, add one, and then take the natural logarithm to obtain the composite environmental disturbance term. All differences are aligned to the same timestamp to ensure dimensional consistency and avoid cross-period mixing; divide the absolute value of the node arrival time difference by the absolute value of the node arrival time difference plus one to obtain the signal propagation delay factor. This factor is used to limit the risk of an excessively small denominator caused by extremely small node arrival time differences and to maintain numerical stability; multiply the composite environmental disturbance term by the signal propagation delay factor, and take the exponent of the product with the negative of the product as the exponent using the natural constant e. The illumination attenuation correction term is obtained by multiplying the Bluetooth signal strength, illumination attenuation correction term, signal propagation delay factor, and cosine value of antenna phase difference in sequence to suppress amplitude anomalies. The cosine value of antenna phase difference is used to preserve coherent direction information and weaken sign flipping caused by phase jumps. The Bluetooth signal compensation values ​​of each Bluetooth Mesh node are synchronized and aligned according to timestamps to construct a Bluetooth signal compensation value sequence. During synchronization, millisecond-level timestamps are used as the reference, and missing points are marked for subsequent unified filtering. Then, a moving average filter is performed on the Bluetooth signal compensation value sequence, and the signal change rate is recalculated. When the signal change rate is lower than the signal change threshold for N consecutive sampling periods, the corresponding time interval is marked as the compensation stable interval, and the original Bluetooth signal strength is replaced with the Bluetooth signal compensation value. Here, N is a positive integer greater than one, and the replacement process only takes effect within the compensation stable interval to avoid over-correction of unstable segments and ensure the continuity and stability of input for subsequent feature extraction stages.

[0037] The specific formula for calculating the Bluetooth signal compensation value is as follows:

[0038] ;

[0039] In the formula, Indicates the Bluetooth signal compensation value. Indicates Bluetooth signal strength. Indicates ambient light intensity. This indicates the ambient light intensity in the previous sampling period. Indicates ambient temperature. This indicates the ambient temperature of the previous sampling period. Indicates ambient humidity. This indicates the ambient humidity during the previous sampling period. Indicates the time difference of node arrival. This indicates the antenna phase difference.

[0040] In this implementation scheme, the composite environmental disturbance term and the signal propagation delay factor jointly drive the illumination attenuation correction term, making the correction amount of Bluetooth signal strength monotonically controllable with changes in environmental abrupt changes and propagation delay, thus avoiding numerical divergence. The antenna phase difference retains coherent direction information through cosine mapping, reducing the reverse interference of phase jumps on amplitude, and making the Bluetooth signal compensation value more stable in terms of directional consistency. The Bluetooth signal compensation value sequence is constructed after synchronous alignment according to millisecond-level timestamps, eliminating the risk of mismatch across nodes and time periods, and ensuring that subsequent calculations are carried out on the same time base. The compensation stability interval is determined by the signal change rate, and the original Bluetooth signal strength is replaced with the Bluetooth signal compensation value only within this interval, which can suppress over-correction and compensation oscillation, forming a continuous and smooth feature input. Overall, the usability and consistency of the Bluetooth signal compensation value are improved, providing a more reliable data foundation for subsequent fingerprint feature sample generation and collaborative positioning calculation.

[0041] Specifically, the steps for extracting real-time collaborative positioning and sensing data by node number and evaluating the signal stability of each node are as follows: First, extract the Bluetooth signal strength, time difference of arrival (TDOA), and antenna phase difference of each node by node number, using a unified millisecond-level timestamp and completing time alignment before extraction to ensure comparability under the same time reference. Second, perform inter-node differential operations on the Bluetooth signal strength of multiple nodes at the same timestamp, calculating the signal strength difference and TDOA between adjacent nodes. The differential order is fixed in ascending order of node numbers to avoid inconsistencies in signs, and the differential results correspond one-to-one with the timestamps for subsequent indexing. Third, combine the signal strength difference and TDOA between adjacent nodes by node number into a signal spatial feature sequence. The sequence structure maintains a fixed column order and includes a timestamp index to prevent fitting bias introduced by feature position drift. Fourth, perform least-squares fitting using the signal spatial feature sequence based on the spatial topological relationship between nodes to calculate the relative distance between nodes. The least-squares matrix uses full-rank constraints and introduces a small regularization term in ill-conditioned cases to ensure solvability. When the fitting residual exceeds the corresponding threshold, backtrack to the data of the previous stable time slice to maintain continuity. Finally, extract the real-time collaborative positioning and sensing data of each node. The absolute value of the node Bluetooth signal compensation value is incremented by one and then the natural logarithm is taken to obtain the amplitude quantization term. Before taking the logarithm, the zero-point is incremented to prevent numerical divergence and maintain monotonicity. The signal amplitude logarithm quantization term, the cosine of the antenna phase difference, and a constant are multiplied sequentially by the signal-to-noise ratio to obtain the signal spatial consistency term. The antenna phase difference is expressed in radians and its range is normalized before taking the cosine to reduce transition effects. The constant, the square of the node arrival time difference, the square of the difference between the relative distance between nodes and the relative distance between nodes in the previous sampling period, the IMU acceleration magnitude, and the IMU angular velocity magnitude are calculated. The signal propagation disturbance term is obtained by sequentially adding the received packet loss rate to the signal propagation disturbance term. Each quantity is taken at the same timestamp and the physical units are kept unchanged to ensure the determinism of the summation order. The signal spatial consistency term is divided by the signal propagation disturbance term to obtain a stable intermediate value. A very small amount is added to the denominator when it is close to zero to ensure the stability of the operation, and the timestamp is recorded for auditing. The stable intermediate value is multiplied by the natural logarithm of the node voltage plus one to obtain the node signal stability evaluation value. The node voltage is incremented by one before taking the logarithm to enhance the resolution of the low voltage area. Finally, a sequence of node signal stability evaluation values ​​bound to the node number and timestamp is output.

[0042] The specific formula for calculating the node signal stability assessment value is as follows:

[0043] ;

[0044] In the formula, This represents the node signal stability assessment value. Indicates real-time Bluetooth signal strength. Indicates the time difference of node arrival. Indicates the antenna phase difference. Indicates the relative distance between nodes. This represents the relative distance between nodes in the previous sampling period. Indicates the magnitude of IMU acceleration. Indicates the magnitude of the IMU angular velocity. Indicates the signal-to-noise ratio. Indicates the packet loss rate. This represents the node voltage.

[0045] In this implementation scheme, a signal spatial feature sequence is constructed by aligning the Bluetooth signal strength with millisecond-level timestamps, the node arrival time difference, and the antenna phase difference. The relative distance between nodes is obtained through a least-squares robust solution. Then, combined with the amplitude quantization term generated by the node Bluetooth signal compensation value, the signal spatial consistency term formed by multiplying the amplitude quantization term by the cosine of the antenna phase difference and the signal-to-noise ratio, and the signal propagation disturbance term formed by summing the square of the node arrival time difference, the square of the relative distance change between nodes, the IMU acceleration magnitude, the IMU angular velocity magnitude, and the received packet loss rate, a stable intermediate value is gradually derived. Finally, the logarithmic mapping of the node voltage is introduced to obtain the node signal stability evaluation value. This makes the observations of different nodes at different times comparable and ranked under the same criteria, significantly enhancing the consistency judgment capability in phase transition scenarios, suppressing the estimation bias caused by topological ill-conditioning, and providing a traceable stability quantization sequence. This provides a directly callable sorting basis and confidence foundation for subsequent feature selection, weight allocation, and positioning inversion.

[0046] Specifically, the steps for constructing a fingerprint feature self-supervised model based on the evaluation results, outputting fingerprint feature samples, and building a fingerprint feature index matrix are as follows: Historical fingerprint feature samples are extracted from the fingerprint feature database and fused with the node signal stability evaluation values ​​calculated from the corresponding collaborative localization sensing data. During fusion, the nodes are aligned one-to-one with the sampling timestamp, using a millisecond-level time base consistent with the preceding preprocessing. Entries with failed alignment are removed to ensure consistent sample labels. A fingerprint feature self-supervised model is constructed using a self-supervised variational autoencoder algorithm. The model input includes aligned fingerprint feature samples and node signal stability evaluation values. The latent space dimension is selected on the validation set using the minimum reconstruction error criterion. During training, a sliding time window is used for batch partitioning, and an early stopping strategy is implemented to avoid overfitting. Signals during different sampling periods are learned through feature reconstruction. The distribution variation pattern is analyzed, using the change in relative distance between nodes as a self-supervised reference term. The feature consistency index output by the model is adaptively adjusted. During adjustment, the change in relative distance between nodes is monotonically mapped to the feature reconstruction residual at the same time stamp to unify the scale, complete the model training, and fix the inference parameters. Real-time sampling periodic collaborative positioning perception data and node signal stability evaluation values ​​are input into the fingerprint feature self-supervised model, and fingerprint feature samples are output. The node number and sampling timestamp are recorded simultaneously during output for subsequent retrieval and backtracking. The collaborative positioning perception data, relative distance between nodes, node signal stability evaluation values, and fingerprint feature samples are indexed and sorted according to the node number and sampling timestamp to establish a fingerprint feature index matrix. The index key adopts the combination of node number and sampling timestamp. New samples are appended in chronological order and the version number is retained to support auditing and incremental updates.

[0047] In this implementation scheme, the fingerprint feature self-supervised model takes collaborative positioning sensing data and node signal stability evaluation values ​​as joint inputs. By introducing the relative distance change between nodes as a self-supervised reference term through feature reconstruction, the feature consistency index can adaptively converge with environmental drift, reducing dependence on external annotations and maintaining feature comparability across sampling periods. The fingerprint feature samples output in real time are strictly bound to the node number and sampling timestamp. The fingerprint feature index matrix organizes new samples with combined indexes, forming a traceable and incrementally updatable retrieval structure, which significantly shortens the latency of online matching and backtracking verification, improves the hit efficiency and inference stability of fingerprint retrieval, and provides a high-confidence feature input basis for subsequent target positioning coordinate calculation.

[0048] Specifically, the steps for establishing a Bluetooth Mesh positioning coordinate system based on the fingerprint feature index matrix and obtaining the target positioning coordinates by combining the collaborative positioning sensing data corresponding to real-time fingerprint feature samples are as follows: Extract the node signal stability evaluation value of each node from the fingerprint feature index matrix and sort them from largest to smallest. During sorting, bind the node number and sampling timestamp to ensure traceability. Define the node with the highest node signal stability evaluation value as the origin, the second highest as the positive X-axis, and the third highest as the node in the XY plane. Establish a Bluetooth Mesh positioning coordinate system. Once the three base nodes are determined, they remain unchanged within the current sliding time window and remain unchanged when the window switches. To maintain coordinate continuity, the coordinates are reassessed using the same rules. Historical fingerprint feature samples from each node are input into the collaborative localization engine to obtain the corresponding node spatial coordinates, which correspond one-to-one with the node numbers in the fingerprint feature index matrix. Then, the Euclidean distance similarity algorithm is used to calculate the feature distance between the real-time fingerprint feature sample and each fingerprint feature sample in the fingerprint feature index matrix. The top M matching samples, ranked from lowest to highest feature distance, are selected as the candidate localization set, and the corresponding node numbers and node spatial coordinates are recorded; where M is a positive integer greater than one. A weighted average is calculated for the node spatial coordinates in the candidate localization reference set, using the reciprocal of the node signal stability evaluation value and the node arrival time difference. The variance is used as the joint weight, and the weights are normalized at the same timestamp to avoid scale bias, resulting in weighted positioning coordinates to form the initial positioning coordinates. Then, the node displacement is integrated and compensated using IMU acceleration and IMU angular velocity data. Before integration, the steady-state segment of the IMU acceleration magnitude and IMU angular velocity magnitude is selected within the previous sliding time window. The difference between the mean IMU acceleration and the gravity reference value in this window is used as the zero bias of IMU acceleration, and the mean IMU angular velocity in this window is used as the zero bias of IMU angular velocity. The gravity reference value refers to the nominal gravity magnitude used to subtract the gravity component from the IMU acceleration in the static state. The acquisition process is to first use the factory calibration value as the initial value, and then... On-site verification and updating of the estimated values ​​under the current environment are performed through static sampling. These values ​​are fixed within the same sliding time window and re-estimated when the window switches. During integration, a step size consistent with the sampling timestamp is used. At the end of each sliding time window, the current zero bias is updated with the mean of the zero bias within the window to form a recursive constraint. When the IMU acceleration magnitude and IMU angular velocity magnitude both fall within a continuous confidence interval constructed based on the historical steady-state mean and standard deviation, a zero velocity update is performed, and the velocity increment in that interval is set to zero. When there is an inconsistency between the displacement obtained by integration and the relative distance between nodes, least squares is used to backtrack and trim the velocity increment to correct the coordinate deviation caused by motion, thus obtaining the target positioning coordinates.

[0049] In this implementation scheme, three base nodes are selected based on their signal stability evaluation values ​​to establish a Bluetooth Mesh positioning coordinate system. This ensures that the origin and reference direction are always anchored on highly stable observations, resulting in more stable coordinate references and continuous switching across windows. The top M matching samples are selected from the fingerprint feature index matrix based on Euclidean distance similarity, and the node spatial coordinates are recorded. This candidate set controls computational latency while maintaining matching accuracy. A weighted average is performed using the reciprocal of the node signal stability evaluation value and the variance of the node arrival time difference as joint weights, which weakens the influence of low-stability and high-fluctuation observations, and significantly reduces the variance of the initial positioning coordinates. IMU acceleration and IMU angular velocity are introduced for displacement integral compensation to eliminate trajectory offset caused by inter-sampling motion, ensuring that the target positioning coordinates remain smooth and traceable even in dynamic scenarios. Overall, a robust link from feature matching to coordinate generation is achieved, significantly reducing the sensitivity of positioning results to noise disturbances and transient fluctuations, and significantly improving the consistency of online updates and repeated measurements.

[0050] Specifically, the steps for evaluating the cooperative consistency of target localization are as follows: Summing the absolute value of the difference between the relative distances between nodes and the relative distances between nodes in the previous sampling period, the node arrival time difference, the IMU acceleration magnitude, and the IMU angular velocity magnitude yields a motion characteristic change term. The summation value is taken at the same millisecond-level timestamp and the physical units remain unchanged to ensure comparability. Using the node signal stability evaluation value as an exponential function of the natural constant e, a signal stability amplification term is obtained. This amplification term is used to increase the effective weight of motion information when stability is high and to suppress sudden noise when stability is low. Multiplying the system motion characteristic change term by the signal stability amplification term yields the cooperative change basis term. The product is numerically continuous and monotonic for subsequent threshold determination. The Bluetooth signal strength is then... The absolute value of the difference between the degree and the Bluetooth signal compensation value is added to one to obtain the signal difference constraint term. This addition is used to avoid unstable ratios caused by extremely small differences. The voltage suppression adjustment term is obtained by taking the exponential function value of the natural constant e with the negative of the node voltage as the exponent and adding one. This term increases the penalty at low voltage and decreases the penalty at high voltage to reflect the power supply status. The communication reliability correction term is obtained by subtracting the received packet loss rate from the constant and adding the minima. The minima is a very small positive real number that is not zero, used to avoid the denominator being zero due to boundary values. The location coordination evaluation value is obtained by dividing the basic term of cooperative change by the product of the signal difference constraint term, the voltage suppression adjustment term, and the communication reliability correction term. After the calculation is completed, the corresponding timestamp and node number are recorded for comparison and traceability.

[0051] The specific formula for calculating the location synergy assessment value is as follows:

[0052] ;

[0053] In the formula, This indicates the positioning synergy assessment value. Indicates the relative distance between nodes. This represents the relative distance between nodes in the previous sampling period. Indicates the time difference of node arrival. Indicates the magnitude of IMU acceleration. Indicates the magnitude of the IMU angular velocity. This represents the node signal stability assessment value. Indicates Bluetooth signal strength. Indicates the Bluetooth signal compensation value. Indicates node voltage. Indicates the packet loss rate. Indicates a minus term.

[0054] In this embodiment, Table 1 is a data table of positioning coordination evaluation values, listing the positioning coordination evaluation values ​​and related calculation data for five target positioning coordinates. Specifically: Target positioning coordinate 1: The relative distance between nodes is 3.20, the relative distance between nodes in the previous sampling period is 3.05, the node arrival time difference is 0.0020, the IMU acceleration magnitude is 0.45, the IMU angular velocity magnitude is 0.12, the node signal stability evaluation value is 1.10, the Bluetooth signal strength is -60.0, the Bluetooth signal compensation value is -58.5, the node voltage is 3.70, the received packet loss rate is 0.04, and the corresponding positioning coordination evaluation value is 0.882. Target positioning coordinates 2: Relative distance between nodes is 2.85, relative distance between nodes in the previous sampling period is 2.80, node arrival time difference is 0.0015, IMU acceleration magnitude is 0.30, IMU angular velocity magnitude is 0.10, node signal stability evaluation value is 0.85, Bluetooth signal strength is -65.0, Bluetooth signal compensation value is -63.0, node voltage is 3.85, packet loss rate is 0.06, and the corresponding positioning cooperation evaluation value is 0.367. Target positioning coordinates 3: Relative distance between nodes is 4.10, relative distance between nodes in the previous sampling period is 3.95, node arrival time difference is 0.0022, IMU acceleration magnitude is 0.60, IMU angular velocity magnitude is 0.20, node signal stability evaluation value is 1.40, Bluetooth signal strength is -57.0, Bluetooth signal compensation value is -56.0, node voltage is 3.60, packet loss rate is 0.03, and the corresponding positioning cooperation evaluation value is 1.937. Target location coordinates 4: Relative distance between nodes is 3.50, relative distance between nodes in the previous sampling period is 3.45, node arrival time difference is 0.0018, IMU acceleration magnitude is 0.25, IMU angular velocity magnitude is 0.08, node signal stability evaluation value is 0.70, Bluetooth signal strength is -62.0, Bluetooth signal compensation value is -60.0, node voltage is 3.75, received packet loss rate is 0.08, and the corresponding positioning coordination evaluation value is 0.272. Target location coordinates 5: Relative distance between nodes is 2.60, relative distance between nodes in the previous sampling period is 2.55, node arrival time difference is 0.0012, IMU acceleration magnitude is 0.35, IMU angular velocity magnitude is 0.15, node signal stability evaluation value is 1.20, Bluetooth signal strength is -59.0, Bluetooth signal compensation value is -57.5, node voltage is 3.65, received packet loss rate is 0.05, and the corresponding positioning coordination evaluation value is 0.751.

[0055] surface Location synergy assessment value data table

[0056] like Figure 4The figure shows the positioning coordination evaluation values ​​of five target coordinates under the same sampling period and their relationship with the coordination threshold. Two colors are used to distinguish the judgment results: bars exceeding the coordination threshold are represented in blue, otherwise in orange. The black dashed line in the figure is the coordination threshold baseline, used to determine whether the coordinate correction process is triggered, visually reflecting the stability of the coordinated positioning state. As can be seen from the figure: the positioning coordination evaluation values ​​of target coordinates 1, 3, and 5 are all higher than the coordination threshold, and the current target positioning coordinates are retained; the positioning coordination evaluation values ​​of target coordinates 2 and 4 are both lower than the coordination threshold, and the Mesh network self-calibration mechanism should be triggered to perform local weighted adjustment based on the relative distance between nodes and the reference node coordinates to correct the current target positioning coordinates. Figure 4 The visual presentation of color partitions and threshold dashed lines intuitively supports the judgment logic of the positioning optimization and self-calibration module of this invention.

[0057] In this implementation scheme, the relative distance between nodes, the relative distance between nodes in the previous sampling period, the time difference of node arrival, the magnitude of IMU acceleration, the magnitude of IMU angular velocity, and the node signal stability evaluation value are coupled as basic terms for coordinated change. Then, a signal difference constraint term consisting of Bluetooth signal strength and Bluetooth signal compensation value, a voltage suppression adjustment term formed by node voltage, and a communication reliability correction term formed by received packet loss rate and minima are jointly constrained. Finally, a positioning coordination evaluation value is output, which unifies motion disturbance, amplitude deviation, power status, and link quality into a single thresholdable quantitative index. This quantitative index has the characteristics of numerical monotonicity and robust denominator, which makes it easy to set a stable coordination threshold and keep the trigger boundary from drifting, thereby improving the accuracy of abnormal triggering, reducing the probability of false correction, and suppressing misleading fluctuations caused by low voltage and high loss rate. Moreover, each judgment is bound to a timestamp and node number, forming a traceable decision basis, ensuring that subsequent coordinate correction steps are executed within a clear and controlled boundary.

[0058] Specifically, the steps for determining whether to perform coordinate correction based on the results, and achieving adaptive positioning optimization, are as follows: Real-time comparison of the positioning coordination evaluation value and the coordination threshold. When the positioning coordination evaluation value is greater than the coordination threshold, the current target positioning coordinates are retained. Millisecond-level timestamps are used for synchronous comparison and recording of the determination results for backtracking. When the positioning coordination evaluation value is less than or equal to the coordination threshold, the Mesh network self-calibration mechanism is triggered. The neighboring node with the highest node signal stability evaluation value is selected as the reference node. The set of neighboring nodes within the current timestamp consists of nodes with valid relative distances between them. The coordinates of the reference node are derived from the established Bluetooth Mesh positioning coordinate system. Local weighted adjustment is performed based on the relative distance between nodes and the reference node coordinates. The weights increase with the increase of the node signal stability evaluation value and decrease with the increase of the residual of the relative distance between nodes. The system specifies that the neighborhood size of the reference node is five, selected in ascending order of relative distance between nodes, and the node signal stability evaluation value is not lower than the current timestamp quantile threshold. The weight function adopts a Gaussian form and uses the square of the residual as the independent variable for exponential decay. Robust suppression uses the Huber kernel to perform a quadratic penalty when the residual does not exceed the threshold and a linear penalty when the residual exceeds the threshold. The iteration starts with the current target positioning coordinates as the initial value, and resets the weights and updates the correction after each solution. The maximum number of iterations is twenty. The convergence criterion is that the Euclidean norm of the coordinate correction is less than one millimeter for two consecutive times and the change in the positioning coordination evaluation value is less than one times ten minus six. After convergence, the system stops and outputs the corrected current target positioning coordinates. If convergence is not achieved but the maximum number of iterations is reached, the final result is output and marked as a sample that needs to be reviewed, and the coordinate record and quality label corresponding to the timestamp are updated simultaneously.

[0059] In this implementation scheme, the real-time comparison between the positioning coordination evaluation value and the coordination threshold is used as the trigger premise to realize the correction of the target positioning coordinates. The correction process has a definite convergence boundary and controllable computational amount, avoiding coordinate oscillation and over-correction. The output current target positioning coordinates and quality label are updated synchronously, forming a traceable closed-loop correction link and improving the robustness and verifiability of online positioning results.

[0060] like Figure 2As shown, the second aspect of the present invention provides a Bluetooth mesh fingerprint compensation positioning device for dynamic environments, comprising: a data acquisition and preprocessing module, an environment compensation and correction module, a feature extraction and positioning calculation module, and a positioning optimization and self-calibration module, wherein: the data acquisition and preprocessing module is used to periodically acquire collaborative positioning sensing data, and perform time synchronization, noise suppression, abrupt change smoothing, anomaly detection, and missing data imputation on the acquired data to obtain preprocessed collaborative positioning sensing data; the environment compensation and correction module is used to extract the preprocessed ambient light intensity and Bluetooth signal strength, determine whether the Bluetooth signal strength has interference anomalies, and compensate the Bluetooth signal strength in combination with the corresponding collaborative positioning sensing data; the feature extraction and positioning calculation module is used to extract real-time collaborative positioning sensing data according to node number, evaluate the signal stability of each node, construct a fingerprint feature self-supervised model based on the evaluation results, output fingerprint feature samples, and construct a fingerprint feature index matrix; the positioning optimization and self-calibration module is used to establish a Bluetooth Mesh positioning coordinate system based on the fingerprint feature index matrix, and obtain the target positioning coordinates in combination with the collaborative positioning sensing data corresponding to the real-time fingerprint feature samples, evaluate the collaborative consistency of target positioning, and determine whether to perform coordinate correction based on the results to achieve adaptive positioning optimization.

[0061] In this implementation scheme, the Bluetooth mesh fingerprint compensation positioning device for dynamic environments achieves closed-loop operation of acquisition, judgment, compensation, feature learning, fingerprint retrieval, and coordinate generation through a continuous link from collaborative positioning sensing data, ambient light intensity compensation, fingerprint feature self-supervised model, fingerprint feature index matrix, Bluetooth Mesh positioning coordinate system to target positioning coordinates. It can still stably output target positioning coordinates under conditions of strong light change, temperature fluctuation, humidity change, and link jitter. It reduces sample distortion caused by nonlinear attenuation of signal-to-noise ratio by relying on the linkage compensation of Bluetooth signal strength and ambient light intensity, uses the node signal stability evaluation value to screen high-quality fingerprint feature samples and improve matching reliability, combines self-supervised training to reduce external annotation dependence and maintain comparability across sampling periods, and triggers self-calibration through threshold consistency judgment to enable rapid convergence of positioning results. Thus, it improves positioning accuracy and verifiability in continuous operation scenarios, reduces online computing latency and maintenance costs, and meets the real-time collaborative positioning requirements under large-scale mesh deployment.

[0062] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0063] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A Bluetooth mesh fingerprint compensation positioning method for dynamic environments, characterized in that, Includes the following steps: S1 periodically collects collaborative positioning and sensing data, performs time synchronization, noise suppression, abrupt change smoothing, anomaly detection and missing data imputation on the collected data, and obtains preprocessed collaborative positioning and sensing data; S2, extract the preprocessed ambient light intensity and Bluetooth signal strength, determine whether the Bluetooth signal strength is abnormally interfered, and compensate the Bluetooth signal strength in combination with the corresponding cooperative positioning and sensing data. S3. Extract real-time collaborative positioning and sensing data according to node number, evaluate the signal stability of each node, construct a fingerprint feature self-supervised model based on the evaluation results, output fingerprint feature samples, and construct a fingerprint feature index matrix. S4. Establish a Bluetooth Mesh positioning coordinate system based on the fingerprint feature index matrix, and obtain the target positioning coordinates by combining the collaborative positioning perception data corresponding to the real-time fingerprint feature samples. Evaluate the collaborative consistency of target positioning, and determine whether to perform coordinate correction based on the results to achieve adaptive positioning optimization.

2. The Bluetooth mesh fingerprint compensation positioning method for dynamic environments according to claim 1, characterized in that: The specific steps for periodically collecting collaborative positioning and sensing data, and performing time synchronization, noise suppression, abrupt change smoothing, anomaly detection, and missing data imputation on the collected data to obtain preprocessed collaborative positioning and sensing data are as follows: A fixed sliding time window is set as a sampling period to collect collaborative positioning and sensing data generated by Bluetooth Mesh nodes in real time during the positioning and coordination process. The collaborative positioning and sensing data includes node number, Bluetooth signal strength, antenna phase difference, node arrival time difference, ambient light intensity, ambient temperature, ambient humidity, node voltage, IMU acceleration, IMU angular velocity, signal-to-noise ratio, and received packet loss rate. For the collected cooperative positioning and sensing data, time synchronization and timing unification are performed using a clock drift estimation algorithm based on Kalman filtering, combined with an adaptive median filtering algorithm for noise suppression and abrupt smoothing. Anomalies in the cooperative positioning and sensing data are identified and removed using a local outlier algorithm to eliminate outliers caused by node failures and signal reflections. Missing time segments in the cooperative positioning and sensing data are interpolated using a K-nearest neighbor-based interpolation algorithm to restore continuous feature sequences. Finally, the cooperative positioning and sensing data are numerically normalized using a minimum-maximum normalization algorithm to unify numerical scales and eliminate dimensional differences.

3. The Bluetooth mesh fingerprint compensation positioning method for dynamic environments according to claim 1, characterized in that: The specific steps for extracting the preprocessed ambient light intensity and Bluetooth signal strength, and determining whether the Bluetooth signal strength shows any interference anomalies, are as follows: The preprocessed ambient light intensity and Bluetooth signal strength are extracted, and an ambient light intensity sequence and a Bluetooth signal strength sequence are constructed in time stamp order. A sliding window difference operation is performed on the ambient light intensity sequence to calculate the rate of change of light intensity in adjacent sampling periods. When the absolute value of the rate of change of light intensity is greater than the light intensity change threshold, the corresponding time interval is marked as a light intensity change interval. Within the light intensity change interval, the variance and rate of change of Bluetooth signal strength are calculated. When the rate of change of signal strength exceeds the signal change threshold and the signal variance exceeds the signal fluctuation threshold, it is determined that the Bluetooth signal strength has an interference anomaly.

4. The Bluetooth mesh fingerprint compensation positioning method for dynamic environments according to claim 1, characterized in that: The specific steps for compensating Bluetooth signal strength by combining corresponding cooperative positioning and sensing data are as follows: Extract the corresponding cooperative positioning sensing data within the illumination change interval, calculate the absolute values ​​of the differences between ambient light intensity and ambient light intensity in the previous sampling period, ambient temperature and ambient humidity in the previous sampling period, add the three absolute values ​​together, add one, and take the natural logarithm to obtain the composite environmental disturbance term; divide the absolute value of the node arrival time difference by the absolute value of the node arrival time difference plus one to obtain the signal propagation delay factor; multiply the composite environmental disturbance term by the signal propagation delay factor, and take the exponent of the product with the negative of the product as the exponent of the natural constant e to obtain the illumination attenuation correction term; multiply the Bluetooth signal strength, illumination attenuation correction term, signal propagation delay factor, and cosine of antenna phase difference in sequence to obtain the Bluetooth signal compensation value; The Bluetooth signal compensation values ​​of each Bluetooth Mesh node are synchronized and aligned according to the timestamp to construct a Bluetooth signal compensation value sequence. A moving average filter is performed on the Bluetooth signal compensation value sequence and the signal change rate is recalculated. When the signal change rate is lower than the signal change threshold for N consecutive sampling periods, the corresponding time interval is marked as the compensation stable interval, and the original Bluetooth signal strength is replaced with the Bluetooth signal compensation value.

5. The Bluetooth mesh fingerprint compensation positioning method for dynamic environments according to claim 1, characterized in that: The specific steps for extracting real-time collaborative positioning and sensing data by node number and evaluating the signal stability of each node are as follows: Extract the Bluetooth signal strength, time difference of arrival, and antenna phase difference of each node according to the node number; perform inter-node differential operation on the Bluetooth signal strength of multiple nodes under the same timestamp to calculate the signal strength difference and time difference of arrival of adjacent nodes; and combine the signal strength difference and time difference of arrival of adjacent nodes into a signal spatial feature sequence according to the node number. Based on the spatial topology between nodes, least squares fitting is performed using the signal spatial feature sequence to calculate the relative distance between nodes. Then, real-time cooperative positioning and sensing data of each node are extracted. The absolute value of the node's Bluetooth signal compensation value is taken, one is added, and the natural logarithm is taken to obtain the amplitude quantization term. The signal amplitude logarithm quantization term, the cosine of the antenna phase difference, and a constant are added to the signal-to-noise ratio in sequence to obtain the signal spatial consistency term. The constant, the square of the node arrival time difference, the square of the difference between the relative distance between nodes and the relative distance between nodes in the previous sampling period, the IMU acceleration magnitude, the IMU angular velocity magnitude, and the received packet loss rate are added in sequence to obtain the signal propagation disturbance term. The signal spatial consistency term is divided by the signal propagation disturbance term to obtain the stable intermediate value. The stable intermediate value is multiplied by the natural logarithm of the node voltage plus one to obtain the node signal stability evaluation value.

6. The Bluetooth mesh fingerprint compensation positioning method for dynamic environments according to claim 1, characterized in that: The specific steps for constructing a fingerprint feature self-supervised model based on the evaluation results, outputting fingerprint feature samples, and constructing a fingerprint feature index matrix are as follows: Historical fingerprint feature samples are extracted from the fingerprint feature database and fused with the node signal stability evaluation values ​​calculated from the corresponding cooperative localization sensing data. A self-supervised fingerprint feature model is constructed by combining the self-supervised variational autoencoder algorithm. The signal distribution variation pattern during different sampling periods is learned through feature reconstruction. The change in relative distance between nodes is used as a self-supervised reference term to adaptively adjust the feature consistency index output by the model, thus completing the model training. Real-time sampling period cooperative localization sensing data and node signal stability evaluation values ​​are input into the fingerprint feature self-supervised model to output fingerprint feature samples. The cooperative localization sensing data, relative distance between nodes, node signal stability evaluation values, and fingerprint feature samples are indexed and sorted according to node number and sampling timestamp to establish a fingerprint feature index matrix.

7. The Bluetooth mesh fingerprint compensation positioning method for dynamic environments according to claim 1, characterized in that: The specific steps for establishing a Bluetooth Mesh positioning coordinate system based on the fingerprint feature index matrix and obtaining the target positioning coordinates by combining the collaborative positioning sensing data corresponding to the real-time fingerprint feature samples are as follows: The node signal stability evaluation value of each node is extracted from the fingerprint feature index matrix and sorted from largest to smallest. The node with the highest node signal stability evaluation value is defined as the origin of the coordinate system, the second highest node is defined in the positive X-axis direction, and the third highest node is defined in the XY plane to establish a Bluetooth Mesh positioning coordinate system. The historical fingerprint feature samples of each node are input into the collaborative positioning calculation engine to obtain the corresponding node spatial coordinates. The Euclidean distance similarity algorithm is used to calculate the feature distance between the real-time fingerprint feature sample and each fingerprint feature sample in the fingerprint feature index matrix. The top M matching samples are selected as candidate localization sets according to the feature distance from low to high, and the corresponding node number and node spatial coordinates are recorded. A weighted average calculation is performed on the spatial coordinates of nodes in the candidate positioning reference set. The reciprocal of the node signal stability evaluation value and the variance of the node arrival time difference are used as joint weights to obtain the weighted positioning coordinates, forming the initial positioning coordinates. Then, IMU acceleration and angular velocity data are used to perform integral compensation on the node displacement to correct the coordinate deviation caused by the motion and obtain the target positioning coordinates.

8. The Bluetooth mesh fingerprint compensation positioning method for dynamic environments according to claim 1, characterized in that: The specific steps for assessing the synergistic consistency of target positioning are as follows: The motion characteristic variation term is obtained by summing the absolute value of the difference between the relative distance between nodes and the relative distance between nodes in the previous sampling period, the node arrival time difference, the magnitude of IMU acceleration, and the magnitude of IMU angular velocity. The signal stability amplification term is obtained by taking the exponential function of the node signal stability evaluation value with respect to the natural constant e. The system motion characteristic variation term is multiplied by the signal stability amplification term to obtain the basic term of cooperative variation. The signal difference constraint term is obtained by adding one to the absolute value of the difference between the Bluetooth signal strength and the Bluetooth signal compensation value. The voltage suppression adjustment term is obtained by taking the exponential function of the natural constant e with respect to the negative of the node voltage and adding one. The communication reliability correction term is obtained by subtracting the received packet loss rate from the constant and adding a minima. The positioning cooperative evaluation value is obtained by dividing the basic term of cooperative variation by the product of the signal difference constraint term, the voltage suppression adjustment term, and the communication reliability correction term.

9. The Bluetooth mesh fingerprint compensation positioning method for dynamic environments according to claim 1, characterized in that: The specific steps for determining whether to perform coordinate correction based on the results to achieve adaptive positioning optimization are as follows: The system compares the positioning coordination evaluation value and the coordination threshold in real time. When the positioning coordination evaluation value is greater than the coordination threshold, the current target positioning coordinates are retained. When the positioning coordination evaluation value is less than or equal to the coordination threshold, the Mesh network self-calibration mechanism is triggered. The neighboring node with the highest node signal stability evaluation value is selected as the reference node. Based on the relative distance between nodes and the coordinates of the reference node, a local weighted adjustment is performed to correct the current target positioning coordinates.

10. A Bluetooth mesh fingerprint compensation positioning device for dynamic environments, characterized in that: include: The module comprises a data acquisition and preprocessing module, an environmental compensation and correction module, a feature extraction and localization calculation module, and a localization optimization and self-calibration module, among which: The data acquisition and preprocessing module is used to periodically acquire collaborative positioning and sensing data, and to perform time synchronization, noise suppression, mutation smoothing, anomaly detection and missing data imputation on the acquired data to obtain preprocessed collaborative positioning and sensing data. The environmental compensation and correction module is used to extract the preprocessed ambient light intensity and Bluetooth signal strength, determine whether the Bluetooth signal strength is abnormally interfered, and compensate the Bluetooth signal strength in combination with the corresponding cooperative positioning and sensing data. The feature extraction and localization calculation module is used to extract real-time collaborative localization sensing data according to node number, evaluate the signal stability of each node, construct a fingerprint feature self-supervised model based on the evaluation results, output fingerprint feature samples, and construct a fingerprint feature index matrix. The positioning optimization and self-calibration module is used to establish a Bluetooth Mesh positioning coordinate system based on the fingerprint feature index matrix, and obtain the target positioning coordinates by combining the collaborative positioning perception data corresponding to the real-time fingerprint feature samples. It evaluates the collaborative consistency of the target positioning, determines whether to perform coordinate correction based on the results, and realizes adaptive positioning optimization.

Images