Bluetooth indoor fusion positioning method based on hybrid filtering and double weight distribution

Abstract

This invention belongs to the field of Bluetooth indoor positioning technology and discloses a Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation. The method first suppresses outliers, local mutations, and random fluctuations in the original RSSI sequence through a multi-level hybrid filtering preprocessing mechanism, improving the stability of RSSI observation data. Then, it constructs a first-level dynamic weight to evaluate the observation quality of different Bluetooth beacons in real time, reducing the impact of low-quality beacons on positioning solutions. Furthermore, it constructs a second-level dynamic weight to adaptively evaluate the output stability and current applicability of different positioning algorithms and fuses the positioning results of multiple single positioning algorithms. Through the above design, a complete positioning process of "signal preprocessing - beacon quality evaluation - multi-algorithm positioning - algorithm weight fusion" is formed, enabling the method of this invention to achieve higher positioning accuracy, better trajectory continuity, and stronger robustness in complex dynamic environments.

Description

Technical Field

[0001] This invention belongs to the field of Bluetooth indoor positioning technology, and particularly relates to a Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation. Background Technology

[0002] Bluetooth Low Energy (BLE) has become one of the important technologies in the field of indoor positioning due to its advantages such as low cost, low power consumption, flexible deployment, and convenient networking.

[0003] Existing BLE indoor positioning systems typically use Received Signal Strength Indicator (RSSI) to characterize the attenuation of wireless signal propagation, and perform ranging based on the relationship between RSSI and propagation distance. They then combine positioning algorithms such as weighted centroid method, trilateration method, or Bayesian positioning method to obtain the target location.

[0004] However, in actual indoor environments, RSSI signals are highly susceptible to multipath effects, shadow fading, wall obstruction, metal reflection, people walking around, and hardware thermal noise, resulting in problems such as large random fluctuations, obvious local spikes, short-term discontinuities, and sudden jumps in the original observation sequence.

[0005] Even if the distance measurement model parameters have been pre-set or calibrated, the original RSSI observations themselves may still contain strong noise, which will cause large errors in the distance estimation results and further affect the accuracy and stability of the target position calculation.

[0006] Besides signal fluctuation issues, existing BLE indoor positioning methods generally suffer from insufficient algorithm adaptability in the positioning calculation stage. Different positioning algorithms exhibit inconsistent performance under different scenarios and observation conditions.

[0007] For example, trilateration has high geometric accuracy under favorable observation conditions, but it is sensitive to ranging errors and the geometry of the base station. Weighted centroid method has some robustness, but its accuracy has a limited upper limit; Bayesian positioning method is adaptable to uncertain observations, but it depends on probability distribution modeling and prior conditions.

[0008] Therefore, in complex and dynamic indoor environments, relying solely on a single positioning algorithm often fails to simultaneously achieve accuracy, stability, and robustness. To improve BLE positioning performance, some solutions have attempted to filter the RSSI input or fuse multiple positioning algorithms at the output. However, most existing solutions still suffer from the following shortcomings: 1. Existing BLE indoor positioning solutions mostly use a single filtering or result correction method for RSSI signals, which makes it difficult to effectively suppress pulse anomalies, local mutations and random fluctuations at the same time, resulting in insufficient stability of ranging and positioning results in complex indoor environments.

[0009] 2. Existing solutions typically lack a dynamic evaluation mechanism for the differences in observation quality among different Bluetooth beacons. Low-quality observation sources may still have a large weight in the positioning solution, thus affecting the overall positioning accuracy.

[0010] 3. Existing multi-algorithm fusion methods mostly use fixed weights, empirical weights, or weighting within a single model, lacking an adaptive weight adjustment mechanism based on temporal stability. This means that locally failed algorithms may still participate in the fusion, reducing the robustness of the localization results.

[0011] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art. Summary of the Invention

[0012] To address the aforementioned technical problems, the present invention aims to propose a Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation, so as to improve the accuracy, stability and robustness of BLE positioning in complex indoor environments.

[0013] To achieve the above objectives, the present invention adopts the following technical solution: The Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation includes the following steps: Step 1. Perform multi-level hybrid filtering preprocessing on the original RSSI sequences of each Bluetooth beacon to obtain the final filtered RSSI sequence and the filtered RSSI representative value of each Bluetooth beacon in each window; Step 2. Then, based on the RSSI stability index, continuity index, intensity rationality index, and historical consistency, a first-level dynamic weight is constructed to evaluate the current observation quality of different Bluetooth beacons; Step 3. Using the filtered RSSI representative value as the basic observation input, the first-level dynamic weight is introduced into three different localization algorithms: weighted centroid localization, trilateration, and Bayesian localization, and the localization results of each algorithm are obtained respectively. Step 4. Based on the residual stability, algorithm applicability, and overall signal quality of each positioning algorithm in continuous time, construct the second-level dynamic weights for each positioning algorithm; and dynamically weight and fuse the position estimation results of each positioning algorithm according to the second-level dynamic weights to obtain the final fused positioning result of the target node.

[0014] The present invention has the following advantages: As described above, this invention discloses a Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation. This method uses a multi-level hybrid filtering preprocessing mechanism to hierarchically suppress outliers, local mutations, and random fluctuations in BLE RSSI signals, improving the stability and usability of the original observation data. This processing provides a more reliable input foundation for subsequent beacon quality evaluation, ranging calculation, and multi-algorithm positioning, reducing the interference of complex indoor environments on positioning results from the source. This invention constructs a first-level dynamic weight allocation mechanism to evaluate the observation quality of different Bluetooth beacons in real time, allowing beacons with stable, continuous, and reasonable RSSI to contribute more to the positioning solution, while reducing the weight of low-quality beacons affected by obstruction, multipath, or abnormal fluctuations. This mechanism can prevent low-quality observation sources from directly participating in equal-weight positioning, thereby improving the reliability of single-positioning algorithm positioning results. Furthermore, this invention further constructs a second-level dynamic weight allocation mechanism to adaptively evaluate the output stability and current applicability of different positioning algorithms, and completes multi-algorithm fusion under overall signal quality constraints. The second-level dynamic weight allocation mechanism works synergistically with the multi-level hybrid filtering and the first-level dynamic weight to achieve a continuous optimization process from "signal preprocessing - beacon quality screening - adaptive fusion of algorithm results". This reduces the impact of abnormal observation sources and local failure algorithms on the final positioning results and improves the accuracy, continuity and robustness of BLE positioning in complex indoor environments. Attached Figure Description

[0015] Figure 1 This is a flowchart of the Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation in an embodiment of the present invention; Figure 2 This is a schematic diagram of the arrangement of Bluetooth beacons in an embodiment of the present invention; Figure 3 This is a flowchart of the multi-stage hybrid filtering preprocessing in an embodiment of the present invention. Detailed Implementation

[0016] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: Example 1 To address the problems of large RSSI signal fluctuations, difficulty in effectively identifying low-quality beacons, insufficient adaptability of single positioning algorithms, and fixed weights in multi-algorithm fusion in complex indoor environments, this embodiment 1 provides a Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation. This method first suppresses outliers, local mutations, and random fluctuations in the original RSSI sequence through a multi-level hybrid filtering preprocessing mechanism, improving the stability of RSSI observation data. Then, a first-level dynamic weight is constructed to evaluate the observation quality of different Bluetooth beacons in real time, reducing the impact of low-quality beacons on positioning solutions. Furthermore, a second-level dynamic weight is constructed to adaptively evaluate the output stability and current applicability of different positioning algorithms, and the positioning results of multiple single positioning algorithms are fused. Through the above design, a complete positioning process of "signal preprocessing - beacon quality evaluation - multi-algorithm positioning - algorithm weight fusion" is formed, enabling the method of this invention to achieve higher positioning accuracy, better trajectory continuity, and stronger robustness in complex dynamic environments.

[0017] like Figure 1 As shown, the Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation includes the following steps: Step 1. Perform multi-level hybrid filtering preprocessing on the original RSSI sequences of each Bluetooth beacon to obtain the filtered RSSI sequence and the filtered RSSI representative value of each Bluetooth beacon in each window.

[0018] Step 1 is used to collect the raw RSSI sequences of each Bluetooth beacon and preprocess them using a conventional hybrid filtering process. The main purpose is to reduce the impact of outliers, local mutations and random fluctuations on subsequent ranging and weight calculation.

[0019] Since the interquartile range detection, median filtering, Kalman filtering, and smoothing used in step 1 are all existing signal preprocessing methods, this invention does not emphasize improvements to the individual filtering algorithms themselves.

[0020] Step 1 serves as a preliminary data cleaning step for dynamic weight allocation and fusion positioning in subsequent steps.

[0021] like Figure 2 As shown, four low-power Bluetooth beacons, AP1, AP2, AP3, and AP4, are deployed within an indoor positioning area.

[0022] Four Bluetooth beacons are fixed around the positioning area, and a receiving terminal is placed on the target to be located to receive the RSSI signals emitted by each Bluetooth beacon in real time. Each Bluetooth beacon continuously broadcasts Bluetooth signals, and the receiving terminal records the four RSSI observations in chronological order, forming four sets of raw RSSI time series data.

[0023] In this embodiment, a sliding window method is used to process the original RSSI sequence, and the length of the sliding window is taken as... That is, there are 20 sampling points in a sliding window.

[0024] Let the first The Bluetooth beacon in the 1st The original RSSI sequences within each sliding window are: ; in Indicates the first The Bluetooth beacon in the 1st The original RSSI sequence within each window; Indicates the first RSSI values ​​at each sampling point; Indicates the length of the sliding window. .

[0025] For each Bluetooth beacon's RSSI sequence within the current window, an anomaly detection method based on interquartile range is first used to identify pulse anomalies, and sampling points outside the valid interval are replaced using the median of the window. Then, an adaptive median filtering method is used to identify local anomalies using the absolute deviation of the median and perform median replacement. Next, a one-dimensional Kalman filter model is used to recursively estimate the processed RSSI sequence to reduce the influence of random noise. Finally, a short-window moving average and smoothed Gaussian filter are performed on the Kalman filter output, where the short-window moving average window length is set to 4, to obtain the final smoothed RSSI sequence.

[0026] like Figure 3 As shown below, the specific process of multi-stage hybrid filtering preprocessing will be explained in detail.

[0027] Step 1.1. First, use the interquartile range method to identify outliers within the window.

[0028] Define interquartile range for: .

[0029] in and These represent the first and third quartiles of the RSSI sequence for this window, respectively. When Exceeding the range At that time, a window-based substitution was used to obtain the first-level processing result. .

[0030] ; in This represents the RSSI value after primary processing; Indicates the anomaly detection coefficient. To find the median.

[0031] Step 1.2. Then, adaptive median filtering is used to correct local abrupt changes. Let the first... The local window centered on each sampling point is its local median and median absolute deviation They are respectively: ; ; when At that time, a local median replacement was used to obtain the secondary processing result. .in Indicates the local anomaly determination coefficient; Represents a very small positive number. Preferred .

[0032] Step 1.3. Then proceed with... Perform a one-dimensional Kalman filter. The recursive process is as follows: ; ; ; .

[0033] in and These are the prior and posterior RSSI state estimates, respectively; and Let represent the prior and posterior estimation error covariances, respectively; Indicates Kalman gain; and These are the process noise covariance and the observation noise covariance, respectively.

[0034] Step 1.4. Finally, perform short-window smoothing on the Kalman filter results to obtain the final filtered RSSI value: .

[0035] in, This represents the final filtered RSSI value; Indicates smoothing weights; This indicates a smooth window.

[0036] No. The Bluetooth beacon in the 1st The final filtered RSSI sequence after step 1 processing within each window is as follows: ; in Indicates the first The Bluetooth beacon in the 1st The final filtered RSSI sequence within each window; This indicates the first [number] in this window. The final filtered RSSI value for each sampling point; Indicates the length of the sliding window.

[0037] The average of the final filtered results within the window is used to obtain the representative RSSI value for subsequent ranging and first-level dynamic weight calculation: .

[0038] in, Indicates the first The Bluetooth beacon in the 1st The filtered RSSI value is represented within each window.

[0039] Through the hybrid filtering process in steps 1.1 to 1.4 above, the original RSSI sequence is converted into a more stable RSSI representative value, providing basic observation input for subsequent beacon quality assessment and multi-algorithm positioning solution.

[0040] Multi-level hybrid filtering preprocessing organically combines interquartile range-based anomaly detection, adaptive median filtering, Kalman filtering, and short-window smoothing to achieve layered suppression of pulse anomalies, local mutations, and random fluctuations in RSSI signals.

[0041] The advantages of implementing multi-stage hybrid filtering preprocessing in this invention are as follows: The first step is to use interquartile range (IQR) to handle outliers across the entire window. Some points in the RSSI are impulse outliers that significantly deviate from the overall window distribution. If these extreme points are not addressed first, they will affect subsequent local median, MAD, and Kalman filter state estimations. Therefore, IQR is used for coarse screening to remove the most obvious outliers.

[0042] The second step is to use adaptive median filtering to correct local abrupt changes. IQR mainly considers the statistical distribution of the entire window, but some outliers are not necessarily global outliers, but rather abrupt changes relative to their local neighborhood. Therefore, a second correction using the local median and MAD can handle local short-term jumps.

[0043] The third step is to perform a one-dimensional Kalman filter. One-dimensional Kalman filtering is suitable for handling relatively continuous random fluctuations, but if outliers and local mutations are not removed beforehand, anomalous RSSIs will directly enter the recursive process, skewing the state estimate. Therefore, placing the Kalman filter after anomaly replacement is more in line with its function of smoothing random noise.

[0044] The fourth step is to perform short-window smoothing. After the previous processing, large anomalies and local mutations in the RSSI sequence have been weakened. The final short-window smoothing is mainly used to suppress residual high-frequency fluctuations and make the RSSI representative value within the window more stable.

[0045] Step 2. Then, based on the RSSI stability index, continuity index, intensity rationality index, and historical consistency, a first-level dynamic weight is constructed to evaluate the current observation quality of different Bluetooth beacons.

[0046] Based on the filtered RSSI sequence and RSSI representative value obtained in step 1, the observation quality of different Bluetooth beacons within the current window is dynamically evaluated, and the evaluation result is converted into a first-level dynamic weight that can be directly used in subsequent positioning calculations.

[0047] During BLE indoor positioning, the RSSI observation quality of different Bluetooth beacons often varies significantly due to differences in spatial location, propagation path, obstruction status, and multipath environment.

[0048] For example, beacons that are close to the target and unobstructed usually have a relatively stable RSSI sequence; while beacons that are obstructed by people, walls, or are subject to strong multipath effects may have RSSI sequences that fluctuate violently, jump briefly, or have abnormal intensity.

[0049] If all beacons are treated with equal weight in the positioning calculation, low-quality beacons will still have a significant impact on the positioning results, leading to ranging errors and position shifts. Therefore, after completing multi-level hybrid filtering preprocessing, a first-level dynamic weight is further constructed to quantitatively evaluate the current observation quality of each Bluetooth beacon. This weight mainly considers four aspects: First, is the RSSI sequence stable within the current window? Second, are the RSSI sequences continuous between adjacent windows? Third, is the RSSI strength within a reasonable range? Fourth, is the current signal quality consistent with historical conditions? These evaluations allow high-quality beacons to receive greater weight in subsequent positioning calculations, while weakening the impact of low-quality beacons.

[0050] Step 2.1. First, calculate the first... The Bluetooth beacon in the 1st The three indicators within a window are the RSSI stability index, continuity index, and intensity rationality index. The process for obtaining these three indicators is as follows: I. Calculation of stability index.

[0051] Stability metrics are used to evaluate the degree of fluctuation in the RSSI sequence within the current window. Generally, within a short time window, if the target's movement distance is small and the propagation environment is relatively stable, the RSSI of the same Bluetooth beacon should not exhibit drastic fluctuations.

[0052] If the RSSI variance is large within the current window, it indicates that the beacon may be affected by multipath interference, occlusion variations, or transient noise, and its observation quality should be reduced accordingly. Definition 1 The Bluetooth beacon in the 1st The RSSI variance within each window is: .

[0053] in Indicates the first The Bluetooth beacon in the 1st The variance of the filtered RSSI sequence within a window; This represents the final filtered RSSI value; Indicates the first The Bluetooth beacon in the 1st The RSSI value is represented by a window filter.

[0054] A stability index is constructed based on the RSSI variance, using the following formula: .

[0055] in Indicates the first The Bluetooth beacon in the 1st Stability metrics within a window, This represents the stability adjustment coefficient.

[0056] From the above formula, it can be seen that when A smaller value indicates that the RSSI fluctuation is relatively small within the current window. Approaching 1; when A larger value indicates significant fluctuations in the RSSI within the current window. Decrease.

[0057] The value can be 2-10 dB, and the specific value can be determined based on the performance of the Bluetooth device, the sampling frequency, and the indoor environment calibration data.

[0058] Therefore, the stability index can reflect the stability of the RSSI observations of the beacon within the current window.

[0059] II. Calculation of continuity indicators.

[0060] The continuity index is used to evaluate the degree of change in the RSSI representative value between the current window and the previous window. In continuous positioning scenarios, the position of the target node usually changes continuously over time, and the RSSI should also have a certain degree of temporal continuity.

[0061] If a beacon exhibits a significant jump between adjacent windows, it may indicate that the beacon has been affected by sudden obstruction, movement of people, or abrupt changes in its propagation path. Definition 1 The Bluetooth beacon in the 1st The change in RSSI between adjacent windows within a window is:

[0062] .

[0063] in, Indicates the first The Bluetooth beacon in the 1st The window and the first The RSSI value between windows represents the amount of change in value; Indicates the first The first Bluetooth beacon in the previous window, i.e. The filtered RSSI value for each window.

[0064] A continuity index is constructed based on the changes in adjacent windows, using the following formula: .

[0065] in Indicates the first The Bluetooth beacon in the 1st Continuous indicators within a window; This represents the continuous adjustment coefficient.

[0066] When the RSSI representative values ​​of adjacent windows change little. A larger value indicates that the beacon observations have good temporal continuity; conversely, A decrease indicates that the beacon may be experiencing a sudden change in propagation conditions, and its subsequent location contribution should be reduced.

[0067] Preferably 3-6dB, when When the value is small, continuity indicators are more sensitive to short-term jumps; when... When the value is large, the continuity index is more sensitive to changes in adjacent windows.

[0068] For the first window, since there is no RSSI representative value for the previous window, we can let: ;in, Indicates the first window within the first window Initial values ​​for the continuity index of a Bluetooth beacon.

[0069] III. Calculation of strength rationality index.

[0070] The strength rationality index is used to evaluate whether the current RSSI representative value is within a reasonable signal strength range.

[0071] When the BLE RSSI is too weak, it usually indicates that the beacon is far away, the obstruction is strong, or the signal reception quality is poor; when the RSSI is significantly outside the normal range, it may also indicate that there is equipment malfunction or local observation distortion.

[0072] Therefore, reasonable constraints need to be placed on the RSSI strength. Let the reasonable range of RSSI strength in a BLE indoor positioning system be: . This indicates the lower limit of the reasonable strength range of RSSI; This indicates the upper limit of the reasonable strength range of RSSI.

[0073] The value can be determined based on the Bluetooth device's transmit power, receiver sensitivity, indoor environmental calibration data, or historical RSSI observation statistics. Preferably... -95 to -80 dBm is acceptable. The range is -45 to -20 dBm.

[0074] Define the deviation of the filtered RSSI representative value from the reasonable intensity range as: ; in Indicates the first The Bluetooth beacon in the 1st RSSI intensity deviation within a window.

[0075] A strength rationality index is constructed based on the strength deviation: .

[0076] in For the first The Bluetooth beacon in the 1st The strength rationality index within each window.

[0077] This represents the strength rationality adjustment coefficient. The preferred value is 5 to 10 dB.

[0078] when When within a reasonable strength range, ,at this time This indicates that the beacon's strength has a high availability. When it deviates from the reasonable strength range, Increase A decrease indicates a decline in the current observation quality of the beacon.

[0079] Step 2.2. Then, the RSSI stability index, continuity index, and strength rationality index are weighted and summed to obtain the... The Bluetooth beacon in the 1st Real-time signal quality rating within each window.

[0080] After obtaining the stability index, continuity index, and intensity rationality index, the three types of indices are weighted and summed to obtain the... The Bluetooth beacon in the 1st Real-time signal quality rating within each window : .

[0081] in Indicates stability index; Indicates a continuous indicator; Indicators representing the rationality of strength; , , These represent the weighting coefficients corresponding to the stability index, continuity index, and strength rationality index, respectively.

[0082] The weighting coefficients of each indicator satisfy the following: ;in, , , All are non-negative numbers.

[0083] In a preferred embodiment, the following may be taken: The stability index has a slightly higher weight because the fluctuation of RSSI within the current window can directly reflect whether the beacon observation is reliable; the continuity index is used to reflect whether there are sudden jumps between adjacent windows; and the intensity rationality index is used to constrain whether RSSI is within the usable range.

[0084] Of course, the above , , The parameter values ​​are not unique. Those skilled in the art can make reasonable adjustments based on the specific indoor environment, sampling frequency, and performance of the Bluetooth device, which will not be elaborated here.

[0085] because , , All arrive The indicators between, therefore Also located arrive between. The larger the value, the better the current observation quality of the Bluetooth beacon; The smaller the value, the worse the current observation quality of the Bluetooth beacon.

[0086] Step 2.3. Construct the first The Bluetooth beacon in the 1st Historical consistency factors within each window are used for historical consistency correction.

[0087] Calculating signal quality scores based solely on the current window may cause weights to change drastically due to accidental disturbances.

[0088] For example, if a beacon's RSSI deteriorates briefly within a single window, but its historical observations have been consistently good, immediately and significantly reducing its weight could lead to unnecessary fluctuations in the positioning results.

[0089] Conversely, if a beacon exhibits poor quality for multiple consecutive windows, its weight should be gradually reduced. Therefore, this invention further introduces a historical consistency correction, ensuring that beacon weights reflect not only current quality but also historical conditions.

[0090] Definition of the first The Bluetooth beacon in the 1st The window was recently Average signal quality score within a historical window for: ; in; Indicates the length of the history window; This indicates the index of the history window.

[0091] When the number of history windows is insufficient In this case, the average number of existing historical windows can be used, that is: ; in, Indicates the current window number.

[0092] For the first window, since there are no history windows, we can set: .

[0093] in, Indicates the first window within the first window Initial values ​​of the historical average signal quality score for each Bluetooth beacon. Among them, Indicates the first The Bluetooth beacon in the 1st Real-time signal quality rating within each window.

[0094] The historical consistency factor is further constructed using the following formula: .

[0095] in Indicates the first The Bluetooth beacon in the 1st Historical consistency factor within a window.

[0096] Indicates the historical consistency adjustment coefficient. The preferred value is 0.1-0.3 dBm.

[0097] When the current quality score is close to the historical average quality score, it indicates that the beacon's current state is consistent with its historical state. The larger the value, the better; otherwise, it indicates that the beacon may have undergone a sudden change. This reduces the weights, thereby suppressing sudden changes in weights.

[0098] Step 2.4. Combining the instantaneous signal quality score and the historical consistency factor, we obtain the... The Bluetooth beacon in the 1st The unnormalized quality weights within each window are expressed by the following formula: ; in Indicates the first The Bluetooth beacon in the 1st Unnormalized quality weights within each window.

[0099] The unnormalized quality weights of all Bluetooth beacons are normalized to obtain the first-level dynamic weights, as shown in the following formula: .

[0100] in, Indicates the first The Bluetooth beacon in the 1st The first level of dynamic weight within each window.

[0101] Indicates the first Unnormalized quality weights for each Bluetooth beacon; Indicates the first Unnormalized quality weights for each Bluetooth beacon; Indicates the Bluetooth beacon summation index; This indicates the total number of Bluetooth beacons participating in the location tracking.

[0102] After normalization, the first-level dynamic weights of each Bluetooth beacon satisfy: ; in, Indicates the first The Bluetooth beacon in the 1st The first level of dynamic weight for each window.

[0103] To avoid instability in normalization caused by all beacon quality weights approaching zero simultaneously, the following form can also be used: ; in, This represents an extremely small positive number to prevent the denominator from being too small.

[0104] The first-level dynamic weights can dynamically reflect the observation reliability of each Bluetooth beacon within the current window.

[0105] When a beacon's RSSI is stable, continuous, and of reasonable strength, its Larger, corresponding The larger the beacon's RSSI, the greater its role in subsequent positioning calculations; when a beacon's RSSI fluctuates significantly, exhibits obvious abrupt changes in adjacent windows, or displays abnormal intensity, its... Decrease, corresponding As the beacon's position decreases, its impact on subsequent positioning results is weakened.

[0106] Therefore, the role of the first-level dynamic weights is not to directly output the final position, but to filter and adjust the observation quality of different Bluetooth beacons before entering multi-algorithm positioning.

[0107] Subsequent weighted centroid localization, trilateration, and Bayesian localization all perform position estimation based on this first-level dynamic weight, thereby reducing the impact of low-quality RSSI observations on the localization results of the single localization algorithm.

[0108] After obtaining the smoothed RSSI sequence, the observation quality of each Bluetooth beacon is quantitatively evaluated. In this embodiment, the signal quality score is obtained by weighting three indicators: stability, continuity, and strength reasonableness, with stability weighted at 0.4, continuity weighted at 0.3, and strength reasonableness weighted at 0.3. Through this weighting method, the signal quality score of each Bluetooth beacon within the current window is obtained. Furthermore, to ensure the smoothness of weight changes over time, a historical consistency correction is introduced into the current score, and the corrected weights are normalized to obtain the dynamic weights of the four Bluetooth beacons within the current window.

[0109] Step 3. Using the filtered RSSI representative value as the basic observation input, the first-level dynamic weights are introduced into three different localization algorithms: weighted centroid localization, trilateration, and Bayesian localization, and the localization results of each algorithm are obtained.

[0110] The purpose of introducing first-level dynamic weights into different positioning algorithms in this invention is to enable each positioning algorithm to perceive the differences in the observation quality of different Bluetooth beacons during the location calculation process. Because the RSSI of different beacons may be affected by obstruction, multipath propagation, and sudden interference in complex indoor environments, their observation reliability varies. If each beacon still participates in the positioning calculation with equal weights, the ranging error generated by low-quality beacons will directly affect the positioning result of a single positioning algorithm.

[0111] Therefore, based on the computational principles of different positioning algorithms, this invention introduces the first-level dynamic weights into the comprehensive weights of weighted centroid positioning, the weighted least squares matrix of trilateration, and the weighted likelihood function of Bayesian positioning, respectively. This allows high-quality beacons to play a greater role in the positioning solution, while weakening the influence of low-quality beacons. In this way, the interference of abnormal RSSI observations on the location results can be reduced during the single-positioning algorithm stage, and a more reliable algorithm input can be provided for the subsequent second-level dynamic weight fusion, thereby improving the accuracy, stability, and robustness of the final fused positioning result.

[0112] Using the filtered RSSI representative value obtained in step 1 as the basic observation input, and combined with the first-level dynamic weight, multiple single-positioning algorithm position estimation results are obtained by using the weighted centroid positioning method, the trilateration method, and the Bayesian positioning method, respectively.

[0113] All of the above positioning methods are commonly used in the field of indoor positioning.

[0114] The focus of this invention is not to reintroduce a single localization algorithm, but to embed the first-level dynamic weights into the calculation process of different algorithms, thereby weakening the influence of low-quality beacons on the localization results of each single localization algorithm.

[0115] Based on the smoothed RSSI sequence and path loss model parameters, distance estimates from the target node to the four Bluetooth beacons are calculated. Then, three single-localization algorithms—weighted centroid localization, trilateration, and Bayesian localization—are used to obtain position estimates. In the weighted centroid localization method, the target position is determined by combining the dynamic weights of the observation layer and the distance attenuation relationship. In the trilateration method, a system of equations is established using the position coordinates of the four Bluetooth beacons and their corresponding distance estimates, and then solved with weights. In the Bayesian localization method, the likelihood value of each candidate point is calculated in the candidate location space, and the candidate point with the highest likelihood value is selected as the localization result. Thus, position estimates from three single-localization algorithms with different mechanisms can be obtained simultaneously.

[0116] Step 3.1. First, convert the filtered RSSI representative value into a distance estimate based on the path loss model: .

[0117] in, Indicates the distance from the target node to the... Distance estimates for each Bluetooth beacon; Indicates a reference distance; Indicates the RSSI at the reference distance; Indicates the path loss index; Indicates the first The filtered RSSI value of a Bluetooth beacon.

[0118] Step 3.2. For weighted centroid localization, the comprehensive centroid weight is defined as: ; in, Indicates the first The beacons are used for the comprehensive weighting of the centroid location; This represents the first level of dynamic weights; Indicates the distance attenuation coefficient; This represents a very small positive number. In one implementation, . For the first The Bluetooth beacon in the 1st The first-level dynamic weight within each window; Indicates the distance from the target node to the... Distance estimates for each Bluetooth beacon.

[0119] The weighted centroid localization result is: .

[0120] in, This indicates the weighted centroid localization result; Indicates the first The coordinates of the blue beacon.

[0121] Step 3.3. For trilateration, let the target node position be... .

[0122] A system of equations is established based on distance constraints, and the equations are then used to determine the system of equations. Linearization with each beacon as a reference can be written as: .

[0123] in, Represents the linearized coefficient matrix; This represents a vector of constant terms.

[0124] Since each distance difference equation is simultaneously related to the reference beacon and the... Related to a beacon, defining the equation weights for: .

[0125] And construct a weighted matrix : .

[0126] The weighted least squares solution for trilateration is: ; in, This represents the position estimation result output by the trilateration method.

[0127] When a matrix does not satisfy the invertibility condition, a pseudo-inverse form can be used to solve it.

[0128] Step 3.4. For Bayesian localization, let the candidate location set be: ; in, Indicates the first Candidate location points, .

[0129] Candidate points Corresponding to the The theoretical RSSI for each beacon is: .

[0130] in Candidate points Corresponding to the Theoretical RSSI for a single beacon; To incorporate the first-level dynamic weights into the Bayesian localization process, a weighted RSSI residual is constructed: .

[0131] in, This represents the weighted RSSI residuals of the candidate locations; Indicates the first The standard deviation of RSSI noise for each beacon.

[0132] The corresponding weighted likelihood function for: .

[0133] The Bayesian localization result is: .

[0134] in, This represents the Bayesian localization result.

[0135] Therefore, in the first The three single-positioning algorithm location estimation results are obtained within one window and are uniformly denoted as: .

[0136] in, This indicates the weighted centroid positioning method. This represents the trilateration method. This represents the Bayesian localization method. The above results will serve as inputs for the second-level dynamic weight calculation and final fusion localization in step 4.

[0137] Step 4. Finally, based on the residual stability, algorithm applicability, and overall signal quality of each positioning algorithm in continuous time, a second-level dynamic weight is constructed for each positioning algorithm; and the position estimation results of each positioning algorithm are dynamically weighted and fused according to the second-level dynamic weight to obtain the final fused positioning result of the target node.

[0138] Step 4 is used to further evaluate the output reliability of different localization algorithms within the current window based on the position estimation results of multiple single localization algorithms obtained in Step 3, and to calculate the corresponding second-level dynamic weights.

[0139] The first-level dynamic weights are applied to the RSSI observation quality of different Bluetooth beacons. After the first-level dynamic weights are applied, the weighted centroid localization method, trilateration method, and Bayesian localization method in step 3 output the position estimation results, respectively.

[0140] However, different positioning algorithms have different computational mechanisms and error sensitivities. Even when using the same RSSI observation data and first-level dynamic weights, their output results may still differ.

[0141] For example, weighted centroid localization is computationally stable and not very sensitive to individual anomalies, but its positioning accuracy has a limited upper limit; trilateration has clear geometric constraints and high accuracy when ranging quality is good and beacon geometry is reasonable, but it is more sensitive to ranging errors and geometric structures; Bayesian localization can handle certain uncertainties through candidate location likelihood evaluation, but it is affected by the resolution of candidate points and the clarity of the likelihood distribution. Therefore, this invention further designs a second-level dynamic weight based on the first-level dynamic weight to evaluate the reliability of the current output results of different localization algorithms, so that stable and applicable algorithms can obtain a larger proportion in the final fusion, and the impact of algorithms with local failures or output anomalies is weakened.

[0142] Assume there is a total In this embodiment, several localization algorithms are used for fusion: .

[0143] in, This represents the total number of positioning algorithms participating in the fusion. The first positioning algorithm is the weighted centroid positioning method. The first positioning algorithm is trilateration, and the second... The localization algorithm is Bayesian localization.

[0144] No. The positioning algorithm in the ... The position estimation results within each window are uniformly recorded as: .

[0145] in Indicates the first The positioning algorithm in the ... Position estimation results within each window; and These represent the x and y coordinates of the positioning result from the algorithm, respectively. Indicates the location algorithm number; Indicates the sliding window number.

[0146] To evaluate the stability of the output of different positioning algorithms within the current window, this invention introduces a reference position.

[0147] Since the position of the target node is usually continuous in time during indoor continuous positioning, the current position should not change drastically without reasonable cause relative to the fused position of the previous moment.

[0148] Therefore, the final fusion positioning result of the previous window is used as the reference position for the current window.

[0149] Definition of the first The reference position within each window is: .

[0150] in, Indicates the first The reference positions within each window are used to evaluate the stability of the algorithm output; Indicates the first The final fused localization result within each window. For the first window, since there is no fused localization result from the previous window, the average of the position results from multiple single localization algorithms can be used as the initial reference position: .

[0151] in, Indicates the initial reference position within the first window; Indicates the first The position estimation result of the localization algorithm within the first window. Alternatively, the weighted centroid localization result can be used as the initial reference position. .

[0152] in, This represents the position estimation result of the weighted centroid localization method within the first window. Both of the above initialization methods can achieve the purpose of this invention, and those skilled in the art can choose according to the actual initialization conditions.

[0153] Step 4 first performs dynamic weighting of the position results from the three single positioning algorithms. Using the final fused positioning result of the previous window as the reference position for the current window, the instantaneous residual between the output of each positioning algorithm and the reference position is calculated. An exponentially weighted recursive method is used to update the residual variance estimate to characterize the continuous output stability of each positioning algorithm. Simultaneously, an unnormalized baseline confidence level is constructed and normalized to obtain the baseline algorithm weights, based on the algorithm applicability factors of each positioning algorithm. Further, the overall signal quality is calculated based on the instantaneous signal quality scores of each Bluetooth beacon within the current window, and an adaptive correction coefficient is constructed to correct the baseline algorithm weights, resulting in a second-level dynamic weight. Finally, the position estimation results of each positioning algorithm are weighted and fused according to the second-level dynamic weight to obtain the final fused positioning result of the target node within the current window. This result is then used as the reference position for the next window to continue participating in subsequent weight updates.

[0154] Step 4 specifically includes the following steps: Step 4.1. First, calculate the residual stability factor for different positioning algorithms.

[0155] After determining the reference position, the deviation between the current output of each positioning algorithm and the reference position is calculated, and this deviation is used as the instantaneous residual of the current window of that algorithm. The positioning algorithm in the ... The instantaneous residual within a window is defined as: .

[0156] in, Indicates the first The positioning algorithm in the ... Instantaneous residuals within a window; Indicates the first The position estimation results of each positioning algorithm within the current window; Indicates the reference position of the current window; This represents the Euclidean distance norm.

[0157] Instantaneous residual Expanded into coordinate form, as shown below: ;in, and They represent the first The x and y coordinates of the output results of a positioning algorithm; and These represent the x-coordinate and y-coordinate of the reference position, respectively.

[0158] The smaller the instantaneous residual, the more consistent the current output of the algorithm is with the fused trajectory of the previous moment; the larger the instantaneous residual, the more likely the current result of the algorithm may have a jump or deviation. It should be noted that this invention does not assume that "the closer to the previous moment, the more accurate it is." Instead, it utilizes the characteristic of the temporal continuity of target position changes in continuous positioning scenarios to regard algorithm results that significantly deviate from the historical fusion trajectory as candidate results with lower credibility.

[0159] Calculating algorithm weights based solely on instantaneous residuals within a single window is susceptible to random errors.

[0160] To reflect the stability of different positioning algorithms within a continuous window, this invention uses an exponentially weighted recursive method to update the squared residuals of each algorithm, thereby obtaining an estimate of the residual variance.

[0161] No. The positioning algorithm in the ... Residual variance estimates within each window Defined as: ; in, This represents the residual variance estimate of the algorithm within the previous window; This represents the instantaneous residual of the algorithm within the current window; Represents the forgetting factor, and The residual variance estimate is used as the residual stability factor.

[0162] For the first window, it can be initialized as follows: ; in, Indicates the first The initial value of the residual variance of each positioning algorithm within the first window.

[0163] Forgetting factor Used to adjust the ratio of historical residual information to current residual information.

[0164] when When the weights are larger, the algorithm's weight changes more smoothly, making it suitable for scenarios with relatively stable motion; when When the weights are smaller, the algorithm weights are more sensitive to changes in the current window, making it suitable for scenarios where the environment changes rapidly.

[0165] By recursively calculating the residual variance, it can be determined whether a certain positioning algorithm maintains stable output within a continuous window.

[0166] If the output results of a certain positioning algorithm are consistently consistent with the reference trajectory, then Smaller; if a positioning algorithm continuously jumps or deviates, then If the weight of a component increases, its subsequent fusion weight should be reduced.

[0167] Step 4.2. Then calculate the algorithm applicability factor for different positioning algorithms.

[0168] The residual variance calculated in step 4.1 above reflects the temporal stability of the algorithm output relative to the reference trajectory, but different positioning algorithms also have different applicable conditions.

[0169] To more fully account for the impact of different algorithms, this invention further introduces an algorithm applicability factor. Used to evaluate the first Whether a positioning algorithm is suitable to be assigned a higher weight within the current window.

[0170] The algorithm applicability factor does not replace the residual variance, but rather, together with the residual variance, determines the algorithm's reliability.

[0171] Among them, residual variance is used to determine whether the algorithm output is stable, and algorithm applicability factor is used to determine whether the algorithm has good applicability under the current beacon weight distribution, geometric structure or likelihood distribution conditions.

[0172] I. Applicability factor of weighted centroid positioning method.

[0173] The weighted centroid localization method mainly relies on the weighted coordinates of multiple Bluetooth beacons. When the first-level dynamic weight distribution is relatively balanced, it indicates that multiple beacons can provide effective constraints, and the weighted centroid result is usually more stable. When the weights are highly concentrated on a few beacons, it indicates that there are not enough effective beacons, and the weighted centroid result is easily influenced by a few beacons.

[0174] Therefore, the applicability factor of the weighted centroid localization method is defined as: .

[0175] in, Indicates the first The positioning algorithm, namely the weighted centroid positioning method, is the [number]th [positioning algorithm]. Applicability factors within each window; Indicates the total number of Bluetooth beacons; Indicates the first The Bluetooth beacon in the 1st The first level of dynamic weight within each window.

[0176] When the weights of each beacon are relatively balanced Smaller A larger value indicates that the weighted centroid localization result is more stable; when the weights are concentrated on a few beacons, Larger The decrease indicates that the weighted centroid localization results are easily affected by a few beacons.

[0177] II. Applicability factor of the trilateration method.

[0178] The solution quality of the trilateration method is closely related to the beacon geometry. If the beacon geometry is reasonable, the linearization matrix conditions are better, and the trilateration solution is more stable. If the beacon geometry is degenerate, such as the beacons being approximately collinear or having an uneven geometric distribution, the trilateration result is prone to amplifying the ranging error.

[0179] Therefore, the applicability factor of the trilateration method is defined as: ; in, Indicates the first The first positioning algorithm, namely trilateration, is the third... Applicability factors within each window; Represents the linearized matrix of trilateral positioning The condition number. When When the value is smaller, it indicates that the geometric conditions of the trilateration equations are better. Larger; when A large value indicates that the trilateration equations are at risk of ill-conditioned or geometrically degenerate. reduce.

[0180] Both of the above forms can be used to characterize the applicability of the trilateration method under current geometric conditions.

[0181] III. Applicability Factor of Bayesian Localization Method.

[0182] Bayesian localization determines the target location by the likelihood values ​​of candidate locations. When the maximum likelihood value is significantly higher than that of other candidate points, it indicates that the distribution of candidate locations has a clear peak, and the Bayesian localization result is relatively reliable. When the likelihood values ​​of multiple candidate locations are close, it indicates that there are multiple solutions or uncertainties in the candidate locations, and the reliability of the Bayesian localization result should be reduced.

[0183] Let the maximum likelihood value within the current window be: ; in, Indicates the first The maximum likelihood value among the candidate locations for Bayesian localization within a window; Represents the set of candidate positions; Indicates candidate location points The likelihood value.

[0184] Let the second highest likelihood value within the current window be: .

[0185] in, Indicates the first The second largest likelihood value in the set of candidate locations for Bayesian localization within a window.

[0186] The applicability factor of Bayesian localization is defined as follows: ; in, Indicates the first The positioning algorithm, namely Bayesian positioning, is the [number]th [positioning algorithm]. Applicability factors within each window; Represents the maximum likelihood value; This represents the second largest likelihood value; This represents the smallest positive number that prevents the denominator from being zero.

[0187] When the difference between the maximum likelihood value and the second maximum likelihood value is large A larger difference indicates that the Bayesian localization result has strong uniqueness; when the difference is small, A smaller value indicates insufficient discrimination of candidate locations, suggesting significant uncertainty in the Bayesian localization results.

[0188] Step 4.3. After obtaining the residual stability factor and algorithm applicability factor for different positioning algorithms, construct the first... The positioning algorithm in the ... Unnormalized baseline confidence within a window.

[0189] The unnormalized baseline confidence of all positioning algorithms is normalized to obtain the baseline algorithm weights.

[0190] Among them, constructing the first The positioning algorithm in the ... Unnormalized benchmark confidence within each window The formula is as follows: ;

[0191] in, Indicates the first The applicability factor of each positioning algorithm within the current window; Indicates the first Residual variance estimates of each positioning algorithm; This represents the smallest positive number that prevents the denominator from being zero.

[0192] As can be seen from the above formula, when the residual variance of a certain algorithm is small and the algorithm applicability factor is large, A larger residual variance indicates that the algorithm's current output is stable and applicable, and it should receive a larger weight in the final fusion; when the residual variance of an algorithm is large or the applicability factor is small, A smaller value indicates that the algorithm's current reliability is low, and its fusion ratio should be reduced.

[0193] Normalize the unnormalized baseline confidence of all positioning algorithms to obtain the baseline algorithm weights: ; in, Indicates the first The positioning algorithm in the ... The baseline algorithm weights within each window; Indicates the first Unnormalized baseline reliability of a positioning algorithm. This indicates the summation index obtained from the localization algorithm; This represents the total number of positioning algorithms involved in the fusion.

[0194] After normalization, the baseline algorithm weights of each localization algorithm satisfy: ; in, Indicates the first The baseline algorithm weights of each positioning algorithm.

[0195] Step 4.4. Construct the overall signal quality adaptive correction coefficient using the real-time signal quality score from Step 2. .

[0196] The weights of the aforementioned benchmark algorithms are primarily determined based on the residual variance and applicability factor of each algorithm. However, when the overall RSSI observation quality is poor, the output results of each positioning algorithm may be significantly affected.

[0197] If we rely entirely on the short-term residual performance of a single positioning algorithm, there may be a situation where a particular positioning algorithm is given excessive weight due to its occasional small residual.

[0198] To avoid this problem, an adaptive correction factor for overall signal quality is constructed using the signal quality score from step 2.

[0199] Definition of the first Average signal quality of all Bluetooth beacons within a window for: ; because lie in arrive Between, therefore Also located arrive between. The larger the value, the better the overall Bluetooth beacon observation quality within the current window; The smaller the value, the worse the overall observation quality within the current window.

[0200] The adaptive correction coefficient is constructed as follows: .in, Indicates the first Adaptive correction coefficient within the window; This indicates the lower limit of the correction factor, and ; This indicates the average signal quality within the current window.

[0201] When the overall signal quality is high Larger Larger values ​​indicate greater trust in the baseline algorithm weights determined by the residual variance and algorithm suitability factor; however, when the overall signal quality is low, Smaller Approximately reduce the differences between the weights of different algorithms as the fusion strategy approaches the lower limit, thereby improving robustness.

[0202] Step 4.5. Based on the baseline algorithm weights and the overall signal quality adaptive correction coefficients, obtain the... The positioning algorithm in the ... The second-level dynamic weights within each window are expressed by the following formula: ; in Indicates the first The positioning algorithm in the ... The second-level dynamic weights within each window; Indicates the first The baseline algorithm weights of each positioning algorithm; This represents the total number of positioning algorithms involved in the fusion.

[0203] As can be seen from the above formula, the second-level dynamic weights consist of two parts: Part One The second part represents the dynamic allocation results based on the algorithm's residual stability and applicability. This represents the equal-weight fusion constraint, used to prevent excessive weight concentration when the overall signal quality is poor.

[0204] when At higher levels, Larger Weights closer to the baseline algorithm That is, they tend to choose algorithms with small residuals and strong applicability. At lower levels, Smaller It is closer to an equal distribution.

[0205] The above design effectively avoids the problem of over-reliance on a single algorithm when overall observations are unreliable.

[0206] The second-level dynamic weights satisfy: .

[0207] Through the above design, the second-level dynamic weight not only considers the residual stability of each positioning algorithm in continuous time, but also the applicability of different positioning algorithms under the current conditions, and further uses the signal quality score obtained in the first-level dynamic weight stage to make overall corrections to the algorithm weights. Therefore, the second-level dynamic weight can dynamically adjust the fusion ratio of multiple positioning algorithms under different indoor environments and signal quality conditions.

[0208] Step 4.6. Based on the second-level dynamic weights in Step 4.5, perform weighted fusion of the positioning results of each positioning algorithm.

[0209] This invention uses a second-level dynamic weighting system to weight and fuse the location estimation results of each positioning algorithm to obtain the final fused positioning result. The formula is as follows: .

[0210] in Indicates the first The final fusion and localization result within each window; For the first The location estimation results of the positioning algorithm.

[0211] The second-level dynamic weights are used to fuse the localization results of multiple single localization algorithms. Compared with fixed weights, empirical weights, or simple averaging, the second-level dynamic weights of this invention have the following functions: First, they can reduce the impact of sudden algorithm jumps on the final result based on the residual changes between each localization algorithm and the reference trajectory; second, they can avoid over-reliance on the corresponding algorithm when the geometry is poor, the beacon weights are overly concentrated, or the Bayesian likelihood distribution is unclear, based on the applicable conditions of different algorithms; third, they can inherit the overall signal quality information reflected by the first-level dynamic weights.

[0212] When the overall RSSI quality is good, the system can more fully utilize the dynamic weight allocation function; when the overall RSSI quality is poor, the system appropriately shrinks towards balanced fusion to improve the stability of the positioning results. Therefore, step 4 in this paper realizes the progression from "beacon observation quality evaluation" to "algorithm output credibility evaluation," so that the final fused positioning result does not only depend on the output of a single algorithm, but can adaptively determine the contribution ratio of each algorithm according to the current signal environment and algorithm performance. When the above parameter settings are adopted, abnormal RSSI fluctuations can be suppressed at the signal preprocessing level, and the influence of low-quality observation sources and locally failed algorithms can be suppressed at the positioning fusion level, thereby improving the accuracy and stability of Bluetooth positioning in complex indoor environments.

[0213] Example 2 This embodiment 2 describes a Bluetooth indoor fusion positioning system based on hybrid filtering and dual weight allocation, which is based on the same inventive concept as the Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation in embodiment 1 above.

[0214] The Bluetooth indoor fusion positioning system based on hybrid filtering and dual weight allocation in this embodiment includes the following modules: The preprocessing module performs multi-level hybrid filtering preprocessing on the original RSSI sequences of each Bluetooth beacon to obtain the final filtered RSSI sequence and the filtered RSSI representative value of each Bluetooth beacon in each window. The first-level dynamic weight construction module constructs first-level dynamic weights based on RSSI stability index, continuity index, intensity rationality index, and historical consistency, which are used to evaluate the current observation quality of different Bluetooth beacons. The single localization algorithm processing module uses the filtered RSSI representative value as the basic observation input and introduces the first-level dynamic weight into three different localization algorithms: weighted centroid localization, trilateration, and Bayesian localization, to obtain the localization results of each localization algorithm. The system also includes a fusion positioning module. Based on the residual stability, applicability, and overall signal quality of each positioning algorithm in continuous time intervals, a second-level dynamic weight is constructed for each positioning algorithm. The position estimation results of each positioning algorithm are dynamically weighted and fused according to the second-level dynamic weight to obtain the final fusion positioning result of the target node.

[0215] It should be noted that any content not mentioned in the above-described functional modules of the system described in Embodiment 2 can be referred to the step description of the corresponding method in Embodiment 1 above, and will not be repeated in detail here.

[0216] Example 3 This embodiment 3 describes a computer device including a memory and one or more processors. Executable code is stored in the memory. When the processor executes the executable code, it implements the steps of the Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation described in embodiment 1 above.

[0217] In this embodiment, the computer device can be any device or apparatus with data processing capabilities, and will not be described in detail here.

[0218] Example 4 This embodiment 4 describes a computer-readable storage medium storing a program that, when executed by a processor, is used to implement the steps of the Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation in embodiment 1 above.

[0219] The computer-readable storage medium can be an internal storage unit of any device or apparatus with data processing capabilities, such as a hard disk or memory, or an external storage device of any device with data processing capabilities, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc.

[0220] Of course, the above description is only a preferred embodiment of the present invention. The present invention is not limited to the above-described embodiments. It should be noted that any equivalent substitutions or obvious modifications made by those skilled in the art under the guidance of this specification fall within the scope of this specification and should be protected by the present invention.

Claims

1. A Bluetooth indoor fusion positioning method based on hybrid filtering and double weight distribution, characterized in that, Includes the following steps: Step 1. Perform multi-level hybrid filtering preprocessing on the original RSSI sequences of each Bluetooth beacon to obtain the final filtered RSSI sequence and the filtered RSSI representative value of each Bluetooth beacon in each window; Step 2. Then, based on the RSSI stability index, continuity index, intensity rationality index, and historical consistency, a first-level dynamic weight is constructed to evaluate the current observation quality of different Bluetooth beacons; Step 3. Using the filtered RSSI representative value as the basic observation input, the first-level dynamic weight is introduced into three different localization algorithms: weighted centroid localization, trilateration, and Bayesian localization, and the localization results of each algorithm are obtained respectively. Step 4. Finally, based on the residual stability, algorithm applicability, and overall signal quality of each positioning algorithm in continuous time, construct the second-level dynamic weights for each positioning algorithm; The location estimation results of each positioning algorithm are dynamically weighted and fused according to the second-level dynamic weights to obtain the final fused positioning result of the target node. Step 4 specifically involves: Step 4.

1. First, calculate the residual stability factor for different positioning algorithms; Step 4.

2. Then calculate the algorithm applicability factor for different positioning algorithms; Step 4.

3. After obtaining the residual stability factor of different positioning algorithms and the algorithm applicability factor, the non-normalized reference reliability of the first positioning algorithm in the second window is constructed, and the formula is as follows: ​​​ ； wherein, represents the applicability factor of the jth positioning algorithm in the current window; represents the residual variance estimate of the jth positioning algorithm, i.e., the residual stability factor, represents a very small positive number to prevent the denominator from being zero;​​ Normalize the unnormalized baseline confidence of all positioning algorithms to obtain the baseline algorithm weights: ； in Indicates the first The positioning algorithm in the ... The baseline algorithm weights within each window; Indicates the first Unnormalized baseline reliability of a positioning algorithm; This indicates the summation index obtained from the localization algorithm; This represents the total number of positioning algorithms participating in the fusion process; After normalization, the baseline algorithm weights of each localization algorithm satisfy: ; in Indicates the first The baseline algorithm weights of each positioning algorithm; Step 4.

4. Construct the overall signal quality adaptive correction coefficient using the real-time signal quality score from Step 2. ; Step 4.

5. Based on the baseline algorithm weights and the overall signal quality adaptive correction coefficients, obtain the... The positioning algorithm in the ... The second-level dynamic weights within each window are expressed by the following formula: ； in Indicates the first The positioning algorithm in the ... The second-level dynamic weights within each window; Indicates the first The baseline algorithm weights of each positioning algorithm; the second-level dynamic weights satisfy: ; Step 4.

6. Based on the second-level dynamic weights in Step 4.5, perform weighted fusion of the positioning results of each positioning algorithm.

2. The Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation according to claim 1, characterized in that, Step 2 specifically involves: Step 2.

1. First, calculate the first... The Bluetooth beacon in the first Within a window, three indicators are considered: RSSI stability index, continuity index, and intensity rationality index. Step 2.

2. Then, the RSSI stability index, continuity index, and strength rationality index are weighted and summed to obtain the... The Bluetooth beacon in the first Real-time signal quality score within each window; Step 2.

3. Construct the first The Bluetooth beacon in the first Historical consistency factor within a window; Step 2.

4. Combining the instantaneous signal quality score and the historical consistency factor, we obtain the... The Bluetooth beacon in the first The unnormalized quality weights within each window are expressed by the following formula: ; in Indicates the first The Bluetooth beacon in the first Unnormalized quality weights within each window; Indicates the first The Bluetooth beacon in the first Real-time signal quality score within each window; Indicates the historical consistency factor; Normalize the unnormalized quality weights of all Bluetooth beacons to obtain the first-level dynamic weights.

3. The Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation according to claim 2, characterized in that, In step 2.1, the process of obtaining the three indicators is as follows: I. Calculation of stability indices; Definition of the first The Bluetooth beacon in the first The RSSI variance within each window is: ； in Indicates the first The Bluetooth beacon in the first The variance of the filtered RSSI sequence within a window; This represents the final filtered RSSI value; Indicates the first The Bluetooth beacon in the first Each window filters the RSSI representative value; Indicates the length of the sliding window; A stability index is constructed based on the RSSI variance, using the following formula: ； in Indicates the first The Bluetooth beacon in the first Stability metrics within a window, Indicates the stability adjustment coefficient; II. Calculation of continuity indicators; Definition of the first The Bluetooth beacon in the first The change in RSSI between adjacent windows within a window is: ； in For the first The Bluetooth beacon in the first The window and the first The RSSI between windows represents the amount of value change; The previous window, i.e., the first The filtered RSSI value for each window; A continuity index is constructed based on the changes in adjacent windows, using the following formula: ； in Indicates the first The Bluetooth beacon in the first Continuous indicators within a window; Indicates the continuous adjustment coefficient; For the first window, since there is no RSSI representative value for the previous window, let: ;in, Indicates the first window within the first window Initial values ​​for the continuity index of each Bluetooth beacon; III. Calculation of strength rationality index; Let the reasonable strength range of RSSI in the BLE indoor positioning system be: ;in, This indicates the lower limit of the reasonable strength range of RSSI; This indicates the upper limit of the reasonable strength range of RSSI; Define the deviation of the filtered RSSI representative value from the reasonable intensity range as: ； in For the first The Bluetooth beacon in the first RSSI intensity deviation within a window; A strength rationality index is constructed based on the strength deviation: ； in For the first The Bluetooth beacon in the first Strength rationality indicators within each window; This represents the strength rationality adjustment coefficient.

4. The Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation according to claim 2, characterized in that, In step 2.2, the formula for calculating the instantaneous signal quality score is as follows: After obtaining the stability index, continuity index, and intensity rationality index, the three types of indices are weighted and summed to obtain the... The Bluetooth beacon in the first Real-time signal quality rating within each window: ； in Indicates stability index; Indicates a continuous indicator; Indicators representing the rationality of strength; , , These represent the weighting coefficients corresponding to the stability index, continuity index, and strength rationality index, respectively. The weighting coefficients of each indicator satisfy the following: ;in , , All are non-negative numbers.

5. The Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation according to claim 2, characterized in that, In step 2.3, the construction process of the historical consistency factor is as follows: Definition of the first The Bluetooth beacon in the first The window was recently Average signal quality score within a historical window for: ;in; Indicates the length of the history window; Indicates the history window index; Indicates the first The Bluetooth beacon in the first Real-time signal quality score within each window; When the number of history windows is insufficient In this case, the average number of existing historical windows can be used, that is: ； in, Indicates the current window number; For the first window, since there is no history window, let: ; in Indicates the first Initial value of the historical average signal quality score for each Bluetooth beacon within the first window; Indicates the first The Bluetooth beacon in the first Real-time signal quality score within each window; The historical consistency factor is further constructed using the following formula: ； in For the first The Bluetooth beacon in the first Historical consistency factor within a window; Indicates the current window, i.e., the [number]th [unit]. Real-time signal quality score for each window; This represents the historical consistency adjustment coefficient.

6. The Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation according to claim 1, characterized in that, Step 3 specifically involves: Step 3.

1. First, convert the filtered RSSI representative value into a distance estimate based on the path loss model: ； in Indicates the distance from the target node to the... Distance estimates for each Bluetooth beacon; Indicates a reference distance; Indicates the RSSI at the reference distance; Indicates the path loss index; Indicates the first The Bluetooth beacon in the first Each window filters the RSSI representative value; Step 3.

2. For weighted centroid localization, the comprehensive centroid weight is defined as: ； in, Indicates the first The beacons are used for the comprehensive weighting of the centroid location; Indicates the first The Bluetooth beacon in the first The first-level dynamic weight within each window; Indicates the distance attenuation coefficient; Represents a very small positive number; Indicates the total number of Bluetooth beacons; For the first The Bluetooth beacon in the first The first-level dynamic weight within each window; Indicates the distance from the target node to the... Distance estimates for each Bluetooth beacon; The weighted centroid localization result is as follows: ； in, This indicates the weighted centroid localization result; Indicates the first The coordinates of the blue beacon; Step 3.

3. For trilateration, let the target node position be... ; A system of equations is established based on distance constraints, and the equations are then used to determine the system of equations. Linearization is performed with each beacon as a reference, written as: ； in Represents the linearized coefficient matrix; Represents a vector of constant terms; Since each distance difference equation is simultaneously related to the reference beacon and the... Related to a beacon, defining the equation weights for: ； in Indicates the first The Bluetooth beacon in the first The first-level dynamic weight within each window; And construct a weighted matrix : ； The weighted least squares solution for trilateration is: ； in This represents the position estimation result output by the trilateration method; Step 3.

4. For Bayesian localization, let the candidate location set be: ； in, Indicates the first Candidate location points, Candidate points Corresponding to the The theoretical RSSI for each beacon is: ； in Candidate points Corresponding to the Theoretical RSSI for a single beacon; To incorporate the first-level dynamic weights into the Bayesian localization process, a weighted RSSI residual is constructed: ； in, This represents the weighted RSSI residuals of the candidate locations; Indicates the first Standard deviation of RSSI noise for each beacon; The corresponding weighted likelihood function for: ； The Bayesian localization result is: ； in, This represents the Bayesian localization result; Therefore, in the first The position estimation results of the three single positioning algorithms are obtained within each window and are uniformly denoted as: ； in This indicates the weighted centroid positioning method. This represents the trilateration method. This represents the Bayesian localization method.

7. The Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation according to claim 1, characterized in that, Step 4.1 specifically involves: After determining the reference position, the deviation between the current output of each positioning algorithm and the reference position is calculated, and this deviation is used as the instantaneous residual of the current window of that algorithm; The positioning algorithm in the ... The instantaneous residual within a window is defined as: ； in Indicates the first The positioning algorithm in the ... Instantaneous residuals within a window; Indicates the first The position estimation results of each positioning algorithm within the current window; Indicates the reference position of the current window; Represents the Euclidean distance norm; Instantaneous residual Expanded into coordinate form, as shown below: ； in, and They represent the first The x and y coordinates of the output results of a positioning algorithm; and These represent the x-coordinate and y-coordinate of the reference position, respectively. ; No. The positioning algorithm in the ... Residual variance estimates within each window Defined as: ； in, This represents the residual variance estimate of the algorithm within the previous window; This represents the instantaneous residual of the algorithm within the current window; Represents the forgetting factor, and The residual variance estimate is used as the residual stability factor. For the first window, initialize as follows: ; in Indicates the first The initial value of the residual variance of each positioning algorithm within the first window.

8. The Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation according to claim 1, characterized in that, Step 4.2 specifically involves: I. Define the applicability factor of the weighted centroid positioning method as: ； in, Indicates the first The positioning algorithm, namely the weighted centroid positioning method, is the [number]th [positioning algorithm]. Applicability factors within each window; Indicates the total number of Bluetooth beacons; Indicates the first The Bluetooth beacon in the first The first-level dynamic weight within each window; II. Define the applicability factor of the trilateration method as: ； in, Indicates the first The first positioning algorithm, namely trilateration, is the third... Applicability factors within each window; Represents the linearized matrix of trilateral positioning condition number; III. Calculate the suitability factor for Bayesian localization; Let the maximum likelihood value within the current window be: ； in, Indicates the first The maximum likelihood value among the candidate locations for Bayesian localization within a window; Represents the set of candidate positions; Indicates candidate location points The likelihood value; This represents the operation of iterating through a set and finding the maximum value. Let the second highest likelihood value within the current window be: ; in, Indicates the first The second largest likelihood value in the set of candidate locations for Bayesian localization within a window; The applicability factor of Bayesian localization is defined as follows: ； in Indicates the first The positioning algorithm, namely Bayesian positioning, is the [number]th [positioning algorithm]. Applicability factors within each window; This represents the maximum likelihood value; This represents the second largest likelihood value; This represents the smallest positive number that prevents the denominator from being zero.

9. The Bluetooth indoor fusion positioning method based on hybrid filtering and dual weight allocation according to claim 1, characterized in that, In step 4.4, the first is defined. Average signal quality of all Bluetooth beacons within a window for: ； in Indicates the first The Bluetooth beacon in the first Real-time signal quality score within each window; Indicates the total number of Bluetooth beacons; The adaptive correction coefficient is constructed as follows: ； in, Indicates the first Adaptive correction coefficient within the window; This indicates the lower limit of the correction factor, and .

Images