Bluetooth positioning precision optimization method based on high-precision positioning

Abstract

The invention discloses a Bluetooth positioning precision optimization method based on high-precision positioning, and the method comprises the steps: collecting Bluetooth signal data, and carrying out the preprocessing of the Bluetooth signal data, and generating a multi-base-station synchronous observation sequence; constructing a Radon operator path constraint module, and generating spatial integral constraint data; introducing energy conservation and variation constraint, and performing inversion to generate a Bluetooth error distribution field; a signal deflection event is detected, positioning uncertainty is evaluated, and risk information is generated; acquiring calibration position information, and locally updating an error field according to an error energy migration principle; and calling the updated error field, compensating and smoothing the positioning result, and outputting a final track. According to the method, the Bluetooth positioning error field capable of being self-updated is generated through construction of the spatial path constraint and inversion, and continuous improvement of the Bluetooth positioning precision and the trajectory stability in a complex dynamic environment is achieved.

Description

Technical Field

[0001] This invention relates to the field of wireless positioning technology, and in particular to a method for optimizing Bluetooth positioning accuracy based on high-precision positioning. Background Technology

[0002] With the development of the Internet of Things (IoT), smart parks, and the Industrial Internet, the demand for real-time location tracking of personnel and assets is constantly growing. Bluetooth Low Energy (BLE) positioning is widely used in indoor positioning, personnel management, and equipment tracking due to its low deployment cost, low terminal power consumption, and good compatibility. Existing Bluetooth positioning technologies typically rely on information such as received signal strength indication (RSI), angle of arrival (Angle of Arrival), or time of arrival (Time of Arrival) to estimate location through fingerprint matching, trilateration, or angle measurement. This can meet basic positioning needs in relatively simple and static environments.

[0003] However, in complex environments such as industrial sites, warehousing and logistics facilities, and large industrial parks, Bluetooth positioning signals are easily affected by factors such as walls, metal equipment, shelves, and personnel movement, resulting in problems such as obstruction, multipath reflection, and signal attenuation and instability. Most existing technologies treat Bluetooth observation results as independent measurements, lacking a holistic model of the continuity of target movement and spatial path constraints. They typically rely on empirical models or simple filtering methods for correction, making it difficult to accurately reflect the propagation characteristics of Bluetooth signals in complex spaces. Some solutions depend on fixed-period calibration or static fingerprint database updates, failing to dynamically adjust according to environmental changes, leading to a decrease in positioning accuracy over time.

[0004] Existing Bluetooth positioning systems have limited ability to determine the reliability of positioning results and struggle to promptly identify abnormal positioning situations caused by sudden environmental changes. Corrections are often only made after significant deviations in the positioning results, affecting the stability and continuity of the positioning. For already acquired calibration location information, existing methods mostly employ point-to-point correction or local replacement, lacking constraints and evolution mechanisms on the overall error distribution. This can easily introduce new discontinuities after local corrections.

[0005] Therefore, how to provide a Bluetooth positioning accuracy optimization method based on high-precision positioning is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a Bluetooth positioning accuracy optimization method based on high-precision positioning. This invention achieves continuous optimization of Bluetooth positioning results by preprocessing and synchronizing multi-base station Bluetooth observation data, combining spatial path constraint construction, error inversion under energy conservation constraints, uncertainty assessment, and adaptive update of the error field. This invention fully utilizes Bluetooth signal observation, spatial path projection constraints, overall signal energy inversion, and dynamic environment perception technologies to uniformly model and correct the propagation characteristics of Bluetooth signals in complex environments. It can effectively suppress positioning errors under conditions of occlusion, multipath reflection, and environmental changes, possessing advantages such as high positioning accuracy, strong result stability, strong adaptability, and wide applicability.

[0007] A Bluetooth positioning accuracy optimization method based on high-precision positioning according to an embodiment of the present invention includes: Collect Bluetooth signal data of the target device at each Bluetooth base station, preprocess the Bluetooth signal data, and form a Bluetooth observation sequence synchronized by multiple base stations; A spatial path constraint generation module based on Radon transform operator is constructed to map the continuous motion information of the target device corresponding to the Bluetooth observation sequence into multi-directional spatial projection data. The motion trajectory of the target device in the positioning area is processed by integral projection to generate integral constraint data. Based on Bluetooth observation sequences and integral constraint data, an energy distribution representation of Bluetooth signals in space is constructed according to energy conservation constraints. By introducing variational constraints, the overall signal energy attenuation between each Bluetooth base station and the target device is inverted to generate a Bluetooth positioning error field. Signal deflection event detection is performed on Bluetooth observation sequences. An improved NanoDet network is used to process image data of the positioning area to identify occluded objects and moving obstacles and to evaluate the uncertainty of the current Bluetooth positioning. When the uncertainty assessment result exceeds the preset threshold, the calibration position information of the target device in the corresponding time window is obtained, and the Bluetooth positioning error field is locally updated based on the calibration position information and the error energy transfer principle. The initial Bluetooth positioning result is obtained based on the current Bluetooth observation sequence. The updated Bluetooth positioning error field is then used to compensate for the error in the initial Bluetooth positioning result. The compensated positioning result is then smoothed to output the optimized target device positioning result.

[0008] Optionally, the Bluetooth signal data includes the received signal strength value, angle information, and timestamp.

[0009] Optionally, the preprocessing of Bluetooth signal data includes time alignment, outlier removal, and scale normalization of the Bluetooth signal data.

[0010] Optionally, the step of performing integral projection processing on the motion trajectory of the target device within the positioning area to generate integral constraint data includes: The Bluetooth observation sequence synchronized by multiple base stations is processed by continuous time slice segmentation. The observation data in adjacent time slices are organized into a set of candidate path segments in chronological order. The start and end times, the set of Bluetooth base stations participating in the observation, and the corresponding observation completeness are recorded for each candidate path segment. A spatial path constraint generation module based on the Radon transform operator is constructed. The spatial path constraint generation module includes a projection direction adaptive generation unit, a Radon integral projection unit, and an integral constraint structured coding unit. The projection direction adaptive generation unit generates a projection direction set based on the spatial geometric distribution of the Bluetooth base stations participating in the observation within the positioning area. The projection direction set includes a direction set corresponding to the direction of the connection between the Bluetooth base stations and a direction set orthogonal to the direction of the connection. The Radon integral projection unit performs Radon transform operator integral projection processing on each candidate path segment under the set of projection directions. Radon transform operator integral projection processing includes: mapping the spatial occupancy representation of the candidate path segment in the positioning area to the linear integral projection result along the given projection direction under the given projection direction, and outputting the projection distribution on the corresponding projection distance axis for each projection direction. The integral constraint structured coding unit performs structured coding on the projection distribution under each projection direction. The structured coding includes: discretizing the projection distance axis into bins, accumulating the projection occupancy intensity in each bin, combining the projection distributions of different projection directions into a multi-direction projection description in a unified format, and adding candidate path segment identifiers and time segment identifiers to the multi-direction projection description. Cross-directional consistency screening is performed based on the multi-directional projection descriptions corresponding to each candidate path segment. Candidate path segments that exhibit discontinuous or contradictory projection distributions under multiple projection directions are eliminated, while the multi-directional projection descriptions that pass the screening are retained as integral constraint data.

[0011] Optionally, generating the Bluetooth positioning error field includes: Based on the Bluetooth observation sequence, the received signal strength indication value of each Bluetooth base station to the target device is extracted in each time slice. The received signal strength indication value is organized according to the Bluetooth base station identifier and timestamp to obtain a multi-base station energy observation dataset. Based on the integral constraint data, the integral constraint data is bound to the multi-base station energy observation dataset within the corresponding time slice according to the time slice identifier and the candidate path segment identifier, so as to obtain the energy observation-path constraint pairing dataset for overall inversion. A spatial discrete representation is established within the positioning area, the positioning area is divided into multiple spatial units, and an energy attenuation state quantity is set for each spatial unit to form an expression of the energy distribution of Bluetooth signals in space. The energy distribution expression is constructed according to the energy conservation constraint, which includes: for the observation of any Bluetooth base station in any time slice, the cumulative result of the energy decay state quantity of the spatial unit traversed along the corresponding propagation path during the propagation process from the Bluetooth base station to the target device is required to be consistent with the energy observation of the Bluetooth base station in the time slice; Under the condition of satisfying the energy conservation constraint, variational constraint conditions are introduced to perform a global inversion of the energy decay state variables. The variational constraint conditions include: The spatial variation of the energy decay state quantity is kept continuous across all spatial units, and abrupt changes between adjacent spatial units are suppressed. This minimizes the overall deviation between the energy observation of each Bluetooth base station in each time slice and the energy decay result accumulated along the propagation path. Based on the overall inversion result, the spatial distribution of the energy decay state quantity of each spatial unit is obtained, and the spatial distribution of the energy decay state quantity is determined as the Bluetooth positioning error field.

[0012] Optionally, the process of identifying occluding objects and moving obstacles and assessing the uncertainty of the current Bluetooth positioning includes: Based on the Bluetooth observation sequence, the received signal strength indication value and angle information of each Bluetooth base station in adjacent time slots are extracted in chronological order. The intensity change and angle change of each Bluetooth base station between adjacent time slots are calculated. The changes of each Bluetooth base station are aggregated according to the time slot identifier to form a multi-base station change sequence. Signal deflection event detection is performed on the multi-base station change sequence. Within the same time slot, when the received signal strength change direction or angle change direction of the preset Bluetooth base station pair is opposite, and the amount of received signal strength change or angle change of the preset Bluetooth base station pair exceeds the preset deflection judgment threshold, a signal deflection event is determined to have occurred in the time slot, and the time slot identifier and the base station identifier of the preset Bluetooth base station pair are recorded. Image data of the localization area is acquired within the time window corresponding to the time slice and time alignment processing consistent with the time slice is performed before being input into the improved NanoDet network. The improved NanoDet network consists of a candidate region generation layer, a temporal consistency suppression layer, and an occlusion risk feature encoding layer. The candidate region generation layer receives the set of Bluetooth base station identifiers corresponding to the signal deflection events, maps the set of Bluetooth base station identifiers to candidate regions of interest within the positioning area, and performs priority detection on the image data of the positioning area within the candidate regions of interest to generate a set of candidate targets. The temporal consistency suppression layer associates the category, spatial location and scale changes of candidate targets based on the candidate target set within adjacent time windows, eliminates candidate targets that cannot form a stable trajectory association within continuous time windows, and retains the detection results of occluded objects and moving obstacles that meet the conditions of positional continuity and scale continuity within continuous time windows. The occlusion risk feature encoding layer encodes risk features for the retained detection results of occluded objects and moving obstacles. The risk feature encoding includes weighted fusion of the number of detected targets, the spatial occupancy of the detected targets in the candidate interest area, the distance of the detected targets relative to the center of the candidate interest area, and the detection confidence, to generate occlusion risk information. The occlusion risk information is then fused with the frequency of signal deflection events and the observation completeness of the multi-base station change sequence to obtain the Bluetooth positioning uncertainty assessment result for the current time slice.

[0013] Optionally, the step of locally updating the Bluetooth positioning error field based on the calibration location information according to the error energy transfer principle includes: When the Bluetooth positioning uncertainty assessment result exceeds the preset threshold, the calibration location information of the target device is obtained within the time window corresponding to the current time slice. The calibration location information is then bound to the Bluetooth observation sequence according to the timestamp to form a calibration sample. The local update region is determined based on the calibration location information. The local update region is defined as the spatial range centered on the spatial location corresponding to the calibration location information and bounded by a preset update radius. The set of error state quantities within the local update region is extracted from the Bluetooth positioning error field. The error energy transfer principle is determined, and the error energy transfer principle includes: The error state quantity of the Bluetooth positioning error field in the local update area is regarded as error energy that can be spatially migrated. The error state quantity at the spatial location corresponding to the calibration location information is adjusted to the preset reference error value. The error energy that is reduced at the calibration location is distributed to the remaining spatial locations in the local update area according to the migration weight related to the spatial distance. According to the error energy transfer principle, the set of error state variables in the local update region is locally updated, including: calculating the transfer weight for each spatial location in the local update region, adding the reduced error energy to the error state variable at the corresponding spatial location according to the transfer weight increment, and performing boundary consistency processing on the error state variables at the boundary of the local update region. Write back the set of error state variables of the local update region after the local update is completed to the Bluetooth positioning error field to obtain the updated Bluetooth positioning error field.

[0014] Optionally, the optimized target device positioning result output includes: Based on the Bluetooth observation sequence, perform Bluetooth initial positioning calculation on the target device within the current time slice to obtain the Bluetooth initial positioning result of the current time slice, and add the current time slice identifier to the Bluetooth initial positioning result; Using the spatial location corresponding to the initial Bluetooth positioning result as the query condition, error compensation information corresponding to the spatial location is retrieved in the updated Bluetooth positioning error field. The error compensation information includes the horizontal coordinate compensation amount and the vertical coordinate compensation amount. The error compensation information is superimposed on the initial Bluetooth positioning result to obtain the error-compensated positioning result. A time slice identifier consistent with the initial Bluetooth positioning result is added to the error-compensated positioning result. The positioning results after error compensation are arranged in chronological order over multiple consecutive time slices to form a positioning trajectory sequence. Based on the motion continuity constraints of the target device, the positioning trajectory sequence is smoothed. The motion continuity constraints include the maximum displacement threshold constraint and the maximum velocity threshold constraint between adjacent time slices. The positioning trajectory sequence after trajectory smoothing is output as the optimized positioning result of the target device.

[0015] The beneficial effects of this invention are: This invention synchronously acquires and preprocesses Bluetooth signal data from multiple base stations and introduces a holistic modeling approach based on spatial path constraints. This allows Bluetooth positioning to move beyond single-point observation corrections and incorporate the continuous motion characteristics of the target device to constrain the positioning process, effectively reducing positioning errors caused by instantaneous signal fluctuations. By uniformly representing the propagation process of Bluetooth signals in space, this invention maintains the continuity and consistency of positioning results in complex spatial structures, improving the stability of Bluetooth positioning in complex environments.

[0016] This invention constructs a Bluetooth positioning error field and introduces energy conservation and continuity constraints to achieve a holistic inversion and expression of positioning errors, transforming error correction from local empirical adjustments to a globally consistent spatial correction mechanism. After acquiring calibration location information, the error field is locally updated using the error energy transfer principle, enabling dynamic calibration without disrupting the overall error structure. This avoids introducing new spatial discontinuities through local corrections, thereby achieving continuous optimization of positioning accuracy.

[0017] This invention, by jointly evaluating Bluetooth signal deflection events and environmental occlusion information, can promptly identify moments of decreased positioning reliability and trigger calibration and update processes when needed. This enables the positioning system to proactively perceive environmental changes and adaptively adjust, maintaining high positioning accuracy and trajectory smoothness over long periods in complex and dynamic scenarios such as industrial sites, warehousing and logistics, and large parks, thereby enhancing the practicality and engineering application value of Bluetooth positioning systems. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a Bluetooth positioning accuracy optimization method based on high-precision positioning proposed in this invention; Figure 2 This diagram illustrates the joint evaluation of positioning uncertainty using signal deflection event detection and an improved NanoDet network, based on a high-precision positioning-based Bluetooth positioning accuracy optimization method proposed in this invention. Detailed Implementation

[0019] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0020] refer to Figure 1 and Figure 2 A method for optimizing Bluetooth positioning accuracy based on high-precision positioning, comprising: Collect Bluetooth signal data of the target device at each Bluetooth base station, preprocess the Bluetooth signal data, and form a Bluetooth observation sequence synchronized by multiple base stations; A spatial path constraint generation module based on Radon transform operator is constructed to map the continuous motion information of the target device corresponding to the Bluetooth observation sequence into multi-directional spatial projection data. The motion trajectory of the target device in the positioning area is processed by integral projection to generate integral constraint data. Based on Bluetooth observation sequences and integral constraint data, an energy distribution representation of Bluetooth signals in space is constructed according to energy conservation constraints. By introducing variational constraints, the overall signal energy attenuation between each Bluetooth base station and the target device is inverted to generate a Bluetooth positioning error field. Signal deflection event detection is performed on Bluetooth observation sequences. An improved NanoDet network is used to process image data of the positioning area to identify occluded objects and moving obstacles and to evaluate the uncertainty of the current Bluetooth positioning. When the uncertainty assessment result exceeds the preset threshold, the calibration position information of the target device in the corresponding time window is obtained, and the Bluetooth positioning error field is locally updated based on the calibration position information and the error energy transfer principle. The initial Bluetooth positioning result is obtained based on the current Bluetooth observation sequence. The updated Bluetooth positioning error field is then used to compensate for the error in the initial Bluetooth positioning result. The compensated positioning result is then smoothed to output the optimized target device positioning result.

[0021] In this embodiment, the Bluetooth signal data includes the received signal strength value, angle information, and timestamp.

[0022] In this embodiment, the preprocessing of Bluetooth signal data includes time alignment, outlier removal, and scale normalization of the Bluetooth signal data.

[0023] In this embodiment, the step of performing integral projection processing on the motion trajectory of the target device within the positioning area to generate integral constraint data includes: The Bluetooth observation sequence synchronized by multiple base stations is processed by continuous time slice segmentation. The observation data in adjacent time slices are organized into a set of candidate path segments in chronological order. The start and end times, the set of Bluetooth base stations participating in the observation, and the corresponding observation completeness are recorded for each candidate path segment. A spatial path constraint generation module based on the Radon transform operator is constructed. The spatial path constraint generation module includes a projection direction adaptive generation unit, a Radon integral projection unit, and an integral constraint structured coding unit. The projection direction adaptive generation unit generates a projection direction set based on the spatial geometric distribution of the Bluetooth base stations participating in the observation within the positioning area. This projection direction set includes a set of directions corresponding to the direction of the connection between the Bluetooth base stations and a set of directions orthogonal to the direction of the connection. Specifically, the generation of the projection direction set involves: Obtain the two-dimensional coordinates of the Bluetooth base stations participating in the observation in the coordinate system of the positioning area, form base station connection lines according to the pairwise combination of base stations, calculate the direction angle corresponding to each base station connection line in the coordinate system, and form a set of directions corresponding to the direction of the Bluetooth base station connection lines. For each direction angle in the direction set, generate its orthogonal direction angle. By offsetting the direction angle with the 90-degree direction, obtain the direction set orthogonal to the direction of the connection to the base station. The set of connecting directions and the set of orthogonal directions are merged and deduplicated, and then sparsified according to a preset angle interval to obtain the set of projection directions. The Radon integral projection unit performs integral projection processing on each candidate path segment using the Radon transform operator under the set of projection directions. The integral projection processing of the Radon transform operator includes: mapping the spatial occupancy representation of the candidate path segment within the positioning region to a linear integral projection result along the given projection direction, and outputting the projection distribution on the corresponding projection distance axis for each projection direction. Specifically, mapping the spatial occupancy representation of the candidate path segment within the positioning region to a linear integral projection result along the given projection direction involves: A spatial discrete grid is established within the positioning area. The spatial positions of candidate path segments in each time slice are mapped to the corresponding grid cells. An occupancy mark or occupancy weight is assigned to the mapped grid cells to obtain the spatial occupancy representation of the candidate path segments. Construct a projection distance axis perpendicular to the projection direction under a given projection direction, and bin and classify each grid cell in the positioning area according to the projection distance to the projection distance axis; The occupancy markers or occupancy weights within each projection distance bin are accumulated to obtain the linear integral projection result along the projection distance axis in the projection direction. The projection distribution in the projection direction is formed by the correspondence between the projection distance and the accumulated occupancy value. The integral constraint structured coding unit performs structured coding on the projection distribution under each projection direction. The structured coding includes: discretizing the projection distance axis into bins, accumulating the projection occupancy intensity in each bin, combining the projection distributions of different projection directions into a multi-direction projection description in a unified format, and adding candidate path segment identifiers and time segment identifiers to the multi-direction projection description. Cross-directional consistency screening is performed based on the multi-directional projection descriptions corresponding to each candidate path segment. Candidate path segments that exhibit discontinuous or contradictory projection distributions under multiple projection directions are eliminated, while the multi-directional projection descriptions that pass the screening are retained as integral constraint data.

[0024] In this embodiment, generating the Bluetooth positioning error field includes: Based on the Bluetooth observation sequence, the received signal strength indication value of each Bluetooth base station to the target device is extracted in each time slice. The received signal strength indication value is organized according to the Bluetooth base station identifier and timestamp to obtain a multi-base station energy observation dataset. Based on the integral constraint data, the integral constraint data is bound to the multi-base station energy observation dataset within the corresponding time slice according to the time slice identifier and the candidate path segment identifier, so as to obtain the energy observation-path constraint pairing dataset for overall inversion. A spatial discrete representation is established within the positioning area, the positioning area is divided into multiple spatial units, and an energy attenuation state quantity is set for each spatial unit to form an expression of the energy distribution of Bluetooth signals in space. The energy distribution expression is constructed according to the energy conservation constraint, which includes: for any observation of any Bluetooth base station in any time slice, the cumulative result of the energy decay state quantity of the spatial unit traversed along the corresponding propagation path during the propagation process from the Bluetooth base station to the target device is required to be consistent with the energy observation of the Bluetooth base station in the time slice. The geometric range of the propagation path is limited by the corresponding integral constraint data in the energy observation-path constraint pairing dataset. Under the condition of satisfying the energy conservation constraint, variational constraint conditions are introduced to perform a global inversion of the energy decay state variables. The variational constraint conditions include: The spatial variation of the energy decay state quantity is kept continuous across all spatial units, and abrupt changes between adjacent spatial units are suppressed. This minimizes the overall deviation between the energy observation of each Bluetooth base station in each time slice and the energy decay result accumulated along the propagation path. Based on the overall inversion result, the spatial distribution of the energy decay state quantity of each spatial unit is obtained, and the spatial distribution of the energy decay state quantity is determined as the Bluetooth positioning error field.

[0025] In this embodiment, the step of identifying obstructing objects and moving obstacles and assessing the uncertainty of the current Bluetooth positioning includes: Based on the Bluetooth observation sequence, the received signal strength indication value and angle information of each Bluetooth base station within adjacent time slots are extracted in chronological order. The intensity change and angle change of each Bluetooth base station between adjacent time slots are calculated. The changes of each Bluetooth base station are aggregated according to the time slot identifier to form a multi-base station change sequence. Specifically, the calculation of the intensity change and angle change of each Bluetooth base station between adjacent time slots is as follows: The received signal strength indication value and angle information of each Bluetooth base station are paired using time slice identifiers to determine the corresponding strength value and angle value of two adjacent time slices under the same Bluetooth base station. The intensity change is obtained by performing a difference operation on the intensity values ​​corresponding to two adjacent time slices, and the angle change is obtained by performing a difference operation on the angle values ​​corresponding to two adjacent time slices. The signs of the intensity change and angle change are determined to obtain the direction of change, and the intensity change, angle change and corresponding direction of change are associated and stored with the time slice identifier and Bluetooth base station identifier; Signal deflection event detection is performed on the multi-base station change sequence. Within the same time slot, when the received signal strength change direction or angle change direction of the preset Bluetooth base station pair is opposite, and the amount of received signal strength change or angle change of the preset Bluetooth base station pair exceeds the preset deflection judgment threshold, a signal deflection event is determined to have occurred in the time slot. The time slot identifier and the base station identifier of the preset Bluetooth base station pair are recorded. The preset deflection judgment threshold includes an intensity deflection threshold and an angle deflection threshold. The intensity deflection threshold is 6dB, and the angle deflection threshold is 15°. Image data of the localization area is acquired within the time window corresponding to the time slice and time alignment processing consistent with the time slice is performed before being input into the improved NanoDet network. The improved NanoDet network consists of a candidate region generation layer, a temporal consistency suppression layer, and an occlusion risk feature encoding layer. The candidate region generation layer receives a set of Bluetooth base station identifiers corresponding to signal deflection events. This set of Bluetooth base station identifiers is mapped to candidate regions of interest within the positioning area. Priority detection is performed on the image data of the positioning area within these candidate regions of interest to generate a set of candidate targets. Specifically, mapping the set of Bluetooth base station identifiers to candidate regions of interest within the positioning area involves: Read the installation location coordinates of each Bluetooth base station in the positioning area coordinate system from the Bluetooth base station identifier set, and construct a base station connection set based on the base station installation location coordinates; The midpoint of the connection between the preset Bluetooth base station pairs where the signal deflection event occurs is taken as the center point of the candidate region of interest, and the spatial range of the candidate region of interest is generated with a preset spatial radius. The spatial ranges generated by multiple base station pairs are then combined. The spatial extent of the candidate region of interest is converted into a pixel region in the image through a pre-defined spatial coordinate-to-image pixel coordinate mapping relationship, thus obtaining the candidate region of interest for priority detection; The temporal consistency suppression layer associates the category, spatial location and scale changes of candidate targets based on the candidate target set within adjacent time windows, eliminates candidate targets that cannot form a stable trajectory association within continuous time windows, and retains the detection results of occluded objects and moving obstacles that meet the conditions of positional continuity and scale continuity within continuous time windows. The occlusion risk feature encoding layer encodes risk features for the retained detection results of occluded objects and moving obstacles. This risk feature encoding includes weighted fusion of the number of detected targets, the spatial occupancy of the detected targets within the candidate interest area, the distance of the detected targets relative to the center of the candidate interest area, and the detection confidence level. This generates occlusion risk information. The occlusion risk information is then fused with the frequency of signal deflection events and the observation completeness of the multi-base station change sequence to obtain the Bluetooth positioning uncertainty assessment result for the current time slice. The detection confidence level is determined as follows: After inputting the localization area image data into the improved NanoDet network, the improved NanoDet network outputs the bounding box and category information of each detected target, as well as the confidence score of the detected target. The confidence score is obtained by combining the category probability output by the network classification branch and the targetness score output by the network regression branch. The confidence score is normalized to a value range of zero to one as the detection confidence.

[0026] In this embodiment, the step of locally updating the Bluetooth positioning error field based on the calibration location information according to the error energy transfer principle includes: When the Bluetooth positioning uncertainty assessment result exceeds the preset threshold, the calibration location information of the target device is obtained within the time window corresponding to the current time slice. The calibration location information is then bound to the Bluetooth observation sequence according to the timestamp to form a calibration sample. The local update region is determined based on the calibration location information. The local update region is defined as the spatial range centered on the spatial location corresponding to the calibration location information and bounded by a preset update radius. The set of error state quantities within the local update region is extracted from the Bluetooth positioning error field, where the preset update radius is 3 meters. The error energy transfer principle is determined, and the error energy transfer principle includes: The error state quantity of the Bluetooth positioning error field in the local update area is regarded as error energy that can be spatially migrated. The error state quantity at the spatial location corresponding to the calibration location information is adjusted to a preset reference error value. The error energy that is reduced at the calibration location is distributed to the remaining spatial locations in the local update area according to the migration weight related to the spatial distance. The preset reference error value is 0.5 meters. According to the error energy transfer principle, the set of error state variables in the local update region is locally updated, including: calculating the transfer weight for each spatial location in the local update region, adding the reduced error energy to the error state variable at the corresponding spatial location according to the transfer weight increment, and performing boundary consistency processing on the error state variables at the boundary of the local update region. Write back the set of error state variables of the local update region after the local update is completed to the Bluetooth positioning error field to obtain the updated Bluetooth positioning error field.

[0027] In this embodiment, the output of the optimized target device positioning result includes: Based on the Bluetooth observation sequence, perform Bluetooth initial positioning calculation on the target device within the current time slice to obtain the Bluetooth initial positioning result of the current time slice, and add the current time slice identifier to the Bluetooth initial positioning result; Using the spatial location corresponding to the initial Bluetooth positioning result as the query condition, error compensation information corresponding to the spatial location is retrieved in the updated Bluetooth positioning error field. The error compensation information includes the horizontal coordinate compensation amount and the vertical coordinate compensation amount. The error compensation information is superimposed on the initial Bluetooth positioning result to obtain the error-compensated positioning result. A time slice identifier consistent with the initial Bluetooth positioning result is added to the error-compensated positioning result. The positioning results after error compensation are arranged in chronological order over multiple consecutive time slices to form a positioning trajectory sequence. Based on the motion continuity constraints of the target device, the positioning trajectory sequence is smoothed. The motion continuity constraints include the maximum displacement threshold constraint and the maximum velocity threshold constraint between adjacent time slices. The positioning trajectory sequence after trajectory smoothing is output as the optimized positioning result of the target device.

[0028] Example 1: To verify the feasibility of this invention in practice, it was applied to an integrated indoor production and warehousing workshop of a manufacturing company. The workshop covers an area of ​​approximately 9,500 square meters and includes multiple production lines, metal processing equipment, rows of storage racks, and several forklift access lanes. Bluetooth Low Energy base stations have been deployed on the workshop ceiling and pillar areas according to the existing system for real-time location tracking and scheduling management of personnel and forklifts.

[0029] Before adopting the method of this invention, the workshop positioning system only calculated the location based on the strength of the Bluetooth received signal. In daily production, due to the dense metal equipment, frequent forklift movement, and personnel flow, the Bluetooth signal propagation path often changes, resulting in significant fluctuations in the positioning results. Especially in areas with dense shelving and during periods of concentrated forklift operation, the positioning trajectory exhibits jitter and short-term deviations, affecting the scheduling system's accurate judgment of the positions of personnel and equipment.

[0030] In this embodiment, the positioning accuracy optimization method of the present invention is introduced without changing the original Bluetooth base station hardware deployment. The system first collects Bluetooth signal data of the target device at each Bluetooth base station, and performs time alignment, outlier removal, and normalization on the data to form a Bluetooth observation sequence synchronized by multiple base stations. Based on this observation sequence, the system constructs continuous motion information of the target device, and maps the motion trajectory of the target device in the workshop into multi-directional spatial projections through a spatial path constraint generation mechanism based on the Radon transform operator, generating spatial path integral constraint data, thereby constraining the reachable path of the target device as a whole.

[0031] Subsequently, the system combines Bluetooth observation sequences with spatial path integral constraint data to construct a Bluetooth signal energy distribution representation within the workshop space. Then, based on energy conservation constraints, it performs an overall inversion of the signal energy attenuation between each Bluetooth base station and the target device, forming a Bluetooth positioning error field. This error field can reflect the stable error distribution characteristics in different areas of the workshop caused by factors such as equipment obstruction and shelving layout.

[0032] During positioning operations, the system performs signal deflection event detection on multi-base station Bluetooth observation sequences and simultaneously collects workshop video image data, inputting it into an improved NanoDet network. Guided by Bluetooth observation anomalies, the improved NanoDet network focuses on occlusion targets such as forklifts, personnel, and large equipment, generating occlusion risk information. The system fuses the occlusion risk information with the signal deflection event information to obtain the Bluetooth positioning uncertainty assessment result for the current time slice. When the assessment result exceeds a threshold, the system acquires the calibration position information within the corresponding time window and locally updates the Bluetooth positioning error field based on the error energy transfer principle, ensuring spatial continuity of error correction.

[0033] During the online positioning phase, the system calculates the initial Bluetooth positioning result based on the Bluetooth observation sequence, calls the updated Bluetooth positioning error field for error compensation, then performs trajectory smoothing on the compensated positioning result, and finally outputs the optimized target device positioning result.

[0034] To verify the effectiveness of the method of this invention, 10 forklifts were selected in the workshop as test subjects and operated continuously for 5 working days. The positioning effects of the traditional Bluetooth positioning method and the method of this invention were compared. Using manually set calibration points in the workshop as references, the positioning error and trajectory stability indicators were statistically analyzed over different time periods.

[0035] Table 1. Comparison of Positioning Effects in Production and Warehousing Workshops

[0036] Table 1 shows that under strong environmental interference during the daytime, the average error of traditional Bluetooth positioning is 2.12 meters, the maximum error is 4.95 meters, and the number of trajectory jumps is 5.8 times / hour, indicating significant fluctuations in positioning results. The method of this invention reduces the average error to 0.98 meters, the maximum error to 2.21 meters, and the trajectory jumps to 1.6 times / hour under the same conditions, improving both positioning accuracy and continuity.

[0037] During the night shift, when the overall environment is relatively stable, traditional Bluetooth positioning has an average error of 1.87 meters and a maximum error of 4.32 meters, with a trajectory jump rate of 4.9 times per hour, still exhibiting significant deviation and jitter. The method of this invention corresponds to an average error of 0.91 meters and a maximum error of 2.05 meters, with a trajectory jump rate of 1.4 times per hour, demonstrating that it can output smoother and more reliable positioning results even in relatively stable scenarios.

[0038] In concentrated forklift operations, occlusion and multipath propagation become more prominent. Traditional Bluetooth positioning suffers an average error of 2.46 meters and a maximum error of 5.28 meters, with trajectory jumps occurring 6.4 times per hour, further reducing stability. The method of this invention, under these highly dynamic conditions, still maintains an average error of 1.05 meters and a maximum error of 2.36 meters, while reducing trajectory jumps to 1.9 times per hour, demonstrating adaptability to dynamic environmental changes and continuous optimization.

[0039] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for optimizing Bluetooth positioning accuracy based on high-precision positioning, characterized in that, include: Collect Bluetooth signal data of the target device at each Bluetooth base station, preprocess the Bluetooth signal data, and form a Bluetooth observation sequence synchronized by multiple base stations; A spatial path constraint generation module based on Radon transform operator is constructed to map the continuous motion information of the target device corresponding to the Bluetooth observation sequence into multi-directional spatial projection data. The motion trajectory of the target device in the positioning area is processed by integral projection to generate integral constraint data. Based on Bluetooth observation sequences and integral constraint data, an energy distribution representation of Bluetooth signals in space is constructed according to energy conservation constraints. By introducing variational constraints, the overall signal energy attenuation between each Bluetooth base station and the target device is inverted to generate a Bluetooth positioning error field. Signal deflection event detection is performed on Bluetooth observation sequences. An improved NanoDet network is used to process image data of the positioning area to identify occluded objects and moving obstacles and to evaluate the uncertainty of the current Bluetooth positioning. When the uncertainty assessment result exceeds the preset threshold, the calibration position information of the target device in the corresponding time window is obtained, and the Bluetooth positioning error field is locally updated based on the calibration position information and the error energy transfer principle. The initial Bluetooth positioning result is obtained based on the current Bluetooth observation sequence. The updated Bluetooth positioning error field is then used to compensate for the error in the initial Bluetooth positioning result. The compensated positioning result is then smoothed to output the optimized target device positioning result.

2. The Bluetooth positioning accuracy optimization method based on high-precision positioning according to claim 1, characterized in that, The Bluetooth signal data includes the received signal strength value, angle information, and timestamp.

3. The Bluetooth positioning accuracy optimization method based on high-precision positioning according to claim 1, characterized in that, The preprocessing of Bluetooth signal data includes time alignment, outlier removal, and scale normalization.

4. The Bluetooth positioning accuracy optimization method based on high-precision positioning according to claim 1, characterized in that, The step of performing integral projection processing on the motion trajectory of the target device within the positioning area to generate integral constraint data includes: The Bluetooth observation sequence synchronized by multiple base stations is processed by continuous time slice segmentation. The observation data in adjacent time slices are organized into a set of candidate path segments in chronological order. The start and end times, the set of Bluetooth base stations participating in the observation, and the corresponding observation completeness are recorded for each candidate path segment. A spatial path constraint generation module based on the Radon transform operator is constructed. The spatial path constraint generation module includes a projection direction adaptive generation unit, a Radon integral projection unit, and an integral constraint structured coding unit. The projection direction adaptive generation unit generates a projection direction set based on the spatial geometric distribution of the Bluetooth base stations participating in the observation within the positioning area. The projection direction set includes a direction set corresponding to the direction of the connection between the Bluetooth base stations and a direction set orthogonal to the direction of the connection. The Radon integral projection unit performs Radon transform operator integral projection processing on each candidate path segment under the set of projection directions. Radon transform operator integral projection processing includes: mapping the spatial occupancy representation of the candidate path segment in the positioning area to the linear integral projection result along the given projection direction under the given projection direction, and outputting the projection distribution on the corresponding projection distance axis for each projection direction. The integral constraint structured coding unit performs structured coding on the projection distribution under each projection direction. The structured coding includes: discretizing the projection distance axis into bins, accumulating the projection occupancy intensity in each bin, combining the projection distributions of different projection directions into a multi-direction projection description in a unified format, and adding candidate path segment identifiers and time segment identifiers to the multi-direction projection description. Cross-directional consistency screening is performed based on the multi-directional projection descriptions corresponding to each candidate path segment. Candidate path segments that exhibit discontinuous or contradictory projection distributions under multiple projection directions are eliminated, while the multi-directional projection descriptions that pass the screening are retained as integral constraint data.

5. The Bluetooth positioning accuracy optimization method based on high-precision positioning according to claim 1, characterized in that, The generation of the Bluetooth positioning error field includes: Based on the Bluetooth observation sequence, the received signal strength indication value of each Bluetooth base station to the target device is extracted in each time slice. The received signal strength indication value is organized according to the Bluetooth base station identifier and timestamp to obtain a multi-base station energy observation dataset. Based on the integral constraint data, the integral constraint data is bound to the multi-base station energy observation dataset within the corresponding time slice according to the time slice identifier and the candidate path segment identifier, so as to obtain the energy observation-path constraint pairing dataset for overall inversion. A spatial discrete representation is established within the positioning area, the positioning area is divided into multiple spatial units, and an energy attenuation state quantity is set for each spatial unit to form an expression of the energy distribution of Bluetooth signals in space. The energy distribution expression is constructed according to the energy conservation constraint, which includes: for the observation of any Bluetooth base station in any time slice, the cumulative result of the energy decay state quantity of the spatial unit traversed along the corresponding propagation path during the propagation process from the Bluetooth base station to the target device is required to be consistent with the energy observation of the Bluetooth base station in the time slice; Under the condition of satisfying the energy conservation constraint, variational constraint conditions are introduced to perform a global inversion of the energy decay state variables. The variational constraint conditions include: The spatial variation of the energy decay state quantity is kept continuous across all spatial units, and abrupt changes between adjacent spatial units are suppressed. This minimizes the overall deviation between the energy observation of each Bluetooth base station in each time slice and the energy decay result accumulated along the propagation path. Based on the overall inversion result, the spatial distribution of the energy decay state quantity of each spatial unit is obtained, and the spatial distribution of the energy decay state quantity is determined as the Bluetooth positioning error field.

6. The Bluetooth positioning accuracy optimization method based on high-precision positioning according to claim 1, characterized in that, The process of identifying occluded objects and moving obstacles and assessing the uncertainty of the current Bluetooth positioning includes: Based on the Bluetooth observation sequence, the received signal strength indication value and angle information of each Bluetooth base station in adjacent time slots are extracted in chronological order. The intensity change and angle change of each Bluetooth base station between adjacent time slots are calculated. The changes of each Bluetooth base station are aggregated according to the time slot identifier to form a multi-base station change sequence. Signal deflection event detection is performed on the multi-base station change sequence. Within the same time slot, when the received signal strength change direction or angle change direction of the preset Bluetooth base station pair is opposite, and the amount of received signal strength change or angle change of the preset Bluetooth base station pair exceeds the preset deflection judgment threshold, a signal deflection event is determined to have occurred in the time slot, and the time slot identifier and the base station identifier of the preset Bluetooth base station pair are recorded. Image data of the localization area is acquired within the time window corresponding to the time slice and time alignment processing consistent with the time slice is performed before being input into the improved NanoDet network. The improved NanoDet network consists of a candidate region generation layer, a temporal consistency suppression layer, and an occlusion risk feature encoding layer. The candidate region generation layer receives the set of Bluetooth base station identifiers corresponding to the signal deflection events, maps the set of Bluetooth base station identifiers to candidate regions of interest within the positioning area, and performs priority detection on the image data of the positioning area within the candidate regions of interest to generate a set of candidate targets. The temporal consistency suppression layer associates the category, spatial location and scale changes of candidate targets based on the candidate target set within adjacent time windows, eliminates candidate targets that cannot form a stable trajectory association within continuous time windows, and retains the detection results of occluded objects and moving obstacles that meet the conditions of positional continuity and scale continuity within continuous time windows. The occlusion risk feature encoding layer encodes risk features for the retained detection results of occluded objects and moving obstacles. The risk feature encoding includes weighted fusion of the number of detected targets, the spatial occupancy of the detected targets in the candidate interest area, the distance of the detected targets relative to the center of the candidate interest area, and the detection confidence, to generate occlusion risk information. The occlusion risk information is then fused with the frequency of signal deflection events and the observation completeness of the multi-base station change sequence to obtain the Bluetooth positioning uncertainty assessment result for the current time slice.

7. The Bluetooth positioning accuracy optimization method based on high-precision positioning according to claim 1, characterized in that, The process of locally updating the Bluetooth positioning error field based on calibration location information according to the error energy transfer principle includes: When the Bluetooth positioning uncertainty assessment result exceeds the preset threshold, the calibration location information of the target device is obtained within the time window corresponding to the current time slice. The calibration location information is then bound to the Bluetooth observation sequence according to the timestamp to form a calibration sample. The local update region is determined based on the calibration location information. The local update region is defined as the spatial range centered on the spatial location corresponding to the calibration location information and bounded by a preset update radius. The set of error state quantities within the local update region is extracted from the Bluetooth positioning error field. The error energy transfer principle is determined, and the error energy transfer principle includes: The error state quantity of the Bluetooth positioning error field in the local update area is regarded as error energy that can be spatially migrated. The error state quantity at the spatial location corresponding to the calibration location information is adjusted to the preset reference error value. The error energy that is reduced at the calibration location is distributed to the remaining spatial locations in the local update area according to the migration weight related to the spatial distance. According to the error energy transfer principle, the set of error state variables in the local update region is locally updated, including: calculating the transfer weight for each spatial location in the local update region, adding the reduced error energy to the error state variable at the corresponding spatial location according to the transfer weight increment, and performing boundary consistency processing on the error state variables at the boundary of the local update region. Write back the set of error state variables of the local update region after the local update is completed to the Bluetooth positioning error field to obtain the updated Bluetooth positioning error field.

8. The Bluetooth positioning accuracy optimization method based on high-precision positioning according to claim 1, characterized in that, The optimized target device positioning results output include: Based on the Bluetooth observation sequence, perform Bluetooth initial positioning calculation on the target device within the current time slice to obtain the Bluetooth initial positioning result of the current time slice, and add the current time slice identifier to the Bluetooth initial positioning result; Using the spatial location corresponding to the initial Bluetooth positioning result as the query condition, error compensation information corresponding to the spatial location is retrieved in the updated Bluetooth positioning error field. The error compensation information includes the horizontal coordinate compensation amount and the vertical coordinate compensation amount. The error compensation information is superimposed on the initial Bluetooth positioning result to obtain the error-compensated positioning result. A time slice identifier consistent with the initial Bluetooth positioning result is added to the error-compensated positioning result. The positioning results after error compensation are arranged in chronological order over multiple consecutive time slices to form a positioning trajectory sequence. Based on the motion continuity constraints of the target device, the positioning trajectory sequence is smoothed. The motion continuity constraints include the maximum displacement threshold constraint and the maximum velocity threshold constraint between adjacent time slices. The positioning trajectory sequence after trajectory smoothing is output as the optimized positioning result of the target device.

Images