A Bluetooth communication status assessment method based on multi-source data fusion
By employing multi-source data fusion and dimensional evolution fitting methods, the problem of difficulty in fusing multi-source time-series data in Bluetooth communication status assessment is solved, enabling accurate and forward-looking assessment of Bluetooth communication status and improving the ability to identify latent degradation and periodic fluctuations.
Patent Information
- Application Number
- CN202611279146.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-21
- Publication Date
- 2026-11-17
AI Technical Summary
Existing Bluetooth communication status assessment methods struggle to integrate multi-source time-series data such as link, frequency hopping channel, device operation, and connection parameters, making it difficult to accurately identify the continuous evolution trajectory of Bluetooth communication status. This results in untimely identification of periodic fluctuations, retransmission maintenance, and connection parameter mismatch.
A multi-source data fusion and dimensionality-upgrading evolution fitting method is adopted. By acquiring Bluetooth communication operation data and preprocessing it, a multi-source basic dataset is generated. The connection stage is divided and continuous segments are organized to generate state time series segments. Multi-source fusion arrangement and delay embedding processing are performed to construct a Koopman dimensionality-upgrading observation dataset. State evolution operators are generated and feature decomposition is performed to achieve comprehensive evaluation.
It improves the accuracy and foresight of Bluetooth communication status assessment, enabling the identification of latent degradation and periodic fluctuations before the Bluetooth connection is lost, reducing false alarm rates, and enhancing the timeliness of abnormal status identification and the accuracy of assessment results.
Smart Images

Figure CN123123986A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wireless communication technology, and in particular to a method for evaluating the status of Bluetooth communication based on multi-source data fusion. Background Technology
[0002] Current Bluetooth communication status assessments typically rely on operational metrics such as signal strength, connection latency, packet loss, retransmission count, and changes in connection parameters, and monitor Bluetooth connection quality using pre-defined judgment rules. This type of method can reflect the current operational status of the Bluetooth communication link and is suitable for scenarios with obvious anomalies such as disconnection, strong interference, and significant packet loss.
[0003] However, in actual operation, Bluetooth communication often experiences situations where the connection remains open and individual indicators do not significantly exceed limits, yet the state gradually deteriorates. Existing methods struggle to integrate multi-source time-series data, including link, frequency hopping channel, device operation, and connection parameters, and also fail to depict the continuous evolution trajectory of the communication state. This leads to untimely identification of periodic fluctuations, retransmission maintenance, connection parameter mismatch, and critical instability before disconnection, affecting the accuracy and predictability of Bluetooth communication state assessment.
[0004] Therefore, how to provide a Bluetooth communication status assessment method based on multi-source data fusion is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0005] One objective of this invention is to propose a Bluetooth communication state assessment method based on multi-source data fusion. This invention utilizes multi-source time-series fusion and dimensional evolution fitting methods to construct a Bluetooth communication state evolution assessment mechanism, which has the advantages of accurate identification and forward-looking early warning.
[0006] A Bluetooth communication status assessment method based on multi-source data fusion according to an embodiment of the present invention includes the following steps: Acquire and preprocess the Bluetooth communication operation data to be evaluated to generate a multi-source basic dataset for Bluetooth communication. The Bluetooth communication multi-source basic dataset is divided into connection stages and continuous segments are organized to generate state time sequence segments. Multi-source fusion and orchestration of state time series segments within the same time window are performed to generate multi-source fused time series data. Based on Takens' embedding theorem, delayed embedding processing is performed on multi-source fused time series data to generate communication trajectory data containing a sequence of phase space state points; A Koopman upscaling observation dataset is constructed based on communication trajectory data, and an evolutionary mapping fit is performed to generate state evolution operators; Eigenvalue decomposition and state mapping are performed on the state evolution operator to generate evolution feature data; A comprehensive evaluation of Bluetooth communication status is performed based on evolutionary feature data to generate Bluetooth communication status evaluation results.
[0007] Optionally, the Bluetooth communication operation data to be evaluated includes Bluetooth communication link data, frequency hopping channel status data, device operation status data, and connection parameter change data. The preprocessing includes time alignment, device attribution marking, connection process marking, collection anomaly removal, and data normalization processing.
[0008] Optionally, the generation of the state timing segment specifically includes: Based on the connection process markers and sampling times in the multi-source basic dataset of Bluetooth communication, the connection stage boundaries of the Bluetooth communication process are identified, and connection stage boundary data is generated. The multi-source basic dataset for Bluetooth communication is segmented according to the connection phase boundaries, and continuously sampled data within the same connection phase are merged to generate continuous candidate segments. The continuous candidate segments are sorted for sampling continuity, device affiliation consistency and connection process consistency, and segment tags and connection stage tags are configured to generate state time sequence segments. The state time sequence segments include segment tags, connection stage tags, sampling time series and multi-source basic data of Bluetooth communication within the segment.
[0009] Optionally, the generation of the multi-source fused time-series data specifically includes: Based on the sampling time sequence and connection stage marker in the state time sequence segment, the Bluetooth communication multi-source basic data in the segment are divided into the same time window and arranged within the window to generate multi-source arranged data within the window. The multi-source data within the window is arranged in the order of source, unified in terms of units, and spliced with state variables to generate a multi-source communication state vector corresponding to each fusion time window. The multi-source communication state vectors are arranged in chronological order according to the fusion time window, and the fragment markers and connection stage markers are retained to generate multi-source fusion time-series data.
[0010] Optionally, the generation of the communication trajectory data specifically includes: Based on Takens' embedding theorem, observation state quantities are selected from the multi-source communication state vector sequence in multi-source fused time series data, and then arranged in time order according to the fused time window sequence to generate a communication observation state sequence. The delay step size is configured based on the sampling time interval and the continuity of state changes of the communication observation state sequence, and the current observation state quantity and historical observation state quantity are extracted from the communication observation state sequence according to the delay step size to generate a delayed observation sequence. The embedding dimension is configured according to the number of historical observations required for the Bluetooth communication state evolution, and the delayed observation sequence is combined according to the delayed coordinate reconstruction method in Takens' embedding theorem to generate phase space state points. The phase space state points are arranged in chronological order according to the fusion time window, and the fragment markers, connection stage markers and state point time positions are preserved to generate a phase space state point sequence. The phase space state point sequence is continuously organized, and the phase space state point sequence that meets the continuous arrangement conditions is used as communication trajectory data. The continuous arrangement conditions include consistent segment labels, consistent connection stage labels, continuous state point time position, consistent delay step size, consistent embedding dimension, and complete delayed observation state quantity.
[0011] Optionally, the generation of the state evolution operator specifically includes: Based on the phase space state point sequence in the communication trajectory data, the original state arrangement and higher-order change arrangement are performed to generate linear and nonlinear observation data. Based on the current and historical observed state quantities in the phase space state points, the time-delayed observation data is generated by arranging them with a delay. Synchronous and delayed changes among observed state quantities from different sources in phase space are combined and arranged to generate interactive observation data; Linear observation data, nonlinear observation data, time-delay observation data, and interactive observation data are collected according to the phase space state points in order to generate the Koopman upgraded observation dataset. According to the order of adjacent state points in the phase space state point sequence, the Koopman upscaling observation data of the previous state point is configured as the evolution input data, and the Koopman upscaling observation data of the next state point is configured as the evolution output data. Evolutionary mapping fitting is performed based on evolutionary input data and evolutionary output data to generate candidate state evolution operators; The candidate state evolution operators are sorted by fragment assignment, connection stage assignment, and operator stability to generate state evolution operators.
[0012] Optionally, the generation of the evolutionary feature data specifically includes: The state evolution operator is separated into evolutionary components and the direction of change is divided to generate operator decomposition data, which includes stable evolution components, periodic fluctuation components, drift evolution components and unstable evolution components. The operator decomposition data is mapped to the state point time position and the corresponding multi-source communication state vector in the phase space state point sequence to generate state mapping data. Based on state mapping data, features of stability, fluctuation, drift and instability in the Bluetooth communication state evolution process are collected to generate candidate evolution feature data, which includes stable evolution features, periodic fluctuation features, continuous drift features and critical instability features. Candidate evolutionary feature data are processed temporally according to fragment labels, connection stage labels, and state point time positions to generate evolutionary feature data. The evolutionary feature data includes fragment labels, connection stage labels, fusion time window labels, evolutionary feature types, and feature change directions.
[0013] Optionally, the generation of the Bluetooth communication status assessment result specifically includes: Based on evolutionary feature data, state assessment evidence is configured and stage associations are performed according to connection stage markers to generate stage state evidence data; Based on the state assessment evidence type and connection phase marker in the phase state evidence data, the Bluetooth communication state category is determined, and Bluetooth communication state category data is generated. The Bluetooth communication status category data and phase status evidence data are organized according to the fragment tag and connection phase tag to generate Bluetooth communication status assessment results.
[0014] The beneficial effects of this invention are: This invention achieves continuous, multi-source fused time-series data by uniformly preprocessing, dividing the connection phases, and merging and orchestrating Bluetooth communication link data, frequency hopping channel status data, device operating status data, and connection parameter change data within the same time window. This enables communication status information, originally scattered across different acquisition sources, to be formed according to the connection process and time sequence. Consequently, Bluetooth communication status assessment is no longer limited to judging signal strength, retransmission count, or connection delay at a single moment, but can be comprehensively analyzed by combining link status, channel status, device operating load, and connection parameter changes, thus improving the completeness and consistency of the Bluetooth communication operating status description.
[0015] This invention applies delayed embedding processing to multi-source fused time-series data based on Takens' embedding theorem, generating communication trajectory data containing a sequence of phase space state points. This allows current and historical observed state variables to participate in communication state reconstruction, providing a more comprehensive characterization of the continuous evolution of Bluetooth communication states over time. Even when the Bluetooth connection is not broken and no individual communication indicators show significant anomalies, this communication trajectory data reflects the continuous changes in response delay, retransmission maintenance, frequency hopping channel fluctuations, and connection parameter adjustments, facilitating early identification of latent degradation and periodic fluctuations.
[0016] This invention further constructs a Koopman dimensionality-upgrading observation dataset based on communication trajectory data, and generates state evolution operators through evolutionary mapping fitting. Then, it performs feature decomposition and state mapping on the state evolution operators to generate evolutionary feature data. By dividing the data into stable evolutionary components, periodic fluctuation components, drift evolutionary components, and unstable evolutionary components, it can distinguish stable operation, periodic fluctuations, continuous drift, and critical instability in Bluetooth communication states. Combined with connection stage markers, it forms Bluetooth communication state evaluation results, thereby improving the timeliness of abnormal state identification, the accuracy of evaluation results, and the foresight of communication maintenance under conditions of unbroken connection. Attached Figure Description
[0017] 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 communication status assessment method based on multi-source data fusion proposed in this invention; Figure 2 This is a flowchart illustrating the generation of state evolution operators for a Bluetooth communication state assessment method based on multi-source data fusion proposed in this invention. Detailed Implementation
[0018] 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.
[0019] refer to Figures 1-2 A Bluetooth communication status assessment method based on multi-source data fusion includes the following steps: Acquire and preprocess the Bluetooth communication operation data to be evaluated to generate a multi-source basic dataset for Bluetooth communication. The Bluetooth communication multi-source basic dataset is divided into connection stages and continuous segments are organized to generate state time sequence segments. Multi-source fusion and orchestration of state time series segments within the same time window are performed to generate multi-source fused time series data. Based on Takens' embedding theorem, delayed embedding processing is performed on multi-source fused time series data to generate communication trajectory data containing a sequence of phase space state points; A Koopman upscaling observation dataset is constructed based on communication trajectory data, and an evolutionary mapping fit is performed to generate state evolution operators; Eigenvalue decomposition and state mapping are performed on the state evolution operator to generate evolution feature data; A comprehensive evaluation of Bluetooth communication status is performed based on evolutionary feature data to generate Bluetooth communication status evaluation results.
[0020] In this embodiment, the Bluetooth communication operation data to be evaluated includes Bluetooth communication link data, frequency hopping channel status data, device operation status data, and connection parameter change data. Preprocessing includes time alignment, device affiliation marking, connection process marking, collection anomaly removal, and data normalization.
[0021] In this embodiment, the generation of state timing segments specifically includes: Based on the connection process markers and sampling times in the multi-source basic dataset of Bluetooth communication, the connection stage boundaries of the Bluetooth communication process are identified, and connection stage boundary data is generated. The connection phase boundaries include connection establishment boundary, service discovery boundary, feature subscription boundary, data transmission boundary, connection parameter update boundary, and reconnection maintenance boundary; the sampling time precision is 1 millisecond to 100 milliseconds; the connection establishment boundary is based on the first valid communication record after the connection state changes from disconnected to connected, and 2 to 10 sampling points are retained before and after the connection establishment boundary; the service discovery boundary is based on the first service information read record to the last service information read end record; the feature subscription boundary is based on the sampling position after feature notification is enabled, feature write is enabled, and feature subscription is confirmed; the data transmission boundary is based on the presence of at least one of data sending and receiving records within 3 to 20 consecutive sampling points; the connection parameter update boundary is based on the sampling position where at least one of the connection interval, slave delay, maximum transmission unit, physical layer rate, and transmission power changes, and 1 to 5 sampling points are retained before and after the connection parameter update boundary; the reconnection maintenance boundary is based on the re-establishment of the connection within 1 to 30 seconds after the disconnection record and the presence of data transmission records within 3 to 20 consecutive sampling points. The multi-source basic dataset for Bluetooth communication is segmented according to the connection phase boundaries, and continuously sampled data within the same connection phase are merged to generate continuous candidate segments. Phase segmentation involves aggregating Bluetooth communication link data, frequency hopping channel status data, device operating status data, and connection parameter change data within the same phase based on device ownership markers, connection process markers, and sampling time. Continuous sampling refers to sampling intervals between adjacent data within the same connection phase that are 1 to 3 times the sampling period. When the sampling interval between adjacent data exceeds 3 times the sampling period, the location of the adjacent data is used as the segment break point. The minimum number of sampling points for consecutive candidate segments is 10 to 50, and the duration of consecutive candidate segments is 100 milliseconds to 60 seconds. When segmenting adjacent connection phases, 1 to 5 overlapping sampling points are retained before and after the connection phase boundary. The sampling continuity, device affiliation consistency, and connection process consistency of continuous candidate segments are sorted out, and segment tags and connection stage tags are configured to generate state time sequence segments. The state time sequence segments include segment tags, connection stage tags, sampling time series, and multi-source basic data of Bluetooth communication within the segment. Sampling continuity refers to the arrangement of data within the same continuous candidate segment in the order of sampling time, with the time interval between adjacent data remaining within an allowable range; when sampling is interrupted, time is repeated, or the order is reversed, it is considered a segment break point. Device ownership consistency means that data within the same consecutive candidate segment belongs to the same Bluetooth communication object, and Bluetooth communication link data, frequency hopping channel status data, device operating status data, and connection parameter change data have consistent device ownership tags. Connection process consistency means that data within the same consecutive candidate segment belongs to the same Bluetooth connection process, and the connection process marker remains consistent. When there is a disconnection and reconnection that does not meet the reconnection maintenance boundary conditions, a jump in the connection process marker, or data from different connection processes are mixed in, the segments are divided into different state timing segments.
[0022] In this embodiment, the generation of multi-source fusion time-series data specifically includes: Based on the sampling time sequence and connection stage marker in the state time sequence segment, the Bluetooth communication multi-source basic data in the segment are divided into the same time window and arranged within the window to generate multi-source arranged data within the window. The multi-source data within a window includes the sampling order, source type, time position within the window, and window affiliation marker for each source's basic data. The sampling order is determined from earliest to latest according to the sampling time within the same fusion time window. The source type is used to indicate the data category to which each source's basic data belongs. The time position within the window is used to indicate the sampling position of each source's basic data relative to the start time of the window. The window affiliation marker is used to aggregate data within the same fusion time window. The multi-source data within the window is arranged in source order, dimensionally unified, and state variables are concatenated to generate a multi-source communication state vector corresponding to each fusion time window. The multi-source communication state vector includes link state variables, frequency hopping channel state variables, device operation state variables, and connection parameter state variables. Link state variables are generated by unifying the units of signal strength, connection interval, response delay, retransmission count, and data arrival interval within the same fusion time window. The frequency hopping channel state quantity is generated by arranging the channel state according to the channel number, frequency hopping order, channel reception strength, channel retransmission record and channel verification failure record within the same fusion time window; The device operating status parameters are generated by arranging the operating status according to the device buffer usage, sending queue length, receiving queue length, and processing callback delay within the same fusion time window; The connection parameter status is generated by arranging the parameter status according to the connection interval update, slave delay update, maximum transmission unit negotiation, physical layer rate switching and transmit power adjustment within the same fusion time window. The multi-source communication state vectors are arranged in chronological order according to the fusion time window, and the fragment markers and connection stage markers are retained to generate multi-source fusion time-series data. The multi-source fusion time-series data includes fragment markers, connection stage markers, fusion time window sequences, and multi-source communication state vector sequences.
[0023] In this embodiment, the generation of communication trajectory data specifically includes: Based on Takens' embedding theorem, observation state quantities are selected from the multi-source communication state vector sequence in multi-source fused time series data, and then arranged in time order according to the fused time window sequence to generate a communication observation state sequence. The observed state quantities include at least one of the following: link state quantities, frequency hopping channel state quantities, device operation state quantities, and connection parameter state quantities; the communication observation state sequence includes the observed state quantities, fusion time window markers, and time positions within the window; the window length of the fusion time window is 100 milliseconds to 1000 milliseconds, the moving step size of adjacent fusion time windows is 50 milliseconds to 500 milliseconds, the number of fusion time windows used for delay embedding processing within each state time segment is not less than 20, and the number of state components of the observed state quantities is 4 to 32; The delay step size is configured based on the sampling time interval and the continuity of state changes of the communication observation state sequence, and the current observation state quantity and historical observation state quantity are extracted from the communication observation state sequence according to the delay step size to generate a delayed observation sequence. State change continuity refers to the seamless transition between changes in the observed state quantities of adjacent fusion time windows in the communication observation state sequence, without any abrupt breaks caused by acquisition interruptions, stage jumps, or the mixing of abnormal data. It is used to determine whether delayed sampling can reflect the continuous evolution of Bluetooth communication state. The delay step size is the number of fusion time windows between two adjacent delayed samplings; the historical observation state quantity is the observation state quantity obtained by backtracking from the current fusion time window according to the delay step size; The embedding dimension is configured according to the number of historical observations required for the Bluetooth communication state evolution, and the delayed observation sequence is combined according to the delayed coordinate reconstruction method in Takens' embedding theorem to generate phase space state points. The generation of phase space state points includes: taking the current fusion time window in the communication observation state sequence as the current reconstruction position, extracting the observation state quantities within the current fusion time window as the current observation state quantities; backtracking backward from the current reconstruction position according to the delay step size, sequentially extracting the observation state quantities within multiple historical fusion time windows as historical observation state quantities; arranging and combining the current observation state quantities and the historical observation state quantities obtained according to the delay step size in the order of delay coordinates to generate a phase space state point; if the total number of current observation state quantities and historical observation state quantities is insufficient to meet the embedding dimension requirement, a phase space state point at the current reconstruction position is not generated; the embedding dimension is the number of delayed observation state quantity groups required to constitute a single phase space state point; The phase space state points are arranged in chronological order according to the fusion time window, and the fragment markers, connection stage markers and state point time positions are retained to generate a phase space state point sequence. The phase space state points in the same phase space state point sequence use the same delay step size and the same embedding dimension, and the number of phase space state points in each phase space state point sequence is not less than 10. The phase space state point sequence is continuously organized, and the phase space state point sequence that meets the continuous arrangement conditions is used as the communication trajectory data. The continuous arrangement conditions include consistent segment labels, consistent connection stage labels, continuous state point time position, consistent delay step size, consistent embedding dimension, and complete delayed observation state quantity. The communication trajectory data includes segment labels, connection stage labels, phase space state point sequence, state point time position, and state point arrangement order.
[0024] In this embodiment, the generation of the state evolution operator specifically includes: Based on the phase space state point sequence in the communication trajectory data, the original state arrangement and higher-order change arrangement are performed to generate linear and nonlinear observation data. Linear observation data is generated from the current and historical observation state quantities at each phase space state point, arranged in the original state order of current first and historical last. The generation of nonlinear observation data includes: multiplying each observed state variable in the linear observation data by itself to generate second-order observation data; calculating the difference between observed state variables of the same source type and adjacent delay positions to generate delay difference observation data; multiplying the delay difference observation data by itself to generate second-order difference observation data; and arranging the second-order observation data, delay difference observation data, and second-order difference observation data according to the phase space state point arrangement order to generate nonlinear observation data. The current and historical observation state quantities in the phase space state points are arranged with delay to generate time-delayed observation data. The time-delayed observation data includes the observation state quantities of the current fusion time window, the observation state quantities of the historical fusion time windows obtained by backtracking according to the delay step size, the delay position markers, and the number of delay intervals. Synchronous and delayed changes among observed state quantities from different sources in phase space state points are combined and arranged to generate interactive observation data. Specifically, within the same phase space state point, current observed state quantities from different sources are selected as synchronous interaction objects; difference and product calculations are performed on the synchronous interaction objects to generate synchronous difference data and synchronous product data; the current observed state quantity and historical observed state quantities from different sources obtained by backtracking according to the delay step size are selected as delayed interaction objects; difference and product calculations are performed on the delayed interaction objects to generate delay difference data and delay product data; the synchronous difference data, synchronous product data, delay difference data, and delay product data are aggregated according to the arrangement order of phase space state points to generate interactive observation data. Linear observation data, nonlinear observation data, time-delay observation data, and interactive observation data are collected according to the phase space state points in order to generate the Koopman upgraded observation dataset. Each phase space state point in the Koopman upscaling observation dataset corresponds to one set of Koopman upscaling observation data. Each set of Koopman upscaling observation data is arranged in the order of linear observation data, nonlinear observation data, time-delay observation data, and interactive observation data, and is fixed at 64 upscaling observation data components. When there are fewer than 64 upscaling observation data components, zero values are added to the last position to reach 64 upscaling observation data components. When there are more than 64 upscaling observation data components, the first 64 upscaling observation data components are retained according to the arrangement order. All sets of Koopman upscaling observation data within the same communication trajectory data use the same number of components and the same arrangement order. According to the order of adjacent state points in the phase space state point sequence, the Koopman upscaling observation data of the previous state point is configured as the evolutionary input data, and the Koopman upscaling observation data of the next state point is configured as the evolutionary output data; the evolutionary input data and the evolutionary output data have the same fragment label, connection stage label, delay step size and embedding dimension. Evolutionary mapping fitting is performed based on evolutionary input data and evolutionary output data to generate candidate state evolution operators; The generation of candidate state evolution operators includes: pairing evolutionary input data and evolutionary output data according to the order of adjacent state points to generate evolutionary mapping samples; establishing a mapping parameter matrix according to the number of up-dimensional observation data components in the evolutionary input data and the number of up-dimensional observation data components in the evolutionary output data, where the column direction of the mapping parameter matrix corresponds to the up-dimensional observation data components in the evolutionary input data, and the row direction corresponds to the up-dimensional observation data components in the evolutionary output data; setting the mapping parameters between input components and output components of the same source type and the same arrangement position as unit mapping values, and setting the remaining mapping parameters to 0 values to generate an initial mapping parameter matrix; for each set of evolutionary mapping samples, multiplying each up-dimensional observation data component in the evolutionary input data with the corresponding mapping parameter in the initial mapping parameter matrix, and accumulating them according to the arrangement position of the evolutionary output data components to generate predicted evolutionary output data; calculating the difference between the predicted evolutionary output data and the evolutionary output data term by term, and accumulating the squared differences of each term to generate a sum of squared differences; for each term in the mapping parameter matrix... For each mapping parameter, the evolutionary input data component corresponding to the column containing the mapping parameter, the difference corresponding to the row containing the mapping parameter, and the step size coefficient are multiplied to generate the parameter adjustment amount. The parameter adjustment amount is then subtracted from the current mapping parameters to generate the adjusted mapping parameters. The step size coefficient is between 0.001 and 0.01. After each adjustment of the mapping parameter matrix, the predicted evolutionary output data is regenerated and the fitted difference sum of squares is recalculated. When the adjusted fitted difference sum of squares is less than the original fitted difference sum of squares, the adjusted mapping parameter matrix is retained. When the adjusted fitted difference sum of squares is not less than the original fitted difference sum of squares, the original mapping parameter matrix is restored and the step size coefficient is halved. When the decrease in the fitted difference sum of squares in three consecutive parameter adjustments does not exceed 0.001 of the previous fitted difference sum of squares, the parameter adjustment is stopped, and the set of mapping parameters with the lowest fitted difference sum of squares is used as the state evolution mapping parameters. The state evolution mapping parameters, fitted difference sum of squares, segment labels, connection stage labels, and phase space state point sequences are bound together to generate candidate state evolution operators. Candidate state evolution operators are processed by segment assignment, connection stage assignment, and operator stability adjustment to generate state evolution operators. The state evolution operators include segment labels, connection stage labels, state evolution mapping parameters, fitted sum of squared differences, and phase space state point sequences. The operator stability sorting includes: continuously arranging candidate state evolution operators under the same segment label and the same connection stage label; when the difference between each item of the state evolution mapping parameter of adjacent candidate state evolution operators does not exceed 0.05, and the change in the sum of squared fitting differences of adjacent candidate state evolution operators does not exceed 0.1 of the sum of squared fitting differences of the previous candidate state evolution operator, the candidate state evolution operator is determined to meet the operator stability requirements; when at least one of the following occurs, such as the difference between each item of the state evolution mapping parameter exceeds 0.05, the change in the sum of squared fitting differences exceeds 0.1 of the sum of squared fitting differences of the previous candidate state evolution operator, the phase space state point sequence has sampling breaks, or the phase space state point sequence has connection stage mixing, the candidate state evolution operator is removed.
[0025] In this embodiment, the generation of evolutionary feature data specifically includes: The state evolution operator is subjected to evolutionary component separation and change direction classification to generate operator decomposition data, which includes stable evolution components, periodic fluctuation components, drift evolution components, and unstable evolution components. Specifically, the state evolution mapping parameters in the state evolution operator are subjected to eigenvalue decomposition and term-by-term change calculation to obtain the evolution direction, evolution intensity, change direction label, offset direction, cumulative offset, and sum of squared differences in the fitted difference for each component of the upgraded observation data. Components with an evolution intensity greater than or equal to 0.95 and less than or equal to 1.05, consistent evolution direction labels between adjacent state points, and sum of squared differences in the fitted difference are selected. Evolution directions with a sum of squares less than or equal to 0 are classified as stable evolution directions, and stable evolution components are generated. In four consecutive adjacent state points, evolution directions where adjacent change directions are marked with three consecutive reverse switchings are classified as repetitive change directions, and periodic fluctuation components are generated. In four consecutive adjacent state points, evolution directions where adjacent change directions are marked with consistent values and the cumulative offset increases successively are classified as persistent offset directions, and drift evolution components are generated. Evolution directions with an evolution intensity greater than 1.05 and a sum of squared differences of fit increasing three consecutive times are classified as amplified change directions, and unstable evolution components are generated. The operator decomposition data is mapped to the state point time position and the corresponding multi-source communication state vector in the phase space state point sequence to generate state mapping data; the state mapping data includes the state point time position, fusion time window marker, observation state quantity source type, decomposition component type and component action position. Based on the state mapping data, the stability, fluctuation, drift and instability of Bluetooth communication state evolution are characterized by feature aggregation, and candidate evolution feature data are generated. The candidate evolution feature data includes stable evolution features, periodic fluctuation features, continuous drift features and critical instability features. Candidate evolutionary feature data are processed temporally according to fragment labels, connection stage labels, and state point time positions to generate evolutionary feature data. Evolutionary feature data includes fragment labels, connection stage labels, fusion time window labels, evolutionary feature types, and feature change directions. The determination of the characteristic change direction includes: determining the characteristic change direction based on the operator decomposition data corresponding to the candidate evolution characteristic data; when the candidate evolution characteristic data comes from a stable evolution component, the characteristic change direction is recorded as the stable change direction; when the candidate evolution characteristic data comes from a periodic fluctuation component, the characteristic change direction is recorded as the repetitive change direction; when the candidate evolution characteristic data comes from a drift evolution component, the characteristic change direction is recorded as the continuous offset direction; when the candidate evolution characteristic data comes from an unstable evolution component, the characteristic change direction is recorded as the amplified change direction.
[0026] In this embodiment, the generation of Bluetooth communication status assessment results specifically includes: Based on evolutionary feature data, state assessment evidence is configured and stage associations are performed according to connection stage markers to generate stage state evidence data; The configuration of state assessment evidence includes: under the same segment label and the same connection stage label, configuring 10 consecutive fusion time windows as one evidence statistical segment; within one evidence statistical segment, when the number of windows where stable evolution characteristics and stable change directions appear simultaneously reaches 6, it is configured as stable state evidence; when the number of windows where periodic fluctuation characteristics and repetitive change directions appear simultaneously reaches 4, it is configured as periodic fluctuation state evidence; when the number of windows where persistent drift characteristics and persistent shift directions appear simultaneously reaches 4, it is configured as latent deterioration state evidence; when the number of windows where critical instability characteristics and amplified change directions appear simultaneously reaches 3, it is configured as critical instability state evidence. Based on the state assessment evidence type and connection phase marker in the phase state evidence data, the Bluetooth communication state category is determined, and Bluetooth communication state category data is generated. The Bluetooth communication state category determination includes: a stable communication state is generated when the state assessment evidence type is stable state evidence and the connection phase is marked as the data transmission phase; a periodic fluctuation state is generated when the state assessment evidence type is periodic fluctuation state evidence and the connection phase is marked as the data transmission phase; a retransmission maintenance state is generated when the state assessment evidence type is latent degradation state evidence and the connection phase is marked as the data transmission phase; a connection parameter mismatch state is generated when the state assessment evidence type is latent degradation state evidence and the connection phase is marked as the connection parameter update phase; and a critical instability state before disconnection is generated when the state assessment evidence type is critical instability state evidence and the connection phase is marked as the reconnection maintenance phase. The Bluetooth communication status category data and phase status evidence data are organized according to the fragment tag and connection phase tag to generate Bluetooth communication status assessment results.
[0027] Example 1: To verify the feasibility of this invention in practice, it was applied to a scenario of evaluating the communication status between Bluetooth handheld terminals and Bluetooth positioning tags in a smart warehouse area. In this scenario, 6 Bluetooth handheld terminals maintained connections with 48 Bluetooth positioning tags. The warehouse area contained 12 rows of metal shelves and 4 mobile handling devices. The tags uploaded location broadcasts and status data every 200 milliseconds. The Bluetooth connection was displayed as connected most of the time, and no direct disconnection alarm occurred. However, when personnel obstructed the view, people walked between shelves, or handling equipment approached, there were instances where response delays gradually increased, retransmission records appeared frequently, and connection parameters were frequently adjusted, yet no single indicator met the abnormal judgment criteria.
[0028] During implementation, 20 minutes of continuous Bluetooth communication operation data to be evaluated were collected with a sampling time accuracy of 10 milliseconds. This yielded data on signal strength, connection interval, response latency, retransmission count, data arrival interval, channel number, frequency hopping order, channel received signal strength, channel check failure records, device buffer usage, transmit queue length, receive queue length, processing callback latency, connection interval update, slave latency update, maximum transmission unit negotiation, physical layer rate switching, and transmit power adjustment records for each connection process. During the data transmission phase, the fusion time window was set to 500 milliseconds, the movement step size was set to 100 milliseconds, and each state time segment retained 60 to 180 consecutive fusion time windows.
[0029] When performing delayed embedding processing on multi-source fused time-series data, the delay step size is set to two fusion time windows, and the embedding dimension is set to four sets of delayed observation state variables. Each phase space state point consists of the current fusion time window and the historical fusion time window data traced back. Subsequently, 64 Koopman upscaling observation data components are generated for each phase space state point. Adjacent state points are configured as evolutionary input data and evolutionary output data, respectively, and evolutionary feature data are formed through state evolution operators. Based on the evolutionary feature data, stable state evidence, periodic fluctuation state evidence, latent degradation state evidence, and critical instability state evidence are configured to generate Bluetooth communication state assessment results.
[0030] Based on 20 minutes of continuous Bluetooth communication operation data collected within the aforementioned smart warehouse area, and using gateway connection logs, terminal communication records, and manual review status as comparison records, a statistical comparison of the traditional static threshold evaluation method and the method of this invention was conducted in the same scenario, yielding the following performance comparison data: Table 1. Performance Comparison of the Invention and Traditional Bluetooth Communication Status Evaluation Methods
[0031] As can be clearly seen from Table 1, the method of the present invention is superior to the traditional method in many indicators.
[0032] The method of this invention improves the latent degradation identification rate from 78.6% to 86.9%, an increase of 8.3%. This is because the invention does not only look at the current signal strength, delay and retransmission count, but reflects the continuous changes in Bluetooth communication status through multi-source fusion of timing data and phase space state point sequences, making it easier to identify slow degradation in the unbroken connection state.
[0033] Regarding the periodic fluctuation recognition rate, the traditional method achieves 81.4%, while the method of this invention achieves 88.2%, representing an improvement of 6.8%. This improvement mainly comes from the delayed embedding processing under the Takens embedding theorem. The current observation state and the historical observation state are combined into communication trajectory data, which can preserve the fluctuation patterns between multiple fusion time windows, thus making the recognition of periodic jitter more stable.
[0034] Regarding the retransmission retention recognition rate, the method of this invention achieves 87.1%, which is 7.3% higher than the 79.8% of the traditional method. This is because the Koopman upscaling observation dataset contains linear observation data, nonlinear observation data, latency observation data, and interactive observation data, which can incorporate the linkage between the number of retransmissions, response latency, and device buffer changes into the evolutionary mapping fitting.
[0035] Regarding the critical instability identification rate, the method of this invention achieves 84.0%, while the traditional method achieves 76.5%, representing an improvement of 7.5%. This improvement stems from the eigenvalue decomposition of the state evolution operator, which can separate the amplified change direction and critical instability characteristics from the state evolution process, eliminating reliance on the result judgment after the disconnection occurs.
[0036] Regarding the false alarm rate, the method of this invention reduces the false alarm rate from 9.8% to 7.1%, a reduction of 2.7%. The reason is that this invention uses stage state evidence data for comprehensive evaluation, and only generates the corresponding state when the evolutionary feature type and feature change direction meet the configuration conditions within the evidence statistics segment, thereby reducing misjudgments caused by single-point fluctuations.
[0037] Regarding the average warning lead time, the method of this invention provides 6.3 seconds, compared to 4.6 seconds for the traditional method, representing an increase of 1.7 seconds. This improvement stems from the earlier characterization of state change trends using communication trajectory data and state evolution operators, enabling the system to identify critical instability tendencies before the Bluetooth connection is lost.
[0038] The above are merely preferred embodiments 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 Bluetooth communication status assessment method based on multi-source data fusion, characterized in that, Includes the following steps: Acquire and preprocess the Bluetooth communication operation data to be evaluated to generate a multi-source basic dataset for Bluetooth communication. The Bluetooth communication multi-source basic dataset is divided into connection stages and continuous segments are organized to generate state time sequence segments. Multi-source fusion and orchestration of state time series segments within the same time window are performed to generate multi-source fused time series data. Based on Takens' embedding theorem, delayed embedding processing is performed on multi-source fused time series data to generate communication trajectory data containing a sequence of phase space state points; Based on communication trajectory data, a Koopman up-dimensional observation dataset is constructed and an evolutionary mapping fit is performed to generate state evolution operators; Eigenvalue decomposition and state mapping are performed on the state evolution operator to generate evolution feature data; A comprehensive evaluation of Bluetooth communication status is performed based on evolutionary feature data to generate Bluetooth communication status evaluation results.
2. The Bluetooth communication status assessment method based on multi-source data fusion according to claim 1, characterized in that, The Bluetooth communication operation data to be evaluated includes Bluetooth communication link data, frequency hopping channel status data, device operation status data, and connection parameter change data. The preprocessing includes time alignment, device affiliation marking, connection process marking, collection anomaly removal, and data normalization.
3. The Bluetooth communication status assessment method based on multi-source data fusion according to claim 1, characterized in that, The generation of the state time sequence segment specifically includes: Based on the connection process markers and sampling times in the multi-source basic dataset of Bluetooth communication, the connection stage boundaries of the Bluetooth communication process are identified, and connection stage boundary data is generated. The multi-source basic dataset for Bluetooth communication is segmented according to the connection phase boundaries, and continuously sampled data within the same connection phase are merged to generate continuous candidate segments. The continuous candidate segments are sorted for sampling continuity, device affiliation consistency and connection process consistency, and segment tags and connection stage tags are configured to generate state time sequence segments. The state time sequence segments include segment tags, connection stage tags, sampling time series and multi-source basic data of Bluetooth communication within the segment.
4. The Bluetooth communication status assessment method based on multi-source data fusion according to claim 1, characterized in that, The generation of the multi-source fused time-series data specifically includes: Based on the sampling time sequence and connection stage marker in the state time sequence segment, the Bluetooth communication multi-source basic data in the segment are divided into the same time window and arranged within the window to generate multi-source arranged data within the window. The multi-source data within the window is arranged in the order of source, unified in terms of units, and spliced with state variables to generate a multi-source communication state vector corresponding to each fusion time window. The multi-source communication state vectors are arranged in chronological order according to the fusion time window, and the fragment markers and connection stage markers are retained to generate multi-source fusion time-series data.
5. The Bluetooth communication status assessment method based on multi-source data fusion according to claim 1, characterized in that, The generation of the communication trajectory data specifically includes: Based on Takens' embedding theorem, observation state quantities are selected from the multi-source communication state vector sequence in multi-source fused time series data, and then arranged in time order according to the fused time window sequence to generate a communication observation state sequence. The delay step size is configured based on the sampling time interval and the continuity of state changes of the communication observation state sequence, and the current observation state quantity and historical observation state quantity are extracted from the communication observation state sequence according to the delay step size to generate a delayed observation sequence. The embedding dimension is configured according to the number of historical observations required for the Bluetooth communication state evolution, and the delayed observation sequence is combined according to the delayed coordinate reconstruction method in Takens' embedding theorem to generate phase space state points. The phase space state points are arranged in chronological order according to the fusion time window, and the fragment markers, connection stage markers and state point time positions are preserved to generate a phase space state point sequence. The phase space state point sequence is continuously organized, and the phase space state point sequence that meets the continuous arrangement conditions is used as communication trajectory data. The continuous arrangement conditions include consistent segment labels, consistent connection stage labels, continuous state point time position, consistent delay step size, consistent embedding dimension, and complete delayed observation state quantity.
6. The Bluetooth communication status assessment method based on multi-source data fusion according to claim 1, characterized in that, The generation of the state evolution operator specifically includes: Based on the phase space state point sequence in the communication trajectory data, the original state arrangement and higher-order change arrangement are performed to generate linear and nonlinear observation data. Based on the current and historical observed state quantities in the phase space state points, the time-delayed observation data is generated by arranging them with a delay. Synchronous and delayed changes among observed state quantities from different sources in phase space are combined and arranged to generate interactive observation data; Linear observation data, nonlinear observation data, time-delay observation data, and interactive observation data are collected according to the phase space state points in order to generate the Koopman upgraded observation dataset. According to the order of adjacent state points in the phase space state point sequence, the Koopman upscaling observation data of the previous state point is configured as the evolution input data, and the Koopman upscaling observation data of the next state point is configured as the evolution output data. Evolutionary mapping fitting is performed based on evolutionary input data and evolutionary output data to generate candidate state evolution operators; The candidate state evolution operators are sorted by fragment assignment, connection stage assignment, and operator stability to generate state evolution operators.
7. The Bluetooth communication status assessment method based on multi-source data fusion according to claim 1, characterized in that, The generation of the evolutionary feature data specifically includes: The state evolution operator is separated into evolutionary components and the direction of change is divided to generate operator decomposition data, which includes stable evolution components, periodic fluctuation components, drift evolution components and unstable evolution components. The operator decomposition data is mapped to the state point time position and the corresponding multi-source communication state vector in the phase space state point sequence to generate state mapping data. Based on state mapping data, features of stability, fluctuation, drift and instability in the Bluetooth communication state evolution process are collected to generate candidate evolution feature data, which includes stable evolution features, periodic fluctuation features, continuous drift features and critical instability features. Candidate evolutionary feature data are processed temporally according to fragment labels, connection stage labels, and state point time positions to generate evolutionary feature data. The evolutionary feature data includes fragment labels, connection stage labels, fusion time window labels, evolutionary feature types, and feature change directions.
8. The Bluetooth communication status assessment method based on multi-source data fusion according to claim 1, characterized in that, The generation of the Bluetooth communication status assessment result specifically includes: Based on evolutionary feature data, state assessment evidence is configured and stage associations are performed according to connection stage markers to generate stage state evidence data; Based on the state assessment evidence type and connection phase marker in the phase state evidence data, the Bluetooth communication state category is determined, and Bluetooth communication state category data is generated. The Bluetooth communication status category data and phase status evidence data are organized according to the fragment tag and connection phase tag to generate Bluetooth communication status assessment results.
