Work card high-precision positioning method integrating multi-source positioning and inertial navigation

Abstract

The invention relates to a multi-source positioning and inertial navigation fused work card high-precision positioning method. The method comprises the following steps: acquiring observation data of at least two heterogeneous wireless positioning modules and inertial navigation data of an IMU (Inertial Measurement Unit) in real time; based on observation data quality evaluation and motion state recognition, a fusion strategy defining the weight of each data source is generated in a self-adaptive mode; performing space-time alignment and fusion on the data according to the strategy, and outputting a positioning result; and finally, performing two-stage compensation and closed-loop feedback on IMU accumulated drift by using the result. According to the invention, through an environment adaptive fusion strategy and closed-loop error compensation, the problems of insufficient positioning precision, poor stability and short endurance in a complex industrial scene are effectively solved, and the positioning effects of high precision, high robustness and low power consumption are realized.

Description

Technical Field

[0001] This application relates to the field of employee badge positioning, and in particular to a high-precision positioning method for employee badges that integrates multi-source positioning and inertial navigation. Background Technology

[0002] In scenarios such as industrial personnel management and factory safety supervision, employee badge positioning technology has become a key support for ensuring production safety and improving management efficiency. Among them, multi-source positioning and inertial navigation fusion technology is widely used in employee badge positioning solution design because it can balance positioning coverage and accuracy. Existing technologies typically use a combination of various heterogeneous wireless positioning modules such as ultra-wideband (UWB), satellite (BeiDou / GPS), Bluetooth, and Wi-Fi with inertial measurement units (IMUs). Multi-source data fusion is achieved through filtering algorithms or fixed weight allocation methods to obtain personnel location information and meet basic positioning requirements.

[0003] However, existing fusion methods have significant drawbacks. First, their fusion strategies (such as weights and filtering parameters) are usually pre-set statically, failing to adapt to complex industrial environments (such as non-line-of-sight propagation and scene switching) and the user's dynamic motion state. This leads to a sharp drop in positioning accuracy when the environment changes abruptly or the motion pattern changes. Second, although the inherent cumulative drift error of inertial navigation is smoothed by filtering, there is a lack of an effective mechanism for long-term, closed-loop feedback compensation using high-confidence fusion results. The error will still accumulate over time, significantly increasing after long-term positioning. Finally, power consumption control is unreasonable. Many methods adopt a mode of activating all positioning modules at all times, failing to achieve dynamic collaborative scheduling of sensors. This results in short battery life for work badges, which cannot meet the long-term use requirements of industrial scenarios.

[0004] In response to the aforementioned technologies, this application proposes a high-precision positioning method for work badges that integrates multi-source positioning and inertial navigation. This method aims to address the problems of poor scene adaptation, unstable accuracy, high power consumption, and insufficient drift suppression in existing technologies, thereby improving the practicality and reliability of work badge positioning in complex industrial scenarios. Summary of the Invention

[0005] To address the aforementioned issues, this application provides a high-precision positioning method for work badges that integrates multi-source positioning and inertial navigation.

[0006] This application provides a high-precision positioning method for work badges that integrates multi-source positioning and inertial navigation, employing the following technical solution: A high-precision positioning method for employee badges, which integrates multi-source positioning and inertial navigation, includes the following steps: S1. Data Acquisition: Real-time acquisition of observation data from at least two heterogeneous wireless positioning modules built into the work badge, as well as inertial navigation data from the inertial measurement unit; the observation data includes ranging or positioning data from various heterogeneous wireless positioning modules. S2. Strategy Generation: Based on the quality assessment of the observation data and the current motion state identified based on the inertial navigation data, an adaptive fusion strategy is generated; the fusion strategy at least defines the weight allocation of the observation data and the inertial navigation data in the subsequent fusion process; S3. Fusion Positioning: Based on the fusion strategy, the observation data and the inertial navigation data are spatiotemporally aligned and fused to output the fusion positioning result; S4. Error Compensation and Feedback: Using the fusion positioning results, the cumulative drift error of the inertial measurement unit is compensated to obtain corrected inertial navigation data, and the corrected inertial navigation data is fed back to the subsequent fusion positioning process.

[0007] Preferably, the compensation for the cumulative drift error of the inertial measurement unit in step S4 is implemented using a two-stage compensation model, including: First-level compensation: A filtering algorithm based on pedestrian kinematic constraints is applied to correct the raw inertial navigation data of the inertial measurement unit in real time to suppress short-term dynamic errors; Secondary compensation: The historical data sequence of the inertial measurement unit and the fused positioning result and the residual sequence generated by the independent position calculation of the inertial measurement unit are input into a pre-trained lightweight neural network model for prediction, and the correction amount for long-term cumulative drift is output. The correction amount output by the secondary compensation is used to correct the inertial navigation data after the primary compensation, and then the corrected inertial navigation data is input into the subsequent fusion positioning process.

[0008] Preferably, the lightweight neural network model is a long short-term memory network model, which is trained through transfer learning: first, it is pre-trained on a public or synthetic inertial navigation dataset, and then fine-tuned using measured data collected from the target work badge in a typical industrial scenario.

[0009] Preferably, the quality assessment in step S2 includes the identification of environmental characteristics, specifically: The channel features of the observed data are extracted, and a classification model is used to determine whether the signal propagation is under non-line-of-sight conditions and / or to identify the current typical environmental scene type; the typical environmental scene type includes at least industrial indoor scene, outdoor open scene and semi-obscured transition scene; The fusion strategy dynamically adjusts the trust weights of different observation data based on the judgment results of the non-line-of-sight conditions and / or the environmental scene type, and adjusts the error correction parameters in the fusion process accordingly.

[0010] Preferably, a sensor co-activation step is further included between step S1 and step S2: A power consumption-accuracy prediction model is constructed, which takes the current fusion strategy, historical positioning accuracy sequence and battery level as input; Based on the output of the power consumption-accuracy prediction model, some of the heterogeneous wireless positioning modules are dynamically activated or put into sleep mode, and the module activation status information is fed back to step S2; when generating the fusion strategy, step S2 makes adjustments based on the module activation status information.

[0011] Preferably, the power consumption-accuracy prediction model is a reinforcement learning model, and its reward function is configured to: give a positive reward during periods when the positioning accuracy is higher than a preset threshold, and apply a negative reward to any activation decision that causes the positioning accuracy to fall below the preset threshold.

[0012] Preferably, the filtering algorithm based on pedestrian kinematic constraints is an improved Kalman filter algorithm, which introduces zero velocity detection and zero angular velocity correction into its state equation, and the process noise covariance matrix is ​​adaptively adjusted according to the motion intensity identified in real time by the inertial navigation data of the inertial measurement unit.

[0013] Preferably, after identifying non-line-of-sight conditions, the affected observation data are corrected using a model, specifically as follows: Based on the non-line-of-sight type and confidence level output by the classification model, a corresponding distance deviation compensation model is selected from the preset error deviation library to compensate the original observation values, and the corrected observation data is then sent to step S3.

[0014] Preferably, the spatiotemporal alignment in step S3 specifically includes: The timestamps of the observation data and the inertial navigation data are aligned using a precision clock synchronization protocol. By using a unified coordinate system transformation matrix, the observation data and the inertial navigation data are transformed to the same coordinate system.

[0015] Preferably, the heterogeneous wireless positioning module is selected from any two or more combinations of ultra-wideband (UWB) positioning modules, satellite positioning modules, Bluetooth positioning modules, and Wi-Fi positioning modules.

[0016] In summary, this application includes at least one of the following beneficial technical effects: 1. Achieved high-precision and robust continuous positioning in complex industrial scenarios: By performing real-time environmental feature recognition and motion state assessment, and dynamically generating adaptive fusion strategies accordingly, this method changes the traditional fixed-parameter mode of fusion algorithms. It can intelligently cope with drastic signal changes from outdoor to indoor environments and from line-of-sight to non-line-of-sight, significantly improving positioning accuracy and stability in dynamic and complex environments, and solving the industry problem of drastic performance degradation of single or static fusion technologies during scene switching.

[0017] 2. Fundamentally and effectively suppresses the cumulative drift error of inertial navigation: An innovative two-stage compensation and closed-loop feedback mechanism combining model filtering and AI learning is proposed. The first-stage compensation corrects short-term dynamic errors in real time based on rigorous kinematic constraints; the second-stage compensation utilizes a lightweight neural network to learn and predict long-term drift patterns; and finally, the high-confidence fused positioning result is fed back to the compensation process. This closed-loop system achieves continuous calibration and suppression of IMU drift, providing a core guarantee for long-term, highly reliable autonomous navigation.

[0018] 3. System-level energy efficiency optimization was achieved while ensuring positioning performance: By introducing a sensor intelligent collaborative activation mechanism based on a predictive model, adaptive decision-making was extended from the data fusion layer to the hardware control layer. This mechanism uses positioning accuracy as a constraint, dynamically schedules the working status of each wireless positioning module, and links with the fusion strategy in real time. As a result, while meeting the accuracy requirements of different scenarios, the overall power consumption of the work badge was significantly reduced, the battery life was extended, and the practicality and economy of the solution were improved. Attached Figure Description

[0019] Figure 1 This is a block diagram of a high-precision positioning method for work badges that integrates multi-source positioning and inertial navigation, as described in Embodiment 1 of this application. Figure 2 This is a block diagram of a high-precision positioning method for work badges that integrates multi-source positioning and inertial navigation, as described in Embodiment 2 of this application. Detailed Implementation

[0020] The following is in conjunction with the appendix Figure 1 This application will be described in further detail.

[0021] This application discloses a high-precision positioning method for work badges that integrates multi-source positioning and inertial navigation. The core of this method is to address the problems of insufficient accuracy, poor stability, and high power consumption associated with single or fixed fusion positioning technologies in complex industrial scenarios. This method operates within a dedicated positioning work badge, which integrates a multi-source positioning module, an inertial measurement unit (IMU), a main control processor, and a communication unit.

[0022] Example 1: Reference Figure 1 A high-precision positioning method for employee badges, which integrates multi-source positioning and inertial navigation, includes the following steps: S1. Data Acquisition: Real-time acquisition of observation data from at least two heterogeneous wireless positioning modules built into the work badge, as well as inertial navigation data from the inertial measurement unit; the observation data includes ranging or positioning data from various heterogeneous wireless positioning modules; the heterogeneous wireless positioning modules are selected from any two or more combinations of ultra-wideband positioning modules, satellite positioning modules, Bluetooth positioning modules and Wi-Fi positioning modules. S2. Strategy Generation: Based on the quality assessment of the observation data and the current motion state identified based on the inertial navigation data, an adaptive fusion strategy is generated; the fusion strategy at least defines the weight allocation of the observation data and the inertial navigation data in the subsequent fusion process; S3. Fusion Positioning: Based on the fusion strategy, the observation data and the inertial navigation data are spatiotemporally aligned and fused to output the fusion positioning result; S4. Error Compensation and Feedback: Using the fusion positioning results, the cumulative drift error of the inertial measurement unit is compensated to obtain corrected inertial navigation data, and the corrected inertial navigation data is fed back to the subsequent fusion positioning process.

[0023] The specific implementation method is explained in detail below: S1. Data Acquisition: The work badge incorporates multiple heterogeneous wireless positioning modules. In this embodiment, an ultra-wideband (UWB) positioning module and a Bluetooth 5.1 positioning module are preferably integrated as the primary wireless positioning sources. In outdoor or near-window areas, satellite positioning modules (such as BeiDou / GPS) can also serve as effective data sources. These modules operate continuously. The UWB module provides high-precision ranging data based on Time Difference of Arrival (TDOA) or two-way ranging; the Bluetooth module provides positioning or area awareness data based on Received Signal Strength Indication (RSSI); and the satellite module provides latitude and longitude coordinates. All raw outputs such as ranging values ​​and coordinates from these wireless positioning modules are collectively referred to as "observation data" in this invention. Simultaneously, the work badge's built-in nine-axis IMU (including a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer) collects raw data at high frequency (e.g., 100Hz), and after preliminary calculation, information such as angular velocity, acceleration, and attitude angles is obtained, collectively referred to as inertial navigation data. The main control processor synchronously collects all the above data through a unified interface (such as SPI or I2C).

[0024] S2, Strategy Generation: This step is crucial for achieving adaptive fusion. It comprises two parallel analysis processes: observation data quality assessment and IMU-based motion state identification.

[0025] 1. Quality Assessment: First, feature extraction is performed on the observed data. For example, features such as the number of multipath components, peak amplitude, and kurtosis are extracted from the impulse response of the UWB channel; the mean, variance, and distribution entropy of signal strength are extracted from the Bluetooth RSSI sequence. Then, these feature vectors are input into a pre-trained classification model. In this embodiment, a random forest classifier is preferably used. This classifier outputs two key judgments: 1) the current signal propagation conditions, divided into line-of-sight (LOS) and non-line-of-sight (NLOS), and provides a confidence level; 2) the current typical environmental scene type, such as open outdoor, metal-intensive workshop, office corridor, etc. The model is trained based on a large number of labeled real-world datasets, ensuring a high recognition rate (over 95%). The typical environmental scene types include at least industrial indoor scenes, open outdoor scenes, and semi-obscured transition scenes, and additional types can be added according to actual usage requirements.

[0026] 2. Motion state recognition: At the same time, the inertial navigation data of the IMU is analyzed. By detecting the amplitude and variance of acceleration and angular velocity, the wearer's motion state, such as "stationary", "walking", "running" or "abnormal shaking", is identified in real time.

[0027] Fusion Strategy Generation: Based on the aforementioned quality assessment results and motion state, the system adaptively generates a fusion strategy. This strategy is a set of dynamic parameters, the core of which defines the weight allocation scheme for the four data sources—UWB observation data, Bluetooth observation data, satellite observation data, and IMU inertial navigation data—at the current moment. For example: In "open outdoor" scenarios and when identified as LOS, the strategy may assign higher weight to satellite data, while IMU data is used for smoothing.

[0028] When a “metal-intensive workshop” is identified as an NLOS, the policy significantly reduces the initial trust weight of the current UWB observation data and increases the weight of the IMU and Bluetooth (used for area constraints).

[0029] When the motion state is "stationary", the strategy introduces a "zero-rate update" constraint to reset the IMU velocity error.

[0030] S3, Fusion Positioning: Data fusion is performed based on the fusion strategy generated in S2. First, spatiotemporal alignment is performed: all observation data and IMU data are aligned to the same time base through hardware clock synchronization or software timestamp interpolation (the software timestamp interpolation is based on a precise clock synchronization protocol, ensuring that the timestamp alignment error is controlled within sub-microseconds). Then, all positioning data are unified to the local Cartesian coordinate system of the work badge deployment area using a preset coordinate transformation matrix. Next, an Extended Kalman Filter (EKF) framework is used as the fusion engine. The dynamic weights defined in S2 are transformed into the observation noise covariance matrix of the corresponding observation source in the EKF (the higher the weight, the smaller the noise covariance setting). Within this framework, IMU data is used for state prediction (position, velocity, attitude), while various aligned and corrected (e.g., NLOS correction) observation data are used for measurement updates, ultimately outputting an optimal fused positioning result.

[0031] S4. Error Compensation and Feedback: This is the core closed loop for achieving long-term high accuracy. Utilizing the high-confidence fused positioning results output by S3, a two-stage compensation is performed to address the inherent cumulative drift error of the IMU: First-level compensation (short-term dynamic error suppression): An improved Kalman filter algorithm is employed. This algorithm, based on the standard EKF, introduces pedestrian kinematic constraints as virtual observations. Specifically, it includes: Zero-speed detection: When it is determined from IMU data that the wearer's feet are stationary upon touching the ground, the velocity state variable is forcibly set to zero, and the corresponding covariance is reduced.

[0032] Zero angular velocity correction: Corrects minor drifts of the gyroscope during periods of rest or steady movement.

[0033] Adaptive process noise: Based on the "motion intensity" (such as acceleration variance) identified by S2, the process noise covariance matrix of the filter is dynamically adjusted, allowing for larger model errors during violent motion and tightening constraints during steady motion.

[0034] Secondary Compensation (Long-Term Cumulative Drift Learning): After primary compensation, errors are not completely eliminated, especially slow drift in attitude angles. This invention innovatively introduces a lightweight long short-term memory (LSTM) network model. The input features of this network include: recent (e.g., 30-second) IMU angular velocity and acceleration sequences, and position residual sequences calculated from fused localization results and pure IMU integration results. After training, the lightweight LSM network model can learn drift patterns under specific motion modes and environments, and output long-term cumulative drift corrections for position and attitude.

[0035] Model Training: A transfer learning strategy was employed. First, the LSTM network was pre-trained on a large, publicly available inertial navigation dataset to learn generalized drift patterns. Then, a small amount (e.g., several hours) of data with real trajectory labels collected from the employee badge within the target factory area was used for fine-tuning, adapting the model to local characteristics such as magnetic field interference and wearing methods. The final model parameters were then embedded in the employee badge's firmware.

[0036] Closed-loop feedback: The correction output from the secondary compensation model is superimposed on the IMU navigation data (such as position and velocity) after primary compensation to generate corrected inertial navigation data. This data forms a closed loop on two levels: First, it directly serves as a more accurate initial value for IMU state prediction in the next S3 fusion cycle; second, the correction information it contains indirectly affects the fusion strategy in the next cycle. For example, when the IMU has been effectively corrected, its weight reliability in the fusion strategy will increase accordingly.

[0037] When the classification model determines that a UWB link is NLOS, the system will call a preset distance bias compensation model (such as a regression model based on NLOS type and signal characteristics) to compensate for the original ranging value of the link, obtain the corrected observation data, and then send it to S3 for fusion, thereby reducing the positive bias caused by NLOS.

[0038] Example 2: This embodiment adds a sensor collaborative activation step to optimize power consumption, based on Embodiment 1.

[0039] Reference Figure 2 After S1 (data acquisition) and before S2 (policy generation), the system executes the sensor collaborative activation step: Model Construction: A power consumption-accuracy prediction model is trained on the backend server. This embodiment preferably uses a deep reinforcement learning model, whose state is defined as: a summary of the current fusion strategy, a recent historical positioning accuracy sequence, and the battery level of the work badge. Its actions are defined as: "activation," "sleep," or "low-power scan" commands for different wireless positioning modules (such as UWB and Bluetooth). Its reward function is carefully designed as: R = α*I (accuracy > threshold) - β*power consumption - γ*I (accuracy ≤ threshold), where I() is the indicator function. This means that the model will receive a positive reward when the positioning accuracy meets the target, but any decision that leads to substandard accuracy will incur a high negative penalty, while power consumption itself is a constant cost.

[0040] Online Decision-Making and Collaboration: The trained model is lightweighted and deployed on the employee badge MCU. During runtime, the model outputs action commands based on its current state (e.g., low battery, stable recent accuracy), such as "turn off the UWB module and activate only Bluetooth for area positioning." This module activation status information is immediately fed back to the policy generation module in S2. The policy generation module adjusts the fusion strategy accordingly. For example, when UWB is in sleep mode, the weight of UWB data in the strategy is automatically set to zero, and the system enters a low-power fusion mode primarily based on IMU and Bluetooth. In this way, the system intelligently allocates hardware resources while ensuring that the positioning accuracy is not lower than the scenario requirements (e.g., 1 meter in safe areas, 0.5 meters in dangerous areas), minimizing overall power consumption. Real-world testing shows that this mechanism can improve the employee badge's battery life by more than 30% in typical usage scenarios.

[0041] Through the specific embodiments described above, the present invention achieves the following significant advancements: High accuracy with environmental adaptability: By sensing environmental changes in real time through the classification model and dynamically adjusting the fusion strategy, it overcomes the problem of performance drop in complex scenarios for fixed parameter algorithms.

[0042] Reliable drift suppression closed loop: The innovative "model filtering + AI learning" dual-level compensation mechanism, combined with the fusion results, performs closed-loop feedback correction on the IMU, fundamentally suppressing long-term drift and achieving stable and reliable continuous positioning.

[0043] System-level energy efficiency optimization: Through reinforcement learning-driven sensor co-activation, an intelligent balance between positioning accuracy and power consumption is achieved, meeting the urgent needs of industrial applications for long battery life.

[0044] Highly practical for engineering applications: All models (classification, LSTM, reinforcement learning) have undergone lightweight design and transfer learning training, and can run stably under limited computing resources. The solution has high feasibility and scalability.

[0045] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on these embodiments, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can still combine, add, delete, or otherwise adjust the features of the various embodiments of the present invention according to the circumstances without conflict or creative effort, thereby obtaining different technical solutions that do not fundamentally depart from the concept of the present invention. These technical solutions also fall within the scope of protection of the present invention.

Claims

1. A high-precision positioning method for work badges that integrates multi-source positioning and inertial navigation, characterized in that, Includes the following steps: S1. Data Acquisition: Real-time acquisition of observation data from at least two heterogeneous wireless positioning modules built into the work badge, as well as inertial navigation data from the inertial measurement unit; the observation data includes ranging or positioning data from various heterogeneous wireless positioning modules. S2. Strategy Generation: Based on the quality assessment of the observed data and the current motion state identified based on the inertial navigation data, an adaptive fusion strategy is generated. The fusion strategy at least defines the weight allocation of the observation data and the inertial navigation data in the subsequent fusion process; S3. Fusion Positioning: Based on the fusion strategy, the observation data and the inertial navigation data are spatiotemporally aligned and fused to output the fusion positioning result; S4. Error Compensation and Feedback: Using the fusion positioning results, the cumulative drift error of the inertial measurement unit is compensated to obtain corrected inertial navigation data, and the corrected inertial navigation data is fed back to the subsequent fusion positioning process.

2. The high-precision positioning method for work badges based on multi-source positioning and inertial navigation fusion according to claim 1, characterized in that, The compensation for the cumulative drift error of the inertial measurement unit in step S4 is implemented using a two-stage compensation model, including: First-level compensation: A filtering algorithm based on pedestrian kinematic constraints is applied to correct the raw inertial navigation data of the inertial measurement unit in real time to suppress short-term dynamic errors; Secondary compensation: The historical data sequence of the inertial measurement unit and the fused positioning result and the residual sequence generated by the independent position calculation of the inertial measurement unit are input into a pre-trained lightweight neural network model for prediction, and the correction amount for long-term cumulative drift is output. The correction amount output by the secondary compensation is used to correct the inertial navigation data after the primary compensation, and then the corrected inertial navigation data is input into the subsequent fusion positioning process.

3. The high-precision positioning method for work badges based on multi-source positioning and inertial navigation fusion according to claim 2, characterized in that, The lightweight neural network model is a long short-term memory network model, which is trained through transfer learning: first, it is pre-trained on a public or synthetic inertial navigation dataset, and then fine-tuned using measured data collected from target work badges in typical industrial scenarios.

4. The high-precision positioning method for work badges based on multi-source positioning and inertial navigation fusion according to claim 1, characterized in that, The quality assessment in step S2 includes the identification of environmental characteristics, specifically: The channel features of the observed data are extracted, and a classification model is used to determine whether the signal propagation is under non-line-of-sight conditions and / or to identify the current typical environmental scene type; the typical environmental scene type includes at least industrial indoor scene, outdoor open scene and semi-obscured transition scene; The fusion strategy dynamically adjusts the trust weights of different observation data based on the judgment results of the non-line-of-sight conditions and / or the environmental scene type, and adjusts the error correction parameters in the fusion process accordingly.

5. The high-precision positioning method for work badges based on multi-source positioning and inertial navigation fusion according to claim 1, characterized in that, Between step S1 and step S2, a sensor co-activation step is also included: A power consumption-accuracy prediction model is constructed, which takes the current fusion strategy, historical positioning accuracy sequence and battery level as input; Based on the output of the power consumption-accuracy prediction model, some of the heterogeneous wireless positioning modules are dynamically activated or put into sleep mode, and the module activation status information is fed back to step S2; when generating the fusion strategy, step S2 makes adjustments based on the module activation status information.

6. The high-precision positioning method for work badges based on multi-source positioning and inertial navigation fusion according to claim 5, characterized in that, The power consumption-accuracy prediction model is a reinforcement learning model, and its reward function is configured to: give a positive reward for periods when the positioning accuracy is higher than a preset threshold, and apply a negative reward to any activation decision that causes the positioning accuracy to fall below the preset threshold.

7. The high-precision positioning method for work badges based on multi-source positioning and inertial navigation fusion according to claim 2, characterized in that, The filtering algorithm based on pedestrian kinematic constraints is an improved Kalman filter algorithm. Its state equation introduces zero velocity detection and zero angular velocity correction, and the process noise covariance matrix is ​​adaptively adjusted according to the motion intensity identified in real time by the inertial navigation data of the inertial measurement unit.

8. The high-precision positioning method for work badges based on multi-source positioning and inertial navigation fusion according to claim 4, characterized in that, After identifying non-line-of-sight conditions, the affected observation data are corrected using a model, specifically as follows: Based on the non-line-of-sight type and confidence level output by the classification model, a corresponding distance deviation compensation model is selected from the preset error deviation library to compensate the original observation values, and the corrected observation data is then sent to step S3.

9. The high-precision positioning method for work badges based on multi-source positioning and inertial navigation fusion according to claim 1, characterized in that, The spatiotemporal alignment in step S3 specifically includes: The timestamps of the observation data and the inertial navigation data are aligned using a precision clock synchronization protocol. By using a unified coordinate system transformation matrix, the observation data and the inertial navigation data are transformed to the same coordinate system.

10. The high-precision positioning method for work badges based on multi-source positioning and inertial navigation fusion according to claim 1, characterized in that, The heterogeneous wireless positioning module is selected from any two or more combinations of ultra-wideband positioning modules, satellite positioning modules, Bluetooth positioning modules, and Wi-Fi positioning modules.

Images