A collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraint

Abstract

The application provides a kind of based on multi-source sensor fusion calibration and space constraint collaborative indoor positioning method, the method comprises the following steps: S1, a plurality of position known low-power Bluetooth anchor points are deployed in target indoor area, by utilizing the real physical displacement provided by IMU Pedestrian Dead Reckoning (PDR) as constraint reference, the RSSI compensation parameter of terminal special is dynamically deduced, this method can adaptively calibrate the antenna gain and receiving sensitivity difference of different models of mobile phone, realizes the wide compatibility of positioning algorithm to all brand mobile phones. Through the collaborative solution of weighted least squares method and Kalman filtering, while eliminating low quality signal interference, the trajectory is smoothed by using time sequence filtering, the positioning jump caused by environmental noise is significantly reduced, and the positioning stability is improved. In addition, the A* path planning algorithm is introduced for space topological constraint, and the positioning result is forced to be corrected to the walkable area of the map, effectively eliminating the phenomenon that the trajectory penetrates the wall or obstacle.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and more specifically, to a collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints. Background Technology

[0002] With the continuous development of IoT and location service technologies, indoor positioning methods based on Bluetooth Low Energy (BLE) Received Signal Strength Indicator (RSSI) have been widely researched and applied due to their low deployment cost and wide compatibility. This technology typically measures the RSSI values ​​of multiple fixed BLE anchor points received by the mobile terminal, estimates the distance using a propagation model, and further employs a multilateral positioning algorithm to calculate the user's location.

[0003] However, in real-world, complex indoor environments, such methods still face several key technical challenges: First, the heterogeneity of terminal devices significantly affects ranging accuracy. Different brands and models of smart mobile terminals vary considerably in Bluetooth antenna design, casing material, RF front-end gain, and the filtering and compensation strategies of the operating system for the raw RSSI signal. This leads to a systematic deviation of 5–15 dBm or even greater in the RSSI values ​​collected at the same physical location, facing the same anchor point. Existing positioning systems generally rely on a unified empirical propagation model (such as the Log-distance path loss model), which is difficult to adapt to the ranging reference drift of various terminal types, thus introducing a non-negligible positioning error.

[0004] Secondly, relying solely on RSSI for distance estimation is susceptible to multipath effects, non-line-of-sight (NLOS) propagation, and dynamic environmental interference. This causes the results of multi-directional positioning (such as least squares or tri-directional intersection) to show significant jumps in the time series, lacking trajectory continuity and stability, which affects user experience and the reliability of subsequent applications.

[0005] Furthermore, while current mainstream temporal filtering methods (such as Kalman filtering and particle filtering) can smooth positioning trajectories to some extent, their modeling is usually limited to the kinematic state space and does not fully integrate the physical semantic information of the indoor environment. Especially in scenarios with structures such as walls, partitions, and enclosed rooms, the filtering output may produce unreasonable positioning results that violate building topology constraints, such as crossing obstacles or entering impassable areas, thus limiting the system's practicality in scenarios such as high-precision navigation and security monitoring. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints, which solves the problem of trajectory passing through walls and achieves high-precision, continuous trajectory and compatibility with different brands of mobile terminals.

[0007] This invention provides a collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints, the method comprising the following steps: S1. Deploy multiple low-power Bluetooth anchors with known locations in the target indoor area. During the system initialization phase, use a standard test terminal to collect the received signal strength index (RSSI) values ​​broadcast by each of the low-power Bluetooth anchors at multiple preset known distances. Based on the collected RSSI values ​​and their corresponding known distances, calibrate the environmental reference parameters of the area offline using a fitting method. Send the environmental reference parameters as prior constants to the user terminal. S2, the user terminal collects data output by its built-in inertial measurement unit (IMU) in real time during operation, and performs pedestrian dead reckoning (PDR) based on the data output by the IMU to calculate the relative displacement between adjacent sampling times, and uses the relative displacement as a displacement constraint. S3, based on the displacement constraint, the ranging model corresponding to the RSSI value received by the user terminal is decoupled and reconstructed. The reference parameters in the ranging model are decomposed into the environmental reference parameters and a terminal-specific compensation constant. Within the sliding time window, for the RSSI values ​​received at different times for the same low-power Bluetooth anchor point, the corresponding distance estimates are calculated using the calibrated ranging model, and a displacement difference constraint is constructed by the relative displacement obtained by PDR and the difference between the distance estimates. By minimizing the objective function composed of all displacement difference constraints, the terminal-specific compensation constant of the current user terminal is dynamically estimated and updated in real time. S4. Using the ranging model calibrated by the terminal-specific compensation constant, the RSSI values ​​of each low-power Bluetooth anchor point received by the user terminal are converted into corresponding distance estimates; and based on multiple distance estimates, the weighted least squares method is used to determine the initial observation position. S5, take the initial observation position as the measurement input, combine it with the motion state obtained by the pedestrian dead reckoning PDR, perform temporal smoothing on the positioning state through Kalman filtering, and output the prior positioning coordinates; S6. Based on a pre-built indoor 2D grid map, determine whether the current prior positioning coordinates are located within an obstacle area, or whether the line connecting them to the previous valid positioning point crosses the obstacle area. If not, output the prior positioning coordinates as the final positioning result. If yes, on the grid map, starting from the previous valid positioning point and ending at the nearest passable grid cell in the neighborhood of the prior positioning coordinates, call the A* algorithm to generate a feasible path. On this path, determine the local line segment closest to the prior positioning coordinates and calculate its vertical projection point to the line segment. Then, based on the current maximum movement distance estimated by PDR, calculate the cumulative path length from the previous valid positioning point along the feasible path to the vertical projection point. If the cumulative path length does not exceed the maximum movement distance, output the vertical projection point as the final positioning result; otherwise, extract a point equal to the maximum movement distance from the previous valid positioning point along the feasible path as the final positioning result. The 2D grid map is composed of regularly arranged grid cells, each cell marked as a passable area or an obstacle area.

[0008] In the collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints described in this invention, step S1 includes the following sub-steps: S11, using a standard test terminal, at multiple preset distance points. Stable RSSI values ​​were collected from the same low-power Bluetooth anchor point at each location to form multiple sets of observation data. ,in ; each of the aforementioned distance points satisfy The distance is 1 meter, and the plurality of preset distance points include at least three specific distances: 1 meter, 3 meters, and 10 meters. S12 uses a logarithmic distance path loss model: ; in, The Gaussian noise with zero mean is ignored during parameter fitting. Indicates the first Distance points Signal strength collected at the location; S13, based on the multiple sets of observation data The distance path loss model is fitted using the least squares method to minimize the loss function. To obtain environmental reference parameters With path loss index Its closed-form solution is: ; ; S14, the calculated environmental reference parameters With path loss index It is stored on the server and sent to the user terminal when the user terminal initiates a location request, serving as a priori constant for the subsequent ranging model.

[0009] In the collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints described in this invention, step S2 includes the following sub-steps: S21, Collect the triaxial acceleration data of the terminal at time t. Calculate the resultant acceleration amplitude , for resultant acceleration amplitude Bandpass filtering and peak detection are performed to identify the continuous acceleration peak intervals as a complete gait cycle, and the operation unit corresponding to this cycle is defined as the k-th step; For step k, extract the maximum value of the resultant acceleration amplitude within the gait period of step k. and minimum value The step size of the Kth step is calculated using the Weinberg compensation model. ,in K is the experimental calibration constant; S22, synchronously acquire triaxial angular velocity data at the same time t. and triaxial geomagnetic field data ; Based on the triaxial angular velocity data With triaxial geomagnetic field data An attitude estimation algorithm that integrates gyroscope and magnetometer is used to calculate the global heading angle corresponding to the k-th step. The attitude estimation algorithm is either the Madgwick filter algorithm or the Mahony filter algorithm. S23, based on the step size of the k-th step obtained in step S21. The global heading angle obtained in step S22 at the k-th step The relative displacement components of PDR at step k are calculated. ,in, This represents the two-dimensional planar displacement increment of the terminal relative to the previous position after completing the k-th step.

[0010] In the collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints described in this invention, step S3 includes the following steps: S31, model the Received Signal Strength Indicator (RSSI) as an environmental reference constant. Terminal-specific compensation constants Path loss term With Gaussian noise linear combination , It is ignored in the calculation process; among them, is the actual distance from the terminal to the Bluetooth Low Energy anchor point, and n is the environmental path loss factor; S32, the PDR relative displacement component obtained in step S23 at step k is... The two-dimensional scale of the terminal at time t relative to the initial position is accumulated. When the terminal is in an adjacent time period and The RSSI values ​​received from the same low-power Bluetooth anchor point are respectively and At that time, the corresponding distance estimate is calculated based on the coordinates. and ;in, The known fixed coordinates of this Bluetooth Low Energy anchor point; S33, using the sliding window length W as the time window, estimates the distance over W consecutive gait cycles. Combined with the modulus of PDR displacement increment Construct the minimum objective function Where j is the step index within the sliding window ( ), To obtain the distance obtained from the RSSI model inversion, solve this minimization problem to get the terminal-specific compensation constant estimate for the k-th step. ; S34, when the sliding window contains RSSI samples from M different Bluetooth Low Energy anchors, step S32 is executed independently for each Bluetooth Low Energy anchor h to obtain its specific compensation parameter estimate. Weights are assigned based on the stability and continuity of the RSSI sequences of each Bluetooth Low Energy anchor point; and satisfy Calculate the final compensation constant. ; S35, the estimated value of the terminal-specific compensation constant. Substituting into the RSSI ranging model, the system performs adaptive calibration of the current terminal's RSSI bias characteristics, enabling continuous optimization of positioning accuracy during the user's natural walking process.

[0011] The collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints described in this invention includes the following sub-steps in step S4: S41, based on the N low-power Bluetooth anchors observed at the current moment, eliminate those that meet the criteria. Low-power Bluetooth anchor points, among which Set a preset signal-to-noise ratio threshold; adjust the remaining low-power Bluetooth anchor points accordingly. The values ​​are sorted from high to low, and the top K low-power Bluetooth anchor points are selected for positioning calculation, where K≥3; S42, based on the terminal-specific compensation constant estimate The calibrated RSSI ranging model calculates the distance estimate for the i-th selected low-power Bluetooth anchor point. ;in ; S43, Let the target terminal location be... The known coordinates of the i-th low-power Bluetooth anchor point are: Then we have the equation of the circle: Selecting the first Bluetooth Low Energy (BLE) anchor point as the baseline BLE anchor point, subtracting the quadratic terms from the circle equations corresponding to the remaining K-1 BLE anchor points yields a system of linear equations: Write it in matrix form: ,in: , ; S44, set weights according to the RSSI values ​​of each Bluetooth Low Energy anchor point. The weight corresponding to the i-th Bluetooth Low Energy anchor point is defined as follows: ,in The distance between the i-th low-power Bluetooth anchor point calculated in step S42 and the terminal to be located is used; an M-order diagonal weight matrix W is formed based on the weights, where the diagonal elements of the diagonal weight matrix W are the weights of the corresponding low-power Bluetooth anchor points, and the off-diagonal elements are 0. The matrix form is as follows: ; S45, Solve the weighted least squares problem to obtain the initial positioning coordinates of the terminal: .

[0012] In the collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints described in this invention; the preset signal-to-noise ratio threshold in step S44 The range of values ​​is It is used to filter out unreliable RSSI measurements caused by low signal-to-noise ratio.

[0013] In the collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints described in this invention; in step S5, the initial positioning coordinates of the terminal are obtained. Then, Kalman filtering and collaborative smoothing are performed, including the following steps: S51, Construct a two-dimensional uniform motion state model and define the state vector. ; Define the state transition equation The state transition matrix ; The sampling interval is... The process noise has a covariance matrix of Q; Setting observation equations The observation matrix is , To observe the noise, whose covariance matrix is ​​R, the following iterative process is executed sequentially: predict: ; Error covariance prediction: Calculating the prior error covariance ; Kalman gain calculation: ; Status Update: ; Error covariance update: ; The final output is the smoothed prior position coordinates. .

[0014] In the collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints described in this invention, the covariance matrix Q in step S51 is a diagonal matrix.

[0015] In the collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints described in this invention, step S6 includes the following sub-steps: S61, the indoor environment is rasterized to create a two-dimensional binary topology map, where passable areas are marked as 0, and walls, obstacles, equipment areas, and restricted areas are marked as 1. The legal output position at the previous time step is defined as... The Kalman filter prior position at the current time is Perform the following legality checks: testing point Whether it is located in a passable area, i.e., the corresponding grid value is 0; Detection line segment Does it intersect with the obstacle grid marked as 1? If both checks pass, the final position is output. If the calibration fails, the calibration ends; otherwise, proceed to step S62. S62, with Starting from, with The nearest valid grid point in the neighborhood or the preset target grid endpoint Using the endpoint as the starting point, run the A* algorithm on the two-dimensional binary topology map to generate a collision-free legal path. ; path nodes All coordinates are the center coordinates of the passable grid cells; S63, in the path Search and Recent local line segment and calculate Obtain the perpendicular projection point of the line segment. ,Right now ;Will As the intermediate position after coordinate adsorption; where Represents the vector dot product. This represents the Euclidean norm.

[0016] In the collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints described in this invention, step S6 further includes the following sub-steps: S64, based on the current speed estimate With safety redundancy Calculate the current period Maximum permissible displacement within ; Calculate from Along the path To the projection point Cumulative path length ,like Then output the final position. Otherwise, from Along the path Cut forward to the length of The point is used as the final position. ;in, For path The above satisfies point.

[0017] This invention presents a collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints. By utilizing the real physical displacement provided by IMU pedestrian dead reckoning (PDR) as a constraint benchmark, it dynamically inversely calculates the terminal-specific RSSI compensation parameters. This method can adaptively calibrate the differences in antenna gain and receiver sensitivity among different mobile phone models, achieving broad compatibility of the positioning algorithm with all brands of mobile phones. Furthermore, through the collaborative solution of weighted least squares and Kalman filtering, low-quality signal interference is eliminated, and time-series filtering smooths the trajectory, significantly reducing positioning jumps caused by environmental noise and improving positioning stability. In addition, the A* path planning algorithm is introduced for spatial topology constraints, forcibly correcting the positioning results to the walkable area on the map, effectively preventing the trajectory from crossing walls or obstacles, ensuring the physical rationality of the trajectory, and greatly improving the reliability of the positioning results and user experience. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating an embodiment of the collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints of the present invention. Figure 2This is a schematic diagram of an indoor office scene with restricted areas, based on the collaborative indoor positioning method of multi-source sensor fusion calibration and spatial constraints of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0021] like Figure 1 The diagram shown is a flowchart illustrating an embodiment of a collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints according to the present invention. The method provides a collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints, the method comprising the following steps: In step S1, multiple low-power Bluetooth anchors with known locations are deployed in the target indoor area. During the system initialization phase, a standard test terminal is used to collect the received signal strength index (RSSI) values ​​broadcast by each of the low-power Bluetooth anchors at multiple preset known distances. Based on the collected RSSI values ​​and their corresponding known distances, the environmental reference parameters of the area are calibrated offline using a fitting method. The environmental reference parameters are then sent to the user terminal as prior constants. In step S2, the user terminal collects data output by its built-in inertial measurement unit (IMU) in real time during operation, and performs pedestrian dead reckoning (PDR) based on the data output by the IMU to calculate the relative displacement between adjacent sampling times, and uses the relative displacement as a displacement constraint. In step S3, based on the displacement constraint, the ranging model corresponding to the RSSI value received by the user terminal is decoupled and reconstructed. The reference parameters in the ranging model are decomposed into the environmental reference parameters and a terminal-specific compensation constant. Within the sliding time window, for the RSSI values ​​received at different times for the same low-power Bluetooth anchor point, the corresponding distance estimates are calculated using the calibrated ranging model, and a displacement difference constraint is constructed by the relative displacement obtained by PDR and the difference between the distance estimates. By minimizing the objective function composed of all displacement difference constraints, the terminal-specific compensation constant of the current user terminal is dynamically estimated and updated in real time. In step S4, the RSSI values ​​of each low-power Bluetooth anchor point received by the user terminal are converted into corresponding distance estimates using the ranging model calibrated by the terminal-specific compensation constant; and the initial observation position is determined by weighted least squares method based on multiple distance estimates. In step S5, the initial observation position is used as the measurement input. Combined with the motion state obtained by the pedestrian dead reckoning (PDR), the positioning state is smoothed by Kalman filtering to output the prior positioning coordinates. In step S6, based on a pre-constructed indoor two-dimensional grid map, which consists of regularly arranged grid cells marked as either passable or obstacle areas, it is determined whether the current prior location coordinates are located within an obstacle area, or whether the line connecting the current prior location coordinates to the previous valid location point crosses an obstacle area. If not, the prior location coordinates are output as the final location result. If so, on the grid map, starting from the previous valid location point and ending at the nearest passable grid cell in the neighborhood of the prior location coordinates, the A* algorithm is called to generate a feasible path. Along this path, a path is determined that meets the prior location coordinates. The system locates the nearest local line segment and calculates its vertical projection point to that line segment. Then, based on the current maximum movement distance estimated by the PDR, it calculates the cumulative path length from the previous valid location point along the feasible path to the vertical projection point. If the cumulative path length does not exceed the maximum movement distance, the vertical projection point is output as the final location result; otherwise, a point equal to the maximum movement distance is extracted from the previous valid location point along the feasible path and output as the final location result. The two-dimensional grid map consists of regularly arranged grid cells, each marked as a passable area or an obstacle area.

[0022] In one embodiment, step S1 includes the following sub-steps: In step S11, a standard test terminal is selected at multiple preset distance points. Stable RSSI values ​​were collected from the same low-power Bluetooth anchor point at each location to form multiple sets of observation data. ,in ; each of the aforementioned distance points satisfy The distance is 1 meter, and the plurality of preset distance points include at least three specific distances of 1 meter, 3 meters and 10 meters; it should be noted that in this application, the distance points are preferably 1m, 2m, 3m, 5m, 7m, 10m, etc.

[0023] In step S12, a logarithmic distance path loss model is used: ; in, The Gaussian noise with zero mean is ignored during parameter fitting. Indicates the first Distance points Signal strength collected at the location; In step S13, based on the multiple sets of observation data The distance path loss model is fitted using the least squares method to minimize the loss function. To obtain environmental reference parameters With path loss index Its closed-form solution is: ; ; It should be noted that in the above formula: The total number of observation samples, i.e., the number of distance-RSSI data pairs collected during the calibration phase (e.g., measured at 5 different locations). ); m is the sample index, from 1 to... , representing the i-th set of observation data; The true distance is the actual physical distance (usually in meters) from the m-th reference point to the Bluetooth Low Energy anchor point. During the calibration phase, this value is known (the location of the reference point can be measured) from a set of observation data. The signal strength received is the actual low-power Bluetooth anchor signal strength measured by the device at the m-th reference point (the unit is usually dBm, which is a negative value, such as -65dBm). The reference range signal strength (RSSI) is a key parameter in the path loss model. Its physical meaning is "received signal strength at a distance of 1 meter from the beacon" (in dBm). It is an unknown parameter that needs to be fitted, reflecting the beacon's own transmit power and the environmental reference loss. The path loss exponent is the core parameter of the path loss model, reflecting the rate attenuation of a signal propagating in an environment. Its value varies depending on the environment (e.g., free space). Indoor environment with multiple obstructions ), which are the unknown parameters that need to be fitted; The loss function (cost function) measures the sum of squared errors between the model-predicted RSSI value and the actual measured RSSI value. The goal is to find... and , making Minimize the value of (i.e., the core objective of least squares); The logarithm of distance to base 10 is the standard form of the path loss model. Signal attenuation is proportional to the logarithm of the propagation distance. The summation symbol represents summation over all... Statistical calculations were performed on the group of observation data.

[0024] In step S14, the calculated environmental baseline parameters are... With path loss index It is stored on the server and sent to the user terminal when the user terminal initiates a location request, serving as a priori constant for the subsequent ranging model.

[0025] In one embodiment, step S2 includes the following sub-steps: In step S21, the triaxial acceleration data of the acquisition terminal at time t is collected. Calculate the resultant acceleration amplitude , for resultant acceleration amplitude Bandpass filtering and peak detection are performed to identify the continuous acceleration peak intervals as a complete gait cycle, and the operation unit corresponding to this cycle is defined as the k-th step; For step k, extract the maximum value of the resultant acceleration amplitude within the gait period of step k. and minimum value The step size of the Kth step is calculated using the Weinberg compensation model. ,in K is the experimental calibration constant; In step S22, triaxial angular velocity data are simultaneously acquired at the same time t. and triaxial geomagnetic field data ; Based on the triaxial angular velocity data With triaxial geomagnetic field data An attitude estimation algorithm that integrates gyroscope and magnetometer is used to calculate the global heading angle corresponding to the k-th step. The attitude estimation algorithm is either the Madgwick filter algorithm or the Mahony filter algorithm. In step S23, based on the step size of the k-th step obtained in step S21... The global heading angle obtained in step S22 at the k-th step The relative displacement components of PDR at step k are calculated. ,in, This represents the two-dimensional planar displacement increment of the terminal relative to the previous position after completing the k-th step.

[0026] In one embodiment, step S3 includes the following steps: In step S31, the Received Signal Strength Indication (RSSI) is modeled as an environmental reference constant. Terminal-specific compensation constants Path loss term With Gaussian noise linear combination , It is ignored in the calculation process; among them, is the actual distance from the terminal to the Bluetooth Low Energy anchor point, and n is the environmental path loss factor; In step S32, the PDR relative displacement component obtained in step S23 at step k is... The two-dimensional scale of the terminal at time t relative to the initial position is accumulated. When the terminal is in an adjacent time period and The RSSI values ​​received from the same low-power Bluetooth anchor point are respectively and At that time, the corresponding distance estimate is calculated based on the coordinates. and ;in, The known fixed coordinates of this Bluetooth Low Energy anchor point; In step S33, distance estimates are generated over W consecutive gait cycles, using a sliding window length W as the time window. Combined with the modulus of PDR displacement increment Construct the minimum objective function Where j is the step index within the sliding window ( ), To obtain the distance obtained from the RSSI model inversion, solve this minimization problem to get the terminal-specific compensation constant estimate for the k-th step. ; In step S34, when the sliding window contains RSSI samples from M different Bluetooth Low Energy anchors, step S32 is executed independently for each Bluetooth Low Energy anchor h to obtain its specific compensation parameter estimate. Weights are assigned based on the stability and continuity of the RSSI sequences of each Bluetooth Low Energy anchor point; and satisfy Calculate the final compensation constant. ; In step S35, the estimated value of the terminal-specific compensation constant is... Substituting into the RSSI ranging model, the system performs adaptive calibration of the current terminal's RSSI bias characteristics, enabling continuous optimization of positioning accuracy during the user's natural walking process.

[0027] In one embodiment, step S4 includes the following sub-steps: In step S41, based on the N low-power Bluetooth anchor points observed at the current moment, those that meet the criteria are eliminated. Low-power Bluetooth anchor points, among which Set a preset signal-to-noise ratio threshold; adjust the remaining low-power Bluetooth anchor points accordingly. The values ​​are sorted from high to low, and the top K low-power Bluetooth anchor points are selected for positioning calculation, where K≥3; it should be noted that K=4~6 is preferred in this application.

[0028] In step S42, based on the estimated value of the terminal-specific compensation constant... The calibrated RSSI ranging model calculates the distance estimate for the i-th selected low-power Bluetooth anchor point. ;in ; In step S43, the target terminal location is set to... The known coordinates of the i-th low-power Bluetooth anchor point are: Then we have the equation of the circle: Selecting the first Bluetooth Low Energy (BLE) anchor point as the baseline BLE anchor point, subtracting the quadratic terms from the circle equations corresponding to the remaining K-1 BLE anchor points yields a system of linear equations: Write it in matrix form: ,in: , ; In step S44, weights are set according to the RSSI values ​​of each Bluetooth Low Energy anchor point. The weight corresponding to the i-th Bluetooth Low Energy anchor point is defined as follows: ,in The distance between the i-th low-power Bluetooth anchor point calculated in step S42 and the terminal to be located is used; an M-order diagonal weight matrix W is formed based on the weights, where the diagonal elements of the diagonal weight matrix W are the weights of the corresponding low-power Bluetooth anchor points, and the off-diagonal elements are 0. The matrix form is as follows: ; In step S45, the weighted least squares problem is solved to obtain the initial positioning coordinates of the terminal: .

[0029] In one embodiment, the preset signal-to-noise ratio threshold in step S44 The range of values ​​is It is used to filter out unreliable RSSI measurements caused by low signal-to-noise ratio.

[0030] In one embodiment, in step S5, the initial positioning coordinates of the terminal are obtained. Then, Kalman filtering and collaborative smoothing are performed, including the following steps: In step S51, a two-dimensional uniform motion state model is constructed, and the state vector is defined. ; Define the state transition equation The state transition matrix ; The sampling interval is... The process noise has a covariance matrix of Q; Setting observation equations The observation matrix is , To observe the noise, whose covariance matrix is ​​R, the following iterative process is executed sequentially: predict: ; Error covariance prediction: Calculating the prior error covariance ; Kalman gain calculation: ; Status Update: ; Error covariance update: ; The final output is the smoothed prior position coordinates. .

[0031] In one embodiment, the covariance matrix Q in step S51 is a diagonal matrix.

[0032] In one embodiment, step S6 includes the following sub-steps: In step S61, the indoor environment is rasterized to create a two-dimensional binary topology map, where passable areas are marked as 0, and walls, obstructions, equipment areas, and restricted areas are marked as 1. The legal output position at the previous time step is defined as... The Kalman filter prior position at the current time is Perform the following legality checks: testing point Whether it is located in a passable area, i.e., the corresponding grid value is 0; Detection line segment Does it intersect with the obstacle grid marked as 1? If both checks pass, the final position is output. If the calibration fails, the calibration ends; otherwise, proceed to step S62. In step S62, with Starting from, with The nearest valid grid point in the neighborhood or the preset target grid endpoint Using the endpoint as the starting point, run the A* algorithm on the two-dimensional binary topology map to generate a collision-free legal path. ; path nodes All coordinates are the center coordinates of the passable grid cells; In step S63, in the path Search and Recent local line segment and calculate Obtain the perpendicular projection point of the line segment. ,Right now ;Will As the intermediate position after coordinate adsorption; where Represents the vector dot product. This represents the Euclidean norm.

[0033] In one embodiment, step S6 further includes the following sub-steps: In step S64, based on the current speed estimate With safety redundancy Calculate the current period Maximum permissible displacement within ; Calculate from Along the path To the projection point Cumulative path length ,like Then output the final position. Otherwise, from Along the path Cut forward to the length of The point is used as the final position. ;in, For path The above satisfies point.

[0034] Specifically, the indoor positioning system used in this embodiment includes several fixed low-power Bluetooth anchor points (as known location reference points), a smart mobile terminal carried by the user (integrating BLE module, IMU inertial measurement unit, magnetometer and other sensors), and a back-end server (used for environmental modeling, parameter calibration and trajectory optimization calculation).

[0035] During the system initialization phase, a two-dimensional grid map is first constructed in the target indoor environment, dividing the space into square grid units with a side length of 0.5m. Through on-site surveys or BIM drawings, each grid unit is labeled with the attribute "0" to indicate a passable area and "1" to indicate an obstacle area (such as walls, fixed furniture, equipment areas, etc.), forming the basis of topological constraints.

[0036] First, this application utilizes the mobile terminal to be located to collect environmental data. Specifically, the mobile terminal receives signals broadcast by at least three Bluetooth Low Energy (BLE) anchor points in the vicinity via its built-in wireless communication module (such as a Bluetooth module), and extracts the MAC address and corresponding Received Signal Strength Indication (RSSI) value of each BLE anchor point. Simultaneously, it uses the mobile terminal's built-in Inertial Measurement Unit (IMU) to collect acceleration and angular velocity data.

[0037] To reduce noise interference in the original signal, this embodiment preprocesses the acquired RSSI sequence. Preferably, Gaussian filtering is used to smooth the RSSI sequence to reduce the impact of multipath effects and random noise in the environment on signal strength, thereby obtaining more stable signal observations.

[0038] Because mobile terminals of different brands and models have significant differences in hardware (such as antenna gain and receiver sensitivity), directly using a general signal propagation model would introduce huge errors. Therefore, this embodiment introduces a terminal-specific compensation constant. It employs a "sliding time window" mechanism for dynamic estimation and calibration.

[0039] The specific implementation process is as follows: Define a sliding time window, within which a terminal-specific compensation constant is assumed. Maintain a constant position. Pedestrian dead reckoning (PDR) is performed using data collected by the inertial measurement unit to calculate the mobile terminal's displacement increment within the specified window. Simultaneously, based on the current RSSI value and preset environmental reference parameters (such as path loss exponent n, reference distance),... Signal strength at Constructing terminal-specific compensation constants Received signal strength indication ranging model:

[0040] in, The Gaussian noise with zero mean is ignored during parameter fitting. Indicates the first Distance points The strong sound signal collected at the location.

[0041] In order to find the optimal solution In this embodiment, the displacement increment calculated by PDR is used as a geometric constraint to construct a minimization objective function. This objective function aims to minimize the error between the distance change calculated based on the RSSI ranging model and the actual displacement increment calculated by PDR. By solving this minimization objective function (e.g., using gradient descent or analytical methods), the specific compensation constant of the current terminal can be estimated in real time. This step enables "calibration while in motion," allowing the system to adaptively eliminate system errors caused by terminal heterogeneity.

[0042] After calibration, the real-time distance between the mobile terminal and each Bluetooth Low Energy anchor point is recalculated using the calibrated received signal strength indicator ranging model. Subsequently, the position is calculated using the weighted least squares method.

[0043] In this embodiment, the weighting strategy is dynamically adjusted based on signal strength: low-power Bluetooth anchors with higher RSSI values ​​have higher weights. This is because strong signals typically mean shorter distances and lower noise interference. In this way, the negative impact of long-distance, weak-signal low-power Bluetooth anchors on positioning results is effectively reduced, improving the accuracy of positioning calculations.

[0044] To further improve the continuity and physical rationality of the positioning trajectory, this embodiment adopts a post-processing strategy that combines extended Kalman filtering (EKF) with map matching.

[0045] The fused positioning results are input into an extended Kalman filter. State and observation equations are constructed using the motion characteristics of indoor positioning to smooth the positioning trajectory and filter out high-frequency jitter.

[0046] A two-dimensional grid topology map is introduced. The map is divided into walkable areas (grid value 1) and obstacle areas (grid value 0). After Kalman filtering update, it is determined whether the estimated location is within the walkable area.

[0047] If the location is valid, it is retained; if the location falls into an obstacle area or a non-physical long-distance jump occurs, the A* algorithm is used to search for a valid path from the valid location at the previous time step to the current observation location, and the positioning result is corrected to the nearest passable grid point on that path.

[0048] First, acceleration and angular velocity data are collected using the inertial measurement unit (IMU) built into the mobile terminal to perform pedestrian dead reckoning (PDR). PDR can provide relatively accurate displacement increments and heading changes over a short period of time, and this physical displacement is considered as a truth constraint.

[0049] Based on this, this embodiment constructs a dynamic calibration mechanism. Due to differences in hardware characteristics such as antenna gain and receiver sensitivity among different brands and models of mobile phones (i.e., terminal heterogeneity), significant deviations occur in RSSI readings at the same location. This method, within a sliding time window, assumes that the terminal-specific compensation constant remains constant, uses the displacement increment calculated by PDR as a geometric constraint, and combines the received RSSI observations from multiple Bluetooth Low Energy (BLE) anchor points to construct a minimization objective function. Through an iterative optimization algorithm, the terminal-specific RSSI compensation parameters are dynamically decoupled and derived. This step effectively eliminates the systematic errors caused by differences in mobile phone hardware, enabling the positioning model to adapt to mobile terminals of any brand.

[0050] After obtaining the calibrated terminal-specific compensation parameters, the original RSSI ranging model is corrected to obtain a more accurate distance estimate between the terminal and the Bluetooth Low Energy anchor point. Subsequently, a weighted least squares method is used to calculate the initial position. In this process, weights are dynamically assigned based on the signal-to-noise ratio or strength of each Bluetooth Low Energy anchor point signal, with anchor points having a higher signal quality receiving a larger weight in the calculation, thus effectively suppressing the influence of weak signal noise.

[0051] Specifically, this embodiment optimizes the positioning solution process by constructing a distance-based weighting mechanism, significantly improving positioning accuracy and the system's anti-interference capability. In step S44, weights are dynamically allocated based on the distance between each anchor point and the terminal, using the reciprocal of the square of the distance as the weight coefficient. This ensures that anchor points closer to the terminal and with relatively more reliable signal quality dominate the calculation, while the weights of anchor points farther away and more susceptible to noise interference are automatically reduced. This mechanism effectively suppresses the negative impact of long-distance measurement errors on the results. Subsequently, in step S45, the weighted least squares problem is solved using the constructed M-order diagonal weight matrix W, which mathematically yields the optimal linear unbiased estimate. Compared to traditional equal-weight algorithms, this scheme can more rationally utilize observation data, outputting stable and high-precision initial positioning coordinates even in complex signal environments. Furthermore, the algorithm has low complexity and is easy to deploy in real-time in practical systems.

[0052] Furthermore, to eliminate jitter in the solution results, a Kalman filter is introduced to smooth the position sequence. The Kalman filter uses the state of the previous time step to predict the current position and updates it by combining the observations of weighted least squares method, outputting an initial smooth trajectory. This process effectively filters out positioning jumps caused by instantaneous signal fluctuations.

[0053] Despite the above processing, the positioning trajectory may still occasionally "pass through walls" or enter non-travelable areas (such as dead corners of rooms or inside walls) due to signal multipath effects or environmental interference. To address this, this embodiment introduces spatial topology constraints based on a two-dimensional grid map.

[0054] Specifically, the indoor environment is constructed as a binary grid map, marking walkable areas and obstacle areas. Using the A* path planning algorithm, with the initial smooth trajectory points output by the Kalman filter as a reference, the optimal path is searched within the walkable area. The A* algorithm, through a heuristic search strategy, forces the localization points to "snap up" or "correct" to a legal path that conforms to the physical environment.

[0055] Through this mechanism, this embodiment not only corrects the accumulated error, but more importantly, ensures the physical rationality of the output trajectory, and completely solves the problem of the trajectory crossing obstacles or appearing in physically unreachable locations.

[0056] In summary, this embodiment uses PDR displacement constraints to deduce terminal compensation parameters, combines weighted least squares and Kalman filtering for collaborative solution, and supplements this with spatial topology constraints from the A* algorithm to form a complete indoor positioning closed-loop system. This system not only significantly improves positioning accuracy but also has broad terminal compatibility, making it suitable for various complex indoor scenarios.

[0057] like Figure 2 As shown, to verify the effectiveness and practicality of the present invention, a mixed-use area on the first floor of an office building was selected as the test scenario for on-site deployment and operation. The area has a total area of ​​24m × 16m, and its spatial layout includes a main corridor extending 24m along the x-axis and 2.2m wide; three offices measuring 4m × 5m are located on the left; two meeting rooms measuring 5m × 6m are located on the right; and a 6m × 4m equipment room is located in the center, defined as a non-accessible area. The entire area was digitally modeled using a 0.2m × 0.2m grid, divided into 9,600 grid units. Based on the on-site survey results, obstacle areas such as walls and equipment rooms were marked as "1", while other accessible areas were marked as "0", forming a spatial constraint map for trajectory correction.

[0058] In this scenario, a total of 10 BLE 5.0 low-power Bluetooth anchors (model: nRF52840DK) were deployed, and all low-power Bluetooth anchors were uniformly configured as follows: With a transmit power of 4dBm and a broadcast interval of 300ms, the Bluetooth Low Energy (BLE) anchors are installed at a height of 2.7m to reduce interference from ground-based metal objects. The BLE anchor locations are carefully planned: B1 to B3 are deployed at the entrances to rooms A and B, and in the center of the adjacent corridor, respectively; B4 and B5 are near the entrances to the equipment room and conference room 2; B6 and B7 cover the area of ​​conference room 1; and B8 to B10 are evenly distributed in open areas. This deployment strategy ensures that any test point can receive signals from at least four valid BLE anchors, meeting the geometric observability requirements for high-precision positioning.

[0059] The coordinates of Bluetooth Low Energy anchor points are shown in the table below:

[0060] Before system operation, offline calibration of environmental baseline parameters is performed. Using a standard test terminal, RSSI data are collected at six typical distance points of 1m, 2m, 3m, 5m, 8m, and 10m. Data is sampled for 30 seconds at each point and the average value is calculated. Based on the logarithmic distance path loss model, the environmental baseline constant is obtained by fitting using the least squares method. and path loss index Meanwhile, to accommodate hardware differences between different brand terminals, the system introduces a dynamic compensation mechanism. Using eight gait cycles as a sliding window, when the number of effective low-power Bluetooth anchor points within the window is no less than three and the RSSI variance is less than 6 dBm², the terminal-specific compensation constant estimate is triggered. Online updates.

[0061] During the positioning process, the user walks within the area using a smart terminal that supports BLE and IMU (such as iPhone 13, Huawei Mate 50, etc.). The system collects IMU data at a frequency of 50Hz and scans BLE signals at a frequency of 2Hz. The PDR (Pedestrian Dead Retrieval) module calculates the relative displacement increment in real time based on the Weinberg step size model (coefficient K=0.43) and the Madgwick heading fusion algorithm (parameter β=0.08). When the user moves from the center of the corridor (approximately coordinates (7.5, 12.0)) towards room B, the system first initializes the terminal compensation using RSSI observations during the initial stationary phase (measured value approximately +1.8dB), and then enters continuous tracking mode.

[0062] In this process, if only the traditional RSSI trilateration method is relied upon, the signal of B2 will be significantly attenuated due to the obstruction of the west wall of room B, while the signal of B3 may produce non-line-of-sight (NLOS) error due to near-field multipath reflection, causing the calculated position to jump to an unreasonable area. In contrast, this invention effectively avoids the above problems through a multi-source information fusion and spatial constraint coordination mechanism: the weighted least squares (WLS) module combines the calibrated RSSI distance estimate and PDR displacement information to output the preliminary observation position; the extended Kalman filter (EKF) further smooths the result to generate the prior positioning point. Finally, the map correction module... Conduct a legality check. Once discovered... If the location is in an obstacle area, or if the line connecting it to the previous legal location point passes through a wall, the system immediately starts the A* path planning algorithm to search for the shortest collision-free path in the passable area. Based on the maximum movement distance estimated by PDR, the system performs vertical projection or length truncation on the path and finally outputs a physically feasible location result that meets motion constraints.

[0063] After a 30-minute continuous walking test covering approximately 180 meters, the average positioning error of this embodiment was only 0.47 meters (95% confidence interval [0.32m, 0.68m]), with no trajectory jumps or "wall-penetrating" events occurring throughout the test. In contrast, the traditional RSSI+least squares method exhibited 7 significant positional jumps and 3 wall-penetrating errors in the same test. More importantly, this method achieved an average positioning error of less than 0.6 meters on four different brand terminals: iPhone 13, Huawei Mate 50, Xiaomi 13, and Samsung S22, fully validating its excellent terminal compatibility and engineering practical value.

[0064] This embodiment fully demonstrates how the present invention achieves high-precision and continuous trajectory seamless positioning in complex indoor environments through the deep fusion of BLE signals, inertial sensing, and digital maps, combined with dynamic terminal compensation and physical space constraint mechanisms.

[0065] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0066] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0067] Therefore, the above description is only a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints, characterized in that, The method includes the following steps: S1. Deploy multiple low-power Bluetooth anchors with known locations in the target indoor area. During the system initialization phase, use a standard test terminal to collect the received signal strength index (RSSI) values ​​broadcast by each of the low-power Bluetooth anchors at multiple preset known distances. Based on the collected RSSI values ​​and their corresponding known distances, calibrate the environmental reference parameters of the area offline using a fitting method. Send the environmental reference parameters as prior constants to the user terminal. S2, the user terminal collects data output by its built-in inertial measurement unit (IMU) in real time during operation, and performs pedestrian dead reckoning (PDR) based on the data output by the IMU to calculate the relative displacement between adjacent sampling times, and uses the relative displacement as a displacement constraint. S3, based on the displacement constraint, the ranging model corresponding to the RSSI value received by the user terminal is decoupled and reconstructed. The reference parameters in the ranging model are decomposed into the environmental reference parameters and a terminal-specific compensation constant. Within the sliding time window, for the RSSI values ​​received at different times for the same low-power Bluetooth anchor point, the corresponding distance estimates are calculated using the calibrated ranging model, and a displacement difference constraint is constructed by the relative displacement obtained by PDR and the difference between the distance estimates. By minimizing the objective function composed of all displacement difference constraints, the terminal-specific compensation constant of the current user terminal is dynamically estimated and updated in real time. S4. Using the ranging model calibrated by the terminal-specific compensation constant, the RSSI values ​​of each low-power Bluetooth anchor point received by the user terminal are converted into corresponding distance estimates; and based on multiple distance estimates, the weighted least squares method is used to determine the initial observation position. S5, take the initial observation position as the measurement input, combine it with the motion state obtained by the pedestrian dead reckoning PDR, perform temporal smoothing on the positioning state through Kalman filtering, and output the prior positioning coordinates; S6. Based on a pre-built indoor 2D grid map, determine whether the current prior positioning coordinates are located within an obstacle area, or whether the line connecting them to the previous valid positioning point crosses the obstacle area. If not, output the prior positioning coordinates as the final positioning result. If yes, on the grid map, starting from the previous valid positioning point and ending at the nearest passable grid cell in the neighborhood of the prior positioning coordinates, call the A* algorithm to generate a feasible path. On this path, determine the local line segment closest to the prior positioning coordinates and calculate its vertical projection point to the line segment. Then, based on the current maximum movement distance estimated by PDR, calculate the cumulative path length from the previous valid positioning point along the feasible path to the vertical projection point. If the cumulative path length does not exceed the maximum movement distance, output the vertical projection point as the final positioning result; otherwise, extract a point equal to the maximum movement distance from the previous valid positioning point along the feasible path as the final positioning result. The 2D grid map is composed of regularly arranged grid cells, each cell marked as a passable area or an obstacle area.

2. The collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints according to claim 1, characterized in that, Step S1 includes the following sub-steps: S11, using a standard test terminal, at multiple preset distance points. Stable RSSI values ​​were collected from the same low-power Bluetooth anchor point at each location to form multiple sets of observation data. ,in ; each of the aforementioned distance points satisfy The distance is 1 meter, and the plurality of preset distance points include at least three specific distances: 1 meter, 3 meters, and 10 meters. S12 uses a logarithmic distance path loss model: ; in, The Gaussian noise with zero mean is ignored during parameter fitting. Indicates the first Distance points Signal strength collected at the location; S13, based on the multiple sets of observation data The distance path loss model is fitted using the least squares method to minimize the loss function. To obtain environmental reference parameters With path loss index Its closed-form solution is: ； ； S14, the calculated environmental reference parameters With path loss index It is stored on the server and sent to the user terminal when the user terminal initiates a location request, serving as a priori constant for the subsequent ranging model.

3. The collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints according to claim 2, characterized in that, Step S2 includes the following sub-steps: S21, Collect the triaxial acceleration data of the terminal at time t. Calculate the resultant acceleration amplitude , for resultant acceleration amplitude Bandpass filtering and peak detection are performed to identify the continuous acceleration peak intervals as a complete gait cycle, and the operation unit corresponding to this cycle is defined as the k-th step; For step k, extract the maximum value of the resultant acceleration amplitude within the gait period of step k. and minimum value The step size of the Kth step is calculated using the Weinberg compensation model. ,in K is the experimental calibration constant; S22, synchronously acquire triaxial angular velocity data at the same time t. and triaxial geomagnetic field data ; Based on the triaxial angular velocity data With triaxial geomagnetic field data An attitude estimation algorithm that integrates gyroscope and magnetometer is used to calculate the global heading angle corresponding to the k-th step. The attitude estimation algorithm is either the Madgwick filter algorithm or the Mahony filter algorithm. S23, based on the step size of the k-th step obtained in step S21. The global heading angle obtained in step S22 at the k-th step The relative displacement components of PDR at step k are calculated. ,in, This represents the two-dimensional planar displacement increment of the terminal relative to the previous position after completing the k-th step.

4. The collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints according to claim 3, characterized in that, Step S3 includes the following steps: S31, model the Received Signal Strength Indicator (RSSI) as an environmental reference constant. Terminal-specific compensation constants Path loss term With Gaussian noise linear combination , It is ignored in the calculation process; among which, is the actual distance from the terminal to the Bluetooth Low Energy anchor point, and n is the environmental path loss factor; S32, the PDR relative displacement component obtained in step S23 at step k is... The two-dimensional scale of the terminal at time t relative to the initial position is accumulated. When the terminal is in an adjacent time period and The RSSI values ​​received from the same low-power Bluetooth anchor point are respectively and At that time, the corresponding distance estimate is calculated based on the coordinates. and ;in, The known fixed coordinates of this Bluetooth Low Energy anchor point; S33, using the sliding window length W as the time window, estimates the distance over W consecutive gait cycles. Combined with the modulus of PDR displacement increment Construct the minimum objective function Where j is the step index within the sliding window ( ), To obtain the distance obtained from the RSSI model inversion, solve this minimization problem to get the terminal-specific compensation constant estimate for the k-th step. ; S34, when the sliding window contains RSSI samples from M different Bluetooth Low Energy anchors, step S32 is executed independently for each Bluetooth Low Energy anchor h to obtain its specific compensation parameter estimate. Weights are assigned based on the stability and continuity of the RSSI sequences of each Bluetooth Low Energy anchor point; and satisfy Calculate the final compensation constant. ; S35, the estimated value of the terminal-specific compensation constant. Substituting into the RSSI ranging model, the system performs adaptive calibration of the current terminal's RSSI bias characteristics, enabling continuous optimization of positioning accuracy during the user's natural walking process.

5. The collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints according to claim 4, characterized in that, Step S4 includes the following sub-steps: S41, based on the N low-power Bluetooth anchors observed at the current moment, eliminate those that meet the criteria. Low-power Bluetooth anchor points, among which The preset signal-to-noise ratio threshold; Press the remaining low-power Bluetooth anchor points The values ​​are sorted from high to low, and the top K low-power Bluetooth anchor points are selected for positioning calculation, where K≥3; S42, based on the terminal-specific compensation constant estimate The calibrated RSSI ranging model calculates the distance estimate for the i-th selected low-power Bluetooth anchor point. ;in ; S43, Let the target terminal location be... The known coordinates of the i-th low-power Bluetooth anchor point are: Then we have the equation of the circle: Selecting the first Bluetooth Low Energy (BLE) anchor point as the baseline BLE anchor point, subtracting the quadratic terms from the circle equations corresponding to the remaining K-1 BLE anchor points yields a system of linear equations: Write it in matrix form: ,in: , ; S44, set weights according to the RSSI values ​​of each Bluetooth Low Energy anchor point. The weight corresponding to the i-th Bluetooth Low Energy anchor point is defined as follows: ,in The distance between the i-th low-power Bluetooth anchor point calculated in step S42 and the terminal to be located is used; an M-order diagonal weight matrix W is formed based on the weights, where the diagonal elements of the diagonal weight matrix W are the weights of the corresponding low-power Bluetooth anchor points, and the off-diagonal elements are 0. The matrix form is as follows: ; S45, Solve the weighted least squares problem to obtain the initial positioning coordinates of the terminal: .

6. The collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints according to claim 5, characterized in that, The preset signal-to-noise ratio threshold in step S44 The range of values ​​is It is used to filter out unreliable RSSI measurements caused by low signal-to-noise ratio.

7. The collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints according to claim 6, characterized in that, In step S5, the initial positioning coordinates of the terminal are obtained. Then, Kalman filtering and collaborative smoothing are performed, including the following steps: S51, Construct a two-dimensional uniform motion state model and define the state vector. ; Define the state transition equation The state transition matrix ； The sampling interval is... The process noise has a covariance matrix of Q; Setting observation equations The observation matrix is , To observe the noise, whose covariance matrix is ​​R, the following iterative process is executed sequentially: predict: ; Error covariance prediction: Calculating the prior error covariance ; Kalman gain calculation: ; Status Update: ; Error covariance update: ; The final output is the smoothed prior position coordinates. .

8. The collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints according to claim 7, characterized in that, In step S51, the covariance matrix Q is a diagonal matrix.

9. The collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints according to claim 1, characterized in that, Step S6 includes the following sub-steps: S61, the indoor environment is rasterized to create a two-dimensional binary topology map, where passable areas are marked as 0, and walls, obstacles, equipment areas, and restricted areas are marked as 1. The legal output position at the previous time step is defined as... The Kalman filter prior position at the current time is ; Perform the following legality checks: testing point Whether it is located in a passable area, i.e., the corresponding grid value is 0; Detection line segment Does it intersect with the obstacle grid marked as 1? If both checks pass, the final position is output. If the calibration fails, the calibration ends; otherwise, proceed to step S62. S62, with Starting from, with The nearest valid grid point in the neighborhood or the preset target grid endpoint Using the endpoint as the starting point, run the A* algorithm on the two-dimensional binary topology map to generate a collision-free legal path. ; path nodes All coordinates are the center coordinates of the passable grid cells; S63, in the path Search and Recent local line segment and calculate Obtain the perpendicular projection point of the line segment. ,Right now ;Will As the intermediate position after coordinate adsorption; where Represents the vector dot product. This represents the Euclidean norm.

10. The collaborative indoor positioning method based on multi-source sensor fusion calibration and spatial constraints according to claim 9, characterized in that, Step S6 further includes the following sub-steps: S64, based on the current speed estimate With safety redundancy Calculate the current period Maximum permissible displacement within ; Calculate from Along the path To the projection point Cumulative path length ,like Then output the final position. Otherwise, from Along the path Cut forward to the length of The point is used as the final position. ;in, For path The above satisfies point.

Images