A high-reliability positioning verification method based on inertial projection and context constraints, a bracelet and a device

Abstract

The application provides a high-reliability positioning verification method based on inertial projection and context constraint, a bracelet and a device, which are applied to a positioning verification bracelet including an inertial measurement module and an external positioning module. The method obtains a predicted position by performing dead reckoning on inertial measurement data, determines a dynamic verification area based on the predicted position, performs first-level verification on external positioning data, compares the positioning data passing the first-level verification with a preset spatial rule to perform second-level verification, and uses the positioning data passing the second-level verification as an observation to calibrate and correct the dead reckoning. The first-level verification, the second-level verification and the calibration correction are solidified in a system on chip in a hardware pipeline form, and the external positioning data must pass each level of verification in turn to reach an output end, so as to fundamentally eliminate the security failure caused by bypassing the verification logic by software.

Description

Technical Field

[0001] This application relates to the field of satellite navigation and inertial navigation fusion positioning technology, and in particular to a highly reliable positioning verification method and device based on inertial projection and context constraints. Background Technology

[0002] In industrial environments, personnel positioning systems typically employ a Global Navigation Satellite System (GNSS) module combined with application-layer software filtering to process positioning data. In this approach, the software layer performs single-frame threshold filtering on the satellite positioning output to eliminate abnormal data. The Inertial Measurement Unit (INS) only provides auxiliary functions and does not participate in the positioning verification process, while spatial constraints are handled offline by the backend server. However, this approach has the following shortcomings: single-frame threshold filtering lacks continuity verification capabilities based on physical motion laws, making it difficult to distinguish between legitimate motion and spurious displacements caused by signal jumps in environments with strong electromagnetic interference or multipath effects; the positioning results are disconnected from physical spatial rules, leading to logically discontinuous trajectories; and the software-implemented verification logic can be bypassed due to vulnerabilities or attacks, failing to meet the integrity requirements of safety-critical scenarios. Summary of the Invention

[0003] To address the shortcomings of existing technologies where positioning verification logic is implemented in software, leading to bypassable verification paths, lack of motion continuity judgment in single-frame threshold filtering, and discontinuity in trajectory logic due to post-processing of spatial constraints, this application provides a highly reliable positioning verification method, wristband, and device based on inertial projection and context constraints. This method, by embedding inertial projection verification, spatial constraint verification, and adaptive calibration into an unbypassable mandatory data path on-chip in a hardware pipeline, fundamentally eliminates the security failure of positioning verification being bypassed by software. Simultaneously, it ensures that external positioning data undergoes dynamic confidence region verification based on physical motion laws and legality verification based on preset spatial rules before reaching the output. Specifically, this application provides the following technical solutions: This application provides a highly reliable positioning verification method based on inertial projection and context constraints, applied to a positioning verification wristband including an inertial measurement module and an external positioning module. The method includes: Dead reckoning is performed on inertial measurement data to obtain the predicted position; Based on the predicted location, a dynamic verification area is determined. External positioning data is compared with the dynamic verification area to perform a first-level verification. Positioning data that passes the first-level verification is compared with preset spatial rules to perform a second-level verification. The dead reckoning is calibrated and corrected using the positioning data that has passed the second-level verification as an observation. The first-level verification, the second-level verification, and the calibration correction are embedded in the on-chip system in a hardware pipeline. The external positioning data must pass through each level of verification in sequence before it can reach the output.

[0004] Optionally, the dynamic verification region is a confidence region centered on the predicted location and dynamically expanded according to the calculated uncertainty. The first-level verification includes: determining the deviation metric between the external positioning data and the predicted location, and determining that the deviation metric passes when it does not exceed the boundary of the confidence region.

[0005] Optionally, the radius of the confidence region is linearly determined based on the external positioning accuracy factor and the free calculation time since the last calibration correction.

[0006] Optionally, the preset spatial rules include: mapping the coordinates of the positioning data that has passed the first-level verification to a spatial index, searching for the access attribute corresponding to the spatial index in a pre-stored spatial attribute table, and determining whether the positioning data is located in a legal access area based on the access attribute.

[0007] Optionally, the calibration correction employs a state estimation fusion algorithm, using the positioning data that has passed the second-level verification as the observation and the predicted state calculated by dead reckoning as the prior estimate, and adaptively adjusting the observation confidence based on external positioning quality indicators.

[0008] Optionally, the state estimation fusion algorithm is an extended Kalman filter, the observation is a position component, and the output of the calibration correction includes a position correction and a velocity correction.

[0009] Optionally, it also includes: when there is no valid external positioning data input, outputting the calculated position after the calibration correction, and attaching a low confidence flag.

[0010] This application also provides a positioning verification wristband, including: The inertial calculation unit is used to perform dead reckoning on inertial measurement data to obtain the predicted position; The first-level verification unit is used to determine the dynamic verification area based on the predicted location, and compare the external positioning data with the dynamic verification area to perform the first-level verification. The second-level verification unit is used to compare the positioning data that has passed the first-level verification with the preset spatial rules to perform the second-level verification. A calibration fusion unit is used to calibrate and correct the dead reckoning using the positioning data that has passed the second-level verification as an observation. The inertial calculation unit, the first-level verification unit, the second-level verification unit, and the calibration fusion unit are connected in series via an on-chip hardware pipeline to form a forced data path. The external positioning data must pass through each level of verification in sequence before reaching the output.

[0011] This application also provides a positioning verification device, including a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0012] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0013] The beneficial effects of this application are as follows: By physically solidifying the three verification stages—inertial projection verification, spatial constraint verification, and adaptive calibration—into a mandatory data path within the on-chip system in a hardware pipeline format, the entire link of external positioning data from entering the chip to reaching the output is verified at the hardware level, eliminating the structural defects in existing software implementation schemes where verification logic can be bypassed. Simultaneously, the first-level verification performs motion consistency judgment on external positioning based on the dynamic confidence region generated by inertial estimation, while the second-level verification performs spatial legality constraints on the positioning results based on preset spatial rules. This dual verification collaboratively ensures that the output trajectory simultaneously satisfies both physical motion continuity and spatial logical rationality. The high-quality data feedback from the second-level verification calibrates the cumulative drift of inertial estimation, forming a closed-loop mechanism for continuous convergence of verification accuracy. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the overall process of the positioning verification method provided in the embodiments of this application.

[0015] Figure 2 This is a schematic diagram of the inertial projection verification subprocess provided in an embodiment of this application.

[0016] Figure 3 This is a schematic diagram of the map constraint verification subprocess provided in an embodiment of this application.

[0017] Figure 4 This is a schematic diagram of the adaptive calibration fusion subprocess provided in an embodiment of this application.

[0018] Figure 5 This is a schematic diagram of the data path structure of the on-chip system hardware pipeline provided in the embodiments of this application.

[0019] Figure 6 This is a schematic diagram illustrating the relationship between the confidence radius and the calculation time, provided for an embodiment of this application.

[0020] Figure 7This is a schematic diagram showing the comparison of position errors before and after calibration, provided in an embodiment of this application.

[0021] Figure 8 This is a schematic diagram of the system state transition and output selection process provided in the embodiments of this application.

[0022] Reference numerals: U-IMU, Inertial Measurement Module; U-GNSS, Global Navigation Satellite System Interface Module; U-INS, Inertial Navigation Calculation Engine; U-PROJ, Inertial Projection Verification Gate; U-MAP, Map Constraint Verification Gate; U-KF, Adaptive Calibration Fusion Engine; U-CTRL, System Controller. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] It should be noted that the ordinal numbers such as "first" and "second" used in this application are only used to distinguish technical features of the same or similar types, and do not indicate the order of time, importance or numerical size between features, nor do they constitute any additional limitation on the technical features referred to.

[0025] In the field of positioning technology, the fusion positioning of satellite navigation systems and inertial navigation systems is the main technical approach for achieving personnel location tracking. This positioning technology is widely used in industrial environments such as ports, chemical industrial parks, large warehousing and logistics centers, underground mines, and construction sites. In these environments, personnel positioning systems need to maintain trajectory continuity and accuracy under complex conditions such as strong electromagnetic interference, multipath effects, and satellite signal blockage; their reliability directly affects the effectiveness of operational safety management. Existing positioning verification schemes are mostly based on application-layer software filtering, which faces limitations in industrial-grade safety-critical scenarios, including bypassed verification logic, lack of physical motion constraints for outlier determination, and the absence of real-time spatial semantic information. Therefore, this application provides a highly reliable positioning verification method, wristband, and device based on inertial projection and contextual constraints.

[0026] The positioning verification method provided in this embodiment is applicable to electronic devices including, but not limited to: wrist-worn positioning bracelets, positioning modules embedded in safety helmets, chest tag-type positioning labels, industrial smartwatches, portable positioning beacons, and positioning terminals integrated into personal protective equipment. These devices share the common features of a built-in inertial measurement module and an external positioning receiver module, and possess on-chip system-level data processing capabilities. This method can be deployed in industrial scenarios requiring real-time personnel positioning assurance, such as port container handling areas, chemical plant reaction unit areas, underground mine tunnels, large logistics sorting centers, and high-altitude construction platforms.

[0027] In this embodiment, the positioning verification wristband includes a System on Chip (SoC), which integrates an inertial measurement unit (U-IMU), a global navigation satellite system interface (U-GNSS), an inertial navigation calculation engine (U-INS), an inertial projection verification gate (U-PROJ), a map constraint verification gate (U-MAP), an adaptive calibration fusion engine (U-KF), and a system controller (U-CTRL).

[0028] The inertial measurement module includes a three-axis accelerometer and a three-axis gyroscope, outputting six-axis inertial measurement data at a 200Hz sampling rate. The Global Navigation Satellite System (GNSS) interface module communicates with an external GNSS receiver chip via a serial peripheral interface to receive positioning data frames. The system-on-a-chip also includes 512KB of on-chip static random access memory (SRAM) for storing a pre-loaded port geographic information digital map lookup table (the map lookup table actually occupies approximately 500KB, with the remaining space used for runtime state caching). The four functional units—U-INS, U-PROJ, U-MAP, and U-KF—are connected in series via an on-chip hardware pipeline as a forced data path, and the units are cascaded together through gated registers. The verification flag signal of the upstream unit directly controls the write enable terminal of the downstream gated register, ensuring that external positioning data must pass through each level of verification at the physical layer before reaching the output. Figure 5 As shown, this hardware pipeline data path constitutes the core architectural feature of this solution.

[0029] During the system power-on initialization phase, the system controller U-CTRL executes a four-step initialization sequence. The system controller reads the port geographic information digital map lookup table from external flash memory via a serial peripheral interface and transfers it to the on-chip static random access memory (SRAM) of the U-MAP. The system controller clears the state vector register and covariance matrix register of the U-INS. The system controller starts the U-GNSS interface and polls for the first valid positioning frame whose horizontal accuracy factor (HDOP) is not greater than a preset threshold. This threshold is typically in the range of 1.5 to 3.0; in this embodiment, it is 2.0. The system controller converts the latitude and longitude of the valid positioning frame into position coordinates in the local northeast-northeast coordinate system via planar projection, writes this into the initial position state of the U-INS, and sets the system status register to running state. For example, in a port environment, after system power-on, if the horizontal accuracy factor of the GNSS output in the first frame is 3.5, in the second frame it is 2.8, and in the third frame it is 1.6, the system completes initialization in the third frame, with a total waiting time of approximately 3 seconds. If the horizontal accuracy factor exceeds the admission threshold within 120 seconds after the system is powered on, the system controller enters the degraded initialization mode and uses the position saved before the most recent shutdown as the initial position.

[0030] like Figure 1 As shown, the positioning verification method provided in this embodiment includes the following steps: S100 performs dead reckoning on inertial measurement data to obtain the predicted position.

[0031] like Figure 1 As shown, the inertial estimation unit U-INS receives six-axis inertial measurement data output by the inertial measurement module U-IMU at a frequency of 200Hz, including three-axis acceleration. and triaxial angular velocity The inertial estimation unit performs strapdown inertial navigation calculations on the six-axis inertial measurement data to obtain the predicted position at the current moment. The strapdown inertial navigation system (SINS) calculation includes three stages: attitude update, velocity update, and position update.

[0032] S110, perform attitude update on the angular velocity data to obtain the current attitude rotation matrix.

[0033] The inertial calculation unit uses a unit quaternion. This represents the attitude of the positioning verification wristband relative to the navigation coordinate system. The inertial estimation unit integrates the three-axis angular velocities according to the quaternion differential equation to update the attitude quaternion, and its first-order update form is: in The integration step size is used in this embodiment. This corresponds to the reciprocal of the 200Hz sampling period of the inertial measurement module. The antisymmetric matrix constructed from the aforementioned three-axis angular velocities is: The inertial calculation unit performs normalization processing on the attitude quaternion after each integration. This eliminates the magnitude drift introduced by numerical integration. The inertial estimation unit generates a rotation matrix from the body coordinate system to the navigation coordinate system based on the normalized attitude quaternions. : For example, when a worker wears a wristband horizontally and walks facing east, the initial posture quaternion is: ,correspond This is the identity matrix, meaning the body coordinate system is aligned with the navigation coordinate system. If the gyroscope measures the angular velocity around the z-axis at a certain moment... After one step-size integration, the quaternion is updated to , corresponding to The heading change. In another embodiment, the attitude update can also be achieved using the rotation vector method, which... As an equivalent rotation vector, and converted into a rotation matrix increment via the Rodriguez formula, this method is equivalent in accuracy to quaternion integration under small angle conditions.

[0034] S120, the acceleration in the body coordinate system is converted into the navigation coordinate system using the rotation matrix and the velocity is obtained by integration.

[0035] The inertial calculation unit utilizes the rotation matrix. acceleration in body coordinate system Converting the acceleration to the navigation coordinate system, subtracting the gravitational component, and integrating over time yields the horizontal velocity in the navigation coordinate system: in For the rotation matrix The Line number For the column elements, the gravitational component exists only in the vertical direction; the velocity integral in the horizontal direction does not need to be subtracted from gravity. For example, when a person walks eastward at a constant speed of 1.2 m / s and wears the wristband horizontally, the accelerometer reading is... ,through The acceleration of the navigation system after conversion is After deducting gravity The horizontal acceleration is zero, and the velocity is maintained. The speed update remains unchanged. In another embodiment, the speed update can also be achieved by simplifying the double integral. In pedestrian-level applications where the speed of human activity is less than 5 m / s, the Coriolis force correction term introduced by the Earth's rotation is ignored. This simplification reduces the hardware computational complexity without significantly different positioning accuracy.

[0036] S130, the current predicted position is obtained by integrating the velocity.

[0037] The inertial calculation unit performs position integration on the velocity in the navigation coordinate system: The predicted location The output of the inertial estimation is passed to the subsequent verification stage. For example, when the initial position is... Furthermore, when a person walks eastward at a constant speed of 1.2 m / s, the predicted position is updated after 1 second (200 steps of integration). That is, it has moved 1.2m eastward. Due to the inherent accelerometer bias and gyroscope drift of the microelectromechanical system's inertial measurement module, the predicted position will accumulate errors as the free calculation time increases. For example, when the accelerometer has When the x-axis is zero biased, the position drift after 60 seconds is: This accumulated drift is eliminated by the subsequent S400 calibration correction mechanism.

[0038] S200, a dynamic verification area is determined based on the predicted location, and the external positioning data is compared with the dynamic verification area to perform a first-level verification; the positioning data that passes the first-level verification is compared with a preset spatial rule to perform a second-level verification.

[0039] like Figure 2 As shown, the first-level verification (i.e., inertial projection verification) is performed by the inertial projection verification gate U-PROJ. The input to the first-level verification includes the predicted position. The external positioning data Its horizontal accuracy factor HDOP, and the free calculation time since the last successful calibration correction. .

[0040] S210, dynamically determine the confidence radius of the dynamic verification region based on the calculated uncertainty.

[0041] The inertial projection verification gate calculates the dynamic confidence radius based on the external positioning accuracy factor and the inertial calculation cumulative uncertainty. : Basic tolerance The dynamic verification region is defined as the predicted position. Centered on the dynamic confidence radius A circular region with radius [missing information]. For example... Figure 6 As shown, the dynamic confidence radius increases linearly with the free calculation time until it reaches its upper limit. The aforementioned basic tolerance The drift extension term reflects the inherent accuracy of external positioning under current satellite geometry. This reflects the increase in cumulative position uncertainty in inertial calculations since the last calibration.

[0042] The confidence factor Typically, the value is in the range of 2 to 4; in this embodiment, it is set to 3, corresponding to the 99.7% confidence interval of a Gaussian distribution. The GNSS single-axis positioning accuracy... Typically, the range is between 1.5m and 5.0m; this embodiment uses 3.0m, corresponding to the horizontal accuracy of a typical open-environment L1 single-frequency GNSS receiver. The inertial-calculated position drift rate... Typically, the range is between 0.1 m / s and 1.0 m / s; in this embodiment, 0.5 m / s is used. This value is based on the zero-bias stability of the accelerometer in a typical consumer-grade microelectromechanical system inertial measurement module (approximately...). The positional error generated by free estimation over 60 seconds under pedestrian dynamic conditions is approximately... The calculated average position error growth rate is... The value is rounded to 0.5 m / s after taking a safety margin. The upper limit of the confidence radius is... Typically, the range is between 30m and 100m; in this embodiment, 50m is used.

[0043] For example, when the horizontal precision factor is 1.8 and the free calculation time is 5 seconds, the basic tolerance is... The drift expansion amount is The dynamic confidence radius is When the free calculation duration reaches 80 seconds, the drift expansion is: , If the value exceeds 50m, it is clipped to 50m by the upper limit of the confidence radius. In another embodiment, the dynamic verification region can also be implemented using an elliptical confidence region, with different confidence radii set along the direction of movement and perpendicular to the direction of movement to accommodate the non-isotropic calculation error distribution. In yet another embodiment, the dynamic verification region can also be implemented using a polygonal fence, aligning the shape of the confidence region with the current channel direction to achieve a more refined spatial adaptation.

[0044] S220, calculate the deviation metric between the external positioning data and the predicted location and compare it with the dynamic verification area.

[0045] The inertial projection verification gate calculates the external positioning data. With the predicted location The deviation metric between them. The deviation metric is a two-dimensional Euclidean distance: The inertial projection verification gate compares the deviation metric with the dynamic confidence radius. When the deviation metric does not exceed the dynamic confidence radius, the external positioning data is determined to have passed the first-level verification, and a first-level verification flag is output. The external positioning data is then transmitted to the downstream map constraint verification gate U-MAP via a gating register. When the deviation metric exceeds the dynamic confidence radius, a rejection is determined, and an output is generated. The external positioning data cannot physically reach the downstream module through the gate register. The inertial projection verification gate uses a squared distance form for comparison in its hardware implementation. To avoid square root calculations.

[0046] For example, when the predicted position is And the external positioning data is When, deviation from measurement If the value is less than the dynamic confidence radius of 18.7m, the judgment is passed. If the external positioning data jumps due to strong electromagnetic interference... Deviation measurement The deviation exceeds the dynamic confidence radius of 24.0m, therefore the decision is rejected, and the abrupt change data is physically isolated from the downstream verification module. In another embodiment, the deviation metric can also be implemented using Mahalanobis distance, incorporating the covariance matrix of the estimation error into the distance calculation to obtain a more accurate multidimensional probability decision boundary.

[0047] The inertial projection verification gate also maintains a continuous rejection counter. Each time the first-level check flag The counter increments automatically, and the first-level verification flag... The counter is reset to zero when the continuous rejection counter exceeds a preset continuous rejection alarm threshold. When this occurs, a continuous external positioning anomaly alarm flag is triggered. The continuous rejection alarm threshold is typically in the range of 5 to 20; in this embodiment, it is set to 10, which corresponds to at least 10 seconds of no valid external positioning input when the external positioning module operates at a 1Hz output frequency.

[0048] S300 compares the positioning data obtained through the first-level verification with preset spatial rules to perform the second-level verification.

[0049] like Figure 3As shown, the second-level verification (i.e., map context constraint verification) is performed by the map constraint verification gate U-MAP. The input to the second-level verification is the positioning data passed from the inertial projection verification gate U-PROJ through the gate register, i.e., the positioning data that has passed the first-level verification. The map constraint verification gate maps the coordinates of the positioning data to a spatial index, searches for the accessibility attribute corresponding to the spatial index in a pre-stored spatial attribute table, and determines whether the positioning data is located in a legal access area based on the accessibility attribute.

[0050] S310, map the coordinates of the positioning data that have passed the first-level verification to the spatial index.

[0051] The map constraint verification gate discretizes the continuous coordinate space under the northeast-northeast coordinate system into an equally spaced orthogonal grid index space. The gridding process is implemented through integer division with rounding down. in The grid resolution is the northeast celestial coordinate of the map origin, aligned with the coordinate origin locked during system initialization. Typically, the resolution is between 0.5m and 2.0m; in this embodiment, it is 1.0m. When the grid resolution is 1.0m, the integer division degenerates into a truncation and rounding operation, which can be completed with zero latency without a divider at the hardware level.

[0052] The map constraint verification gate calculates the access address of the on-chip static random access memory based on the grid index. The spatial attribute table adopts a row-major storage layout, and the address calculation formula is as follows: ,in This represents the number of grid cells in the eastward direction of the map. In this example, a port span of 2000m in the eastward direction corresponds to... A northward span of 1000m corresponds to The total number of grids is .

[0053] The map constraint verification gate performs boundary checks before address calculation. or or or If the location data is determined to be outside the map coverage area, output directly. And trigger the boundary crossing alarm.

[0054] For example, when the location data that passes the first level of verification is The origin of the map is At that time, the grid index is , The on-chip memory access address is In another embodiment, the coordinate mapping can also be implemented using a multi-level quadtree index to accommodate irregularly shaped port coverage areas, but this requires multiple memory accesses and the hardware implementation complexity continues to increase. In yet another embodiment, the spatial index can also be implemented using point inclusion determination of vector fences, directly comparing coordinates with polygon boundaries to determine the area to which the coordinates belong.

[0055] S320, search for the access attribute corresponding to the space index in the pre-stored space attribute table.

[0056] The spatial attribute table is a port geographic information digital map lookup table pre-stored in on-chip static random access memory. Each grid cell in the map lookup table stores a 2-bit regional attribute label. The encoding is defined as: Indicates the passage area. This indicates a restricted area. The buffer zone is defined as follows: the passageway corresponds to the walkway and work area where personnel can move freely, the restricted area corresponds to the gantry crane rails, the interior of the container yard, and the hazardous materials storage area, and the buffer zone is the transition zone between the passageway and the restricted area.

[0057] The map lookup table uses a bit-packed storage format, with each byte storing the attribute labels of four adjacent grids. The map constraint verification gate calculates the byte address based on the linear address. and byte offset After reading the corresponding byte from the on-chip static random access memory, the attribute label of the current mesh is extracted through shift and mask operations. .

[0058] The total storage capacity of the map lookup table in this embodiment is: bit bit The port geographic information digital map is provided by the port safety management department with raw vector data, which is converted into the bit-packed format according to the grid resolution by an offline rasterization tool and then burned into the external flash memory of the positioning verification wristband.

[0059] For example, for address byte address is The bit offset is The map constraint verification gate reads 1 byte of data from on-chip memory address 140538, extracts the lowest 2 bits, and obtains... That is, the passage area.

[0060] S330, determine whether the location data is located in a legal passage area based on the passage attribute.

[0061] The map constraint verification gate is based on the found region attribute labels. Execute the passage determination logic: When the attribute label is "accessible area", the location data is determined to be located in a legal accessible area, and a second-level verification flag is output. The positioning data is then transmitted to the calibration fusion engine U-KF via a gating register. When the attribute label indicates a restricted area, the positioning data is determined to be located in an illegal area, and an output is generated. The positioning data cannot reach the downstream calibration fusion engine at the physical level.

[0062] When the attribute label is a buffer, the map constraint verification gate performs further determination by combining prior trajectory information and movement direction. The map constraint verification gate maintains a prior attribute register. Record the region attributes of the most recently verified location point that passed the second-level verification. The buffer determination logic is as follows: when... If the area is a passageway or buffer zone and the current direction of movement is not towards the restricted area, it is determined to be an accidental entry into the buffer zone due to external positioning drift, and passage is permitted; when When a person is moving towards a restricted area even though the area is a passage or buffer zone, it is determined that they are approaching the restricted zone, and the movement is refused, triggering an alert; when When the area is a restricted zone (corresponding to the first frame after system initialization or an abnormal recovery scenario), entry will be denied and an alarm will be triggered. The direction of motion is obtained by normalizing the velocity vector output by the inertial calculation unit. The nearest forbidden zone normal vector pre-encoded with the buffer grid Perform a dot product operation; if the dot product value is not greater than zero, it is determined that the direction is not towards the forbidden zone. The nearest forbidden zone normal vector... During the offline rasterization stage, the unit gradient direction pointing to the nearest forbidden zone boundary is calculated for each buffer grid and quantized into an 8-direction code (3 bits). This normal vector code is stored using an auxiliary normal vector lookup table, which contains only grid entries with the attribute label "buffer" and is indexed by the linear address of the buffer grid. After the map constraint verification gate determines a grid as a buffer in the main attribute table, it queries the auxiliary normal vector lookup table using the current grid address to obtain the corresponding 3-bit direction code, then decodes it into a unit vector to perform a dot product check.

[0063] For example, when a person walks normally in a port passageway, the location data falls within a grid with the attribute of "passage area". The data passes normally. However, when the external positioning data deviates by 3 meters due to multipath effects and falls into the restricted area of ​​the gantry crane track, even though the data has passed the first level of verification (the deviation is within the confidence radius), the map constraint verification gate identifies that the location cannot be physically occupied by legitimate personnel and determines... Rejection will be issued if personnel walk along the edge of the passage and the external positioning drift occasionally touches the buffer zone, and the direction of movement is along the passage (parallel to the boundary of the restricted area). The map constraint verification gate determines that the drift is normal and allows it to pass.

[0064] In another embodiment, the preset spatial rules can also be implemented using a continuous distance field, where each grid stores the distance value to the nearest restricted area boundary, and progressive determination is made using a distance threshold. In yet another embodiment, the preset spatial rules can also be implemented using a quadtree spatial index, supporting non-uniform resolution region division to adapt to sub-regions with different precision requirements in the port.

[0065] S400, using the positioning data that has passed the second-level verification as an observation, calibrates and corrects the dead reckoning.

[0066] like Figure 4 As shown, the calibration correction is performed by the adaptive calibration fusion engine U-KF. The calibration fusion engine receives the second-level verification flag output by the map constraint verification gate. At that time, the positioning data that has passed the second level of verification As an observation, the predicted state from the dead reckoning is used as a priori estimate, and the measurement update of the state estimation fusion algorithm is performed to correct the cumulative drift of the inertial reckoning unit.

[0067] S410 calculates the fusion gain based on the predicted covariance and observation noise.

[0068] The calibration fusion engine employs a state estimation fusion algorithm to optimally weight and fuse inertial predictions with external positioning observations. The state vector of the state estimation fusion algorithm is defined as follows: It includes three components: position, velocity, and attitude deviation. The prediction step of the state vector is implicitly completed by the dead reckoning of the inertial calculation unit, and the calibration fusion engine only manages the propagation of the prediction covariance and the state correction during measurement updates.

[0069] The calibration fusion engine performs predictive covariance propagation using the inertial calculation unit's calculation cycle: in For the linearized state transition matrix, This is the process noise matrix. Each component reflects the disturbance of the state caused by the noise of the inertial measurement module within one integration step. The position process noise density... Typically, it falls within the range of 0.005 to 0.05; in this embodiment, it is taken as 0.01 m / The noise density of the velocity process. Typically, it is in the range of 0.05 to 0.5; in this embodiment, it is taken as 0.1 (m / s). .

[0070] When the calibration fusion engine receives a valid observation, the observation is a two-dimensional position. Observation matrix Only the position component of the state vector is selected. The calibration fusion engine adaptively adjusts the observation noise matrix based on external positioning quality indicators. ,in The observed noise standard deviation Compared with the basic tolerance calculated in the first level of verification Consistently, the larger the horizontal precision factor, the greater the observation noise, and the less the calibration fusion engine trusts the observation.

[0071] The calibration fusion engine calculates the innovation covariance. And thus calculate the fusion gain matrix. .because for The inverse of a symmetric positive definite matrix can be calculated directly using analytical formulas without the need for iterative algorithms.

[0072] For example, when the predicted covariance location component is Observation noise (correspond When, the new information covariance Fusion gain in position direction This gain value reflects the characteristic that the fusion result is biased towards the inertial estimation prediction when the uncertainty of inertial estimation is small and the external positioning noise is large.

[0073] S420, using the observation residuals and the fusion gain to correct the state estimation.

[0074] The calibration fusion engine calculates the observation residuals. This is the difference between the valid observation and the predicted position. The calibration fusion engine uses the fusion gain and the observation residual to calculate the state correction. The correction amount is then superimposed on the predicted state to obtain the posterior state estimate: The state correction includes position correction. Speed ​​correction and attitude deviation correction The calibration fusion engine writes the position correction and velocity correction to the status register of the inertial calculation unit via the feedback bus, and the attitude deviation correction corrects the attitude quaternion in the form of small-angle quaternion multiplication.

[0075] For example, when the predicted position is The legal observation is At that time, the observation residual was The position correction amount is , The corrected position estimate is: It is located near the predicted value and slightly closer to the observed value.

[0076] S430, update the predicted covariance and feed the correction amount back to the inertial calculation unit.

[0077] The calibration fusion engine updates the posterior covariance matrix according to the Josephus-stable form: The Josephus form guarantees that the covariance matrix remains positive definite under finite-length floating-point operations. After the measurement update is completed, the calibration fusion engine simultaneously performs two reset operations: resetting the free-evaluation timer. Resetting the counter will result in a continuous rejection count being zeroed out. Reset to zero.

[0078] For example, the location covariance after measurement update from Down to The corresponding location uncertainty decreased from 5.6m to 5.3m, demonstrating the effect of fused observations in improving estimation accuracy. For example... Figure 7 As shown, the inertial estimation position error before calibration increases cumulatively over time, and the position error is effectively compressed after each measurement update.

[0079] When there is no valid observation input (second-level verification flag) (If the value remains 0 or the external positioning module has no output), the calibration fusion engine only performs predictive covariance propagation and does not perform a measurement update step. The position component in the predictive covariance continuously increases with the free calculation time, reflecting an increase in estimation uncertainty. For example, after 10 consecutive seconds without valid observations, the position uncertainty increases from 5.3m to approximately 5.8m. When the position covariance increases to the square of the upper limit of the confidence radius... At that time, the calibration fusion engine triggers a degradation flag.

[0080] In this embodiment, the state estimation fusion algorithm employs an Extended Kalman Filter (EKF). In another embodiment, the state estimation fusion algorithm can also be implemented using an unscented Kalman Filter, employing a deterministic sampling strategy to avoid linearization approximation errors, making it suitable for highly nonlinear dynamic scenarios. In yet another embodiment, the calibration correction can also be achieved through weighted average fusion, using fixed weight coefficients to weight and sum the inertial calculated position and the external positioning position. This method has the lowest hardware complexity but cannot provide optimal fusion weights.

[0081] S500, when there is no valid external positioning data input, outputs the calculated position after calibration and correction and attaches a low confidence flag.

[0082] like Figure 8 As shown, the output selection state machine in the system controller U-CTRL comprehensively determines the current working mode based on the verification flags at each level and the free calculation time, and selects the corresponding positioning output result. The output selection state machine includes three working modes: normal mode, external positioning missing mode, and degraded mode.

[0083] S510 determines the current working mode based on the verification flag and the estimated duration.

[0084] The mode transition logic of the output selection state machine is as follows: when the continuous rejection counter... The continuous rejection alarm threshold has been reached. Or the external positioning module has no output for more than the preset timeout period. At this time, the state machine switches from normal mode to external location missing mode. When the free calculation duration... Exceeding the preset downgrade time threshold Or the location component of the predicted covariance exceeds When the external location is missing, the state machine switches from the degraded mode. When a frame of valid data that passes both the first-level and second-level checks is received, the state machine returns to the normal mode.

[0085] The timeout period Typically, the time threshold is between 3 and 10 seconds; in this embodiment, it is set to 5 seconds. The degradation time threshold is... Typically, the timeframe is between 30s and 120s. In this embodiment, we take 60s, which corresponds to the moment when the cumulative drift calculated by inertia is approximately 18m.

[0086] S520, Select the output positioning result and generate a confidence score according to the working mode.

[0087] The output selection state machine selects the output based on the current operating mode: In normal mode, the fusion position of the calibration fusion engine is output. In the external positioning missing mode, the estimated position of the inertial estimation unit is output. It also includes a low-confidence flag. In degraded mode, it outputs the last known valid position. A zero-confidence flag is attached, and an alarm signal is triggered simultaneously. The last known valid position is the most recent second-level verification flag. Location data that is latched in time.

[0088] The system controller calculates the confidence score based on the posterior covariance matrix output by the calibration fusion engine. : Location uncertainty The confidence level is 100 when the position uncertainty is zero, and 100% when the position uncertainty reaches the upper limit of the confidence radius. The confidence level is 0. The upper limit of the confidence level is constrained to 60 in the external location missing mode and forced to 0 in the degraded mode.

[0089] For example, the location covariance in normal mode hour, , External location data was lost for 30 seconds. , In downgrade mode The system controller synchronously raises the alarm pin to drive the wristband vibration motor to alert the operator.

[0090] Based on the same inventive concept, this application also provides a positioning verification wristband. The positioning verification wristband includes: The inertial estimation unit is used to perform dead reckoning on inertial measurement data to obtain the predicted position. The inertial estimation unit integrates a hardware multiply-accumulate array and a state register group, performing a three-step strapdown inertial navigation system (SINS) solution pipeline at a frequency of 200Hz, consisting of attitude update, velocity update, and position update, and outputting the predicted position and velocity at the current moment. The inertial estimation unit employs a double-buffered state register structure to ensure that each stage of the pipeline consistently reads the previous moment's state.

[0091] The first-level verification unit is used to determine a dynamic verification area based on the predicted location and compare the external positioning data with the dynamic verification area to perform the first-level verification. The first-level verification unit includes a confidence radius calculation pipeline and a distance comparison pipeline, which operate in parallel. The confidence radius calculation pipeline calculates the dynamic confidence radius based on the external positioning accuracy factor and the free calculation time, while the distance comparison pipeline calculates the squared distance between the external positioning data and the predicted location. The output flag signal of the first-level verification unit directly controls the write enable terminal of the downstream gate register.

[0092] The second-level verification unit compares the positioning data that passed the first-level verification with preset spatial rules to perform a second-level verification. The second-level verification unit includes grid index calculation logic, an on-chip static random access memory (SRAM) read interface, and attribute determination logic. The grid index calculation logic converts continuous coordinates into discrete grid addresses, and the attribute determination logic performs a passage determination based on the found region attribute labels. The output flag signal of the second-level verification unit controls the write enable pin of the gate register of the calibration fusion unit.

[0093] The calibration fusion unit is used to calibrate and correct the dead reckoning using positioning data that has passed the second-level verification as observations. The calibration fusion unit includes a matrix operation core, a covariance register set, and a feedback interface. The matrix operation core implements... Matrix analytic inversion Matrix multiplication and Symmetric matrix update. The feedback interface writes the state correction amount into the state register of the inertial calculation unit and synchronously resets the free calculation timer.

[0094] The inertial estimation unit, the first-level verification unit, the second-level verification unit, and the calibration fusion unit are connected in series via an on-chip hardware pipeline to form a mandatory data path. External positioning data must pass through each level of verification sequentially before reaching the output. The gating registers between the units are physically driven by upstream verification flags, ensuring the verification path is unbypassable at the hardware level.

[0095] In another embodiment, the inter-stage connections between units in the hardware pipeline can also be achieved through cascading FIFO buffers. The FIFO buffer is positioned between adjacent verification units. The upstream unit writes the verified positioning data and its verification flag to the write port of the FIFO buffer, and the downstream unit retrieves data from the read port of the FIFO buffer. The write enable signal of the FIFO buffer is directly driven by the upstream verification flag. When the verification flag is rejected, the write enable is invalid, and data cannot enter the FIFO buffer at the physical level. The non-empty flag of the FIFO buffer is connected to the start signal of the downstream unit; the downstream unit only starts processing when there is valid data in the buffer. This implementation, while maintaining the unbypassable verification path, accommodates the processing rate differences between upstream and downstream units through inter-stage data buffering, making it suitable for application scenarios where the computational latency of each verification unit is uneven.

[0096] Based on the same inventive concept, this application also provides a positioning verification device. The positioning verification device includes a processor and a memory. The memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described positioning verification method. The processor may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components, and can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The memory may be a read-only memory, a random access memory, flash memory, a hard disk, or a solid-state drive, or may include any combination of the above-described memories. The positioning verification device may also include a communication interface for data communication with an external positioning module and an inertial measurement module. When the processor executes the computer program stored in the memory, the positioning verification device can perform the above-mentioned positioning verification method, including: performing dead reckoning on inertial measurement data to obtain a predicted position; determining a dynamic verification area based on the predicted position; comparing external positioning data with the dynamic verification area to perform a first-level verification; comparing the positioning data that passes the first-level verification with a preset spatial rule to perform a second-level verification; and using the positioning data that passes the second-level verification as an observation to calibrate and correct the dead reckoning.

[0097] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit described above can be implemented in hardware.

[0098] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A highly reliable positioning verification method based on inertial projection and context constraints, applied to a positioning verification wristband including an inertial measurement module and an external positioning module, the method comprising: Dead reckoning is performed on inertial measurement data to obtain the predicted position; Based on the predicted location, a dynamic verification area is determined. External positioning data is compared with the dynamic verification area to perform a first-level verification. Positioning data that passes the first-level verification is compared with preset spatial rules to perform a second-level verification. The dead reckoning is calibrated and corrected using the positioning data that has passed the second-level verification as an observation. The first-level verification, the second-level verification, and the calibration correction are embedded in the on-chip system in a hardware pipeline. The external positioning data can only reach the output terminal after passing through each level of verification in sequence.

2. The method according to claim 1, characterized in that, The dynamic verification region is a confidence region centered on the predicted position and dynamically expanded according to the calculated uncertainty. The first-level verification includes: Determine the deviation metric between the external positioning data and the predicted location, and determine the pass if the deviation metric does not exceed the boundary of the confidence region.

3. The method according to claim 2, characterized in that, The radius of the confidence region is linearly determined based on the external positioning accuracy factor and the free calculation time since the last calibration correction.

4. The method according to claim 1, characterized in that, The preset space rules include: The coordinates of the location data that have passed the first-level verification are mapped to a spatial index. The access attribute corresponding to the spatial index is searched in the pre-stored spatial attribute table. Based on the access attribute, it is determined whether the location data is located in a legal access area.

5. The method according to claim 1, characterized in that, The calibration correction employs a state estimation fusion algorithm, using the positioning data that has passed the second-level verification as the observation and the predicted state calculated by dead reckoning as the prior estimate, and adaptively adjusting the observation confidence based on external positioning quality indicators.

6. The method according to claim 5, characterized in that, The state estimation fusion algorithm is an extended Kalman filter, the observation is a position component, and the output of the calibration correction includes a position correction and a velocity correction.

7. The method according to claim 1, characterized in that, Also includes: When there is no valid external positioning data input, the calculated position after calibration and correction is output, along with a low confidence flag.

8. A positioning and verification wristband, characterized in that, include: The inertial calculation unit is used to perform dead reckoning on inertial measurement data to obtain the predicted position; The first-level verification unit is used to determine the dynamic verification area based on the predicted location, and compare the external positioning data with the dynamic verification area to perform the first-level verification. The second-level verification unit is used to compare the positioning data that has passed the first-level verification with the preset spatial rules to perform the second-level verification. A calibration fusion unit is used to calibrate and correct the dead reckoning using the positioning data that has passed the second-level verification as an observation. The inertial calculation unit, the first-level verification unit, the second-level verification unit, and the calibration fusion unit are connected in series via an on-chip hardware pipeline to form a forced data path. The external positioning data can only reach the output end after passing through each level of verification in sequence.

9. A positioning verification device, characterized in that, It includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images