Beidou and UWB dual-mode cooperative rail transit positioning method

Abstract

This invention discloses a dual-mode (BeiDou and UWB) collaborative positioning method for rail transit, specifically relating to the field of fusion positioning technology. The method includes the following steps: based on the spatial coverage status of the two types of signals along the train path, a cross-modal distortion discrimination array is constructed to identify trajectory abrupt change regions and dynamically adjust the signal fusion weights; abrupt change points are corrected using a constrained backtracking trajectory fitting algorithm, and pseudo-trajectory points are eliminated by combining the spatial topology of the positioning base station; time-series analysis is performed on the changes in signal fusion weights, and if frequent signal jump trends exist, the relevant fusion weights are lowered to the minimum tolerance threshold to improve trajectory continuity and positioning robustness. This invention achieves accurate identification of trajectory abrupt change regions by constructing a cross-modal distortion discrimination array, effectively corrects abnormal trajectory points by combining dynamic adjustment of fusion weights and a constrained backtracking trajectory fitting algorithm, and suppresses signal jump trends through a time-series analysis mechanism, thereby improving the stability of trajectory fusion and positioning accuracy.

Description

Technical Field

[0001] This invention relates to the field of integrated positioning technology, and more specifically, to a rail transit positioning method that utilizes both BeiDou and UWB dual-mode coordination. Background Technology

[0002] In the process of rail transit train operation control and route management, the continuity, accuracy, and robustness of trajectory positioning directly affect dispatching safety and train operation efficiency. With the gradual integration of BeiDou navigation and ultra-wideband (UWB) positioning technologies, rail transit systems are moving towards higher-precision positioning capabilities. Especially in complex terrain environments (such as tunnels, elevated sections, and mixed platform areas), dual-mode positioning collaboration has become the mainstream direction for improving system robustness. However, in actual deployment, traditional signal fusion mechanisms face various challenges, especially in scenarios involving dynamic signal switching and route changes, where errors in weight determination and path identification can easily occur, leading to serious consequences.

[0003] On the one hand, in the overlapping section of the train's path from the elevated section to the ground area and then into the tunnel area, the BeiDou signal, although in a weakening phase, has not yet completely attenuated. Meanwhile, the ultra-wideband signal, due to the recent connection of the base station and its initial instability, can lead to a situation where the fusion algorithm fails to accurately perceive this transitional state and incorrectly maintains the BeiDou signal as the dominant signal. This results in trajectory points exhibiting high-speed "segment jumps" that do not match the actual low-speed operation, forming a typical "logical jump point" pseudo-trajectory. Such jump trajectories pose significant risks in actual scheduling, potentially causing the intelligent control system to misjudge the train's current position and speed, issuing control commands prematurely or delayed. This can disrupt normal station entry rhythms, cause false speed limit triggering, and even, on some lines employing automatic platform screen door linkage mechanisms, lead to serious safety hazards such as premature or misaligned opening of platform doors.

[0004] On the other hand, during the train's arrival at a station, if the ultra-wideband positioning reference base station is located at the edge of the platform and is affected by non-structural factors such as interference sources carried by passengers in a short period of time, it is prone to instantaneous abnormal coordinate feedback. At the same time, the BeiDou signal may be significantly attenuated or even interrupted due to the obstruction of the train structure. Under this dual interference, traditional fusion algorithms are prone to inferring trends based on historical stable trajectories, mistakenly judging the current abnormal signal as a continuation of the inertial trend, thus outputting a "ghost path" that deviates from the true trajectory. Such false trajectories may not only cause the rail transit management system to incorrectly determine that the train has not yet entered the station or has deviated from the set path, but may also lead to the false triggering of emergency stopping strategies. Once multiple trains are continuously affected by similar interference, the rail transit control system may enter a protective deadlock state, requiring manual intervention to resolve, and in severe cases, it may cause the entire line's scheduling to be interrupted. In addition, "ghost trajectories" may also cause train-to-ground signal mismatch, interfere with the gating mechanism, and increase passenger safety risks. Therefore, this invention proposes a rail transit positioning method that uses BeiDou and UWB dual-mode collaboration to solve the above problems. Summary of the Invention

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A dual-mode (BeiDou and UWB) collaborative positioning method for rail transit includes the following steps:

[0007] Based on the spatial coverage status of BeiDou positioning signals and ultra-wideband positioning signals in the train operation path, the signal stability value, historical drift trend parameters and trajectory change amplitude are obtained, and a cross-modal distortion discrimination array is generated.

[0008] The cross-modal distortion discrimination array identifies whether the current position contains a trajectory change region. If the identification result is that there is a trajectory change, the trajectory curvature change and velocity vector difference between the current trajectory point and three consecutive historical trajectory points are calculated. Based on the calculation results, the fusion weight ratio of Beidou positioning signal and ultra-wideband positioning signal is adjusted to suppress the amplification of positioning error caused by trajectory change.

[0009] After the fusion weight adjustment is completed, the trajectory mutation point is corrected by the constraint backtracking trajectory fitting algorithm. The offset value of the current trajectory point is calculated by the historical trajectory fitting curve. The spatial topology of the ultra-wideband positioning base station is combined to determine whether the trajectory point is located at the pseudo trajectory source location. If the offset value exceeds the error threshold, or the trajectory point is located at the pseudo trajectory source location, the trajectory point is removed from the fusion trajectory sequence.

[0010] Perform time series analysis on the changes in signal fusion weights over n consecutive seconds during trajectory fusion. If the analysis results show a frequent trend of signal channel jumps, then lower the corresponding fusion weights to the minimum tolerance threshold.

[0011] In a preferred embodiment, the steps of constructing a cross-modal distortion discrimination array include:

[0012] Based on the geographical segment type of the train's current location in the running path, the spatial coverage status, signal strength fluctuation range and historical signal obstruction records of BeiDou positioning signal and ultra-wideband positioning signal in the segment are extracted to form a multi-source spatial sensing dataset.

[0013] Based on the multi-source spatial sensing dataset, the signal-to-noise ratio stability value, historical signal drift trend slope, and trajectory change acceleration curve of the BeiDou signal and the ultra-wideband signal at this location are calculated respectively, and the corresponding signal state vector is constructed.

[0014] The signal state vector is mapped to a unified spatiotemporal grid coordinate system, and fused according to the correlation of signal changes and the risk level of trajectory continuity to generate a cross-modal trajectory consistency distortion discrimination array.

[0015] In a preferred embodiment, fusing signals based on correlation with signal variation and risk level of trajectory continuity refers to:

[0016] Based on the signal state vector in a unified spatiotemporal grid coordinate system, the stability difference, drift direction consistency and curvature change rate between BeiDou positioning signal and ultra-wideband positioning signal at adjacent trajectory points are calculated, and a signal variation correlation factor matrix is ​​generated. Each matrix unit corresponds to a grid region, which is used to quantify the spatiotemporal variation characteristics of the grid region.

[0017] The change amplitude of each unit in the signal change correlation factor matrix is ​​compared with the preset continuity threshold range. If the change amplitude is in the corresponding low, medium or high range, the corresponding low risk level label, medium risk level label or high risk level label is assigned respectively, forming a trajectory continuity risk level label group corresponding to the grid area.

[0018] The matrix units of the signal variation correlation factor matrix and the trajectory continuity risk level label group are linked and integrated according to the spatial correspondence of the grid area. The matrix units with medium risk level labels or high risk level labels are recorded as distortion-sensitive units, and a cross-modal distortion discrimination array is generated accordingly.

[0019] In a preferred embodiment, the step of identifying whether the current location contains a trajectory abrupt change region includes:

[0020] Extract the grid cell corresponding to the current train position from the cross-modal distortion discrimination array, and read the label level of the grid cell in the trajectory continuity risk level label group;

[0021] The label level of the grid cell is compared with the label levels of its eight neighboring grid cells. If there are three or more neighboring cells with high risk levels, the current grid position is determined as the central candidate area of ​​the trajectory change region.

[0022] Signal trend consistency detection is performed on the spatial boundary of the central candidate area. If the detection result shows that there is a sign flip or amplitude change between the current and adjacent time series of the signal stability difference value, the current position is finally confirmed to contain a trajectory change region.

[0023] In a preferred embodiment, when the identification result indicates the presence of a trajectory abrupt change, the following operations are performed:

[0024] Extract the position data of the current trajectory point and its three previous historical trajectory points in a unified spatiotemporal grid coordinate system, calculate the curvature change and velocity vector difference between each pair of adjacent trajectory points, and form a trajectory perturbation feature group based on the change trend between the four points.

[0025] The maximum curvature change and the maximum velocity vector difference in the trajectory disturbance feature group are normalized to obtain the curvature disturbance index and the velocity disturbance index. Based on the cumulative change frequency of the two disturbance indices in the past three trajectory points, a change level score is assigned to each. The two scores are weighted and superimposed according to a preset ratio coefficient to generate a fusion offset coefficient. By comparing the difference between the current fusion offset coefficient and the standard reference value, the fusion weight gain that the current trajectory point should adjust towards the BeiDou positioning signal or the ultra-wideband positioning signal is determined, and the fusion weight ratio is updated.

[0026] In a preferred embodiment, the fusion weight update of the current trajectory point is performed according to the following process:

[0027] If the fusion offset index is less than the preset first threshold, the existing fusion weight ratio remains unchanged;

[0028] If the fusion offset index is between the preset first threshold and the preset second threshold, the current fusion weight ratio will be linearly offset towards the dominant signal direction by a preset ratio constant.

[0029] If the fusion offset index is greater than the preset second threshold, the fusion weight gain is calculated according to the following nonlinear formula: subtract the preset second threshold from the current fusion offset index, multiply by the set nonlinear adjustment coefficient, and use it as the fusion weight gain value of the current trajectory point; add this weight gain value to the current weight value of the current dominant signal, and normalize it again by summation to generate the updated fusion weight ratio.

[0030] In a preferred embodiment, correcting trajectory abrupt change points using a constraint backtracking trajectory fitting algorithm includes the following steps:

[0031] Extract the fused positioning data of the current trajectory point, including the trajectory mutation point, and its four consecutive historical trajectory points, construct the trajectory evolution window, and mark the mutation point location as the backtracking starting point;

[0032] The velocity change magnitude, spatial curvature continuity, and signal fusion confidence trend are calculated in chronological order for historical trajectory points in the trajectory evolution window, and fitting constraint boundaries are constructed based on these, including trajectory continuity preservation boundaries, fusion signal trend preservation boundaries, and maximum offset suppression boundaries.

[0033] Under the premise of satisfying all fitting constraints, the coordinates of the abrupt change point position are adjusted for the first time by using the acceleration change trend before and after the trajectory. If the correction result that meets the offset tolerance cannot be generated under all constraints, the compensatory backtracking correction operation is performed, that is, the local trajectory segment is reconstructed by inserting intermediate transition points, and the reconstruction result is written into the fused trajectory sequence.

[0034] In a preferred embodiment, determining whether a trajectory point is located at the pseudo-trajectory source location based on the spatial topology of the ultra-wideband positioning base station means:

[0035] Obtain all the ultra-wideband positioning base station numbers and their fixed spatial coordinate information used within the preset time window of the current trajectory point, calculate the Euclidean spatial distance from the current trajectory point to each base station, and draw connecting edges to each base station with the current point as the center node to form a polygonal star topology graph centered on the trajectory point.

[0036] Using the normal coverage radius of each base station as the boundary, multiple circular coverage areas are drawn around the central trajectory point, and the overlapping areas of each coverage area are marked as cross buffer zones; if the current trajectory point is in the cross area of ​​three or more coverage areas, the trajectory point is determined to be located at the pseudo trajectory source location.

[0037] In a preferred embodiment, performing time series analysis includes the following steps:

[0038] Extract the fusion weight sequence of the current trajectory point and several trajectory points before and after it, and calculate the number of switching times of the dominant signal and the weight fluctuation amplitude of the Beidou positioning signal and the ultra-wideband positioning signal in the sequence respectively.

[0039] If any signal channel experiences three or more dominant signal switchings within the time period, or if the weight fluctuation exceeds the set stability threshold, the fusion weight of that signal channel will be reduced to the minimum tolerance threshold.

[0040] If both the BeiDou positioning signal and the ultra-wideband positioning signal meet any of the following conditions within the same continuous time period: three or more dominant signal switchings occur, or the weight fluctuation amplitude simultaneously exceeds the set stability threshold, then the backup trajectory compensation strategy will be activated.

[0041] The technical effects and advantages of this invention are as follows:

[0042] This invention constructs a cross-modal distortion discrimination array to accurately identify spatiotemporal consistency anomalies during the fusion of BeiDou and UWB positioning signals, thereby improving early perception and response accuracy in areas of trajectory abrupt change. Traditional rail transit positioning methods often rely on a single type of positioning signal source. In complex scenarios such as elevated sections, tunnel entrances, and areas obstructed by buildings, positioning signal distortion, short-term loss, or fluctuations are prone to occur, leading to a sharp drop in positioning accuracy. This invention, based on the spatial coverage status of BeiDou and UWB positioning signals along the train's running path, acquires multi-dimensional state features, including signal stability values, historical drift trend parameters, and trajectory change amplitudes, and then generates a cross-modal distortion discrimination array. This array can dynamically extract whether the current position is in a trajectory abrupt change area, providing a structured judgment basis for the dynamic adjustment of subsequent fusion weights. This enables trajectory fusion to move beyond relying on single signal fluctuations and instead proactively identify abnormal positioning signal behavior, thereby improving the intelligence and environmental adaptability of trajectory fusion.

[0043] This invention achieves rapid correction and optimization of trajectory abrupt changes by combining a dynamic adjustment mechanism of fusion weight ratios with a constraint backtracking trajectory fitting algorithm, effectively improving the smoothness of trajectory continuity and the stability of positioning results. After identifying trajectory abrupt change regions, this invention extracts the trajectory curvature change and velocity vector difference between the current trajectory point and three consecutive historical trajectory points, using these as the basic calculation factors for the trajectory disturbance feature group. Based on this calculation result, the fusion weight ratio of BeiDou positioning signals and ultra-wideband positioning signals is adjusted, giving the dominant signal a higher weight during abrupt changes to offset trajectory drift caused by error amplification in non-dominant signals. After the fusion weight adjustment is completed, the constraint backtracking trajectory fitting algorithm is further used to correct the abrupt changes. The offset value of the current trajectory point is calculated using the historical trajectory fitting curve, and the spatial topology of the ultra-wideband positioning base station is used to determine whether the trajectory point is a source of a false trajectory. If the offset value exceeds the error threshold or the trajectory point is confirmed as a false trajectory point, it can be promptly removed to prevent erroneous trajectory data from entering the fused trajectory sequence, ensuring the structural integrity and traceability of the trajectory in dynamic fluctuation scenarios.

[0044] This invention introduces a time-series analysis mechanism for signal channel weight changes during trajectory fusion, enabling real-time identification and suppression of frequent signal channel jump trends. This enhances the robustness and stability of trajectory fusion, making it particularly suitable for operating environments with frequent signal fluctuations. During train operation, BeiDou positioning signals and ultra-wideband positioning signals may experience frequent switching of dominant signals due to external environmental factors such as obstruction, electromagnetic interference, or changes in equipment status. If not suppressed, this can lead to instability in the fused trajectory. This invention performs time-series analysis on the signal fusion weight changes over several consecutive seconds. By analyzing the number of dominant signal switches and the amplitude of weight fluctuations between BeiDou and ultra-wideband signals, it determines whether a jump trend exists. Once any signal channel is found to have experienced three or more dominant signal switches within this time period, or its fusion weight fluctuation exceeds a set stability threshold, the fusion weight of that channel is reduced to the minimum tolerance threshold, thereby temporarily weakening its interference with the overall trajectory. This strategy not only avoids fusion oscillations caused by signal instability but also provides a basis for triggering subsequent backup mechanisms, effectively supporting the stable evolution of trajectory fusion and making it suitable for rail transit positioning tasks. Attached Figure Description

[0045] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0046] Figure 1 This is a schematic diagram of a rail transit positioning method using BeiDou and UWB dual-mode coordination, as described in this invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0048] Reference Figure 1 The following examples were obtained:

[0049] Example 1: A dual-mode (BeiDou and UWB) collaborative positioning method for rail transit includes the following steps: Based on the spatial coverage status of BeiDou and UWB positioning signals along the train's route, signal stability values, historical drift trend parameters, and trajectory change amplitudes are acquired, and a cross-modal distortion discrimination array is generated. By performing multi-dimensional quantification of the coverage, stability, and historical change trends of BeiDou and UWB positioning signals in different operating sections, the system can establish a basic data structure that fully reflects the quality differences of multi-source signals before trajectory fusion begins. The cross-modal distortion discrimination array plays a unified role in carrying characteristic information such as signal stability, drift direction, and trajectory continuity risk level, providing a clear spatiotemporal reference for subsequent trajectory change identification. This array not only improves the sensitivity of trajectory anomaly detection but also ensures reliable pre-judgment basis for multi-source signal fusion, reducing the interference of pseudo-trajectories and jump points on fusion quality from the source.

[0050] The system uses a cross-modal distortion discrimination array to identify whether the current location contains a trajectory abrupt change region. If the identification result indicates the existence of a trajectory abrupt change, it calculates the change in trajectory curvature and the velocity vector difference between the current trajectory point and three consecutive historical trajectory points. Based on the calculation results, it adjusts the fusion weight ratio of the BeiDou positioning signal and the ultra-wideband positioning signal to suppress the amplification of positioning errors caused by trajectory abrupt changes. Utilizing the risk information marked in the cross-modal distortion discrimination array, the system quickly determines whether trajectory abrupt changes exist in the current region during trajectory fusion, preventing the system from continuing to use the default fusion weights in abnormal sections, which would further amplify the error. By calculating the change in curvature and the velocity vector difference between consecutive trajectory points, the intensity of trajectory disturbance can be accurately quantified, providing a sufficient mathematical basis for adjusting the fusion weight ratio. When a trajectory abrupt change occurs, the dynamic adjustment of the fusion weight ratio can quickly establish the dominant signal source between the BeiDou positioning signal and the ultra-wideband positioning signal, ensuring that the trajectory remains smooth and reliable in abnormal sections, providing a stable input for subsequent trajectory correction.

[0051] After adjusting the fusion weights, a constraint-backtracking trajectory fitting algorithm is used to correct trajectory abrupt changes. The offset value of the current trajectory point is calculated using historical trajectory fitting curves, and the spatial topology of the ultra-wideband positioning base station is considered to determine whether the trajectory point is located at the source of a pseudo-trajectory. If the offset value exceeds the error threshold, or the trajectory point is located at the source of a pseudo-trajectory, the trajectory point is removed from the fused trajectory sequence. Further backtracking correction is performed on trajectory abrupt changes after the fusion weights are adjusted, enabling the trajectory to recalibrate its current position based on historical trajectory trends and acceleration changes when encountering strong disturbances. Calculating the offset value using historical trajectory fitting curves quantifies the degree of deviation of the abrupt change point from the true trajectory. Combining this with the spatial topology of the ultra-wideband positioning base station identifies whether the trajectory point is located in high-risk pseudo-trajectory areas such as base station coverage intersections. If the offset is determined to be too large or located at the source of a pseudo-trajectory, the trajectory point is removed to prevent erroneous trajectories from entering the fused trajectory sequence, thus ensuring the continuity and reliability of the overall trajectory output and providing the system with highly robust anti-interference capabilities.

[0052] Time series analysis is performed on the changes in signal fusion weights over a continuous n-second period during trajectory fusion. If the analysis shows a frequent trend of signal channel jumps, the corresponding fusion weight is lowered to the minimum tolerance threshold. Real-time stability checks are performed on the signal fusion process by monitoring the number of jumps and the amplitude of fluctuations in the fusion weights over a continuous time period. If a frequent trend of signal channel jumps occurs, it indicates that the corresponding signal has unstable characteristics or sudden interference within that time window. In this case, lowering its fusion weight to the minimum tolerance threshold can effectively prevent the signal from continuing to interfere with the trajectory fusion results.

[0053] The steps for constructing a cross-modal distortion discrimination array include: extracting spatial environment labels corresponding to the current geographical segment of the train's operating path based on the current location, and classifying and labeling the geographical segment, such as labeling it as "elevated section," "underground tunnel section," "entrance / exit buffer zone," or "platform stopping section," etc. Within each segment, coverage data generated by BeiDou positioning signals and ultra-wideband positioning signals during actual operation are collected, including spatial coverage status indicators such as whether the signal can be continuously received, the stability of the receiving frequency, and the duration of signal interruption. Simultaneously, within the same segment, the upper and lower limits of signal strength during the past seventy consecutive positioning processes are recorded, and the fluctuation range is statistically analyzed; signal obstruction event records within each time window are extracted from the equipment operation log, including information such as the frequency of obstruction, the duration of obstruction, and the recovery time. Through the above operations, the spatial coverage status, signal strength fluctuation range, and historical signal obstruction records of BeiDou positioning signals and ultra-wideband positioning signals in this segment are extracted, and jointly indexed by time series and geographical location to form a multi-source spatial sensing dataset.

[0054] Based on a multi-source spatial sensing dataset, stability and trend indicators of BeiDou and UWB signals at the current location are calculated separately. In the stability calculation, the variance of the signal-to-noise ratio (SNR) within a sliding time window is introduced as the SNR stability value. Based on continuous signal quality data sampled ten times per second, the trajectory of signal strength variation for each signal channel is fitted, thus deriving the historical signal drift trend slope. In the trajectory change calculation, a spatial path vector model is constructed from the five most recent trajectory points in chronological order. The rate of change of velocity per unit time is extracted, and its second derivative is used to form an acceleration curve to assess whether there are acceleration disturbances or inertial fluctuations in the trajectory. Through the above process, a corresponding signal state vector is constructed. This vector includes multi-dimensional features such as the SNR stability value, drift trend slope, trajectory acceleration parameters, and their direction weights, used to comprehensively characterize the operational status characteristics of BeiDou and UWB positioning signals within the current spatial segment.

[0055] The signal state vector constructed in the previous step is mapped to a unified spatiotemporal grid coordinate system. This involves embedding the current trajectory point's location into a 3D positioning grid with a grid resolution of one meter, using latitude, longitude, and timestamp coordinates. The numerical features corresponding to the signal state vector are then propagated to adjacent grid cells through interpolation, forming a dense signal state map. During the mapping process, each grid cell is assigned a unique spatial code and time label, while simultaneously recording its corresponding BeiDou and ultra-wideband signal states. Directional fusion processing is performed on blurred boundary areas, thus achieving continuous projection of signal vector features in space. Based on the grid structure, fusion is performed according to the correlation of signal variations and the risk level of trajectory continuity to generate a cross-modal distortion discrimination array. This array represents a spatial risk mapping map covering the entire target path segment, providing spatiotemporal distribution data for subsequent key steps such as trajectory abrupt change region identification, fusion weight adjustment, and pseudo-trajectory removal.

[0056] The fusion based on signal variation correlation and trajectory continuity risk level refers to: extracting the positioning signal difference features between each trajectory point and its predecessor based on the signal state vector within a unified spatiotemporal grid coordinate system and the trajectory temporal relationship. For BeiDou positioning signals and ultra-wideband positioning signals, the signal stability difference value is calculated, which is the absolute value of the difference between the signal-to-noise ratio stability values ​​of two adjacent points; the drift direction consistency value is obtained by calculating the cosine similarity of the angle between the signal drift trend vectors of two adjacent points, and the closer the value is to one, the higher the directional consistency between the two points; the curvature change rate is calculated by constructing a trajectory curvature model through a three-point sliding window to calculate the rate of curvature change of the current trajectory segment. The above three types of data are filled into the spatial grid cells corresponding to the trajectory points according to the trajectory temporal structure to form a three-dimensional index set, and further generate a signal variation correlation factor matrix. Each cell in this matrix corresponds one-to-one with a specific grid region to quantify the spatiotemporal variation characteristics of that grid region.

[0057] Statistical analysis is performed on the variation amplitude of each unit in the signal variation correlation factor matrix, extracting the maximum, minimum, and average rate of change of each unit within the past five-second time window. These variation amplitudes are then numerically compared with preset trajectory continuity risk assessment threshold ranges. Three risk grading threshold segments are set, for example, a stable range of 0 to 0.15, a slightly disturbed range of 0.15 to 0.35, and a high-intensity disturbed range above 0.35. Each matrix unit is classified and assessed: if the variation amplitude falls within the lowest range, it is assigned a low-risk label; if it falls within the middle range, it is assigned a medium-risk label; and if it falls within the highest range, it is assigned a high-risk label, thus forming a trajectory continuity risk level label group corresponding to the grid area.

[0058] A spatial mapping relationship is established between the matrix cells of the signal variation correlation factor matrix and the trajectory continuity risk level label group, linking the position of each cell in the former with the corresponding risk label in the latter. To ensure matching accuracy, a position hash mapping function is used to bind the matrix cell number to the grid coordinates one by one. Then, all grid cells are traversed sequentially, and cells with medium-risk or high-risk level labels are selected and marked as target analysis objects, recording their coordinate index, risk type, and historical jump trend, among other auxiliary information. The selected grid cells are recorded as distortion-sensitive cells and embedded in the trajectory-aware path map as high-sensitivity areas of priority during trajectory fusion. Finally, using the distribution structure of distortion-sensitive cells as a framework, a cross-modal distortion discrimination array is generated to support anomaly removal and dynamic adjustment of signal channel weights in the subsequent trajectory fusion process.

[0059] The steps for identifying whether the current location contains a trajectory abrupt change region include: extracting the grid cell corresponding to the current train position from the spatial grid structure provided by the cross-modal distortion discriminant array, and reading the label level of the grid cell in the trajectory continuity risk level label group. To ensure high accuracy in the extraction process, a method of jointly matching grid cells by latitude, longitude, and trajectory point timestamps is adopted, so that the current location can accurately correspond to the spatial number in the discriminant array. For example, if a train trajectory point is located in an area 0.9 meters northeast of the previous position with an elevation change of less than 0.1 meters, then the point will be mapped to a one-dimensional array position with trajectory grid number XYZ, thereby accurately reading the label level of the grid cell in the trajectory continuity risk level label group, such as "low risk level label," "medium risk level label," or "high risk level label." This operation provides a quantifiable initial risk basis for subsequent trajectory abrupt change region judgment.

[0060] To determine whether trajectory risks exhibit a spatial clustering trend, the label level of the current grid cell needs to be compared with the label levels of its eight neighboring grid cells. To ensure spatial continuity in the comparison process, a nine-grid neighborhood structure is introduced, treating the current grid cell as the center position, with the surrounding eight grid cells corresponding to its left, right, front, back, and four diagonal directions. For example, in sections where train trajectory changes frequently, multiple consecutive grid cells in a certain direction may be labeled as "high-risk level". If three or more adjacent cells with high-risk level labels exist consecutively in this nine-grid neighborhood, it indicates that the spatial region exhibits a trend of rapidly decreasing stability and increased signal drift. In this case, the current grid position is identified as a candidate center for a trajectory abrupt change region, and its coordinate information, risk diffusion direction, and regional density distribution of risk levels are recorded internally for boundary analysis in the next sub-step.

[0061] To determine whether a candidate central region truly belongs to a trajectory abrupt change area, signal trend consistency detection needs to be performed on the spatial boundary of the candidate central region. In this process, a maximum of two layers of grid cells are extended from the candidate central region in the four main directions and four diagonal directions to construct a boundary detection zone. For all grid cells within this detection zone, the signal stability difference value between the current trajectory point and the previous trajectory point, as well as the trend of this difference value's change at adjacent time points, are compared. If the detection results show a sign flip in the signal stability difference value between the current and adjacent time points (i.e., from positive to negative or vice versa), it indicates a sudden change in signal trend. Furthermore, if the amplitude change of the signal stability difference value exceeds a preset abrupt change threshold, such as a rapid increase from 0.15 to above 0.5, it indicates that the signal has experienced highly discontinuous fluctuations within a short period. This trend can visually reflect whether the trajectory point is affected by high noise interference, source drift, or abrupt path changes. This process ensures that risk assessment does not rely on a single time point but captures true trajectory abrupt change patterns through continuous time difference checks.

[0062] When high-risk level labels cluster in the spatial neighborhood and boundary signals exhibit a clear trend abrupt change, the current location can be definitively confirmed to contain a trajectory abrupt change region. After confirmation, labeling information is generated for this trajectory point, including risk level, direction of abrupt change, degree of abrupt change, and deviation from historical trajectory changes. This point serves as input data for subsequent fusion weight adjustments, trajectory correction, and pseudo-trajectory removal. This type of determination not only improves the sensitivity of trajectory anomaly detection but also provides a dual verification mechanism for trajectory abrupt change identification across time and space, thereby ensuring that subsequent trajectory processing strategies are implemented on a more reliable basis.

[0063] When the identification result indicates a trajectory abrupt change, the following operations are performed: The position data of the current trajectory point and its three preceding historical trajectory points within a unified spatiotemporal grid coordinate system are extracted, and the curvature change and velocity vector difference between each pair of adjacent trajectory points are calculated. To ensure the continuity and spatiotemporal consistency of trajectory disturbance identification, this implementation selects four consecutive trajectory points as the analysis window, where the first trajectory point is the current trajectory point, and the subsequent three trajectory points are the penultimate to third-to-last historical trajectory points. Based on the trajectory point coordinates, the curvature change between the three pairs of points is calculated, reflecting the smoothness of the trajectory path by analyzing the angle change between trajectory polygonal segments. For example, if the trajectory shows a straight trend between the past two points but suddenly reverses at the current point, the curvature change will significantly increase. Simultaneously, the velocity vector difference between each pair of points is calculated and quantified using the velocity magnitude and direction difference to reflect whether there is abrupt acceleration or rapid deceleration in the trajectory motion. The above calculation results together constitute a set of trajectory disturbance feature parameters, forming a trajectory disturbance feature set, which serves as the input basis for subsequent normalization and hierarchical evaluation.

[0064] The maximum curvature change and the maximum velocity vector difference in the trajectory disturbance feature group are normalized to obtain the curvature disturbance index and the velocity disturbance index, respectively. This step uses a maximum-minimum normalization method, with the historical disturbance extreme values ​​within the recent trajectory segment as the upper and lower limits of normalization, ensuring that the disturbance index is continuously distributed within the [0,1] interval. For example, when the maximum curvature change is 0.85 and the historical extreme value is 0.9, the normalized curvature disturbance index is 0.944; when the maximum velocity vector difference is 1.2 m / s and the historical extreme value is 2.0 m / s, the velocity disturbance index is 0.6. Subsequently, segmented scoring is performed based on the cumulative frequency of these two disturbance indices over the past three trajectory points. The scoring criteria are based on the number of times the disturbance index exceeds 0.7, assigning a change level score of one to three points according to the frequency. This frequency-based disturbance assessment mechanism allows abrupt trends to be identified from a continuous perspective, rather than solely based on a single instantaneous change, enhancing the temporal robustness of trajectory analysis.

[0065] Two change level scores are weighted and superimposed according to a preset ratio coefficient to generate a fusion offset coefficient reflecting the severity of the current trajectory change. The ratio coefficient is set based on the linear fitting results of the experimental path data. Typically, the curvature disturbance score is weighted at 0.6 and the velocity disturbance score at 0.4 to more sensitively reflect the path deviation risk caused by directional changes. For example, if the curvature disturbance score is 2 and the velocity disturbance score is 1, then the fusion offset coefficient is 2×0.6+1×0.4=1.6. This fusion offset coefficient serves as a key basis for weight adjustment and is compared with a standard benchmark value. The standard benchmark value can be set according to different operating states. Under normal operating conditions, it is set to 1.0, indicating that the current trajectory disturbance level is still within the normal range of change. If the fusion offset coefficient is significantly higher than this benchmark value, it indicates that the current trajectory point is affected by strong disturbance, and the weight should be adjusted to the side with higher signal stability first. By comparing the difference between the current fusion offset coefficient and the standard benchmark value, the fusion weight gain that the current trajectory point should adjust towards the BeiDou positioning signal or ultra-wideband positioning signal direction is determined, and the fusion weight ratio is updated.

[0066] The current trajectory point fusion weight update is performed according to the following process. Its core purpose is to dynamically adjust the fusion weight ratio of the BeiDou positioning signal and the ultra-wideband positioning signal based on trajectory disturbances after abrupt trajectory changes occur, thereby improving trajectory continuity and positioning reliability. To ensure that the weight update has a clear execution basis in engineering, this implementation defines the "dominant signal" as: among the BeiDou positioning signal and the ultra-wideband positioning signal at the current trajectory point, the signal source with higher signal stability, smoother historical drift trend parameter changes, and consistent offset direction in the trajectory disturbance feature group, should have its weight enhanced during the update process. The specific update process includes the following steps: if the fusion offset index is less than a preset first threshold, the existing fusion weight ratio remains unchanged. After the trajectory disturbance feature group has been calculated by the previous module, the fusion offset index of the current trajectory point can be obtained. The fusion offset index reflects the degree of trajectory disturbance and is usually expressed in the range of zero to two. In most normal operating scenarios, the trajectory change is relatively smooth, and the fusion offset index will be lower than the first threshold. For example, when a train is in a deceleration phase before entering a station, with minimal signal obstruction and stable trajectory changes, the fusion offset index might be 0.25. If it is lower than a preset first threshold (e.g., the first threshold is set to 0.4), it indicates that the trajectory disturbance is insufficient to affect the trajectory smoothness. In this case, there is no need to change the fusion weight ratio between the BeiDou positioning signal and the ultra-wideband positioning signal. Maintaining the existing fusion weight ratio can avoid unnecessary weight fluctuations caused by minor disturbances, ensuring the continuity and stability of trajectory fusion and providing a stable benchmark for subsequent judgments.

[0067] If the fusion offset index is between a preset first threshold and a preset second threshold, the current fusion weight ratio is linearly offset towards the direction of the dominant signal by a preset constant. If the previous step determines that the fusion offset index has not triggered the first type of situation, it will be further determined whether it is between the first threshold and the second threshold. This range usually represents that the trajectory is in a state of gradual disturbance, such as the change in trajectory curvature gradually approaching the abrupt threshold from the normal range, while the velocity vector difference shows a slight jump within one second. At this time, it is necessary to make a slight adjustment to the fusion weight ratio according to the direction of the dominant signal. This implementation adopts a linear offset mechanism, that is, the weight is linearly increased by a preset constant. For example, when the linear offset constant is set to 0.1, if the dominant signal is the BeiDou positioning signal, the fusion weight of the BeiDou positioning signal will increase by 0.1, while the fusion weight of the ultra-wideband positioning signal will decrease by 0.1, and then be renormalized after execution. This linear offset strategy enables a gradual correction effect when the trajectory disturbance is in the moderate change range, avoiding secondary disturbances caused by sudden weight jumps.

[0068] If the fusion offset index is greater than the preset second threshold, the fusion weight gain is calculated according to the following nonlinear formula: subtract the preset second threshold from the current fusion offset index, multiply by the set nonlinear adjustment coefficient, and use this as the fusion weight gain value for the current trajectory point. When the fusion offset index exceeds the second threshold, it means that the trajectory disturbance shows a significant abrupt change trend. For example, the curvature change of the current trajectory point has exceeded 0.7, and the velocity vector difference has reached a jump of more than one meter per second within one second. At this time, linear offset alone is insufficient to achieve effective correction, so a nonlinear gain is used for weight adjustment. Taking a practical example: if the fusion offset index is 1.6, the second threshold is 1.0, and the nonlinear adjustment coefficient is 0.8, then the fusion weight gain value is (1.6 minus 1.0) multiplied by 0.8, finally obtaining 0.48. This gain value will be used to strengthen the dominant signal. The use of nonlinear calculation method enables the weight adjustment to quickly establish the dominant signal advantage in high-disturbance scenarios, ensuring that the trajectory can be forcibly pulled back to the normal trend.

[0069] The weight gain value is added to the current weight value of the dominant signal and then normalized by summation to generate an updated fusion weight ratio. After obtaining the fusion weight gain value, this gain value is applied to the dominant signal. For example, if the current BeiDou positioning signal weight is 0.5, the ultra-wideband positioning signal weight is 0.5, and the calculated fusion weight gain value is 0.48, then the updated BeiDou positioning signal weight becomes 0.98, and the ultra-wideband positioning signal weight is adjusted to 0.02. Subsequently, to ensure the mathematical consistency of the fusion weight ratio, both need to be normalized by summation. This process ensures that the dominant signal has an absolute advantage in abrupt changes, allowing the trajectory to quickly correct itself along a direction closer to the true trajectory, ultimately generating an updated fusion weight ratio for the next trajectory point positioning fusion.

[0070] When a trajectory abrupt change region is detected during trajectory fusion, and the fusion weight adjustment between the BeiDou positioning signal and the ultra-wideband positioning signal has been completed, a constrained backtracking trajectory fitting algorithm is used to correct the trajectory abrupt change point to eliminate the impact of the trajectory abrupt change on the continuity and accuracy of subsequent positioning. This fitting algorithm comprehensively considers the historical evolution characteristics of the trajectory, the signal fusion trend, and the maximum tolerable offset range to achieve fine-grained repair of the abrupt change point. Specifically, it includes the following steps: extracting the fused positioning data of the current trajectory point (including the trajectory abrupt change point) and its four consecutive historical trajectory points, constructing a trajectory evolution window, and marking the abrupt change point location as the backtracking starting point. After the trajectory abrupt change point is detected, this trajectory point needs to be used as the starting node to construct the backtracking window. In this embodiment, the trajectory evolution window is constructed as follows: tracing back four consecutive historical trajectory points from the current trajectory point, extracting the fused positioning data of these trajectory points in a unified spatiotemporal grid coordinate system, including their corresponding latitude and longitude coordinates, fusion weight ratio, trajectory disturbance index, and velocity vector and curvature change information between them and the previous point. This trajectory evolution window contains five consecutive trajectory points to ensure sufficient sample support for the trajectory evolution trend. The mutation point is clearly marked as the backtracking starting point, which is a key location for subsequent trajectory reconstruction. The repair operation only applies to the data within this starting point and its affected segment.

[0071] For historical trajectory points within the trajectory evolution window, the velocity variation amplitude, spatial curvature continuity, and signal fusion confidence trend are calculated sequentially over time. These are used to construct fitting constraint boundaries, including trajectory continuity preservation boundaries, fusion signal trend preservation boundaries, and maximum offset suppression boundaries. After establishing the trajectory evolution window, the evolution trend of the trajectory points within it needs to be evaluated, and multidimensional constraint boundaries that the repair process must satisfy are formed accordingly. First, the trajectory continuity preservation boundary requires that the velocity change rate and curvature change between adjacent trajectory points must satisfy a steady increasing or decreasing relationship; its boundary value can be set by the time moving average of the trajectory disturbance exponent. Second, the fusion signal trend preservation boundary requires that the fusion weights of the BeiDou positioning signal and the ultra-wideband positioning signal in the trajectory points should show a stable trend; the boundary condition is that the weight variation amplitude must not exceed the set confidence stability interval (e.g., less than 0.3). Third, the maximum offset suppression boundary requires that the maximum spatial offset between the fitting result and the historical trajectory trend must not exceed the error tolerance upper limit (e.g., three meters) to prevent the repaired points from deviating from the normal trajectory. The three boundary conditions mentioned above together constitute the rigid constraint framework of the fitting algorithm, ensuring that the repair result is not only mathematically feasible, but also logically reasonable in terms of trajectory evolution.

[0072] Under the premise of satisfying all fitting constraints, an initial attempt is made to adjust the coordinates of the abrupt change point using the acceleration change trend before and after the trajectory. If a correction result that meets the offset tolerance cannot be generated under all constraints, a compensatory backtracking correction operation is performed, i.e., reconstructing the local trajectory segment by inserting intermediate transition points. Based on the limitations of the trajectory evolution window and the fitting constraint boundaries, an initial attempt is made to adjust the coordinates of the abrupt change point by minimizing the trajectory acceleration abrupt change. The rate of change of velocity, angle turning amplitude, and curvature continuity difference between the two trajectory points before and after the abrupt change point are calculated, and the optimal coordinate position that meets the constraints is found for replacement and repair. If, after multiple fitting attempts, the adjusted coordinates still cause the trajectory disturbance index or offset to exceed the maximum offset suppression boundary, the single-point repair method is considered infeasible, and a compensatory backtracking correction operation needs to be initiated. This operation reconstructs local trajectory segments by inserting intermediate transition points. Specifically, between the trajectory points before and after the abrupt change point, one or two intermediate trajectory points with smooth velocity and angle change characteristics are generated. The interpolation method can be acceleration spline fitting or bilateral curvature guided interpolation, and the interpolation points are guaranteed to form a stable connection within the trajectory continuity maintenance boundary, thereby reconstructing a smooth transition local trajectory segment.

[0073] The reconstructed coordinates or interpolated points replace the original abrupt trajectory points, and the results are written into the fused trajectory sequence to update the fitting state of the current trajectory. After the trajectory abrupt point repair operation is completed, the correction results, whether obtained through coordinate fine-tuning or interpolation reconstruction, must be formally written into the fused trajectory sequence. Specifically, the original fused positioning data at the abrupt point is overwritten and updated, and the velocity vector, curvature, fusion weight, and other auxiliary attribute values ​​of the connecting edges of adjacent trajectory points are recalculated to ensure the logical integrity and continuity of the fused trajectory sequence. If an intermediate transition point is inserted, all subsequent trajectory points need to be reordered and numbered, and the relative reference benchmark of subsequent trajectory points is updated synchronously. After writing the results, the current trajectory state is marked as "fitting complete," and the repair history information of this trajectory segment is stored in the trajectory evolution window cache as an important reference for subsequent judgment of trajectory evolution patterns. This method not only eliminates the risk of trajectory distortion caused by abrupt points but also ensures the physical reliability and stable availability of trajectory data through reasonable fitting boundaries and buffering mechanisms.

[0074] The pseudo-track source location refers to the area where the trajectory point's position shifts, distorts, or drifts due to factors such as multipath reflection of ultra-wideband signals, overlapping critical coverage, or signal interference. This is commonly found in areas where the coverage boundaries of multiple base stations intersect. The specific judgment method includes the following steps: Obtain the numbers of all ultra-wideband positioning base stations used within a preset time window of the current trajectory point and their fixed spatial coordinates; calculate the Euclidean spatial distance from the current trajectory point to each base station; and draw connecting edges to each base station with the current point as the central node, forming a polygonal star-shaped topology map centered on the trajectory point. This step aims to establish the spatial connection structure between the current trajectory point and surrounding base stations. First, using the positioning history recorded by the trajectory fusion module, extract all ultra-wideband positioning base station numbers that participated in the positioning calculation within a preset time window (e.g., two seconds before and after), typically ranging from three to five. Then, call the base station spatial deployment database to obtain the three-dimensional fixed spatial coordinates (including longitude, latitude, and elevation) of the corresponding base stations. Based on these coordinate data, calculate the three-dimensional Euclidean spatial distance from the current trajectory point to each base station to ensure that the constructed structure is consistent with the actual spatial layout. After distance calculation, spatial connection edges are drawn to all base stations with the current trajectory point as the center of the graph, generating a "polygonal star topology graph centered on the trajectory point". This topology graph is used to subsequently determine whether the spatial geometric relationship of the trajectory point is consistent with the pseudo-trajectory structure.

[0075] Using the normal coverage radius of each base station as the boundary, multiple circular coverage areas are drawn around the central trajectory point, and the overlapping areas of each coverage area are marked as cross buffer zones. Based on the established polygonal star topology, a spatial geometric model is further introduced to determine whether the trajectory point is located in a possible pseudo-trajectory interference area. First, based on the set standard coverage radius (e.g., ten meters, fifteen meters, etc.) of each ultra-wideband positioning base station, a circular sensing area within the coverage radius is drawn with the coordinates of each base station as the center. This coverage radius is generally given by site measurement data or engineering deployment strategy. Subsequently, all circular coverage areas are projected onto a unified plane and overlap calculations are performed to mark the "cross buffer zones" formed by the overlap of any three or more circular areas. The "cross buffer zone" referred to in this invention refers to a high-risk area in which signal strength interference superposition, multiple path reflections, or received signal confusion easily occur due to the combined effect of multiple base station signals, which often induces trajectory drift or multiple abrupt changes.

[0076] To determine if the current trajectory point is within the aforementioned cross-buffer zone, if it falls within the intersection of three or more base station coverage areas, it can be considered to have a high probability of pseudo-trajectory drift risk, and the trajectory point can be preliminarily determined to be located at the pseudo-trajectory source location. After constructing the cross-buffer zone, it is necessary to determine whether the current trajectory point falls within that area in geometric space. First, the two-dimensional coordinates of the current trajectory point are projected onto the coverage area analysis plane, and its spatial position is determined based on its relationship with the boundaries of each cross-buffer zone. If the center point of the trajectory point is simultaneously located in the overlapping part of three or more circular coverage areas, it indicates that the point is highly likely to be affected by multi-source signal interference, and there is a risk of abnormal phenomena such as signal overlap, jitter, and delayed paths. According to the definition standard of this invention, the trajectory point can be preliminarily determined to be the pseudo-trajectory source location and marked as a suspected pseudo-trajectory point candidate for use in subsequent offset value linkage elimination strategies.

[0077] The determination result of the pseudo-trajectory source location is jointly judged with the offset value calculated from the historical trajectory fitting curve. If the offset value of the current trajectory point exceeds the error threshold and its spatial location is within the cross buffer of three or more base stations, the trajectory point is removed from the fused trajectory sequence, and the trajectory reconstruction mechanism is triggered. This step achieves the final removal of pseudo-trajectory points and purification of the trajectory sequence. After the spatial determination of the pseudo-trajectory source location has been completed, the offset value obtained from the historical trajectory fitting curve is used for further judgment. The offset value is calculated as the shortest vertical distance from the current trajectory point to the historical trajectory fitting curve. If its absolute value is greater than the set maximum offset error tolerance (e.g., 3.5 meters) and it has been marked as a pseudo-trajectory candidate point located within the cross buffer, then the point can be confirmed as a positioning distortion point. At this time, the point is immediately removed from the trajectory fusion sequence, and subsequent compensation operations such as trajectory segment fitting, intermediate point interpolation, or trajectory extension are re-initiated according to the fusion mechanism settings to ensure the overall physical continuity, logical rationality, and data reliability of the trajectory.

[0078] Time series analysis is performed on the changes in signal fusion weights over a continuous time period, and the fusion weight ratio is dynamically adjusted based on the results. This mechanism can not only identify abnormal jumps in the BeiDou positioning signal and the ultra-wideband positioning signal within a short period of time, but also automatically trigger a backup trajectory compensation strategy to maintain the continuity and reliability of the trajectory output when both channels experience instability simultaneously. The process includes the following steps: extracting the fusion weight sequence of the current trajectory point and several trajectory points before and after it, and calculating the number of dominant signal switching and weight fluctuation amplitude of the BeiDou positioning signal and the ultra-wideband positioning signal in this sequence. This step first constructs a time window centered on the current trajectory point on the time axis, with a window length of n consecutive seconds of sampling range, where n can be three, five, or ten seconds, set according to the sensitivity of the trajectory disturbance. Within this window, each trajectory point records the fusion weight ratio, forming a time series. By extracting point by point, the fusion weight sequence of the BeiDou positioning signal and the fusion weight sequence of the ultra-wideband positioning signal can be formed. Subsequently, two types of indicators were calculated for the sequences of the two positioning signals: the first was the number of dominant signal switching times, which counts the number of times the fusion weight changed its "dominance" between the two signals. For example, if the BeiDou positioning signal weight was once dominant and then overtaken by the ultra-wideband positioning signal, it was counted as one switching. The second was the weight fluctuation amplitude, which is the difference between the maximum and minimum weight values ​​within a time window, used to measure overall stability. For example, if the BeiDou positioning signal weight decreased from 0.7 to 0.3 and then increased to 0.8 within three seconds, the fluctuation amplitude was 0.5. This step provides a quantitative basis for subsequent identification of abnormal jump trends.

[0079] If any signal channel experiences three or more dominant signal switchings within a given time period, or if the weight fluctuation exceeds a set stability threshold, the fusion weight of that signal channel is reduced to the minimum tolerance threshold. After obtaining the number of dominant signal switchings and the weight fluctuation amplitude, these two indicators are compared with preset thresholds. For example, the threshold for the number of dominant signal switchings can be set to three, and the threshold for the weight fluctuation amplitude to 0.45. When the number of dominant signal switchings for a BeiDou positioning signal or an ultra-wideband positioning signal reaches or exceeds three within the same time window, or if the weight fluctuation amplitude is greater than the threshold, the signal channel can be determined to be in a short-term unstable state. For example, if a BeiDou positioning signal experiences four repeated weight jumps within five seconds, it indicates that its short-term signal quality is insufficient to support trajectory fusion. In this case, the fusion weight of the BeiDou positioning signal needs to be directly reduced to the minimum tolerance threshold (e.g., 0.1), and the fusion weight of the ultra-wideband positioning signal needs to be adjusted to a ratio close to one, so that trajectory fusion is not affected by unstable signals. This step has a clear connection with the previous step, using the statistical results from the previous step to derive the judgment criteria, and then making suppressive adjustments to the weights in this step.

[0080] When both the BeiDou positioning signal and the ultra-wideband positioning signal meet any of the following conditions within the same continuous time period: three or more dominant signal switches occur, or the weight fluctuation amplitude simultaneously exceeds the set stability threshold, a backup trajectory compensation strategy is activated. When the above judgment occurs not only in a single signal channel but simultaneously in both the BeiDou positioning signal and the ultra-wideband positioning signal, it indicates that the fused weight sequence as a whole has entered a severely unstable state. For example, when both types of signals generate four dominant signal switches within the same five-second window, and the fluctuation amplitude is greater than 0.5, trajectory fusion cannot effectively rely on either signal channel. In this case, single-channel suppression alone cannot restore trajectory stability. To solve this problem, this invention provides a backup trajectory compensation strategy as a trajectory continuity guarantee mechanism under dual-channel instability conditions. Its core idea is to temporarily use historical trajectory trends to replace the current distorted positioning results to avoid trajectory breakpoints, jumps, or abnormal drifts.

[0081] The backup trajectory compensation strategy includes the following steps: extracting five consecutive historical trajectory points from the current trajectory point to construct a trajectory sliding window; and generating backup trajectory points using interpolation fitting based on the velocity, direction, and acceleration trends of the historical trajectories, and writing these backup trajectory points into the fused trajectory sequence. Specifically, the compensation strategy includes the following steps: First, extracting five historical trajectory points from the current trajectory point to form a structured trajectory sliding window; Second, calculating the velocity, direction, and acceleration trends of the trajectory points within the window in chronological order, for example, using three-point velocity difference to obtain the acceleration trend; Third, generating backup trajectory points for the current trajectory point using trajectory interpolation methods (such as acceleration-guided interpolation or curvature-smoothing spatial interpolation); Fourth, writing these backup trajectory points into the fused trajectory sequence as temporary replacements for the current trajectory point, and continuously monitoring signal stability within the next time window; Fifth, once the BeiDou positioning signal or ultra-wideband positioning signal stabilizes, the backup trajectory compensation strategy is exited and the normal fusion process resumes. This strategy enables the trajectory to remain continuous, smooth, and directionally consistent even when both channels are unstable, demonstrating significant anti-interference capabilities and engineering value.

[0082] It should be noted that the UWB positioning signal in this invention only participates in collaborative fusion in certain trajectory segments, and the signal status is determined and dynamically weighted through a "cross-modal distortion discrimination array." For example, in scenarios with complex signals or where BeiDou signals are easily attenuated, such as when transitioning from elevated to ground level, entering or exiting tunnels, or near platforms, UWB base stations are only deployed at key nodes. In typical "indoor or semi-indoor" scenarios such as locomotive depots, parking lines, maintenance workshops, and station areas, UWB becomes the dominant signal source. In most open areas along the route, BeiDou positioning still dominates, and UWB does not undertake the main positioning task in these areas, avoiding unnecessary costs and deployment burdens. Therefore, this invention has the flexibility to deploy on demand based on environmental characteristics, and does not require full-line deployment, thus avoiding the technical limitations of UWB in long-distance, unobstructed areas. This invention utilizes the high-precision advantage of UWB, and the main tasks involved do not require continuous UWB trajectory acquisition, but only need to provide accurate reference points in "key node areas." Therefore, the short-range characteristics of UWB do not constitute a limitation, but rather improve the accuracy control capability.

[0083] While UWB does present inconveniences in long-distance deployment along railway lines, this invention does not rely on UWB for dominant positioning across the entire area. The invention leverages the accuracy advantages of UWB in key areas through dynamic weighting, signal state awareness, trajectory correction, and pseudo-trajectory elimination. It is applicable to integrated deployment, discontinuous deployment, and cost-effective deployment schemes, possessing significant engineering practical value. Deploying UWB base stations in key areas such as locomotive depots, maintenance workshops, station entrances, and tunnel entrances can solve the pain points of traditional BeiDou trajectory changes and misjudgments, significantly enhancing the safety of rail transit. This invention possesses a complete engineering foundation, clear application scenarios, and ample practicality in rail transit positioning systems. Even with the physical limitations of UWB in current technologies, it still demonstrates a high degree of system integration capability and practical deployment feasibility.

[0084] Acceleration-guided interpolation refers to the process of analyzing the acceleration trends of forward trajectory points during backup trajectory compensation or trajectory correction to deduce the estimated position of the current trajectory point, thus achieving continuous trajectory interpolation. This method is suitable for train operation scenarios where there are obvious acceleration and deceleration patterns between trajectory points, and has strong temporal consistency and physical logic. Specifically, firstly, in the backup trajectory generation stage, five consecutive historical trajectory points preceding the current trajectory point are extracted, the velocity change values ​​between adjacent trajectory points are calculated, and then the acceleration value sequence between the three points is calculated. Secondly, regression analysis is performed on this acceleration sequence to obtain an acceleration trend model, such as determining whether the acceleration is continuously increasing, decreasing, or exhibiting oscillations. Thirdly, based on this acceleration trend and the time interval between trajectory points, the position offset value at the next moment is predicted, i.e., the interpolation result of the current trajectory point is generated. Finally, this interpolation result is written into the trajectory sequence as a backup trajectory point and participates in subsequent trajectory evaluation and updates. This method, while ensuring the physical continuity of the trajectory, introduces a matching mechanism between the acceleration direction and the velocity trend, which can effectively avoid the path abrupt change problem caused by simple spatial position interpolation. It is particularly suitable for typical sections with significant acceleration or deceleration behavior during train operation, such as station entry, station exit, and slope changes.

[0085] The curvature-smoothing spatial interpolation method refers to using the spatial curvature change of the trajectory as the interpolation control basis during the trajectory interpolation process. By analyzing the curvature between trajectory segments, it achieves spatial smooth completion of trajectory points. This method is suitable for scenarios where trajectory points have a nonlinear spatial distribution or transitional turns, effectively ensuring the geometric continuity and turning rationality of the trajectory. The specific implementation of this method includes the following steps: First, extract the spatial coordinate information of the current trajectory point and its four consecutive preceding trajectory points, and map these points to a unified spatiotemporal grid coordinate system. Second, calculate the turning angle between each pair of adjacent trajectory points, and further derive the curvature value sequence. Third, use a sliding window method to perform first-order smoothing on the curvature sequence to eliminate abrupt interference and identify the growth or decay trend of continuous curvature change segments. Fourth, construct a spatial circular arc fitting model based on this trend, and determine the fitting position of the current trajectory point on the spatial curve based on this model, thereby generating the interpolation result of the backup trajectory point. The trajectory generated by this method not only has high numerical continuity but also excellent curve consistency in spatial geometry, making it particularly suitable for complex path structures such as turning sections, tunnel exit sections, and merging lane sections.

[0086] This invention involves multiple thresholds with different functional orientations, such as a stability threshold, an error threshold, a preset first threshold corresponding to the fusion offset index, a preset second threshold corresponding to the fusion offset index, a dominant signal switching frequency threshold, and a minimum tolerance threshold. To ensure the robustness and generalization ability of the method during execution, the setting of these thresholds follows these principles: Statistical distribution analysis principle: First, statistical modeling is performed on a large amount of actual train trajectory data to extract features such as trajectory continuity indicators, signal fusion weight change indicators, and speed and acceleration distributions under normal operating conditions, constructing a multivariate statistical distribution model. Subsequently, based on the quantile ranges of various indicators, reasonable upper and lower bounds are determined, and the distribution boundary within a 90% confidence interval is selected as the initial setting basis for "stability thresholds," etc. For example, if the fluctuation amplitude of the signal fusion weight does not exceed 0.45% in 90% of the samples, then 0.5% is set as the stability threshold. Functional sensitivity verification principle: Through a large number of scenario playback simulations and real trajectory injection tests, the sensitivity of different threshold settings to trajectory output is verified. For example, in the backup trajectory compensation triggering conditions, the number of times the dominant signal is switched is set to two, three, and four for comparison. Finally, the three times with the lowest misjudgment rate and the highest triggering accuracy are selected as the final setting. Similarly, the error threshold is also set based on the boundary of the influence of trajectory deviation on train control accuracy, and is generally taken as half of the maximum acceptable deviation of the positioning system as the warning value.

[0087] The above algorithms or formulas are all dimensionless and numerical calculations, and the results are obtained by software simulation based on a large amount of collected data to obtain the most recent real-world results. The preset parameters are set by those skilled in the art according to the actual situation.

[0088] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0089] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0090] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0091] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology 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 dual-mode (BeiDou and UWB) collaborative positioning method for rail transit, characterized in that, Includes the following steps: Based on the spatial coverage status of BeiDou positioning signals and ultra-wideband positioning signals in the train operation path, the signal stability value, historical drift trend parameters and trajectory change amplitude are obtained, and a cross-modal distortion discrimination array is generated. The cross-modal distortion discrimination array identifies whether the current position contains a trajectory change region. If the identification result is that there is a trajectory change, the trajectory curvature change and velocity vector difference between the current trajectory point and three consecutive historical trajectory points are calculated, and the fusion weight ratio of Beidou positioning signal and ultra-wideband positioning signal is adjusted according to the calculation result. After the fusion weight adjustment is completed, the trajectory mutation point is corrected by the constraint backtracking trajectory fitting algorithm. The offset value of the current trajectory point is calculated by the historical trajectory fitting curve. The spatial topology of the ultra-wideband positioning base station is combined to determine whether the trajectory point is located at the pseudo trajectory source location. If the offset value exceeds the error threshold, or the trajectory point is located at the pseudo trajectory source location, the trajectory point is removed from the fusion trajectory sequence. Perform time series analysis on the changes in signal fusion weights over n consecutive seconds during trajectory fusion. If the analysis results show a frequent trend of signal channel jumps, then lower the corresponding fusion weights to the minimum tolerance threshold. The steps for constructing a cross-modal distortion discriminant array include: Based on the geographical segment type of the train's current location in the running path, the spatial coverage status, signal strength fluctuation range and historical signal obstruction records of BeiDou positioning signal and ultra-wideband positioning signal in the segment are extracted to form a multi-source spatial sensing dataset. Based on the multi-source spatial sensing dataset, the signal-to-noise ratio stability value, historical signal drift trend slope, and trajectory change acceleration curve of the BeiDou signal and the ultra-wideband signal at this location are calculated respectively, and the corresponding signal state vector is constructed. The signal state vector is mapped to a unified spatiotemporal grid coordinate system, and fused according to the correlation of signal changes and the risk level of trajectory continuity to generate a cross-modal trajectory consistency distortion discrimination array. The steps to identify whether the current location contains a region of abrupt change in trajectory include: Extract the grid cell corresponding to the current train position from the cross-modal distortion discrimination array, and read the label level of the grid cell in the trajectory continuity risk level label group; The label level of the grid cell is compared with the label levels of its eight neighboring grid cells. If there are three or more neighboring cells with high risk levels, the current grid position is determined as the central candidate area of ​​the trajectory change region. Perform signal trend consistency detection on the spatial boundary of the central candidate area. If the detection result shows that there is a sign flip or amplitude change between the current and adjacent time series of the signal stability difference value, the current position is finally confirmed to contain a trajectory change region. Correcting trajectory abrupt change points using a constrained backtracking trajectory fitting algorithm includes the following steps: Extract the fused positioning data of the current trajectory point, including the trajectory mutation point, and its four consecutive historical trajectory points, construct the trajectory evolution window, and mark the mutation point location as the backtracking starting point; The velocity change magnitude, spatial curvature continuity, and signal fusion confidence trend are calculated in chronological order for historical trajectory points in the trajectory evolution window, and fitting constraint boundaries are constructed based on these, including trajectory continuity preservation boundaries, fusion signal trend preservation boundaries, and maximum offset suppression boundaries. Under the premise of satisfying all fitting constraints, the coordinates of the abrupt change point position are adjusted for the first time by using the acceleration change trend before and after the trajectory. If the correction result that meets the offset tolerance cannot be generated under all constraints, the compensatory backtracking correction operation is performed, that is, the local trajectory segment is reconstructed by inserting intermediate transition points, and the reconstruction result is written into the fused trajectory sequence.

2. The rail transit positioning method using BeiDou and UWB dual-mode coordination according to claim 1, characterized in that, Fusion based on the correlation of signal changes and the risk level of trajectory continuity refers to: Based on the signal state vector in a unified spatiotemporal grid coordinate system, the stability difference, drift direction consistency and curvature change rate between BeiDou positioning signal and ultra-wideband positioning signal at adjacent trajectory points are calculated, and a signal variation correlation factor matrix is ​​generated, with each matrix unit corresponding to a grid region. The change amplitude of each unit in the signal change correlation factor matrix is ​​compared with the preset continuity threshold range. If the change amplitude is in the corresponding low, medium or high range, the corresponding low risk level label, medium risk level label or high risk level label is assigned respectively, forming a trajectory continuity risk level label group corresponding to the grid area. The matrix units of the signal variation correlation factor matrix and the trajectory continuity risk level label group are linked and integrated according to the spatial correspondence of the grid area. The matrix units with medium risk level labels or high risk level labels are recorded as distortion-sensitive units, and a cross-modal distortion discrimination array is generated accordingly.

3. The rail transit positioning method using BeiDou and UWB dual-mode coordination according to claim 2, characterized in that, When the identification result indicates a sudden change in trajectory, perform the following operations: Extract the position data of the current trajectory point and its three previous historical trajectory points in a unified spatiotemporal grid coordinate system, calculate the curvature change and velocity vector difference between each pair of adjacent trajectory points, and form a trajectory perturbation feature group based on the change trend between the four points. The maximum curvature change and the maximum velocity vector difference in the trajectory disturbance feature group are normalized to obtain the curvature disturbance index and the velocity disturbance index. Based on the cumulative change frequency of the two disturbance indices in the past three trajectory points, they are assigned change level scores respectively. The two scores are weighted and summed according to a preset ratio coefficient to generate a fusion offset coefficient; By comparing the difference between the current fusion offset coefficient and the standard reference value, the fusion weight gain that the current trajectory point should be adjusted towards the BeiDou positioning signal or the ultra-wideband positioning signal is determined, and the fusion weight ratio is updated.

4. The rail transit positioning method using BeiDou and UWB dual-mode coordination according to claim 3, characterized in that, The fusion weight update for the current trajectory point is performed according to the following process: If the fusion offset index is less than the preset first threshold, the existing fusion weight ratio remains unchanged; If the fusion offset index is between the preset first threshold and the preset second threshold, the current fusion weight ratio will be linearly offset towards the dominant signal direction by a preset ratio constant. If the fusion offset index is greater than the preset second threshold, the fusion weight gain is calculated according to the following nonlinear formula: subtract the preset second threshold from the current fusion offset index, multiply by the set nonlinear adjustment coefficient, and use it as the fusion weight gain value of the current trajectory point. The weight gain value is added to the current weight value of the current dominant signal, and then normalized by summation to generate an updated fusion weight ratio.

5. A rail transit positioning method using BeiDou and UWB dual-mode coordination as described in claim 4, characterized in that, Determining whether a trajectory point is located at the pseudo-trajectory source location based on the spatial topology of the ultra-wideband positioning base station refers to: Obtain all the ultra-wideband positioning base station numbers and their fixed spatial coordinate information used within the preset time window of the current trajectory point, calculate the Euclidean spatial distance from the current trajectory point to each base station, and draw connecting edges to each base station with the current point as the center node to form a polygonal star topology graph centered on the trajectory point. Using the normal coverage radius of each base station as the boundary, multiple circular coverage areas are drawn around the central trajectory point, and the overlapping areas of each coverage area are marked as cross buffer zones; if the current trajectory point is in the cross area of ​​three or more coverage areas, the trajectory point is determined to be located at the pseudo trajectory source location.

6. The rail transit positioning method using BeiDou and UWB dual-mode coordination according to claim 5, characterized in that, Performing time series analysis involves the following steps: Extract the fusion weight sequence of the current trajectory point and several trajectory points before and after it, and calculate the number of switching times of the dominant signal and the weight fluctuation amplitude of the BeiDou positioning signal and the ultra-wideband positioning signal in the sequence respectively. If any signal channel experiences three or more dominant signal switchings within the same time window, or if the weight fluctuation exceeds the set stability threshold, the fusion weight of that signal channel will be reduced to the minimum tolerance threshold.

7. A rail transit positioning method using BeiDou and UWB dual-mode coordination as described in claim 6, characterized in that, If both the BeiDou positioning signal and the ultra-wideband positioning signal meet any of the following conditions within the same continuous time period: three or more dominant signal switchings occur, or the weight fluctuation amplitude simultaneously exceeds the set stability threshold, then the backup trajectory compensation strategy will be activated.

Images