A cross-layer mislocalization tunnel topology correction method and system

Abstract

The application relates to the technical field of underground personnel positioning, in particular to a roadway topology correction method and system for cross-layer mispositioning. The method comprises the following steps: acquiring a multi-source prior information set of an underground roadway space, performing multi-dimensional modeling and passable area calibration on the roadway space based on the multi-source prior information set, and generating an underground operation three-dimensional topology graph containing geometric contours and topological connectivity; acquiring a preliminary positioning result and human body motion inertia data at a current time, performing current intra-layer trajectory anomaly detection and cross-layer mispositioning detection in combination with the underground operation three-dimensional topology graph, and generating a joint determination information set; based on the joint determination information set, performing differential calculation and correction for the current intra-layer mispositioning and cross-layer mispositioning, and generating a set of corrected reliable positioning results. The application can effectively solve the positioning error problem caused by signal layering or topological ambiguity, and improve the accuracy and reliability of the positioning result.

Description

Technical Field

[0001] This application relates to the field of underground personnel positioning technology, and in particular to a method and system for correcting cross-layer mispositioning in tunnel topology. Background Technology

[0002] Underground construction environments such as mines and tunnels are complex, with tunnels typically distributed in multiple layers, connected by ramps, connecting tunnels, and elevator shafts. Existing personnel positioning systems deployed underground primarily rely on radio ranging technology to obtain miners' spatial coordinates, supplemented by inertial sensor data for auxiliary calculations. In practice, the system first collects basic geometric data of the tunnels to construct a spatial model, then receives signals from positioning tags in real time to calculate the current position, and uses sensors such as accelerometers and gyroscopes to monitor the movement of personnel. When a jump in position coordinates or an abnormal movement trajectory is detected, existing technologies typically smooth or correct the positioning results based on preset distance thresholds or simple geometric rules to generate a continuous movement trajectory for monitoring by the dispatch center.

[0003] However, in the existing technology, facing the unique multi-layered three-dimensional tunnel structure of underground construction operations, due to the overlapping coverage of signals in the vertical direction and the dense distribution of tunnel centerlines, the positioning system has difficulty in accurately distinguishing whether a person is on the current floor or an adjacent floor, and it is also difficult to identify complex branch paths. This results in the positioning results often showing phenomena such as passing through walls, jumping floors, or deviating significantly from the actual movement trajectory, making the final output positioning information unreliable and unable to meet the requirements of underground safety monitoring for high-precision continuous trajectory. Summary of the Invention

[0004] This application provides a method and system for correcting roadway topology for cross-layer mislocation, in order to solve the above-mentioned problems.

[0005] In a first aspect, this application provides a method for correcting roadway topology for cross-layer mislocation, the method comprising: A multi-source prior information set of underground tunnel space is obtained. Based on the multi-source prior information set, multi-dimensional modeling of tunnel space and calibration of passable areas are performed to generate a three-dimensional topological map of underground operations that includes geometric contours and topological connectivity. Obtain the preliminary positioning results and human motion inertial data at the current moment, and combine them with the three-dimensional topology map of the underground operation to perform trajectory anomaly detection and cross-layer mispositioning detection within the current layer, and generate a joint judgment information set containing the suspected positioning status and mispositioning type; Based on the joint determination information set, differential calculation and correction are performed for mislocation within the current layer and mislocation across layers to generate a set of corrected reliable positioning results.

[0006] Through the above technical solutions, a structured representation of physical space is achieved, effectively avoiding unreasonable positioning phenomena such as wall penetration and floor skipping. Combining human motion inertial data enhances the understanding of people's true movement intentions, solving the problem that a single sensor cannot distinguish between walking on a slope and elevator movement. Furthermore, different projection strategies are adopted according to the type of mispositioning: intra-layer mispositioning projects along the topological direction to the centerline, while cross-layer mispositioning re-matches the target layer node, achieving targeted and accurate correction. Moreover, the piecewise linear interpolation method included in the traversable area calibration handles the gradual change in tunnel width, ensuring a close fit between the initial symmetrical boundary and the actual wall contour, further improving calibration accuracy and the robustness of the correction effect. Finally, through the synergistic effect of the above technical features, this application can support real-time online correction with low computational load, allowing deployment on embedded positioning terminals or edge computing nodes, meeting underground explosion-proof and power consumption requirements.

[0007] Optionally, the process of generating the three-dimensional topological map of the underground operation includes: The multi-source prior information set includes the roadway centerline coordinate sequence, roadway width information, and roadway spatial adjacency relationships; Based on the coordinate sequence of the lane centerline and the spatial adjacency relationship of the lane, multi-dimensional modeling of the lane space is performed to generate a lane network skeleton with topological connectivity. Based on the tunnel width information, passable areas are marked, and inaccessible areas occupied by walls, equipment, and support columns are eliminated. The tunnel network skeleton and passable areas are then integrated into the three-dimensional topology map of the underground operation.

[0008] The above technical solution enables the automated construction of high-precision 3D topological maps from discrete mapping data. By combining the coordinate sequence of the tunnel centerline with the spatial adjacency of the tunnels, a network skeleton reflecting the underground connectivity logic is constructed, solving the logic verification problem caused by the lack of topological relationships in traditional methods. Based on this, tunnel width information is introduced and piecewise linear interpolation is performed, allowing the generated passable areas to adapt to the gradual changes in tunnel width, improving the model's fitting accuracy to complex underground environments. Furthermore, Boolean difference operations are used to eliminate fixed obstacles and smooth the boundaries, eliminating false passable areas caused by equipment occupancy or sharp modeling corners, ensuring the authenticity of passable areas. This deep integration of geometric contours and topological connectivity provides mandatory spatial constraints for subsequent cross-layer mislocation detection, enabling the system to accurately identify abnormal positioning points that, although their coordinate values ​​are reasonable, violate topological connectivity or fall into obstacle areas, thereby significantly reducing the mislocation rate caused by signal crosstalk or geometric ambiguity.

[0009] Optionally, the multi-dimensional modeling of the tunnel space includes: Based on the coordinate sequence of the roadway centerline, each roadway is abstracted into a directional polyline segment with directional attributes, and the three-dimensional coordinates of the polyline on each directional polyline segment are recorded. Based on the spatial adjacency relationship of the tunnels, the three-dimensional coordinates of the polylines belonging to the same branch or inter-layer connection are topologically linked to form an undirected connected graph; Each side of the undirected connected graph corresponds to a complete road segment that is actually passable, and the length and orientation characteristics of the complete road segment are stored.

[0010] The above technical solution achieves a precise mapping of tunnel space from geometric discretization to topological networking. Specifically, by abstracting tunnels into directional polyline segments with directional attributes, the spatial morphological details of the tunnels are preserved, providing a data foundation for high-precision geometric matching. Based on this, topological links are established according to the spatial adjacency relationships of the tunnels, integrating scattered geometric units into a globally connected undirected graph, solving the problem of disconnected or logically isolated tunnels in traditional methods. Furthermore, by storing length and direction features on the edges of the undirected connected graph, rich semantic information is endowed to the topological structure, enabling subsequent topological reachability checks not only to determine whether it is reachable but also to determine how to reach it and whether the path is reasonable by combining kinematic features. This modeling approach, which deeply integrates geometry and topology, improves the robustness of positioning correction in complex, multi-layered underground environments and avoids mispositioning and missed detections caused by missing topological structures.

[0011] Optionally, the process of marking accessible areas includes: Using the tunnel width information as a half-width offset, the tunnel centerline coordinate sequence is shifted point by point to the wall position to generate an initial symmetrical boundary. Based on the initial symmetrical boundary, the area occupied by the fixed obstacle is subjected to Boolean difference operation from inside the initial symmetrical boundary to eliminate the traversable area, and the sharp corners of the boundary left after elimination are rounded to form a passable polygonal area. For cases where the width of the tunnel gradually changes along the centerline, a piecewise linear interpolation method is used to calculate different half-width offsets point by point based on the tunnel width information, so that the initial symmetrical boundary closely fits the actual wall contour.

[0012] The above technical solution enables the automated construction of high-precision passable areas. By generating an initial symmetrical boundary using the tunnel width information as a half-width offset, the basic geometric shape of the tunnel space is established. Based on this, Boolean difference operations are used to remove areas occupied by fixed obstacles, achieving precise trimming of non-passable spaces and ensuring the safety of the positioning results. Furthermore, by rounding the sharp corners of the remaining boundaries after removal, the interference of geometrical abrupt changes on subsequent algorithms is eliminated, improving the smoothness and usability of the boundaries. Simultaneously, for complex scenarios with gradually changing tunnel widths, a piecewise linear interpolation method is introduced to dynamically calculate different half-width offsets, allowing the initial symmetrical boundary to adaptively fit the real wall contour, solving the problem of large errors in traditional fixed-width modeling in variable-width tunnels. These interconnected steps jointly construct a high-precision passable polygonal area that conforms to the tunnel design geometry while also taking into account the actual distribution and morphological changes of obstacles underground, providing reliable spatial data support for subsequent topology-constrained positioning correction.

[0013] Optionally, the process of generating the joint determination information set includes: The preliminary positioning results and the human motion inertial data are obtained to perform trajectory anomaly detection within the current layer and cross-layer mispositioning detection. The current layer trajectory anomaly detection outputs the current layer matching confidence level by analyzing the degree of matching between the preliminary positioning trajectory and the roadway geometric contour. The cross-layer mislocation detection outputs the cross-layer suspicion level through topology reachability review and motion pattern-driven inter-layer transition detection. The joint judgment information set is generated based on the current layer matching credibility and the cross-layer suspicion level.

[0014] The above technical solution enables comprehensive perception and accurate diagnosis of complex underground positioning environments. By acquiring preliminary positioning results and human motion inertial data, a dynamic and static data input foundation is constructed. Based on this, the positioning trajectory is matched in real-time with the roadway geometry using current-layer trajectory anomaly detection, effectively filtering out coordinate jitter within the layer caused by signal reflection. Simultaneously, through topological reachability review and motion pattern-driven analysis in cross-layer mispositioning detection, cross-layer jumps are rigorously verified from both logical connectivity and physical motion laws. The joint generation of current-layer matching credibility and cross-layer suspicion level allows the system not only to detect positioning errors but also to accurately distinguish whether the error type stems from local geometric mismatch or global hierarchical disorder. This mechanism, which combines geometric constraints and topological logic, and mutually verifies static position and dynamic behavior, improves the accuracy and robustness of mispositioning detection, providing data support for solving positioning failures caused by signal crosstalk and topological ambiguity.

[0015] Optionally, the trajectory anomaly detection within the current layer includes: Extract the local tunnel geometric contour centered on the preliminary positioning result, and extract the motion trajectory segment within a preset time window from the historical trajectory sequence; By analyzing whether the continuous change direction of the motion trajectory segment is consistent with the direction of the roadway in the local roadway geometric contour; If the movement trajectory segment makes a sharp turn at a large angle that does not conform to the direction of the roadway, or if the trajectory point falls outside the passable area, it is determined that the trajectory within the current layer is abnormal.

[0016] By combining the above technical solutions with the dual mechanisms of direction change detection and space boundary crossing detection, it is possible to comprehensively cover wall-penetrating errors caused by signal multipath reflection and reverse errors caused by topological ambiguity. This ensures that only positioning results that conform to the laws of physical motion and are located within the legal space can pass the verification, thereby improving the reliability of a single positioning moment.

[0017] Optionally, the topology reachability review includes: Map the preliminary positioning results at the current moment onto the three-dimensional topology map of the underground operation to determine its candidate node; Retrieve the verified historical motion nodes from the previous moment and check whether there is a continuous path from the historical motion nodes along the walkable edges in the topology graph that can reach the candidate nodes. If no continuous path exists, output the topology unreachable result; If a continuous path exists, calculate the path length and output the topology reachability result.

[0018] The above technical solution enables deep logical verification of positioning results. By mapping preliminary positioning results to candidate nodes and performing topological path search with historical movement nodes, a rigorous spatiotemporal consistency constraint is constructed. If a path does not exist, it is directly determined to be unreachable, quickly intercepting cross-layer or wall-penetrating errors. If a path exists, the path length is further calculated, providing data support for dual verification using inertial data. This graph-based reachability verification mechanism solves the deficiency of existing technologies that rely solely on geometric distance matching and cannot identify logical discontinuities, improving the anti-interference capability and trajectory continuity of personnel positioning systems in multi-level tunnel environments.

[0019] Optionally, the motion pattern-driven interlayer transition detection includes: Analyze the posture change characteristics and air pressure change characteristics in the aforementioned human motion inertial data; Monitor whether the pitch angle of the miner shows a continuous and regular upward or downward trend within a continuous time window, and whether the air pressure value shows a stable rise or fall or a step change consistent with the change in altitude. If both indicate the existence of continuous ramp motion or elevator-like up-and-down motion, then the output inter-level transition confidence level is that the inter-level transition has definitely occurred. If only the pitch angle trend changes significantly while the air pressure does not change significantly, or if only the air pressure changes significantly while the pitch angle does not change significantly, it is determined that inter-layer transition may have occurred. If neither changes significantly, it is determined that no interlayer transition has occurred.

[0020] Through the above technical solutions, a highly robust inter-level transition detection mechanism was constructed. The pitch angle feature utilizes human kinematics to accurately capture the body's tilt state during walking on a slope, solving the problem of insufficient resolution of a single barometer under slow altitude changes. The air pressure change feature utilizes atmospheric physics principles to accurately capture abrupt altitude changes during elevator ascent and descent, resolving the ambiguity of a single attitude sensor failing to distinguish between horizontal turns and vertical ascent and descent. The combined use of these two features forms a complementary verification loop: only when the physical attitude and altitude change are logically consistent is it confirmed as a genuine inter-level transition. This dual-feature fusion strategy reduces the misjudgment rate caused by sensor noise, local airflow disturbances, or atypical miner movements, ensuring accurate differentiation between genuine inter-level transitions and signal artifacts in complex underground multi-layered environments. This provides a decision-making basis for subsequent differential calculations, thereby avoiding unreasonable jumps in positioning trajectories and improving the overall reliability of the personnel positioning system.

[0021] Optionally, the process of generating the trusted location result set includes: Based on the mislocation type in the joint determination information set, perform differential resolution; If the mislocation type is mislocation within the current layer, the preliminary positioning result is projected along the topologically accessible direction onto the passable polygon area and constrained to the nearest lane centerline coordinate sequence; If the mislocation type is the cross-layer mislocation, then based on the inter-layer conversion detection result, the location result is re-matched to the topology node corresponding to the target layer and the passable area; The corrected positioning results and timestamps are output to form the trusted positioning result set.

[0022] The above technical solutions achieve precise repair for different mislocation mechanisms. Through classification processing based on a joint decision information set, intra-layer mislocation utilizes geometric projection to revert to the centerline, while cross-layer mislocation utilizes topological rematching to revert to the correct layer. These two methods work together to solve both the multipath interference problem in the plane and the signal crosstalk problem in three-dimensional space. By leveraging the dual constraints of topological reachability verification and motion pattern detection, it is ensured that the corrected trajectory geometrically conforms to the tunnel contour, logically conforms to connectivity, and physically matches human motion characteristics. Ultimately, the output set of reliable positioning results reduces positioning errors and improves the robustness and usability of the personnel positioning system in complex underground environments.

[0023] Secondly, this application provides a roadway topology correction system for cross-layer mislocation, the system comprising: The topology map generation module is used to acquire a multi-source prior information set of underground tunnel space, and based on the multi-source prior information set, to perform multi-dimensional modeling of tunnel space and identification of passable areas, and generate a three-dimensional topology map of underground operations that includes geometric contours and topological connectivity. The cross-layer determination module is used to obtain the preliminary positioning results and human motion inertial data at the current moment, and combine them with the three-dimensional topology map of the underground operation to perform trajectory anomaly detection and cross-layer mispositioning detection within the current layer, and generate a joint determination information set containing the suspected positioning status and mispositioning type. The feasible positioning module is used to perform differential calculation and correction on intra-layer mispositioning and cross-layer mispositioning based on the joint judgment information set, and generate a set of corrected reliable positioning results. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a schematic diagram illustrating an application scenario provided in one embodiment of this application; Figure 2 A flowchart illustrating a method for correcting cross-layer mislocation in a roadway, as provided in an embodiment of this application; Figure 3 This is a schematic diagram of a roadway topology correction system for cross-layer mislocation provided in an embodiment of this application. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0027] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0028] The embodiments of this application will now be described in further detail with reference to the accompanying drawings.

[0029] Figure 1 This application provides an illustration of an application scenario. In the process of cross-layer mispositioning correction, the method provided in this application can effectively solve the positioning error problem caused by signal cross-layering or topological ambiguity, and improve the accuracy and reliability of the positioning results.

[0030] Specifically, the method provided in this application can be applied to any server, where the server interacts with the underground personnel precise positioning system to obtain the multi-source prior information set, preliminary positioning results and human motion inertial data provided by the underground personnel precise positioning system, and finally outputs the reliable positioning result set to relevant personnel.

[0031] The specific implementation method can be referred to in the following embodiments, wherein the data mentioned in the embodiments are only for reference and examples, so that relevant personnel can better understand them.

[0032] Figure 2 This is a flowchart illustrating a method for correcting cross-layer mislocation in a lane topology according to an embodiment of this application. The method of this embodiment can be applied to the server in the above scenario. (See also...) Figure 2 The process, specifically the implementation steps, are as follows: Example 1: Underground working environments are complex, with tunnels typically distributed in multiple layers, connected by ramps, connecting tunnels, and elevator shafts. Existing personnel positioning systems deployed underground often face the problem of signal cross-layer mispositioning. This means that UWB or Wi-Fi signal coverage overlaps vertically, and when a miner is on a particular tunnel level, the positioning system may incorrectly map the positioning result to an adjacent level, causing the positioning trajectory to cross walls or skip levels. Simultaneously, due to the dense centerlines of underground tunnels, when positioning nodes are near intersections, distance-based positioning algorithms struggle to distinguish which branch or level the miner is currently on, leading to intra-level mispositioning caused by geometric topological ambiguity. Furthermore, when relying solely on inertial navigation for positioning, sensor drift is a concern, making it impossible to reliably determine the precise timing of inter-level transitions. Existing methods often lack topological constraints and fail to fully utilize the tunnel topological connectivity for logical verification, resulting in low detection rates and poor correction effects for mispositioning.

[0033] To address the aforementioned problems, this application provides a method for correcting roadway topology for cross-layer mislocation, the method comprising: Step 1: Obtain a multi-source prior information set of underground tunnel space. Based on the multi-source prior information set, perform multi-dimensional modeling of tunnel space and labeling of passable areas to generate a three-dimensional topological map of underground operations that includes geometric contours and topological connectivity. The multi-source prior information set refers to the basic data set used to describe the static spatial characteristics of underground roadways, specifically including the roadway centerline coordinate sequence, roadway width information, and roadway spatial adjacency relationships. The roadway centerline coordinate sequence is obtained through underground surveying equipment (such as a total station or 3D laser scanner) or by vectorizing pre-stored underground operation CAD design drawings to obtain 3D coordinate data of turning points. Roadway width information comes from mine design documents and actual on-site measurements; for example, a transport roadway is designed to be 4.5 meters wide, while some connecting roadways are 2.0 meters wide. Roadway spatial adjacency relationships are formed by an adjacency matrix based on the connection relationships between different roadways (such as junctions and inter-layer connections) marked on the underground operation mining design plan. The underground operation 3D topology map is a digital model generated based on the above multi-source prior information set through two core processing steps: multi-dimensional roadway spatial modeling and traversable area calibration. This model not only possesses topological connectivity but also integrates traversable areas after removing inaccessible areas occupied by walls, equipment, and support pillars, thus forming a high-precision map with geometric contour constraints.

[0034] Specifically, in the multi-dimensional modeling of the tunnel space, each tunnel is abstracted into a directional polyline segment with directional attributes, based on the coordinate sequence of the tunnel centerline, and the three-dimensional coordinates of the polyline on each directional polyline segment are recorded. Based on the spatial adjacency of the tunnels, the three-dimensional coordinates of polylines belonging to the same junction or inter-layer connection are topologically linked to form an undirected connected graph. Each side of the graph corresponds to a complete, actually passable tunnel segment, and the length and orientation characteristics of that segment are stored. When calibrating the passable area, the tunnel width information is used as the half-width offset, and the offset is made point-by-point along both sides of the tunnel centerline coordinate sequence to the wall position, generating an initial symmetrical boundary. For cases where the tunnel width gradually changes along the centerline, a piecewise linear interpolation method is used to calculate different half-width offsets point-by-point based on the tunnel width information, ensuring that the initial symmetrical boundary closely matches the actual wall contour. For example, if a section of a tunnel has a centerline length of 30 meters and a width that gradually changes from 4.0 meters to 3.5 meters, then a half-width offset is calculated every meter along the centerline (2.0 meters for the 1st meter, 1.9 meters for the 15th meter, and 1.75 meters for the 30th meter), generating a continuously varying boundary along the centerline. Subsequently, based on the initial symmetrical boundary, Boolean difference operations are performed on the areas occupied by fixed obstacles (such as support pillars, water pumps, and ventilation doors) within the initial symmetrical boundary to eliminate impassable areas. The sharp corners of the remaining boundary after elimination are then rounded off, ultimately forming a passable polygonal area. This combination of multi-dimensional modeling and region calibration allows for the construction of a three-dimensional topology map that reflects both the tunnel connectivity logic and physical access constraints, providing a mandatory spatial constraint basis for subsequent positioning and correction.

[0035] Step 2: Obtain the preliminary positioning results and human motion inertial data at the current moment, and combine them with the three-dimensional topology map of the underground operation to perform trajectory anomaly detection within the current layer and cross-layer mispositioning detection, and generate a joint judgment information set containing the suspected positioning status and mispositioning type; The preliminary positioning results are obtained in real time from the spatial coordinates of the tag worn by the miner, calculated by underground radio positioning base stations (such as UWB base stations supporting TDOA or TOF ranging). The ranging accuracy is usually 10-30 cm, but it is easily affected by signal crosstalk in multi-layered roadways. The human motion inertial data is collected by the explosion-proof inertial measurement unit (IMU) worn by the miner, including data from a three-axis accelerometer, a three-axis gyroscope, and a barometer. The sampling frequency is recommended to be no less than 50 Hz, and the barometer resolution can reach 0.01 hPa. The joint judgment information set is a comprehensive judgment result output after parallel execution of trajectory anomaly detection within the current layer and cross-layer mispositioning detection. It includes the positioning suspicion status (such as normal, suspicious, abnormal) and the mispositioning type (such as intra-layer deviation, cross-layer jump).

[0036] Specifically, the current layer trajectory anomaly detection analyzes the degree of matching between the preliminary positioning trajectory and the roadway geometry to output the current layer's matching confidence level. This process extracts the local roadway geometry centered on the preliminary positioning result and extracts motion trajectory segments within a preset time window from historical trajectory sequences. It analyzes whether the continuous change direction of the motion trajectory segments is consistent with the roadway orientation in the local roadway geometry. If the motion trajectory segment shows a sharp turn at a large angle that does not conform to the roadway orientation, or if the trajectory point falls outside the passable area, it is determined as a trajectory anomaly within the current layer. For example, if a miner is walking in a north-south oriented transport roadway, and the positioning result shows that they suddenly move 90 degrees westward within 5 seconds with no forks in the roadway, or if the positioning point falls within the equipment marking area, an anomaly marker is triggered. Cross-layer mispositioning detection outputs the cross-layer suspicion level through topology accessibility review and motion pattern-driven inter-layer transition detection. Topology reachability review maps the preliminary positioning results at the current moment onto the 3D topology map of the underground operation to identify candidate nodes. It retrieves verified historical movement nodes from the previous moment and checks if there is a continuous path leading to the candidate nodes. If no continuous path exists (e.g., jumping directly from the -300m level to the -400m level without inter-level edge connection), a topology unreachable result is output. Motion pattern-driven inter-level transition detection analyzes the posture and air pressure change characteristics in human motion inertial data. It monitors whether the miner's pitch angle shows a continuous and regular upward or downward trend within a continuous time window, and whether the air pressure value shows a stable rise or fall or a step change consistent with altitude changes. If both indicate continuous slope movement or elevator-like rise and fall, an inter-level transition is determined to have occurred; if only a single feature changes significantly, it is determined to have occurred; if neither changes, it is determined not to have occurred. For example, when a miner enters an elevator, a jump in the barometer reading within 3 seconds corresponds to an ascent of 42 meters with a stable pitch angle, indicating an elevator-like ascent or descent. Conversely, if walking on a slope, a continuous decrease in pitch angle and a slow drop in air pressure indicate slope-related movement. By fusing radio positioning data with IMU sensor data and employing a dual verification mechanism of topological logic and motion behavior characteristics, different types of positioning anomalies can be accurately identified, generating a highly reliable joint judgment information set.

[0037] Step 3: Based on the joint decision information set, perform differential calculation and correction for mislocation within the current layer and mislocation across layers, and generate a set of corrected reliable positioning results.

[0038] Differential recalculation refers to implementing targeted mathematical correction strategies based on the specific mislocation type identified in the joint judgment information set. If the mislocation type is intra-layer mislocation, it indicates that the positioning point deviates from the true path on the horizontal plane but the layer is correct. In this case, the preliminary positioning result is projected along the topologically accessible direction to the passable polygon area and constrained to the coordinate sequence of the nearest lane centerline. If the mislocation type is cross-layer mislocation, it indicates that the positioning point has made an erroneous jump in the vertical dimension. In this case, based on the inter-layer transformation detection result, the positioning result is re-matched to the topological node and passable area corresponding to the target layer. The reliable positioning result set is a combination of corrected positioning coordinates and the current timestamp, used for subsequent trajectory display and safety monitoring.

[0039] Specifically, for intra-layer mislocation, such as when the initial location result is determined to be outside the traversable area and topology accessibility indicates that the point should be near the centerline, the system calculates the distance from that point to the nearest inflection point on the roadway centerline and uses the coordinates of that inflection point as the corrected location result, thereby eliminating lateral deviations caused by signal multipath effects. For cross-layer mislocation, such as when topology accessibility review finds no topological connection between the layer where the initial location result is located and the layer where the historical movement node is located, and motion pattern detection does not capture inter-layer transitions, the system determines that a cross-layer mislocation has occurred and re-matches the location result to the layer where the historical movement node is located, projecting it along the boundary of the traversable area of ​​that layer. Through this differentiated correction strategy, the problem of a single correction method being ineffective for all types of mislocation is avoided, ensuring that both intra-layer geometric deviations and cross-layer logical jumps can be accurately corrected. This improves the reliability and continuity of the location results in complex underground environments and avoids unreasonable jumps in personnel trajectories.

[0040] Example 2: In one optional embodiment, this application provides a process for generating a three-dimensional topological map of underground operations, the method comprising: Step 1: The multi-source prior information set includes the roadway centerline coordinate sequence, roadway width information, and roadway spatial adjacency relationships; The multi-source prior information set refers to the basic data set used to construct the underground space digital model. Its sources mainly include measured data from underground total stations or 3D laser scanning, vectorized data from pre-stored underground operation CAD design drawings, and mine mining design documents. The roadway centerline coordinate sequence is a set of vertex coordinates with three-dimensional spatial attributes (x, y, z), used to abstractly express the geometric direction and spatial location of the roadway. Roadway width information can refer to the half-width or full-width value of the roadway cross-section; it can be a design standard value or a sequence of measured values ​​varying along the roadway length. Roadway spatial adjacency relationships are topological data describing the connection logic between different roadway segments, usually stored in the form of adjacency matrices or edge lists, clearly identifying the connectivity status of key nodes such as junctions and inter-layer connecting roadways. These three types of information together constitute the data foundation for subsequent modeling and calibration. The centerline coordinate sequence provides the skeletal support, the adjacency relationships establish the connectivity logic, and the width information provides the actual spatial boundary constraints of the model. By integrating the above multi-source information, a high-precision geometric and topological dual-constraint environment can be provided for subsequent mislocation detection.

[0041] Step 2: Based on the coordinate sequence of the tunnel centerline and the spatial adjacency relationship of the tunnel, perform multi-dimensional modeling of the tunnel space to generate a tunnel network skeleton with topological connectivity. Multidimensional modeling of tunnel space refers to the process of transforming discrete surveying data into a graph-theoretic structure with logical connections. Specifically, this process uses the tunnel centerline coordinate sequence as a basis, abstracting each actual tunnel segment into a directional polyline segment with directional attributes, and recording the three-dimensional coordinates of each polyline segment. Subsequently, based on the spatial adjacency relationships of the tunnels, the three-dimensional coordinates of polylines belonging to the same junction or inter-level connection are topologically linked to form an undirected connected graph. In this undirected connected graph, each edge corresponds to a complete, actually passable tunnel segment, and the length and orientation characteristics of that complete tunnel segment are stored. For example, when two tunnels intersect at a certain coordinate point, the system identifies that point as a shared node and establishes a connection between the corresponding polyline segments of the two tunnels at that node, thus ensuring that the generated tunnel network skeleton is logically continuous and connected. This modeling method not only preserves the geometric shape of the tunnels but, more importantly, explicitly expresses the accessibility between tunnels, providing a logical basis for subsequent judgments on whether wall penetration or floor skipping has occurred.

[0042] Step 3: Based on the tunnel width information, mark the passable areas, remove the inaccessible areas occupied by walls, equipment and support columns, and integrate the tunnel network skeleton and passable areas into a three-dimensional topology map of underground operations.

[0043] The identification of passable areas refers to the process of expanding the actual physical passable space based on the tunnel network framework using width information and eliminating obstacle interference. Specifically, firstly, using the tunnel width information as a half-width offset, the offset is made point by point along both sides of the tunnel centerline coordinate sequence to the wall position, generating an initial symmetrical boundary. Then, based on the initial symmetrical boundary, the area occupied by fixed obstacles (such as support columns, water pumps, air doors, etc.) is subjected to Boolean difference operation from inside the initial symmetrical boundary to eliminate traversable areas. The sharp corners of the remaining boundary after elimination are rounded to form a passable polygonal area. For cases where the tunnel width gradually changes along the centerline, a piecewise linear interpolation method is used to calculate different half-width offsets point by point according to the tunnel width information, so that the initial symmetrical boundary closely matches the actual wall contour. For example, if a transport tunnel is 30 meters long and its width gradually changes from 4.0 meters at the entrance to 3.5 meters at the exit, the system will calculate a dynamic half-width offset every 1 meter along the centerline (2.0 meters for the 1st meter, 1.875 meters for the 15th meter, and 1.75 meters for the 30th meter), thus generating a boundary profile that continuously changes with location, rather than a simple rectangular expansion. Finally, the passable polygonal areas, after obstacle removal and smoothing, are spatially integrated with the aforementioned tunnel network skeleton to obtain a three-dimensional topological map of the underground operation containing precise geometric contours and rigorous topological connectivity. This map not only defines the spatial range that miners can reach but also clarifies which areas are impassable due to obstacles, thus providing a precise spatial criterion for determining whether the initial positioning results fall into illegal areas in subsequent steps.

[0044] Example 3: In one possible implementation, this application embodiment provides a method for multi-dimensional modeling of tunnel space, the method comprising the following steps: Step 1: Based on the coordinate sequence of the tunnel centerline, abstract each tunnel into a directional polyline segment with directional attributes, and record the three-dimensional coordinates of the polyline on each directional polyline segment; In this context, a directional polyline segment refers to a set of consecutive straight line segments that discretize a continuous, curved, or straight tunnel path in the actual physical space. The directional attribute characterizes the permissible flow direction for miners within the tunnel or the geometric extension trend of the tunnel, typically defined by a vector pointing from the starting coordinates to the ending coordinates. The three-dimensional coordinates of the polyline include the X, Y, and Z components of each inflection point in the local coordinate system of the underground operation, where the Z component directly reflects the tunnel's depth or elevation. These coordinate data originate from underground total station survey data or vectorized underground CAD design drawings. For example, a 100-meter-long transport tunnel containing two bends can be abstracted into three directional polyline segments, recording the three-dimensional coordinates of its starting point, turning point, and ending point (e.g., P1(10,20,-300), P2(50,20,-300), P3(50,60,-300)), thus transforming the continuous tunnel morphology into a discrete geometric unit that can be stored and computed by a computer. This abstraction method can reduce the data complexity of subsequent topology calculations while preserving the key geometric features of the tunnel.

[0045] Step 2: Based on the spatial adjacency relationship of the tunnels, the three-dimensional coordinates of the polylines belonging to the same branch or inter-layer connection are topologically linked to form an undirected connected graph; The spatial adjacency relationship of lanes describes the connection logic between different lane segments, including the merging relationship of planar intersections and the connection relationship of vertical inter-level connecting lanes. Topological links refer to establishing logical connection edges between the endpoints of multiple directional polyline segments sharing the same physical space node (such as the center point of a fork or the interface of a connecting lane), transforming originally independent geometric segments into a connected graph structure. An undirected connected graph is a graph theory data structure where nodes represent the intersection or endpoints of lanes, edges represent lane segments connecting these nodes, and the default direction of edge travel is bidirectional (unless specifically marked as one-way). Specifically, when the endpoint coordinate distance of two directional polyline segments is detected to be less than a preset threshold (e.g., 0.5 meters), they are determined to belong to the same fork and merged into a single topological node. For example, at a three-way intersection at a -300-meter level, three lane polyline segments with different directions converge. The system maps the common endpoint of these three segments to the same graph node Node_K and establishes three edges between Node_K and its adjacent nodes, thus forming a local connected subgraph. With the help of this topology linking mechanism, the system can build a global road network model covering the entire mine, providing the infrastructure for subsequent path search and accessibility analysis.

[0046] Step 3: Each edge of the undirected connected graph corresponds to a complete road segment that can actually be traveled, and the length and direction characteristics of the complete road segment are stored at the same time.

[0047] A complete road segment refers to a continuous passageway between two adjacent topological nodes without any other branches or obstacles. The length feature is obtained by calculating the sum of the Euclidean distances of all directional polyline segments constituting the edge, used to quantify the maximum possible displacement of a miner on that segment. The orientation feature is determined by calculating the angle between the spatial vector from the starting node to the ending node of the segment and a reference coordinate system (such as true north), used to describe the direction of the roadway's extension. These feature data are stored as attribute fields in the edge structure of the undirected connected graph. For example, for an edge connecting nodes Node_A and Node_B, if it consists of two polyline segments with lengths of 20 meters and 30 meters respectively, the total stored length is 50 meters; if its overall extension direction is 30 degrees east of north, the stored orientation feature is 30°. By storing these refined features, the system can not only determine whether a location is reachable during subsequent trajectory matching, but also verify whether the direction of movement is consistent with the roadway orientation, thereby effectively identifying reverse drift or erroneous jumps caused by signal multipath effects.

[0048] Example 4: In one embodiment, this application provides a method for identifying traversable areas, the method comprising the following steps: Step 1: Using the tunnel width information as the half-width offset, offset point by point along both sides of the tunnel centerline coordinate sequence to the wall position to generate the initial symmetrical boundary; The tunnel width information refers to the transverse dimensions of the tunnel as defined in the tunnel design drawings or measured data, typically represented as a sequence of width values ​​along the tunnel's extension direction. The half-width offset, which is half the tunnel width information, is used to determine the theoretical distance from the centerline to both side walls. This step aims to quickly construct the theoretical geometric profile of the tunnel based on its centerline. Specifically, the system iterates through each discrete point in the tunnel centerline coordinate sequence, reads the corresponding tunnel width value, calculates its half-width, and then extends this half-width distance to the left and right sides perpendicular to the tangent direction of that point, obtaining two boundary points. Connecting all left boundary points forms the left wall profile, and connecting all right boundary points forms the right wall profile; together, they constitute the initial symmetrical boundary. For example, if the length of a tunnel centerline segment is 30 meters, and the tunnel width is constant at 4.0 meters, then the half-width offset is 2.0 meters. The system takes a sampling point every 0.5 meters along the centerline and offsets it 2.0 meters to each side to generate the initial boundary point set. By using this point-by-point offset method, abstract centerline data can be transformed into two-dimensional or three-dimensional spatial regions with actual physical meaning, providing a basic geometric framework for subsequent region calibration.

[0049] Step 2: Based on the initial symmetrical boundary, perform Boolean difference operation on the area occupied by the fixed obstacle from inside the initial symmetrical boundary to eliminate the impassable area, and round off the sharp corners of the boundary left after elimination to form a passable polygonal area. The area occupied by fixed obstacles can refer to a permanent or semi-permanent space within an underground roadway that is inaccessible to personnel, including but not limited to support pillars, pump units, ventilation doors, and areas with dense cable hooks. Boolean difference is a set operation logic used to subtract the area occupied by fixed obstacles from the total area enclosed by the initial symmetrical boundary. This step aims to accurately define the actual safe space for miners' movement and eliminate invalid positioning areas caused by equipment occupancy. Specifically, the system first obtains the geometric model (such as rectangle, circle, or polygon) of the fixed obstacle in the roadway coordinate system and maps it to the plane of the initial symmetrical boundary. Then, it performs the operation of subtracting the obstacle area from the initial area. At this point, sharp concave corners or discontinuous polygonal segments often form at the intersection of the obstacle edge and the initial boundary. To accommodate the continuous characteristics of personnel movement and avoid numerical oscillations at sharp corners in subsequent projection algorithms, the system smooths the sharp corners of the boundary formed after obstacle removal, for example, by using Bézier curve fitting or circular transition algorithms to replace sharp corners with smooth curves, ultimately generating a closed and smooth passable polygonal area. For example, when there is a circular support pillar with a diameter of 0.8 meters in the center of the tunnel, the Boolean difference operation will remove a circular hole in the center of the initial rectangular area and generate right-angled sharp corners at the intersection of the circular edge and the original rectangular boundary. After smoothing, these sharp corners are replaced by arcs with a radius of 0.2 meters, making the boundary of the generated passable area more consistent with the natural curvature of human movement trajectory. This improves the geometric accuracy and topological rationality of the passable area model, avoiding positioning points falling inside obstacles or jumping at unreasonable sharp corners.

[0050] Step 3: For cases where the width of the tunnel gradually changes along the centerline, a piecewise linear interpolation method is used to calculate different half-width offsets point by point based on the tunnel width information, so that the initial symmetrical boundary closely fits the actual wall contour.

[0051] In this context, a gradual change in roadway width along the centerline can refer to situations where the roadway is not a straight, uniformly wide section, but rather has a flared end, a contraction section, or a section with non-linear width variations due to geological factors. Piecewise linear interpolation is a numerical approximation technique used to estimate the width value at any location between known discrete width measurement points. This step addresses the distortion problem in modeling variable-width roadways, ensuring that the generated initial symmetrical boundary dynamically adapts to the actual roadway shape. Specifically, the system divides the roadway centerline into several linear segments, with the endpoints of each segment storing the measured roadway width value. For any point within a segment, the system linearly calculates the instantaneous width of that point based on its proportion to the distances from the endpoints, thus obtaining the dynamic half-width offset for that point. For example, in a connecting tunnel section 20 meters long, the entrance width is 4.0 meters, and the exit width gradually decreases to 2.0 meters. If the average value is taken directly without interpolation, it will lead to errors such as the boundary shrinking inward at the entrance and expanding outward at the exit. After using piecewise linear interpolation, at a distance of 5 meters from the entrance (i.e., the 1 / 4 position), the system automatically calculates a width of 3.5 meters and a half-width offset of 1.75 meters. The boundary points generated accordingly can accurately fit the actual inclined wall. Through this dynamic adjustment mechanism, the initial symmetrical boundary is no longer a rigid parallel line, but an adaptive curve that can closely fit the contour of the actual wall, thus providing a geometric benchmark for subsequent Boolean difference calculations and trajectory matching, and improving the spatial modeling accuracy in complex tunnel environments.

[0052] Example 5: In another embodiment, this application provides a process for generating a joint determination information set, which includes the following steps: Step 1: Obtain preliminary positioning results and human motion inertial data, and perform trajectory anomaly detection within the current layer and cross-layer misposition detection; The initial positioning result can refer to the sequence of spatial coordinates of the miner obtained in real time through an underground wireless positioning system (such as UWB, Wi-Fi, or ZigBee). This coordinates can originate from TDOA or TOF ranging calculations between the base station and the tag. Human motion inertial data can refer to the triaxial acceleration, triaxial angular velocity, and air pressure values ​​collected by the inertial measurement unit (IMU) worn by the miner, used to characterize the miner's posture changes and height displacement. This step aims to provide a multi-dimensional input data foundation for the subsequent dual detection mechanism. Specifically, the system synchronously reads the wireless positioning coordinates and IMU sensor data at preset time windows (e.g., 100ms or 200ms), forming multi-source data frames with timestamps. Intra-layer trajectory anomaly detection focuses on geometric spatial matching analysis, while cross-layer mispositioning detection focuses on the rationality review of logical topology and motion patterns. Both are executed in parallel to cover different types of positioning error scenarios. By synchronously acquiring and parallel processing multi-source data, it is possible to simultaneously capture geometric trajectory drift and logical hierarchical shifts in complex downhole environments, laying a data foundation for generating a highly robust decision information set.

[0053] Step 2: Trajectory Anomaly Detection within the Current Layer. By analyzing the degree of matching between the preliminary positioning trajectory and the geometric contour of the roadway, the matching confidence level of the current layer is output. The current layer matching reliability is a quantitative indicator used to characterize the degree of consistency between the preliminary positioning trajectory and the actual passable area and direction of the roadway within a local range. This reliability is obtained by extracting the local roadway geometric contour centered on the preliminary positioning result and comparing its direction and position with the motion trajectory segments within a preset time window in the historical trajectory sequence. Specifically, the system first retrieves the roadway centerline direction vector and the passable polygon boundary near the current node from the underground operation 3D topology map; then, it calculates the angle between the continuously changing direction vector of the motion trajectory segment and the roadway direction vector, and determines whether the trajectory point falls inside the passable polygon area. If there is a large angle deviation between the trajectory direction and the roadway direction (e.g., a sharp turn of more than 45 degrees in a road section without intersections), or if the trajectory point coordinates fall outside the passable area (e.g., falling into a wall or equipment occupation area), it is judged as a geometric mismatch, and the current layer matching reliability is reduced; conversely, if the trajectory is smooth and completely within the passable area, a high reliability is output. For example, when a miner is walking in a north-south oriented transport roadway, if the preliminary positioning results show that his trajectory suddenly deviates 90 degrees to the west within 5 seconds and there are no branch roads at that point, the system will identify this as a sharp turn anomaly and output a low matching confidence level. Thus, this step can effectively identify intra-layer coordinate drift caused by multipath effects or signal obstruction, ensuring that the positioning results conform to the physical geometric constraints of the roadway.

[0054] Step 3: Cross-layer mislocation detection outputs the cross-layer suspicion level through topology reachability inspection and motion pattern-driven inter-layer transition detection; The cross-layer suspicion level indicates the likelihood of a non-physical cross-layer jump in the current positioning result. Its generation relies on dual confirmation through logical path verification and physiological movement characteristics. Topology reachability review maps the current preliminary positioning result onto a 3D underground topology map to identify candidate nodes and checks whether a continuous path exists from a verified historical movement node in the previous time step along a passable edge in the topology map to reach that candidate node. If no continuous path exists (e.g., a direct jump from the -300m layer to the -400m layer without any connecting tunnels), a topology inaccessibility result is output, increasing the cross-layer suspicion level. Motion pattern-driven inter-layer transition detection analyzes the posture change characteristics (pitch angle) and air pressure change characteristics in human motion inertial data, monitoring whether miners exhibit a continuous and regular upward or downward trend within a continuous time window and whether air pressure values ​​are consistent with altitude changes. If both the pitch angle and air pressure changes indicate continuous slope movement or elevator-like vertical movement, then an inter-floor transition is confirmed. If the positioning result does not reflect this inter-floor change, a misjudgment may occur. However, if the positioning result jumps without corresponding motion characteristics, it is considered a high-risk cross-floor mispositioning. For example, if a step change in barometer reading within 3 seconds corresponds to a 40-meter height difference, while the pitch angle remains horizontal, this matches elevator characteristics. If the positioning result jumps directly from one floor to another without this air pressure and attitude characteristic support, then a very high level of suspicion for cross-floor movement is output. Through cross-validation of topological logic and motion physical characteristics, this step can accurately distinguish between genuine inter-floor movement and false floor jumps caused by signal crosstalk.

[0055] Step 4: Generate a joint judgment information set based on the current layer's matching credibility and cross-layer suspicion level.

[0056] The joint judgment information set is a comprehensive data structure integrating intra-layer geometric matching status and cross-layer logical suspicion status. It includes a location suspicion status identifier and a specific classification of mislocation types. This information set is generated by weighting and combining the current layer matching confidence value and the cross-layer suspicion level value, or by using logical threshold judgment, according to a preset fusion strategy. Specifically, when the current layer matching confidence is below a first threshold and the cross-layer suspicion level is below a second threshold, it is judged as an intra-layer mislocation; when the cross-layer suspicion level is above the second threshold, regardless of the intra-layer matching degree, it is preferentially judged as a cross-layer mislocation suspicion; if both indicators are within the normal range, the location is judged as reliable. This information set not only records the current abnormal state but also clarifies the cause category of the abnormality (such as geometric deviation or topological disconnection), thus providing a clear decision basis for subsequent differential resolution calculations. Through this multi-dimensional information fusion, the system can avoid the limitations of a single criterion, reduce the false alarm rate and false negative rate, and ensure that the generated judgment information set can accurately guide the selection of subsequent correction algorithms.

[0057] Example 6: In another optional embodiment, this application provides a method for detecting trajectory anomalies within the current layer, the method comprising the following steps: Step 1: Extract the local tunnel geometric contour centered on the preliminary positioning results, and extract the motion trajectory segments within the preset time window from the historical trajectory sequence; The preliminary positioning result can refer to the miner's current spatial coordinates calculated in real time by an underground wireless positioning system (such as UWB or Wi-Fi). The local tunnel geometry profile refers to the sequence of tunnel centerline coordinates and the corresponding boundary of the passable polygonal area, extracted within a set radius centered on the preliminary positioning result; it characterizes the physical spatial constraints surrounding the miner's current location. The historical trajectory sequence is a set of topologically validated positioning points stored in memory over a past period. The preset time window is a duration set based on the miner's normal walking speed, for example, 3 to 5 seconds, used to extract short trajectory segments that reflect the continuous movement trend of the personnel. By extracting data within this time window from the historical trajectory sequence, a movement trajectory segment containing multiple consecutive coordinate points can be obtained, reflecting the miner's actual movement path at the most recent moment.

[0058] Specifically, the system first reads the preliminary positioning result P_t at the current time t, and then retrieves the nearest tunnel centerline segment and the accessible boundaries on both sides of the tunnel in the underground operation 3D topology map, forming a local tunnel geometric outline. Simultaneously, the system backtracks the historical positioning data from t-Δt to time t, extracting the motion trajectory segment S_history. For example, when the preset time window is 4 seconds, if the sampling frequency is 10Hz, the most recent 40 consecutive positioning points are extracted to form the trajectory segment. This extraction method allows for the acquisition of sufficient data samples to determine the stability of the motion direction while ensuring real-time computation, providing a foundation for subsequent directional consistency analysis.

[0059] Step 2: Analyze whether the continuous change direction of the motion trajectory segment is consistent with the direction of the roadway in the local roadway geometric outline; The continuously changing direction refers to the instantaneous velocity direction sequence obtained after vector difference processing of the motion trajectory segment, reflecting the miner's direction of movement over a short period of time. The tunnel orientation refers to the tangent direction or extension direction of the tunnel centerline in the local tunnel geometry, representing the unique legal physical travel axis within that area. The core logic of this step is to compare the vector characteristics of the actual movement of personnel with the geometric constraints provided by the environment. Specifically, the system calculates the direction angle of the line connecting two adjacent points in the motion trajectory segment and statistically analyzes its average or principal direction within the time window; simultaneously, it extracts the orientation angle of the centerline of the local tunnel geometry in the current area. If the angle deviation between the two is less than a preset tolerance threshold (e.g., ±30°), it is determined that the directions are consistent; conversely, if the deviation is significant, it indicates that the motion trajectory violates the physical extension law of the tunnel. Furthermore, considering that underground tunnels may have curves, the tunnel orientation can be a segmented calculated local tangent direction. The system maps the motion trajectory segment onto the local tunnel geometry and compares the angle between the trajectory tangent and the tunnel centerline tangent point by point. For example, in a straight north-south tunnel (direction angle 0°), if the extracted motion trajectory segment shows that the miner is continuously moving eastward (direction angle 90°) for 3 seconds, it is determined to be a direction inconsistency. This direction consistency analysis can effectively filter out the random fluctuations of the positioning point caused by multipath effects, because random fluctuations usually have no fixed direction, while mispositioning often manifests as a systematic deviation that is completely contrary to the tunnel direction.

[0060] Step 3: If the movement trajectory segment makes a sharp turn at a large angle that does not conform to the direction of the roadway, or if the trajectory point falls outside the passable area, it is determined that the trajectory in the current layer is abnormal.

[0061] A sharp turn at a large angle refers to a sudden and drastic change in the direction angle of a movement trajectory segment within a short period of time, resulting in a significant deviation from the local tunnel orientation. For example, in a straight tunnel without branching points, a sudden change in trajectory direction exceeding 90° is physically impossible for a pedestrian and is therefore identified as an anomaly. "Outside the passable area" refers to the space outside the boundaries of the generated passable polygon area, including tunnel walls, support pillars, equipment occupancy areas, and other inaccessible areas. Determining an anomaly in the current layer's trajectory can mean the system outputs a Boolean flag or an anomaly level signal, indicating that the current preliminary positioning result is unreliable. For instance, if a miner is walking in a 4-meter-wide transport tunnel and the preliminary positioning result suddenly jumps to 2 meters outside the tunnel wall, or the trajectory shows the miner making a sudden 180° turn in a straight tunnel, the system will immediately trigger an anomaly determination for the current layer's trajectory.

[0062] Example 7: In one optional implementation, this application provides a topology reachability verification method, the method comprising: Step 1: Map the preliminary positioning results at the current moment onto the 3D topology map of the underground operation to determine its candidate node; This step aims to discretize the continuous spatial coordinates output by the radio positioning system into logical nodes in the topology graph, enabling connectivity analysis at the graph theory level. The initial positioning result is typically a three-dimensional coordinate (x, y, z) calculated by a UWB base station or Wi-Fi receiver, while the underground operation's three-dimensional topology graph is an undirected connected graph constructed from the coordinate sequence of the tunnel centerline and adjacency relationships. The mapping process specifically includes: searching the topology graph for the node with the closest Euclidean distance to the current coordinates, or projecting the current coordinates onto the nearest tunnel edge (i.e., a complete tunnel segment), thereby determining the logical node to which that location belongs as a candidate node. For example, when the initial positioning result is coordinates P(105.2, 203.4, -300.5), the system searches the pre-stored topology node database and finds that it has the smallest spatial distance to node numbered Node42 and is located within the traversable polygon area associated with that node; therefore, Node42 is determined as a candidate node for the current moment. This mapping mechanism can transform noisy continuous positioning data into discrete states that conform to the geometric constraints of the roadway, providing accurate start and end conditions for subsequent path search.

[0063] Step 2: Retrieve the verified historical movement nodes from the previous time step and check whether there is a continuous path from the historical movement nodes along the walkable edges in the topology graph that can reach the candidate nodes. This step utilizes the principle of motion continuity in time series to verify whether the current location state and the previous state are physically connected. Historical motion nodes refer to reliable location nodes confirmed after anomaly detection and correction at the previous sampling time, representing the miner's true location at time t-1. The checking process is based on graph search algorithms (such as Depth-First Search (DFS), Breadth-First Search (BFS), or A* algorithm). Starting with historical motion nodes and ending with candidate nodes, it traverses all possible paths in a subgraph composed of walkable edges. If the algorithm finds a sequence of walkable edges connecting the start and end points, a continuous path is determined to exist; otherwise, if all branches are unreachable, or if the path is blocked by topological barriers such as tunnel walls or enclosed areas, a continuous path is determined not to exist. For example, if the historical motion node is located at Node10 at -300 meters, and the current candidate node is mapped to Node88 at -400 meters, and there are no topological edges connecting the two layers within the current time window corresponding to ramps or elevator shafts, the algorithm determines that a continuous path does not exist. This process effectively utilizes the physical connectivity rules of the tunnels, eliminating false positioning caused by signal multipath effects or cross-layer interference, such as wall penetration or instantaneous movement.

[0064] Step 3: If no continuous path exists, output the topology unreachable result; This step is a logical response to path search failures, used to mark a serious topological conflict in the current preliminary location result. When the above check results show that there is no legal path between the start and end points, the system generates a topologically unreachable result, which typically includes a Boolean flag (Flag=False) and an error type code. Outputting a topologically unreachable result means that the preliminary location result at the current moment has a very high probability of cross-layer mislocation or long-distance jump, because it violates the physical law that miners must walk along the tunnel continuously within a given time. For example, when it is detected that a miner jumps from one end of a transport tunnel to a closed blind tunnel at the other end within 1 second, with no connecting path in between, this result is output immediately. This result will serve as an important basis for subsequent differential calculations, triggering the cross-layer mislocation correction process to prevent erroneous trajectories from being recorded or displayed.

[0065] Step 4: If a continuous path exists, calculate the path length and output the topology reachability result.

[0066] This step, based on confirming topological connectivity, further quantifies the theoretical distance of movement between two points to enhance the robustness of the judgment. Once a continuous path is confirmed, the system not only outputs a topological reachability result (Flag=True) but also accumulates the geometric lengths of all edges constituting the path to obtain the theoretical shortest path length or the actual walking path length from historical movement nodes to candidate nodes. The path length can be compared with the mileage calculated from human motion inertial data. If the deviation is within a preset threshold, the reliability of the positioning is further confirmed. For example, if the calculated topological path length is 85 meters, and the displacement obtained by accelerometer integration within the same time period is approximately 82 meters, the two are highly consistent, and a topological reachability result including path length information is output. By introducing the quantitative indicator of path length, not only can reachability be determined, but also its reasonableness can be evaluated, thereby reducing the misjudgment rate caused by the complexity of the topological structure and ensuring that the output positioning result conforms to both geometric constraints and kinematic laws.

[0067] Example 8: In one possible implementation, this application provides a motion mode-driven inter-layer transition detection method, which includes the following steps: Step 1: Analyze the posture change characteristics and air pressure change characteristics in the human motion inertia data; Human motion inertial data refers to the set of motion parameters collected in real time by the inertial measurement unit (IMU) worn by the miner, specifically including triaxial acceleration, triaxial angular velocity, and absolute air pressure. Attitude change characteristics mainly refer to the pitch angle reflecting the miner's tilt state. This characteristic is obtained by integrating the angular velocity data output by the gyroscope and then correcting for gravity vectors using accelerometer data. Air pressure change characteristics refer to the ambient atmospheric pressure value collected by a high-precision barometer, which fluctuates regularly with altitude. Attitude change characteristics are used to characterize whether the miner is walking on a slope, while air pressure change characteristics are used to characterize whether the miner has experienced vertical altitude displacement. For example, when a miner walks upwards on a slope with an inclination angle of 15°, the pitch angle of their device will remain stable in the positive range, while the barometer reading will gradually decrease with increasing altitude. When the miner rides an elevator, the pitch angle is close to zero, but the barometer reading will experience a large, abrupt change in a short period. By extracting these two types of features simultaneously, we can provide multi-dimensional physical evidence for subsequent judgments on inter-layer transitions.

[0068] Step 2: Monitor whether the pitch angle of the miner shows a continuous and regular upward or downward trend within a continuous time window, and whether the air pressure value shows a stable rise or fall or a step change consistent with the change in altitude. The continuous time window can refer to a preset sampling duration, such as a sliding window of 3 to 10 seconds, used to filter instantaneous sensor noise and capture motion trends. The monitoring process includes linear fitting or trend analysis of the pitch angle sequence within the window to determine whether its slope exceeds a preset threshold and whether the direction remains consistent, thereby identifying a continuous and regular upward or downward trend. Simultaneously, differential calculation or step detection is performed on the air pressure value sequence within the window to determine whether its change matches the theoretical altitude change (based on the standard atmospheric pressure formula conversion), thereby identifying stable ascents and descents or step changes. Specifically, if the pitch angle continuously changes from -2° to -15° within the window, and the air pressure value simultaneously decreases from 982.5 hPa to 979.0 hPa, both are considered to indicate an altitude change; if the air pressure value abruptly changes from 985.2 hPa to 980.1 hPa within 3 seconds while the pitch angle remains near 0°, it is considered a step change.

[0069] Step 3: If both indicate the existence of continuous ramp motion or elevator-like up-and-down motion, then the output inter-floor transition confidence level is determined to be that the inter-floor transition has occurred. Inter-level transition confidence is an indicator that quantifies the probability of a floor change occurring at the current moment. A confirmed inter-level transition indicates that the system is confidently confirming that the miner is walking along a ramp or moving vertically through an elevator shaft. Continuous ramp movement corresponds to a scenario where the pitch angle and air pressure change continuously in the same direction, i.e., the pitch angle is significantly non-zero and the air pressure changes monotonically with altitude; elevator-like vertical movement corresponds to a scenario where the pitch angle is close to zero and the air pressure changes rapidly in a step. When either of these two scenarios is confirmed by both features simultaneously, the system determines that the miner has indeed undergone physical inter-level displacement. For example, in the ramp scenario, the continuous negative pitch angle and the continuous decrease in air pressure corroborate each other, ruling out misjudgments caused solely by equipment tilt; in the elevator scenario, the drastic change in air pressure and the horizontal stability of the posture corroborate each other, ruling out misjudgments caused solely by air pressure fluctuations. This dual confirmation mechanism effectively distinguishes between genuine inter-level movement and positioning jumps caused by signal multipath effects or obstruction.

[0070] Step 4: If only the pitch angle trend changes significantly while the air pressure does not change significantly, or if only the air pressure changes significantly while the pitch angle does not change significantly, then it is determined that inter-layer transition may have occurred. In some cases, a possible anomaly is detected by the system in a single dimension, but lacks corroboration from another dimension. In such cases, the authenticity of the inter-level transition is questionable and requires further verification in conjunction with topological accessibility checks. A situation where only the pitch angle changes significantly while the air pressure remains relatively constant typically occurs when miners in horizontal tunnels bend over sharply, climb local obstacles, or experience strong vibrations affecting equipment. In these situations, although there is a change in posture, there is no substantial elevation displacement. A situation where only the air pressure changes significantly while the pitch angle remains relatively constant may stem from localized air pressure fluctuations caused by the underground ventilation system, barometer drift due to sudden temperature changes, or minor swaying during elevator startup that has not yet been detected by the attitude sensors. In this state, the system does not directly perform cross-level correction but marks the state as pending suspicion to prevent incorrect jumps to adjacent floors due to noise from a single sensor. For example, when a miner briefly bends down in a low tunnel, a large fluctuation in the pitch angle but constant air pressure is detected as possible rather than certain, thus avoiding misinterpreting a normal bending motion as downhill walking.

[0071] Step 5: If there is no significant change in either, it is determined that no interlayer transition has occurred.

[0072] The condition of "no significant change" can refer to a slight fluctuation in the pitch angle near zero without a sustained trend, while the air pressure remains stable or fluctuates within the normal environmental noise range. This determination implies that the miner is currently walking in a straight tunnel within the same horizontal layer, or is stationary, with no need for vertical displacement. In this case, any significant jumps in the Z-axis coordinate or changes in floor markings in the preliminary positioning results are highly suspected to be cross-layer mispositioning caused by signal crosstalk, rather than the miner's actual movement. For example, if a miner is walking in a straight line in a transport tunnel at the -300-meter level, with the pitch angle maintained within ±1° and the air pressure fluctuation less than 0.05 hPa, and the positioning system output shows that the miner has instantly moved to the -400-meter layer, the system will, based on the determination that no inter-layer transition has occurred and combined with the topology inaccessibility result, decisively identify the positioning result as mispositioning and initiate the correction process.

[0073] Example 9: In one optional implementation, this application provides a process for generating a trusted location result set, the method comprising the following steps: Step 1: Based on the mislocation type in the joint determination information set, perform differential resolution; The differential resolution algorithm refers to dynamically selecting the corresponding geometric correction algorithm or topology rematching strategy based on the mislocalization classification labels carried in the joint decision information set generated in the previous steps. Mislocalization types mainly include two scenarios: intra-layer mislocalization and cross-layer mislocalization. This step aims to avoid using a single correction logic to handle all localization deviations, thereby preventing secondary errors caused by strategy mismatch. Specifically, the system first parses the joint decision information set to identify whether the root cause of the localization anomaly stems from multipath effects or geometric ambiguity of the signal in a single plane, or from signal cross-layering or hierarchical mapping errors in the vertical direction. For example, when the joint decision information set marks an intra-layer trajectory anomaly with a low cross-layer suspicion level, the system automatically calls the intra-layer projection correction subroutine; conversely, if it marks a high cross-layer suspicion level and topology unreachable, the cross-layer rematching subroutine is called. This type-based branching mechanism ensures that the correction action accurately corresponds to the error source, improving the accuracy of localization recovery.

[0074] Step 2: If the mislocation type is mislocation within the current layer, project the preliminary location result along the topologically accessible direction to the passable polygon area and constrain it to the coordinate sequence of the nearest roadway centerline; In this context, mispositioning within the current layer typically manifests as the location coordinates deviating from the actual tunnel path traveled by the miner on the horizontal plane, while the vertical elevation information remains largely correct. Projecting the preliminary positioning result along the topologically accessible direction can refer to finding the entry point into the passable polygonal region by using the preliminary positioning result as a starting point and traversing the direction of the normal to the connected edge of the node containing that point in the 3D topological map of the underground operation, or the direction of the shortest path. The passable polygonal region is the effective activity space generated based on tunnel width information and obstacle removal. Constraining to the nearest tunnel centerline coordinate sequence can refer to further calculating the Euclidean distance from the positioning point to each polyline segment in the tunnel centerline coordinate sequence after the positioning point falls into the passable region, and selecting the foot of the perpendicular or the nearest neighbor point on the polyline segment with the smallest distance as the final correction coordinates. This process utilizes the prior knowledge that the tunnel centerline is the main trajectory of the miner's movement. For example, if the initial positioning results show that the miner is located at the edge of the passable area of ​​the transport tunnel or even slightly beyond the boundary (due to signal reflection), but this point is topologically connected to historical movement nodes, the system vertically projects this point back to the tunnel centerline, assuming that the miner is actually walking near the tunnel centerline. This not only corrects for instantaneous positioning jumps but also ensures the smoothness and rationality of the output trajectory in terms of geometry.

[0075] Step 3: If the mislocation type is cross-layer mislocation, then based on the inter-layer conversion detection results, re-match the location results to the topology node and passable area corresponding to the target layer; Cross-layer mislocation refers to the positioning system incorrectly calculating the miner to an adjacent floor or a non-connected spatial area. Re-matching to the target floor relies on the inter-layer transition detection results driven by the aforementioned movement pattern. The target floor is the physical level of the miner inferred from air pressure change trends and pitch angle characteristics. Specifically, if the inter-layer transition detection results indicate that the miner is ascending a slope or descending an elevator, the system will forcibly correct the Z-axis coordinates of the initial positioning result to the elevation range of the target floor and search for candidate nodes with topological connections to historical movement nodes from the previous moment in the underground 3D topology map of that target floor. Subsequently, the positioning result is mapped to the passable area associated with that candidate node. For example, when a miner is detected descending from the -300-meter level to the -400-meter level via a connecting roadway, but the initial UWB positioning result is still stuck at the -300-meter level, the system determines the target level to be the -400-meter level based on the barometer's step descent data and the continuous negative change in the pitch angle. The positioning coordinates are then re-matched to the road segment connected to the entrance node in the -400-meter level topology map. This processing method effectively solves the problem of wall penetration or level skipping caused by overlapping vertical signal coverage, ensuring the logical consistency of personnel's position within the three-dimensional tunnel network.

[0076] Step 4: Output the corrected positioning results and timestamps to form a reliable positioning result set.

[0077] The corrected positioning result refers to the spatial coordinates (X, Y, Z) obtained after the aforementioned differential calculations, conforming to the topological constraints and kinematic laws of the tunnel. The timestamp identifies the precise moment corresponding to the corrected result, typically consistent with or synchronized to the initial positioning result at the microsecond level. Forming a reliable positioning result set means encapsulating a series of corrected coordinates at consecutive time points in chronological order for use by upper-level applications. The output of this step directly serves as the final data source for the underground personnel positioning system, used for real-time trajectory display, electronic fence alarms, and emergency rescue command. For example, the system outputs a data packet containing corrected coordinates and a millisecond-level timestamp every 100 milliseconds; the sequence of these data packets constitutes the reliable positioning result set. By introducing timestamps, not only can the miner's movement sequence be reconstructed, but it can also be used in subsequent analysis to calculate instantaneous velocity and acceleration, further verifying the physical authenticity of the trajectory. This result provides a highly reliable data foundation for subsequent mine safety monitoring, avoiding false alarms or missed alarms caused by initial positioning drift.

[0078] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0079] Figure 3 A schematic diagram of a roadway topology correction system for cross-layer mislocation provided in an embodiment of this application is shown below. Figure 3 As shown, the tunnel topology correction system 300 for cross-layer mislocation in this embodiment includes: a topology map generation module 301, a cross-layer determination module 302, and a feasible positioning module 303.

[0080] The topology map generation module 301 is used to acquire a multi-source prior information set of underground tunnel space, perform multi-dimensional modeling of tunnel space and marking of passable areas based on the multi-source prior information set, and generate a three-dimensional topology map of underground operation that includes geometric contours and topological connectivity. The cross-layer determination module 302 is used to obtain the preliminary positioning results and human motion inertial data at the current moment, and combine them with the underground operation three-dimensional topology map to perform trajectory anomaly detection and cross-layer mispositioning detection within the current layer, and generate a joint determination information set containing the suspected positioning status and mispositioning type. The feasible positioning module 303 is used to perform differential calculation and correction on intra-layer mispositioning and cross-layer mispositioning based on the joint judgment information set, and generate a set of corrected reliable positioning results.

[0081] The system in this embodiment can be used to execute the methods of any of the above embodiments, and its implementation principle and technical effect are similar, so they will not be described again here.

Claims

1. A method for correcting roadway topology for cross-layer mislocation, characterized in that, include: A multi-source prior information set of underground tunnel space is obtained. Based on the multi-source prior information set, multi-dimensional modeling of tunnel space and calibration of passable areas are performed to generate a three-dimensional topological map of underground operations that includes geometric contours and topological connectivity. Obtain the preliminary positioning results and human motion inertial data at the current moment, and combine them with the three-dimensional topology map of the underground operation to perform trajectory anomaly detection and cross-layer mispositioning detection within the current layer, and generate a joint judgment information set containing the suspected positioning status and mispositioning type; Based on the joint determination information set, differential calculation and correction are performed for mislocation within the current layer and mislocation across layers to generate a set of corrected reliable positioning results. The process of generating the joint determination information set includes: The preliminary positioning results and the human motion inertial data are obtained to perform trajectory anomaly detection within the current layer and cross-layer mispositioning detection. The current layer trajectory anomaly detection outputs the current layer matching confidence level by analyzing the degree of matching between the preliminary positioning trajectory and the roadway geometric contour. The cross-layer mislocation detection outputs the cross-layer suspicion level through topology reachability review and motion pattern-driven inter-layer transition detection. The joint judgment information set is generated based on the current layer matching credibility and the cross-layer suspicion level; The motion mode-driven interlayer transition detection includes: Analyze the posture change characteristics and air pressure change characteristics in the aforementioned human motion inertial data; Monitor whether the pitch angle of the miner shows a continuous and regular upward or downward trend within a continuous time window, and whether the air pressure value shows a stable rise or fall or a step change consistent with the change in altitude. If both indicate the existence of continuous ramp motion or elevator-like up-and-down motion, then the output inter-level transition confidence level is that the inter-level transition has definitely occurred. If only the pitch angle trend changes significantly while the air pressure does not change significantly, or if only the air pressure changes significantly while the pitch angle does not change significantly, it is determined that inter-layer transition may have occurred. If neither changes significantly, it is determined that no interlayer transition has occurred.

2. The method according to claim 1, characterized in that, The process of generating the three-dimensional topology map of the underground operation includes: The multi-source prior information set includes the roadway centerline coordinate sequence, roadway width information, and roadway spatial adjacency relationships; Based on the coordinate sequence of the lane centerline and the spatial adjacency relationship of the lane, multi-dimensional modeling of the lane space is performed to generate a lane network skeleton with topological connectivity. Based on the tunnel width information, passable areas are marked, and inaccessible areas occupied by walls, equipment, and support columns are eliminated. The tunnel network skeleton and passable areas are then integrated into the three-dimensional topology map of the underground operation.

3. The method according to claim 2, characterized in that, The process of performing multi-dimensional modeling of the tunnel space includes: Based on the coordinate sequence of the roadway centerline, each roadway is abstracted into a directional polyline segment with directional attributes, and the three-dimensional coordinates of the polyline on each directional polyline segment are recorded. Based on the spatial adjacency relationship of the tunnels, the three-dimensional coordinates of the polylines belonging to the same branch or inter-layer connection are topologically linked to form an undirected connected graph; Each side of the undirected connected graph corresponds to a complete road segment that is actually passable, and the length and orientation characteristics of the complete road segment are stored.

4. The method according to claim 2, characterized in that, The process of marking passable areas includes: Using the tunnel width information as a half-width offset, the tunnel centerline coordinate sequence is shifted point by point to the wall position to generate an initial symmetrical boundary. Based on the initial symmetrical boundary, the area occupied by the fixed obstacle is subjected to Boolean difference operation from inside the initial symmetrical boundary to eliminate the traversable area, and the sharp corners of the boundary left after elimination are rounded to form a passable polygonal area. For cases where the width of the tunnel gradually changes along the centerline, a piecewise linear interpolation method is used to calculate different half-width offsets point by point based on the tunnel width information, so that the initial symmetrical boundary closely fits the actual wall contour.

5. The method according to claim 4, characterized in that, The current layer trajectory anomaly detection includes: Extract the local tunnel geometric contour centered on the preliminary positioning result, and extract the motion trajectory segment within a preset time window from the historical trajectory sequence; By analyzing whether the continuous change direction of the motion trajectory segment is consistent with the direction of the roadway in the local roadway geometric contour; If the movement trajectory segment makes a sharp turn at a large angle that does not conform to the direction of the roadway, or if the trajectory point falls outside the passable area, it is determined that the trajectory within the current layer is abnormal.

6. The method according to claim 4, characterized in that, The topology reachability review includes: Map the preliminary positioning results at the current moment onto the three-dimensional topology map of the underground operation to determine its candidate node; Retrieve the verified historical motion nodes from the previous moment and check whether there is a continuous path from the historical motion nodes along the walkable edges in the topology graph that can reach the candidate nodes. If no continuous path exists, output the topology unreachable result; If a continuous path exists, calculate the path length and output the topology reachability result.

7. The method according to claim 4, characterized in that, The process of generating the trusted location result set includes: Based on the mislocation type in the joint determination information set, perform differential resolution; If the mislocation type is mislocation within the current layer, the preliminary positioning result is projected along the topologically accessible direction onto the passable polygon area and constrained to the nearest lane centerline coordinate sequence; If the mislocation type is the cross-layer mislocation, then based on the inter-layer conversion detection result, the location result is re-matched to the topology node corresponding to the target layer and the passable area; The corrected positioning results and timestamps are output to form the trusted positioning result set.

8. A roadway topology correction system for cross-layer mislocation, characterized in that, Applied to the method as described in any one of claims 1-7, comprising: The topology map generation module is used to acquire a multi-source prior information set of underground tunnel space, and based on the multi-source prior information set, to perform multi-dimensional modeling of tunnel space and identification of passable areas, and generate a three-dimensional topology map of underground operations that includes geometric contours and topological connectivity. The cross-layer determination module is used to obtain the preliminary positioning results and human motion inertial data at the current moment, and combine them with the three-dimensional topology map of the underground operation to perform trajectory anomaly detection and cross-layer mispositioning detection within the current layer, and generate a joint determination information set containing the suspected positioning status and mispositioning type. The feasible positioning module is used to perform differential calculation and correction on intra-layer mispositioning and cross-layer mispositioning based on the joint judgment information set, and generate a set of corrected reliable positioning results.

Images