An industrial asset space-time trajectory tracking method and system based on multi-source heterogeneous positioning data fusion

Abstract

The application provides an industrial asset space-time trajectory tracking method and system based on multi-source heterogeneous positioning data fusion. Among them, the edge node on the industrial asset collects multi-source heterogeneous asset positioning data to infer signal propagation path, extracts the geometric characteristics of the terrain reflecting surface and selects the polarization direction of the LoRa and satellite communication link; the edge node is divided into a federal learning group according to the geographical area, the positioning data is transmitted in the group through the polarization direction, the space-time feature segment of the industrial asset trajectory is generated, and is broadcast to the adjacent federal learning group; according to the space-time feature segment broadcast by the adjacent group, combined with the signal coverage blind area information, the topological extension is carried out on the broken area of the cross-group trajectory, and a continuous trajectory prediction path is generated; based on the signal attenuation difference of different terrain areas, the continuous prediction path is integrated, and a space-time trajectory tracking strategy is generated. The technical scheme provided by the application improves the reliability of industrial asset trajectory tracking in complex terrain.

Description

Technical Field

[0001] This application relates to the field of trajectory tracking technology, and in particular to a method and system for spatiotemporal trajectory tracking of industrial assets based on the fusion of multi-source heterogeneous positioning data. Background Technology

[0002] In remote mining areas, oil fields, and other regions without base station coverage, industrial assets (such as transport vehicles and inspection equipment) often need to be moved and operated in complex terrains (mountains, canyons). Such scenarios require asset tracking systems to have high-precision positioning and low-latency communication capabilities, and to maintain stable data transmission under conditions such as multipath interference and extreme weather. At the same time, it is necessary to solve the problem of trajectory breakage caused by terrain obstruction.

[0003] Current solutions for this requirement mainly employ a tracking system that combines low-Earth orbit satellite communication with inertial navigation-assisted positioning. This system achieves wide-area communication coverage through low-Earth orbit satellites, combines acceleration and angular velocity data collected by an inertial navigation module (IMU) to calculate the relative position of assets, and uses satellite signal strength to compensate for the accumulated errors of inertial navigation, while reducing the transmission load of satellite links through data compression technology to adapt to low bandwidth conditions.

[0004] The existing solutions have the following drawbacks: Low-Earth orbit satellite communication has significant transmission delays (on the order of seconds), which cannot meet the real-time requirements of dynamic trajectory tracking, especially in scenarios of rapid asset movement or emergency obstacle avoidance, which can easily lead to trajectory distortion; Inertial navigation relies on integral calculations, and positioning errors continue to accumulate when there is no satellite signal correction for a long time, and the error is amplified in complex terrain, leading to trajectory deviation; Satellite signals are severely attenuated in rain and dusty weather, and data compression may lose key positioning features, resulting in trajectory breaks or missed detection of critical events. Summary of the Invention

[0005] This application provides a method and system for spatiotemporal trajectory tracking of industrial assets based on the fusion of multi-source heterogeneous positioning data, in order to solve the problem of insufficient reliability in the existing technology of industrial asset trajectory tracking.

[0006] Firstly, this application provides a method for spatiotemporal trajectory tracking of industrial assets based on the fusion of multi-source heterogeneous positioning data, including:

[0007] By collecting multi-source heterogeneous asset positioning data through edge nodes deployed on industrial assets, the geometric features of the terrain reflection surface in the signal propagation path are inferred from the asset positioning data based on the changes in the location of the industrial assets, and the polarization direction of the LoRa and satellite communication link is selected based on the geometric features. The polarization direction includes the vertical polarization direction and the horizontal polarization direction.

[0008] The edge nodes are divided into federated learning groups according to geographical regions. Within the federated learning group, the asset location data is transmitted through the polarization direction. Based on the transmission results, spatiotemporal feature segments of industrial asset trajectories are generated, and the spatiotemporal feature segments are broadcast across groups to adjacent federated learning groups through the LoRa and satellite communication links.

[0009] Based on the spatiotemporal feature segments broadcast in the adjacent federated learning groups, and combined with the signal coverage blind zone information corresponding to the polarization direction, the fracture region of the cross-group industrial asset trajectory is topologically extended, and a continuous trajectory prediction path is generated based on the extended fracture region.

[0010] Based on the signal attenuation differences in different terrain regions under the polarization direction, the continuous trajectory prediction paths of each federated learning group are integrated, and a spatiotemporal trajectory tracking strategy is generated based on the integrated continuous trajectory prediction paths.

[0011] Optionally, the step of dividing the edge nodes into federated learning groups according to geographical regions, transmitting the asset location data within the federated learning group via the polarization direction, generating spatiotemporal feature segments of industrial asset trajectories based on the transmission results, and broadcasting the spatiotemporal feature segments across groups to adjacent federated learning groups via the LoRa and satellite communication link includes:

[0012] Based on the geometric features corresponding to the polarization direction, the edge nodes are divided into federated learning groups that cover the continuous region of the same reflective surface. The boundary of the federated learning group is determined by the straight-line penetration range of the signal propagation path under the polarization direction.

[0013] Within the federated learning group, the transmission parameters of the asset location data are directionally matched based on the polarization direction. According to the matching results, edge nodes that form a stable transmission relationship with the polarization direction are selected. The asset location data is then divided into multiple data segments in descending order of transmission stability through the edge nodes.

[0014] Calculate the time interval and position offset between the multiple data segments, combine the parts that are continuous in time and whose position offset is less than a preset distance threshold into a local trajectory sequence, and generate industrial asset trajectory feature segments containing spatiotemporal correlation based on the movement direction and speed change patterns of the local trajectory sequence.

[0015] The industrial asset trajectory feature segments are sorted according to the signal coverage priority corresponding to the polarization direction, and the sorted feature segments are broadcast to adjacent federated learning groups in order of coverage range from near to far via the LoRa and satellite communication links.

[0016] Optionally, the geometric features of the terrain reflective surface in the signal propagation path are extracted from the asset location data to infer changes in the industrial asset location, and the polarization direction of the LoRa-satellite communication link is selected based on the geometric features. The polarization direction includes a vertical polarization direction and a horizontal polarization direction, including:

[0017] The edge nodes synchronously collect both type I and type II positioning data of the industrial assets. The type I positioning data includes the continuous position coordinates of the industrial assets, and the type II positioning data includes the movement direction and speed changes of the industrial assets.

[0018] The movement trajectory direction of the industrial asset is determined based on the continuous position coordinates. An initial straight line is drawn along the movement trajectory direction. The initial straight line is then corrected for curvature based on the movement direction and speed changes. A signal propagation path matching the terrain occlusion distribution is generated based on the corrected initial straight line.

[0019] The height difference of the terrain reflective surfaces and the density distribution of obstacles are measured along the signal propagation path. Reflective surfaces with a height difference greater than the upper limit of the difference are marked as strong reflection areas, and reflective surfaces with an obstacle density distribution greater than a preset density are marked as multipath interference areas. The geometric boundary features of the strong reflection areas and the multipath interference areas are extracted.

[0020] Based on the tilt angle and distribution range in the geometric boundary features, the vertical polarization direction of the LoRa communication link is selected to weaken the reflection interference of the tilt angle, and the horizontal polarization direction of the satellite communication link is selected to penetrate the distribution range.

[0021] Optionally, the step of performing topological extension of the fracture region of the cross-group industrial asset trajectory based on the spatiotemporal feature segments broadcast in the adjacent federated learning groups, combined with the signal coverage blind zone information corresponding to the polarization direction, and generating a continuous trajectory prediction path based on the extended fracture region includes:

[0022] Based on the signal coverage blind zone information corresponding to the polarization direction, the starting point and ending point of the fracture region of the cross-group industrial asset trajectory are determined, and the starting point and ending point are the position coordinates of the first and last valid trajectory points in the adjacent federated learning group, respectively.

[0023] From the spatiotemporal feature segments broadcast by the adjacent federated learning groups, trajectory segments that match the starting point and the ending point are selected. The movement direction of the trajectory segments is superimposed with the shape of the signal coverage blind zone corresponding to the polarization direction to form a candidate extension path.

[0024] The curvature of the candidate extension path is dynamically adjusted according to the signal reflection intensity distribution in the fracture area, and the adjusted curvature is used as a constraint to generate an arc-shaped extension path that fits the boundary of the signal coverage blind zone.

[0025] The arc-shaped extension path and the starting point and ending point are smoothly connected to form a continuous arc-shaped extension path. After deleting the part of the continuous arc-shaped extension path that does not overlap with the signal coverage area of ​​the polarization direction, a continuous trajectory prediction path is generated.

[0026] Optionally, generating industrial asset trajectory feature segments containing spatiotemporal correlations based on the movement direction and speed variation patterns of the local trajectory sequence includes:

[0027] Calculate the difference in the movement direction of adjacent trajectory points in the local trajectory sequence, and mark the continuous trajectory segments whose difference in movement direction is less than the upper limit of the difference as directional stable segments;

[0028] Extract a set of trajectory points with consistent velocity change trends from the stable directional segment, and divide the set of trajectory points into acceleration and deceleration segments according to the direction of increase or decrease of the velocity change trends.

[0029] The parts of the acceleration and deceleration segments that have the same direction of movement and continuous speed trend are combined into basic feature units.

[0030] Based on the spatiotemporal distribution density of the basic feature units, the boundaries of adjacent basic feature units are seamlessly spliced ​​together to form an industrial asset trajectory feature segment containing spatiotemporal correlation.

[0031] Optionally, the step of selecting the vertical polarization direction of the LoRa communication link to reduce reflection interference from the tilt angle based on the tilt angle and distribution range in the geometric boundary features, and selecting the horizontal polarization direction of the satellite communication link to penetrate the distribution range, includes:

[0032] Based on the direction and magnitude of the tilt angle in the geometric boundary features, the vertical polarization direction of the LoRa communication link is selected, and the vertical polarization direction is orthogonal to the tilt angle.

[0033] The horizontal polarization direction of the satellite communication link is determined based on the continuous distribution length of obstacles within the distribution range of the geometric boundary features, wherein the horizontal polarization direction is parallel to the main extension direction of the distribution range;

[0034] Based on the change in tilt angle during the movement of industrial assets, the angle offset of the vertical polarization direction is corrected so that the orthogonal relationship is updated synchronously with the tilt state of the terrain reflector surface.

[0035] Based on the density change of obstacles in the distribution range, the coverage width of the horizontal polarization direction is adjusted so that the horizontal polarization direction penetrates the distribution range.

[0036] Optionally, the step of integrating the continuous trajectory prediction paths of each federated learning group based on the signal attenuation differences in different terrain regions under the polarization direction, and generating a spatiotemporal trajectory tracking strategy based on the integrated continuous trajectory prediction paths, includes:

[0037] According to the terrain reflector type corresponding to the polarization direction, the continuous trajectory prediction paths of each federated learning group are divided into flat terrain paths and complex terrain paths, and the signal attenuation of the flat terrain path is lower than that of the complex terrain path.

[0038] The sections with the smallest signal strength attenuation in the flat terrain path and the sections with the largest signal reflection intensity fluctuations in the complex terrain path are extracted and marked as stable communication areas and dynamic compensation areas, respectively.

[0039] The trajectory prediction paths in the stable communication zone are arranged in ascending order of signal attenuation, while the trajectory prediction paths in the dynamic compensation zone are arranged in descending order of signal reflection intensity fluctuation. A trajectory sequence with priority superposition is formed based on the results of the two arrangements.

[0040] Based on the spatial overlap between the stable communication zone and the dynamic compensation zone in the trajectory sequence, the continuous trajectory prediction paths of each federated learning group are integrated to generate dynamic tracking rules covering mixed terrain areas.

[0041] Secondly, this application provides an industrial asset spatiotemporal trajectory tracking system based on multi-source heterogeneous positioning data fusion, comprising:

[0042] The acquisition module is used to acquire multi-source heterogeneous positioning data through edge nodes deployed on industrial assets, infer the signal propagation path based on the changes in the location of industrial assets in the asset positioning data, extract the geometric features of the terrain reflection surface in the signal propagation path, and select the polarization direction of the LoRa and satellite communication link according to the geometric features.

[0043] The transmission module is used to divide the edge nodes into federated learning groups according to geographical regions, transmit the asset location data within the federated learning group through the polarization direction, generate spatiotemporal feature segments of industrial asset trajectories locally based on the transmission results, and broadcast the spatiotemporal feature segments across groups to adjacent federated learning groups through the LoRa and satellite communication link.

[0044] The extension module is used to extend the topology of the fracture region of the cross-group industrial asset trajectory based on the spatiotemporal feature segments broadcast in the adjacent federated learning groups and the signal coverage blind zone information corresponding to the polarization direction, and generate a continuous trajectory prediction path based on the extended fracture region.

[0045] The verification module is used to dynamically adapt the signal attenuation characteristics of the polarization direction under extreme weather conditions to the update cycle of the continuous trajectory prediction path, and verify the spatiotemporal continuity of the polarization direction switching and the update cycle.

[0046] The generation module is used to integrate the continuous trajectory prediction paths of each federated learning group based on the signal attenuation differences of different terrain regions under the polarization direction, and generate a spatiotemporal trajectory tracking strategy that integrates multi-terrain dynamic communication features based on the integrated continuous trajectory prediction paths.

[0047] Thirdly, embodiments of this application provide a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement the spatiotemporal trajectory tracking method for industrial assets based on multi-source heterogeneous positioning data fusion as described in the first aspect above.

[0048] Fourthly, embodiments of this application provide a computer storage medium storing a computer program, which, when executed by a computer, implements a method for tracking the spatiotemporal trajectory of industrial assets based on the fusion of multi-source heterogeneous positioning data as described in the first aspect.

[0049] In this embodiment, multi-source heterogeneous asset location data is collected by edge nodes deployed on industrial assets. Geometric features of terrain reflective surfaces in the signal propagation path are inferred from the asset location data based on changes in industrial asset location. The polarization direction of the LoRa-satellite communication link is selected based on these geometric features. The edge nodes are divided into federated learning groups according to geographical regions. Within each federated learning group, the asset location data is transmitted via the polarization direction. Spatiotemporal feature segments of the industrial asset trajectory are generated based on the transmission results and broadcast across groups to adjacent federated learning groups via the LoRa-satellite communication link. Based on the broadcast spatiotemporal feature segments in adjacent federated learning groups and combined with signal coverage blind spot information corresponding to the polarization direction, the fracture regions of the cross-group industrial asset trajectories are topologically extended. A continuous trajectory prediction path is generated based on the extended fracture regions. Based on the signal attenuation differences in different terrain regions under the polarization direction, the continuous trajectory prediction paths of each federated learning group are integrated. A spatiotemporal trajectory tracking strategy is generated based on the integrated continuous trajectory prediction path.

[0050] The technical solution of this application has the following beneficial effects:

[0051] By inferring the geometric features of terrain reflectors and selecting polarization directions using multi-source heterogeneous positioning data, the problem of signal multipath interference and attenuation under complex terrain is solved, enabling dynamic optimization of communication links. Based on polarization directions, federated learning groups are divided and spatiotemporal feature fragments are broadcast across groups to construct a distributed collaborative framework, improving the transmission efficiency and integrity of trajectory data in base station-free areas. By combining polarization direction blind zone information to extend broken trajectories, trajectory interruptions caused by terrain occlusion are compensated, and continuous predicted paths are generated, enhancing the robustness of trajectory tracking under complex terrain. By integrating the differences in signal attenuation across multiple terrains to generate a global tracking strategy, it adapts to the dynamic communication characteristics of mixed terrains such as mountains and canyons, improving the environmental adaptability of the tracking system.

[0052] Furthermore, the boundaries of federated learning groups are delineated based on the geometric features of the terrain reflective surfaces corresponding to the polarization direction. Stable edge nodes are selected through directional matching, and asset positioning data is segmented according to transmission stability. Local trajectory sequences are generated based on temporal continuity and positional offset constraints, spatiotemporal correlation feature segments are extracted, and cross-group broadcasting is performed after prioritizing coverage according to polarization direction. Through dynamic partitioning and stable segmented transmission of federated learning groups driven by polarization direction, the packet loss rate of data transmission under complex terrain is reduced. The priority broadcasting mechanism based on spatiotemporal correlation feature segments enables efficient collaboration of cross-group trajectory data, effectively solving the problems of limited communication resources and trajectory fragmentation in remote areas.

[0053] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0054] 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.

[0055] Figure 1 A flowchart of a spatiotemporal trajectory tracking method for industrial assets based on multi-source heterogeneous positioning data fusion provided in this application is shown;

[0056] Figure 2 This paper presents a schematic diagram of the structure of an industrial asset spatiotemporal trajectory tracking system based on multi-source heterogeneous positioning data fusion provided in this application.

[0057] Figure 3 A schematic diagram of the structure of a computing device provided in this application is shown. Detailed Implementation

[0058] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0059] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0060] Researchers have found that in remote mining areas, oil fields, and other areas without base station coverage, existing industrial asset tracking solutions struggle to overcome multipath interference, communication interruptions, and trajectory breaks caused by complex terrain. Furthermore, traditional methods suffer from uncontrollable signal attenuation and low efficiency in location data collaboration under extreme weather conditions. To address these shortcomings, this paper proposes a spatiotemporal trajectory tracking method for industrial assets based on multi-source heterogeneous location data fusion. Specifically, it constructs a distributed trajectory data collaboration framework among federated learning groups by dynamically adapting the communication link polarization direction selection mechanism to terrain reflection characteristics. Combined with signal coverage blind spot compensation and multi-terrain attenuation difference fusion techniques, it generates continuous spatiotemporal trajectory prediction paths. This method can suppress multipath interference, repair trajectory breaks, and improve tracking accuracy and system robustness under complex terrain and dynamic weather conditions in environments without base station coverage.

[0061] The technical solution of this application can be applied to mobile asset tracking scenarios in remote mining areas / oil fields and other areas without base station coverage.

[0062] 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 embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0063] Figure 1 This application provides a flowchart of a method for spatiotemporal trajectory tracking of industrial assets based on multi-source heterogeneous positioning data fusion, as shown in the embodiments of this application. Figure 1 As shown, the method includes:

[0064] 101. Collect multi-source heterogeneous asset positioning data by edge nodes deployed on industrial assets, extract the geometric features of the terrain reflection surface in the signal propagation path from the asset positioning data to infer the changes in the location of industrial assets, and select the polarization direction of the LoRa and satellite communication link according to the geometric features, wherein the polarization direction includes the vertical polarization direction and the horizontal polarization direction.

[0065] In this step, the multi-source heterogeneous asset positioning data includes satellite positioning coordinates, acceleration / angular velocity data from the inertial navigation module, and radio frequency signal strength data, which are used to comprehensively infer the asset location and terrain features. The geometric features of the terrain reflector are physical parameters such as the tilt angle, height difference, and density distribution of obstacles in the signal propagation path, which are used to determine the polarization direction switching strategy.

[0066] In this embodiment, firstly, edge nodes synchronously collect satellite positioning, IMU motion data, and radio frequency signal strength, and generate a time-aligned asset trajectory sequence after removing abnormal jump points; secondly, based on the movement direction and speed changes of adjacent points in the asset trajectory sequence, an initial straight-line propagation path is drawn, and the curvature of the straight-line propagation path is corrected by combining IMU angular velocity data to match the actual terrain occlusion distribution; thirdly, the obstacle height difference (laser ranging data) and density distribution (signal reflection intensity fluctuation) are measured along the corrected propagation path, and strong reflection areas (height difference > 5 meters) and multipath interference areas (density > 3 obstacles / 10 meters) are marked; finally, LoRa communication link vertical polarization is used for strong reflection areas to suppress tilt reflection interference, and satellite communication link horizontal polarization is switched for multipath interference areas to enhance penetration capability.

[0067] Suppose that in an iron ore mining area, the edge nodes on the transport truck collect GPS coordinates (accuracy ±2 meters) and IMU angular velocity data to infer the signal propagation path when it travels along the canyon. After path correction, it is found that the slope angle of the mountain on the right is 35° and the obstacle density on the left is 4 obstacles / 10 meters (rubble pile). Based on this, LoRa vertical polarization is selected for the right path, and satellite horizontal polarization is switched for the left path.

[0068] 102. Divide the edge nodes into federated learning groups according to geographical regions, transmit the asset positioning data within the federated learning group through the polarization direction, generate spatiotemporal feature segments of industrial asset trajectories based on the transmission results, and broadcast the spatiotemporal feature segments across groups to adjacent federated learning groups through the LoRa and satellite communication link.

[0069] In this step, the federated learning group is a geographical region group divided according to the signal coverage capability of the polarization direction, and the edge nodes within the group share the same terrain reflection features.

[0070] In this embodiment, firstly, based on the signal straight-line penetration distance under the polarization direction (e.g., 500 meters for vertical polarization coverage), the geographical area is divided into multiple fan-shaped area groups, and the group boundaries are aligned with the geometric features of the terrain reflective surface. Secondly, within the aligned groups, edge nodes with signal strength fluctuations of <10% in the current polarization direction are selected as relay nodes, and asset positioning data is divided into high-priority feature segments (signal strength > -90dBm) and low-priority feature segments according to the transmission stability of the relay nodes. Thirdly, the time interval (<1 second) and position offset (<5 meters) between adjacent points are calculated for the high-priority feature segment data to obtain the velocity change trend of the continuous trajectory segment (acceleration within ±0.5m / s²). Finally, the high-priority feature segments and their velocity change trends are broadcast to the adjacent end group every 30 seconds via the LoRa link, and the low-priority segments are broadcast to the distant end group every 5 minutes via the satellite link.

[0071] For example, continuing the previous example, within the federated learning group (vertical polarization coverage area) on the west side of the iron ore mine, three stable relay nodes are selected to divide the truck's trajectory data from 10:00 to 10:05 into five segments, three of which meet the high stability condition (signal strength -85dBm); the data is broadcast in real time to the adjacent eastern group via LoRa, and simultaneously uploaded to the central server via satellite.

[0072] 103. Based on the spatiotemporal feature segments broadcast in the adjacent federated learning groups, and combined with the signal coverage blind zone information corresponding to the polarization direction, the fracture region of the cross-group industrial asset trajectory is topologically extended, and a continuous trajectory prediction path is generated based on the extended fracture region.

[0073] In this step, topology extension is a method for filling in broken regions based on the characteristics of adjacent trajectory groups, generating a predicted path that bypasses the blind zone.

[0074] In this embodiment, firstly, the start and end points of trajectory breaks are marked based on the difference in timestamps of feature segments broadcast by adjacent federated learning groups being greater than 30 seconds or the position offset being greater than 50 meters; secondly, the movement direction (angle difference < 15°) and average velocity (difference < 2m / s) of feature segments on both sides of the fracture area formed by the start and end points of the trajectory breaks are extracted, and three candidate paths are generated along the signal coverage boundary of the polarization direction; thirdly, the obstacle density (laser point cloud data) on the candidate paths is measured, and the path with the lowest density is selected as the extension path; finally, the slope of the extension path is matched with the fracture endpoint of the cross-group industrial asset trajectory (difference < 5°), and the part exceeding the coverage range of the polarization direction is deleted to generate a continuous trajectory prediction path.

[0075] For example, continuing the previous example, the truck entered the blind spot of the canyon at 10:06, causing the trajectory to break. After receiving the feature segments broadcast by the east group, the west group generated 3 candidate paths based on the movement directions of both sides (both 10° east of north). The lidar detected that the obstacle density of the middle path was 1 obstacle / meter (the lowest). This path was selected and fused to generate a continuous trajectory prediction path.

[0076] 104. Based on the signal attenuation differences in different terrain regions under the polarization direction, integrate the continuous trajectory prediction paths of each federated learning group, and generate a spatiotemporal trajectory tracking strategy based on the integrated continuous trajectory prediction paths.

[0077] In this step, the signal attenuation difference is the attenuation gradient of signal strength as the terrain changes under different polarization directions.

[0078] In this embodiment, firstly, the continuous trajectory prediction paths of each federated learning group are marked as high / low reliability paths according to terrain type (mountain / plain) and polarization direction signal attenuation difference (e.g., mountains > 0.15dB / m); secondly, high reliability paths are sorted in ascending order of signal attenuation, and low reliability paths are sorted in descending order of signal attenuation fluctuation amplitude (standard deviation < 2dB); thirdly, in the path intersection and overlap area (e.g., mountain-plain boundary), a weighted fusion strategy is used to merge the sorted paths to form a spatiotemporal trajectory tracking strategy; finally, the weight allocation is adjusted every 15 minutes according to real-time weather data (rain attenuation coefficient).

[0079] For example, continuing the previous example, the mountain path on the west side (vertical polarization attenuation of 0.18 dB / m) and the plain path on the east side (horizontal polarization attenuation of 0.08 dB / m) overlap at the entrance of the mining area. The tracking path is generated by merging the paths with a high reliability weight of 0.7 and a low reliability weight of 0.3. During rainy weather, the weights are adjusted to 0.6:0.4 to compensate for the impact of rain attenuation.

[0080] Steps 101-104 suppress multipath interference in complex terrain by using a polarization direction selection mechanism that dynamically adapts to terrain reflection characteristics; improves communication efficiency in base station-free areas by dividing into federated learning groups and prioritizing data transmission; repairs trajectory breaks by using topology extension technology based on signal coverage blind spots; and enhances the system's environmental adaptability through a multi-terrain attenuation difference fusion strategy, ultimately achieving high-precision and high-reliability tracking of mobile assets in remote mining areas.

[0081] To address the issues of low transmission efficiency and fragmented trajectories of industrial asset trajectory data in remote mining areas / oil fields and other areas without base station coverage due to terrain obstruction and limited communication resources, some embodiments involve dividing the edge nodes into federated learning groups based on geographical regions, transmitting asset location data within each federated learning group via the polarization direction, generating spatiotemporal feature segments of the industrial asset trajectory based on the transmission results, and broadcasting these spatiotemporal feature segments across groups to adjacent federated learning groups via the LoRa and satellite communication link. This includes:

[0082] 201. Based on the geometric features corresponding to the polarization direction, the edge nodes are divided into federated learning groups that cover the continuous region of the same reflective surface. The boundary of the federated learning group is determined by the straight-line penetration range of the signal propagation path under the polarization direction.

[0083] In step 201, the federated learning group is a geographical region group divided based on the same terrain reflection characteristics, and the edge nodes within the group share similar signal propagation environments. The straight-line penetration range is the farthest effective distance at which a wireless signal can propagate in a straight line under the polarization direction.

[0084] In this embodiment, firstly, the tilt angle of the terrain reflective surface under the current polarization direction (measured by the lidar of the edge nodes) and the obstacle density (based on the statistical analysis of radio frequency signal reflection intensity fluctuations) are extracted, and areas with a tilt angle > 25° and an obstacle density > 3 per 10 meters are marked as steep slope high-density areas; secondly, based on the physical characteristics of the signal propagation path under the polarization direction (e.g., the straight-line penetration distance of vertical polarization in a mountain scene is 300 meters), the edge nodes are divided into multiple federated learning groups based on the geometric boundary of the steep slope high-density areas, ensuring that all edge nodes in the group are within the effective coverage area of ​​the same polarization direction; finally, the group boundaries are dynamically adjusted through geofencing technology, and the scope of the federated learning group is redefined when the movement of industrial assets causes changes in the terrain reflective surface.

[0085] 202. Within the federated learning group, directional matching is performed on the transmission parameters of the asset location data based on the polarization direction. Edge nodes that form a stable transmission relationship with the polarization direction are selected according to the matching results. The asset location data is divided into multiple data segments in descending order of transmission stability through the edge nodes.

[0086] In step 202, directional matching involves adjusting data transmission parameters (such as transmission angle and frequency) to create an optimal angle between the signal propagation direction and the polarization direction.

[0087] In this embodiment, firstly, the signal strength of each edge node within the federated learning group is tested at different transmission angles (e.g., the antenna angle is adjusted in 5° increments, and the received signal strength RSSI is measured), and nodes with an angle error of <3° with the current polarization direction are selected as candidate nodes; secondly, the data transmission success rate of the candidate nodes within 10 minutes is calculated, nodes with a success rate <95% are removed, and the remaining nodes are marked as stable transmission nodes; finally, based on the real-time signal strength of the stable transmission nodes (e.g., -85dBm is the high stability threshold), the asset positioning data is divided into high-stability data segments (signal strength ≥ -85dBm) and low-stability data segments (signal strength < -85dBm), and priority labels are assigned to the high-stability data segments.

[0088] 203. Calculate the time interval and position offset between the multiple data segments, combine the parts that are continuous in time and whose position offset is less than a preset distance threshold into a local trajectory sequence, and generate an industrial asset trajectory feature segment containing spatiotemporal correlation based on the movement direction and speed change pattern of the local trajectory sequence.

[0089] In step 203, time continuity means that the time interval between adjacent trajectory points is ≤1 second. Position offset is the difference in straight-line distance between adjacent trajectory points.

[0090] In this embodiment, firstly, the trajectory points in the high-stability data segment are sorted by timestamp, the time interval between adjacent points is calculated, and breakpoints with an interval > 1 second are removed; secondly, the position offset of the remaining points is calculated (e.g., using the Euclidean distance formula), and if the offset > 5 meters, it is marked as an anomaly and deleted; thirdly, the movement direction of the processed local trajectory sequence (time interval ≤ 1 second and offset ≤ 5 meters) is calculated (by obtaining the azimuth angle through the coordinate difference of adjacent points), and the velocity change trend is statistically analyzed (absolute acceleration value ≤ 0.5 m / s²); finally, the local trajectory sequences with consistent direction (azimuth angle fluctuation < 10°) and continuous velocity (acceleration fluctuation < 0.3 m / s²) are merged into industrial asset trajectory feature segments, and the feature segments contain position, direction, and velocity labels.

[0091] 204. Sort the industrial asset trajectory feature segments according to the signal coverage priority corresponding to the polarization direction, and broadcast the sorted feature segments to the adjacent federated learning groups in order of coverage range from near to far through the LoRa and satellite communication links.

[0092] In step 204, the signal coverage priority is sorted according to the coverage capability of the polarization direction in different terrains.

[0093] In this embodiment, based on the signal coverage capability corresponding to the polarization direction, the industrial asset trajectory feature segment is divided into short-range coverage features and long-range coverage features. Based on the straight-line penetration distance of the signal propagation path under the polarization direction, the short-range and long-range coverage features are sorted by coverage range, with the short-range features sorted from shortest to longest penetration distance and the long-range features sorted from longest to shortest penetration distance. The sorted short-range coverage features are broadcast via LoRa communication link in order of coverage range from near to far, while the sorted long-range coverage features are broadcast via satellite communication link in order of coverage range from far to near. Based on the coverage feature reception integrity feedback from neighboring federated learning groups, the sorting weights of the short-range and long-range coverage features are dynamically adjusted to match the subsequent broadcast order with the signal coverage capability.

[0094] Here is a specific example:

[0095] Suppose that in an iron ore mining area, a transport truck travels from a canyon (steep slope angle 38°, vertical polarization penetration distance 300 meters) to a scree plain (horizontal polarization penetration distance 800 meters). The edge nodes execute the following process: Step 201: Based on the geometric characteristics of the vertical polarization direction in the canyon (slope angle > 25°, obstacle density 4 / 10 meters), the canyon area is classified as Federated Learning Group A, and the group boundary is determined by the vertical polarization straight-line penetration range (300 meters); the plain area is classified as Group B, and the group boundary is determined by the horizontal polarization penetration range (800 meters). Step 202: Select 3 stable nodes in group A (vertical polarization direction angle error < 2°, transmission success rate 98%), and divide the truck trajectory data from 10:00 to 10:05 into 4 segments, of which 3 segments are high-stability data segments (signal strength -80dBm) and 1 segment is a low-stability segment (signal strength -92dBm); Step 203: Extract continuous trajectory sequences (time interval ≤ 0.8 seconds, position offset ≤ 4 meters) from the high-stability data segments to generate industrial asset trajectory feature segments with directional fluctuation < 8° and consistent speed trend. In step 204, the canyon area feature segments are first marked as near-range coverage features (vertical polarization penetration distance 300 meters), and the plain area segments are marked as long-range coverage features (horizontal polarization penetration distance 800 meters). The near-range features are sorted from shortest to longest penetration distance (300 meters → 350 meters → 400 meters), and the long-range features are sorted from longest to shortest penetration distance (800 meters → 750 meters → 700 meters). The near-range features are broadcast to adjacent group B via LoRa in the order from near to far (300 meters → 350 meters → 400 meters), and the long-range features are broadcast by satellite in the order from far to near (800 meters → 750 meters → 700 meters). Finally, group B reports that the 350-meter segment of the near-range features is lost (signal strength < -90 dBm), and the subsequent broadcasts will increase the weight of the 350-meter segment to the highest priority for priority retransmission; the 700-meter segment of the long-range features is received completely and the original order is maintained.

[0096] Steps 201-204 dynamically divide federated learning groups based on the geometric features of terrain reflective surfaces to optimize communication resource allocation; select stable nodes based on polarization direction angle matching to reduce data transmission packet loss rate under complex terrain; generate high-precision trajectory feature segments through spatiotemporal continuity constraints to suppress abnormal interference; and use a priority-driven cross-group broadcast mechanism to adapt to the differences in signal coverage across various terrains. Ultimately, this achieves highly reliable tracking of industrial asset trajectories in base station-free iron ore mining areas, solving the trajectory breakage problem caused by multipath reflection and signal attenuation.

[0097] To address the issues of signal propagation path distortion, severe multipath interference, and poor communication link selection adaptability caused by complex terrain in industrial assets in remote mining areas / oil fields and other areas without base station coverage, some embodiments extract the geometric features of the terrain reflection surface in the signal propagation path from the asset location data, and select the polarization direction of the LoRa and satellite communication link based on the geometric features. The polarization direction includes a vertical polarization direction and a horizontal polarization direction, including:

[0098] 301. The edge nodes synchronously collect primary and secondary positioning data of industrial assets. The primary positioning data includes the continuous position coordinates of the industrial assets, and the secondary positioning data includes the movement direction and speed changes of the industrial assets.

[0099] In step 301, one type of positioning data is the latitude and longitude coordinates of the industrial asset at continuous timestamps, used to describe its location trajectory. The other type of positioning data is the real-time movement direction and speed changes of the industrial asset, collected by the inertial navigation module (IMU).

[0100] In this embodiment, firstly, a type of positioning data (1 coordinate point per second) is collected at a frequency of 1Hz using the GPS module of the edge node, while a type of positioning data (including the rate of change of orientation angle and acceleration) is collected at a frequency of 10Hz using the IMU; secondly, coordinate jump points (distance difference between adjacent points > 20 meters) in the type of positioning data are filtered out, and the type of positioning data is timestamped (interpolated to 1Hz); finally, the processed two types of positioning data are bound together into a time-synchronized trajectory dataset for subsequent path inference.

[0101] 302. Determine the movement trajectory direction of the industrial asset based on the continuous position coordinates, draw an initial straight line along the movement trajectory direction, and correct the curvature of the initial straight line according to the movement direction and speed changes, and generate a signal propagation path that matches the terrain occlusion distribution based on the corrected initial straight line.

[0102] In step 302, the curvature correction is to adjust the curvature of the initial straight path according to the changes in the direction of motion and speed, and to match the actual terrain occlusion.

[0103] In this embodiment, firstly, five consecutive coordinate points are extracted from the first type of positioning data, and their average movement direction (e.g., 28° east of due north) is calculated. An initial straight path is then drawn along this direction. Secondly, based on the rate of change of the directional angle (e.g., 5° per second) and acceleration (e.g., 0.3 m / s²) in the second type of positioning data, a curvature correction (e.g., adding a 2° deflection angle for every 10 meters of path) is superimposed on the initial straight line. Finally, based on the comparison between the corrected path and the terrain point cloud data scanned by the lidar, invalid path segments blocked by mountains / obstacles are deleted, and a signal propagation path matching the terrain is generated.

[0104] 303. Measure the height difference of the terrain reflective surfaces and the obstacle density distribution along the signal propagation path, mark the reflective surfaces with a height difference greater than the upper limit of the difference as strong reflection areas, mark the reflective surfaces with an obstacle density distribution greater than a preset density as multipath interference areas, and extract the geometric boundary features of the strong reflection areas and the multipath interference areas.

[0105] In step 303, the height difference is the vertical drop between obstacles on both sides of the signal propagation path. The obstacle density distribution is the number of obstacles per unit distance.

[0106] In this embodiment, firstly, the terrain on both sides is scanned using a lidar along the corrected signal propagation path, and the height difference of obstacles is measured. Areas with a height difference greater than 5 meters are marked as strong reflection zones. Secondly, the number of obstacles (such as piles of rubble or mine pits) in every 10-meter interval along the path is counted, and intervals with a density greater than 3 obstacles per 10 meters are marked as multipath interference zones. Finally, the geometric boundaries (such as a tilt angle of 35° and a distribution length of 200 meters) between the strong reflection zones (height difference greater than 5 meters) and the multipath interference zones are extracted to form a terrain geometric boundary feature map.

[0107] 304. Based on the tilt angle and distribution range in the geometric boundary features, select the vertical polarization direction of the LoRa communication link to weaken the reflection interference of the tilt angle, and select the horizontal polarization direction of the satellite communication link to penetrate the distribution range.

[0108] In step 304, the tilt angle is the angle between the terrain reflecting surface and the horizontal plane. The distribution range is the continuous coverage length of the multipath interference zone along the signal propagation path.

[0109] In this embodiment, firstly, the complementary angle between the tilt angle of the strong reflection zone and the vertical polarization direction is calculated. If the complementary angle is less than 60°, LoRa vertical polarization is selected to weaken reflection interference. Secondly, the penetration distance required by the satellite horizontal polarization signal is calculated for the distribution range of the multipath interference zone (e.g., 200 meters). If the satellite link coverage capability is ≥240 meters, horizontal polarization is switched. Finally, terrain changes are monitored in real time, and polarization direction reselection is triggered when the tilt angle or distribution range exceeds the threshold.

[0110] Here is a specific example:

[0111] Assuming a mining area is in an iron ore mining region, a transport truck travels along a canyon (with an average slope of 38°) towards the mining site. The edge node executes the following process: Step 301: Collect Class I positioning data (GPS coordinates, accuracy ±2 meters) and Class II data (IMU azimuth rate of change 4° / second, acceleration 0.4m / s²) of the truck from 10:00 to 10:05, and generate a synchronous trajectory dataset after filtering; Step 302: Calculate the average movement direction as 32° east of due north based on the first 5 coordinate points, and draw an initial straight path; Combine the IMU azimuth rate of change (4° / second) to add a 3° rightward correction to the path every 20 meters, generating a curved path that bypasses the right side of the mountain (7-meter height difference from lidar scanning); Step 303: Measure the height difference of the left side of the mountain along the corrected path, which is 9 meters (>5-meter threshold), and mark it as a strong reflection zone. Meanwhile, 4 obstacles were detected every 10 meters within the right path interval (>3 obstacles / 10-meter threshold), which were marked as multipath interference areas. The tilt angle of the strong reflection area was extracted as 38° (calculated by laser point cloud normal), and the distribution length of the multipath interference area was 180 meters. In step 304, for the tilt angle of the strong reflection area as 38°, LoRa vertical polarization (complementary angle 52° < 60°) was selected. For the distribution length of the multipath interference area as 180 meters, the required distance for satellite horizontal polarization was calculated as 216 meters (180 × 1.2). Since the satellite currently covers 240 meters, the horizontal polarization was switched.

[0112] Steps 301-304 generate a terrain-matched signal propagation path through multi-source positioning data, accurately extract strong reflection and multipath interference features, and dynamically select the polarization direction based on physical parameters. This significantly reduces multipath reflection interference and signal attenuation in complex terrains such as iron ore mining areas, improves communication link stability and asset tracking accuracy, and solves the problems of trajectory distortion and communication interruption caused by terrain obstruction in base station-free areas.

[0113] To address the issue of fragmented industrial asset trajectories caused by signal coverage blind spots in remote mining areas / oil fields and other areas without base station coverage, some embodiments involve topologically extending the fragmented regions of cross-group industrial asset trajectories based on spatiotemporal feature segments broadcast in adjacent federated learning groups, combined with signal coverage blind spot information corresponding to the polarization direction, and generating continuous trajectory prediction paths based on the extended fragmented regions, including:

[0114] 401. Based on the signal coverage blind zone information corresponding to the polarization direction, determine the starting point and ending point of the fracture region of the cross-group industrial asset trajectory, wherein the starting point and ending point are the position coordinates of the first and last valid trajectory points in the adjacent federated learning group, respectively.

[0115] In step 401, the starting point and the ending point are the first and last valid trajectory points in the trajectory break region, determined by the latest and earliest position coordinates within the adjacent federated learning group.

[0116] In this embodiment, firstly, based on a signal coverage blind zone map (pre-mapped or generated in real time) of polarization direction (e.g., vertical polarization), the interval where the signal strength is consistently below -100dBm in the industrial asset trajectory is located; within this interval, the first and last valid points of the trajectory data are obtained from adjacent federated learning groups (e.g., group A and group B). If the time interval between the end point timestamp of group A and the start point timestamp of group B is >30 seconds and the position offset is >50 meters, it is marked as a fracture region; finally, the boundary of the fracture region is defined with the coordinates of the end point of group A as the starting point and the coordinates of the start point of group B as the ending point.

[0117] 402. Select trajectory segments that match the starting point and the ending point from the spatiotemporal feature segments broadcast by the adjacent federated learning groups, and directionally superimpose the movement direction of the trajectory segments with the shape of the signal coverage blind zone corresponding to the polarization direction to form a candidate extension path.

[0118] In step 402, directional overlay is to geometrically align the movement direction of the trajectory segment with the shape of the blind zone to generate candidate paths.

[0119] In this embodiment, firstly, trajectory segments matching the movement direction and speed of the starting and ending points are selected from the spatiotemporal feature segments broadcast by adjacent federated learning groups. The matching requirements are a direction difference of <8° and a speed difference of <1m / s. Secondly, the boundary azimuth angle of the blind zone shape is extracted (e.g., if the blind zone is elongated, the main direction is northeast-southwest). The movement direction of the matched trajectory segments is superimposed with the main direction of the blind zone (e.g., the trajectory direction is superimposed at 25° due north east and the main direction of the blind zone is superimposed at 30° northeast to generate a path with an angle of 55°). Finally, three candidate extension paths (distributed at 5° intervals) are generated along the superposition direction to cover the passable area of ​​the blind zone.

[0120] 403. Dynamically adjust the curvature of the candidate extension path according to the signal reflection intensity distribution in the fracture area, and generate an arc-shaped extension path that fits the boundary of the signal coverage blind zone with the adjusted curvature as a constraint.

[0121] In step 403, the signal reflection intensity distribution is the gradient change of signal reflection intensity at different locations within the blind zone. The curvature adjustment dynamically changes the path curvature based on the reflection intensity gradient to avoid high-reflection interference areas.

[0122] In this embodiment, firstly, the average signal reflection intensity of every 10-meter interval on the candidate extension path is measured (e.g., -90dBm in the edge area and -110dBm in the center area), and a reflection intensity distribution map is generated; secondly, areas with a reflection intensity decrease gradient > 5dB / 10 meters in the reflection intensity distribution map are marked as high interference areas, and the curvature to be detoured in the high interference areas is calculated (e.g., a 3° deflection is added to every 10 meters of the path); finally, the curvature of the candidate extension path is adjusted according to the curvature to generate an arc-shaped extension path that fits the boundary of the blind zone (within 5 meters of the boundary).

[0123] 404. Smoothly connect the arc-shaped extension path and the starting point and ending point to form a continuous arc-shaped extension path. Delete the portion of the continuous arc-shaped extension path that does not overlap with the signal coverage area of ​​the polarization direction to generate a continuous trajectory prediction path.

[0124] In step 404, the endpoint smooth connection is to control the difference between the slope of the endpoint of the arc path and the slope of the start and end points of the fracture area to within 5°.

[0125] In this embodiment, firstly, the slope of the movement direction between the endpoint of the arc-shaped extension path and the starting and ending points of the fracture area is calculated (e.g., the slope of the starting point is 25° and the slope of the arc-shaped endpoint is 28°). If the difference is greater than 5°, a transition path segment (e.g., a 2-meter-long buffer segment) is inserted. Secondly, the distance between the arc-shaped path and the polarization direction coverage boundary is measured, and segments exceeding the 300-meter coverage range are deleted. Finally, the remaining path after the above processing is spliced ​​with the original trajectory to generate a continuous trajectory prediction path.

[0126] Here is a specific example:

[0127] Suppose that in an iron ore mining area, a transport truck is traveling from group A (canyon vertical polarization coverage area) to group B (plain horizontal polarization coverage area) in a federated learning group. Due to signal obstruction by the mountain in the middle of the canyon, it enters a signal blind zone (vertical polarization signal strength -105dBm). The following process is executed: Step 401: Based on the end point of group A (10:05:00, coordinates X1, Y1) and the starting point of group B (10:05:35, coordinates X2, Y2), mark the break area (time interval 35 seconds, position offset 60 meters); Step 402: Filter the broadcast data of group B to find the trajectory that matches the direction of the end point of group A (22° east of due north). The fragment (direction difference 3°, velocity difference 0.8m / s) is superimposed with the blind zone northeast-southwest main direction (30°) to generate a candidate extension path (angle 28°); step 403 measures the reflection intensity of the candidate path, the intensity in the central area is -110dBm (gradient decrease 8dB / 10m), adjusts the curvature of the path (4° deflection every 10m), and generates an arc-shaped extension path 3m from the boundary of the blind zone; step 404 inserts the endpoint (slope 25°) and the start and end points (slope 22°) of the arc-shaped path into a 2-meter buffer section, deletes the far 10-meter path that exceeds the vertical polarization coverage of 300m, and finally generates a continuous trajectory prediction path.

[0128] Steps 401-404 use polarization direction signal coverage blind zone to dynamically locate trajectory breakage areas, combine adjacent group trajectory direction matching and blind zone shape superposition to generate candidate paths, adaptively adjust path curvature based on reflection intensity distribution and delete over-coverage segments, and finally achieve high-precision trajectory repair in base station-free iron ore mountain areas, significantly reducing the risk of communication interruption and trajectory breakage caused by complex terrain.

[0129] To address the issue of discontinuous and low-precision generation of spatiotemporal feature segments of industrial asset trajectories in remote mining areas / oil fields and other areas without base station coverage due to abrupt changes in motion state, some embodiments include generating industrial asset trajectory feature segments containing spatiotemporal correlations based on the movement direction and speed variation patterns of the local trajectory sequence, including:

[0130] 501. Calculate the difference in movement direction between adjacent trajectory points in the local trajectory sequence, and mark the continuous trajectory segments whose difference in movement direction is less than the upper limit of the difference as directionally stable segments;

[0131] In step 501, the directional stable segment is the trajectory interval where the difference in the movement direction of continuous trajectory points is less than a preset threshold.

[0132] In this embodiment of the application, firstly, the azimuth angle of the movement direction of every two adjacent trajectory points in the local trajectory sequence is calculated (e.g., the tangent value is obtained through coordinate difference), and the absolute value of the difference is calculated; secondly, the trajectory sequence is traversed, and the interval with more than 3 consecutive points and the absolute value of the direction difference is less than 10° is marked as the direction stable segment, and the slope of the first and last points of the direction stable segment is smoothed (e.g., a transition point is inserted to eliminate abrupt changes).

[0133] 502. Extract a set of trajectory points with consistent velocity change trends from the stable directional segment, and divide the set of trajectory points into acceleration and deceleration segments according to the direction of increase or decrease of the velocity change trends.

[0134] In step 502, the acceleration segment and the deceleration segment are the intervals in which the velocity change trend of the continuous trajectory points is consistent.

[0135] In this embodiment, firstly, the velocity difference between adjacent trajectory points (e.g., velocity of the later point minus velocity of the earlier point) is calculated in the directional stable segment. If the difference is greater than 0.2 m / s², it is marked as an acceleration trend; if it is less than -0.2 m / s², it is marked as a deceleration trend. Secondly, trajectory points with three or more consecutive acceleration trends are merged into an acceleration segment, and points with three or more consecutive deceleration trends are merged into a deceleration segment. Finally, velocity interpolation (e.g., linear interpolation) is performed on the first and last points of the acceleration and deceleration segments to eliminate velocity jumps within the segments.

[0136] 503. Combine the portions of the acceleration and deceleration segments that have the same direction of movement and continuous speed trend into basic feature units;

[0137] In step 503, the azimuth fluctuation of trajectory points within the same finger segment in the same direction of movement is less than 5°. The velocity trend is continuous when the absolute value change of acceleration / deceleration is less than 0.1 m / s².

[0138] In this embodiment, firstly, the intervals in the acceleration and deceleration segments where the direction of movement fluctuates by less than 5° (e.g., 12° ± 3° east of due north) are extracted; secondly, the standard deviation of the acceleration within the segment is calculated (e.g., < 0.1 m / s²), and intervals where the absolute value of acceleration / deceleration changes by less than 0.1 m / s² are selected; finally, the intervals of the acceleration and deceleration segments that meet the conditions of direction and velocity continuity are merged into basic feature units, and isolated points with abrupt changes in direction or trend are deleted.

[0139] 504. Based on the spatiotemporal distribution density of the basic feature units, the boundaries of adjacent basic feature units are seamlessly spliced ​​together to form an industrial asset trajectory feature segment containing spatiotemporal correlation.

[0140] In step 504, the spatiotemporal distribution density is the number of basic feature units per unit time or space. Seamless stitching means that the difference between the azimuth angle and velocity of adjacent unit boundaries is less than a preset threshold.

[0141] In this embodiment, firstly, the distribution density of basic feature units in time and space is statistically analyzed (e.g., one unit every 50 meters in a bend area of ​​a mining area), and high-density areas are marked. Secondly, the azimuth difference and velocity difference are calculated for the boundary points of adjacent units in the high-density areas. If the azimuth difference is <3° and the velocity difference is <0.5m / s, they are directly spliced. If the threshold is exceeded, transition trajectory points are inserted (e.g., intermediate points are generated by linear interpolation) to achieve smooth splicing. Finally, the discontinuous parts of direction or velocity in the spliced ​​trajectory are deleted to generate industrial asset trajectory feature segments containing spatiotemporal correlation.

[0142] Here is a specific example:

[0143] Assuming a transport truck is traveling along a canyon bend in an iron ore mining area, the edge nodes execute the following process: Step 501: Calculate the direction difference between adjacent points from the truck's local trajectory sequence from 10:00 to 10:05, and select 10 consecutive points (all direction differences < 8°) as the directionally stable segment (15° ± 5° east of due north); Step 502: Identify the acceleration segment (10:00:30-10:01:00, speed increases from 2 m / s to 4 m / s) and the deceleration segment (10:02:00-10:02:30) within the directionally stable segment. Step 503: Extract the interval (10:00:40-10:00:50) from the acceleration section where the directional fluctuation is <4° and the standard deviation of acceleration is <0.08m / s², and merge it with similar intervals in the deceleration section to form basic feature units; Step 504: Statistically distribute one basic feature unit every 30 meters in the curve area, directly splice adjacent units (azimuth difference 2°, speed difference 0.3m / s), and insert two transition points for units with large differences (azimuth difference 6°) to generate industrial asset trajectory feature segments.

[0144] Steps 501-504 extract high-precision basic feature units through directional stability screening and velocity trend segmentation. Combined with a spatiotemporal density-driven seamless stitching mechanism, continuous spatiotemporally related trajectory feature segments are generated in complex scenarios such as curves and steep slopes in mining areas. This significantly reduces trajectory jumps and distortions caused by sudden changes in motion state, and improves the smoothness and reliability of asset tracking in base station-free areas.

[0145] To address the issues of lagging polarization adaptation and uncontrollable signal attenuation caused by dynamic terrain changes in communication links for industrial assets in remote mining areas / oil fields and other areas without base station coverage, some embodiments include selecting the vertical polarization direction of the LoRa communication link based on the tilt angle and distribution range in the geometric boundary features to reduce reflection interference from the tilt angle, and selecting the horizontal polarization direction of the satellite communication link to penetrate the distribution range, including:

[0146] 601. Based on the direction and magnitude of the tilt angle in the geometric boundary features, select the vertical polarization direction of the LoRa communication link, wherein the vertical polarization direction and the tilt angle are orthogonal.

[0147] In step 601, the orthogonal relationship means that the vertical polarization direction forms a 90° angle with the terrain slope surface in order to minimize reflection interference.

[0148] In this embodiment, firstly, the terrain reflective surface is scanned by lidar, the surface normal direction is fitted, and the tilt angle between the normal and the horizontal plane (e.g., 35° northeast direction) is calculated; secondly, the orthogonal direction of the vertical polarization direction of the LoRa communication link (i.e., 60° northwest) is determined based on the direction of the tilt angle (30° northeast), and the angular deviation between the actual polarization direction and the orthogonal direction is calculated (e.g., if the current polarization direction is 55° northwest, the deviation is 5°); finally, the LoRa antenna angle is adjusted to 60° northwest to make the vertical polarization direction strictly orthogonal to the terrain tilt surface.

[0149] 602. Based on the continuous distribution length of obstacles in the distribution range of the geometric boundary features, determine the horizontal polarization direction of the satellite communication link, wherein the horizontal polarization direction is parallel to the main extension direction of the distribution range;

[0150] In step 602, the continuous distribution length is the distance the obstacle continuously covers along the signal propagation path. The main extension direction is the dominant extension orientation of the obstacle distribution area.

[0151] In this embodiment, firstly, the density of obstacles along the signal propagation path is statistically analyzed (e.g., 3 rock piles per 10 meters), and continuous intervals with a density > 2 per 10 meters (e.g., a 200-meter gravel belt) are marked; secondly, the main extension direction of the obstacle distribution in the continuous interval is fitted using the least squares method (e.g., due east to due west, azimuth angle 90°); finally, the horizontal polarization direction of the satellite communication link is adjusted to be parallel to the main extension direction (i.e., the horizontal polarization direction is adjusted to 90°) to ensure that the signal penetrates the obstacle belt.

[0152] 603. Based on the change in the tilt angle during the movement of industrial assets, correct the angular offset of the vertical polarization direction so that the orthogonal relationship is updated synchronously with the tilt state of the terrain reflector surface.

[0153] In step 603, the angle change is the dynamic change in terrain tilt angle caused by the movement of industrial assets. The angle offset is the adjustment of the vertical polarization direction relative to the orthogonal direction.

[0154] In this embodiment, firstly, the changes in terrain tilt angle are monitored in real time (e.g., lidar data is updated every 5 seconds). If the angle change exceeds 2° (e.g., from 35° to 37°), the orthogonal direction is recalculated (60° Northwest → 63° Northwest). Secondly, based on the deviation between the current vertical polarization direction and the latest orthogonal direction (e.g., 3°), the antenna angle is adjusted at a rate of 1° per second until the deviation is zero. Finally, the signal reflection intensity of the adjusted antenna angle is verified (e.g., the reflection intensity decreases by 5dB) to confirm the orthogonal relationship is effective.

[0155] 604. Based on the density change of obstacles in the distribution range, adjust the coverage width of the horizontal polarization direction so that the horizontal polarization direction penetrates the distribution range.

[0156] In step 604, the density change is the increase or decrease in the number of obstacles per unit distance within the obstacle distribution range. The coverage width adjustment expands or contracts the coverage area of ​​the horizontally polarized signal based on the density change.

[0157] In this embodiment, firstly, the density change of obstacles in the distribution range is statistically analyzed (e.g., the density of gravel belts increases from 3 obstacles / 10 meters to 4 obstacles / 10 meters), and the additional penetration distance required is calculated (for every additional obstacle / 10 meter, the coverage width is expanded by 10 meters); secondly, the beamwidth of the satellite's horizontally polarized antenna is adjusted (e.g., from 20° to 22°) to expand the coverage range to 220 meters; finally, the signal penetration capability after expansion is verified (e.g., the packet loss rate decreases from 10% to 6%) to confirm the adaptability of the coverage width.

[0158] Here is a specific example:

[0159] Assuming a transport truck travels along a canyon (mountain inclination angle 35° northeast) towards an obstacle-dense area (200-meter scree belt, main extension direction due east-due west) in an iron ore mining area, the edge node executes the following process: In step 601, the lidar scans the mountain surface, fits the inclination angle of 35° northeast, and adjusts the LoRa vertical polarization direction to 60° northwest (orthogonal direction), reducing signal reflection interference by 8dB; In step 602, the main direction of the scree belt obstacle distribution is identified as due east-due west (azimuth angle 90°), and the satellite horizontal polarization direction is adjusted to 90°, reducing the packet loss rate from 12% to 7%; In step 603, the truck's movement causes the mountain inclination angle to increase to 38°, the orthogonal direction is recalculated to 63° northwest, and the LoRa antenna is adjusted to the new direction at a rate of 1° per second, maintaining reflection interference <-90dBm; In step 604, the scree belt density increases to 4 obstacles / 10 meters, the satellite horizontal polarization coverage width is expanded to 220 meters, and the packet loss rate is further reduced to 5%.

[0160] Steps 601-604 dynamically adjust the vertical and horizontal polarization directions by real-time monitoring of terrain tilt angle and obstacle distribution. This enables adaptive matching of communication links and terrain reflection characteristics in complex scenarios such as iron ore mountainous areas, significantly suppressing multipath interference and signal attenuation, and improving the integrity and reliability of industrial asset tracking data in dynamic environments.

[0161] To address the issue of static tracking rules and insufficient environmental adaptability in industrial asset trajectories in remote mining areas / oil fields and other areas without base station coverage due to differences in communication characteristics across diverse terrains, some embodiments involve integrating continuous trajectory prediction paths from various federated learning groups based on signal attenuation differences in different terrain regions under the polarization direction, and generating a spatiotemporal trajectory tracking strategy based on the integrated continuous trajectory prediction paths, including:

[0162] 701. According to the terrain reflector type corresponding to the polarization direction, the continuous trajectory prediction path of each federated learning group is divided into flat terrain path and complex terrain path, wherein the signal attenuation of the flat terrain path is lower than that of the complex terrain path.

[0163] In step 701, the terrain reflector type is a terrain category classified according to the polarization direction signal attenuation, including flat terrain and complex terrain.

[0164] In this embodiment, firstly, the polarization direction signal attenuation of the trajectory prediction paths within each federated learning group is statistically analyzed (e.g., by calculating the average value using historical communication data). Paths with attenuation ≤ 0.1 dB / m are marked as flat terrain paths (e.g., gravel plains in mining areas), and paths with attenuation > 0.1 dB / m are marked as complex terrain paths (e.g., steep slopes in canyons). Secondly, based on the geographical coordinates of the paths, flat and complex terrain paths are grouped according to their spatial distribution to ensure that the signal attenuation characteristics of paths within the same group are consistent.

[0165] 702. Extract the section with the smallest signal strength attenuation in the flat terrain path and the section with the largest signal reflection intensity fluctuation in the complex terrain path, and mark them as the stable communication area and the dynamic compensation area, respectively.

[0166] In step 702, the signal strength attenuation is the slope of signal strength decay with distance on flat terrain paths. The signal reflection intensity fluctuation is the standard deviation of signal strength on complex terrain paths.

[0167] In this embodiment, firstly, the signal strength attenuation slope per 100 meters is calculated for flat terrain paths, and the top 20% of segments with the smallest slope (e.g., a decrease of 1 dB per 100 meters) are selected; secondly, the signal strength standard deviation is calculated for complex terrain paths, and the top 20% of segments with the largest fluctuations (e.g., standard deviation > 8 dB) are selected; finally, the selected flat segments are marked as stable communication areas, the complex segments are marked as dynamic compensation areas, and terrain type labels are added to each segment.

[0168] 703. Arrange the trajectory prediction paths in the stable communication zone in ascending order of signal attenuation, and arrange the trajectory prediction paths in the dynamic compensation zone in descending order of signal reflection intensity fluctuation amplitude, and form a priority superimposed trajectory sequence based on the two arrangement results.

[0169] In step 703, the signal attenuation is ordered in ascending order for flat terrain paths, based on the attenuation per 100 meters. The fluctuation amplitude is ordered in descending order for complex terrain paths, based on the standard deviation of signal strength.

[0170] In this embodiment, firstly, the flat paths in the stable communication zone are sorted in ascending order of attenuation, with priority given to low attenuation paths (e.g., paths with 1 dB attenuation per 100 meters); secondly, the complex paths in the dynamic compensation zone are sorted in descending order of fluctuation amplitude, with priority given to high fluctuation paths (e.g., paths with a standard deviation of 8 dB); finally, the two sorting results are superimposed with preset weights (70% for flat paths and 30% for complex paths) to generate a trajectory sequence with comprehensive priority.

[0171] 704. Based on the spatial overlap of the stable communication area and the dynamic compensation area in the trajectory sequence, integrate the continuous trajectory prediction paths of each federated learning group to generate dynamic tracking rules covering multi-terrain mixed areas.

[0172] In step 704, spatial overlap is the coverage ratio of the stable communication area and the dynamic compensation area within the same geographical range. The dynamic tracking rule is a path fusion strategy generated based on priority trajectory sequences and terrain overlap features.

[0173] In this embodiment, firstly, the geographical coordinate overlap range between the stable communication zone and the dynamic compensation zone is calculated (e.g., the overlapping area is within 500 meters of the mine entrance); secondly, within the geographical coordinate overlap range, a weighted fusion strategy (70% weight for the stable zone and 30% for the dynamic zone) is used to integrate the continuous trajectory prediction path to generate a hybrid path, while the non-overlapping area directly uses the priority path; finally, the weight allocation of the weighted fusion strategy is dynamically adjusted according to real-time weather data (e.g., rain attenuation coefficient) to generate tracking rules covering multi-terrain mixed areas.

[0174] Here is a specific example:

[0175] Suppose that in a remote iron ore mining area, a transport truck travels from a gravel plain (flat terrain, vertical polarization attenuation 0.08 dB / m) to a steep canyon slope (complex terrain, horizontal polarization attenuation 0.25 dB / m) and executes the following process: In step 701, the plain path of Federated Learning Group A is marked as flat terrain (attenuation 0.08 dB / m), and the canyon path of Group B is marked as complex terrain (attenuation 0.25 dB / m); Step 702 extracts the 10 km segment with the lowest signal attenuation slope (0.7 dB / m per 100 m) from the plain path. B) Marking the stable communication zone, the 5-kilometer segment with the largest fluctuation (standard deviation 9dB) from the canyon path is marked as the dynamic compensation zone; in step 703, the stable communication zone is sorted by attenuation (0.7dB→0.9dB→1.1dB), and the dynamic compensation zone is sorted by fluctuation (9dB→8dB→7dB), generating a priority-overlapping trajectory sequence; in step 704, in the overlapping area at the mine entrance (2 km at the plain-canyon boundary), the path is fused with a weight of 7:3; in the non-overlapping area inside the canyon, the priority complex path is directly used. The finally generated dynamic tracking rule reduces the communication packet loss rate of trucks in mixed terrain by 40%, and the trajectory prediction error is <2 meters.

[0176] Steps 701-704 use terrain reflection features to drive path classification and priority fusion. In mixed terrain scenarios such as iron ore mountainous areas, the system adaptively balances communication stability in flat areas with anti-interference capability in complex areas, generating environment-sensitive dynamic tracking rules. This significantly improves the continuity of industrial asset trajectories and system robustness in areas without base station coverage.

[0177] Figure 2 This application provides a schematic diagram of the structure of an industrial asset spatiotemporal trajectory tracking system based on multi-source heterogeneous positioning data fusion, as shown in the embodiment of the present application. Figure 2 As shown, the system includes:

[0178] The acquisition module 21 is used to acquire multi-source heterogeneous positioning data through edge nodes deployed on industrial assets, infer the signal propagation path based on the changes in the location of industrial assets in the asset positioning data, extract the geometric features of the terrain reflection surface in the signal propagation path, and select the polarization direction of the LoRa and satellite communication link according to the geometric features.

[0179] Transmission module 22 is used to divide the edge nodes into federated learning groups according to geographical regions, transmit the asset positioning data within the federated learning group through the polarization direction, generate spatiotemporal feature segments of industrial asset trajectories locally based on the transmission results, and broadcast the spatiotemporal feature segments across groups to adjacent federated learning groups through the LoRa and satellite communication link.

[0180] Extension module 23 is used to extend the topology of the fracture region of cross-group industrial asset trajectory based on the spatiotemporal feature segments broadcast in the adjacent federated learning groups and the signal coverage blind zone information corresponding to the polarization direction, and generate a continuous trajectory prediction path based on the extended fracture region.

[0181] The verification module 24 is used to dynamically adapt the signal attenuation characteristics of the polarization direction under extreme weather conditions to the update cycle of the continuous trajectory prediction path, and verify the spatiotemporal continuity of the polarization direction switching and the update cycle.

[0182] The generation module 25 is used to integrate the continuous trajectory prediction paths of each federated learning group according to the signal attenuation differences of different terrain regions under the polarization direction, and generate a spatiotemporal trajectory tracking strategy that integrates multi-terrain dynamic communication features based on the integrated continuous trajectory prediction paths.

[0183] Figure 2 The aforementioned industrial asset spatiotemporal trajectory tracking system based on multi-source heterogeneous positioning data fusion can perform... Figure 1 The implementation principle and technical effects of the industrial asset spatiotemporal trajectory tracking method based on multi-source heterogeneous positioning data fusion described in the illustrated embodiment will not be repeated here. The specific operation methods of each module and unit in the industrial asset spatiotemporal trajectory tracking system based on multi-source heterogeneous positioning data fusion in the above embodiments have been described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0184] In one possible design, Figure 2 The industrial asset spatiotemporal trajectory tracking system based on multi-source heterogeneous positioning data fusion, as shown in the embodiment, can be implemented as a computing device, such as... Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0185] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 32.

[0186] The processing component 32 is used for the above Figure 1 The embodiment describes a method for tracking the spatiotemporal trajectory of industrial assets based on the fusion of multi-source heterogeneous positioning data.

[0187] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0188] Storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0189] Of course, computing devices may also include other components, such as input / output interfaces, display components, communication components, etc.

[0190] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc.

[0191] The communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0192] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0193] This application also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The embodiment shown is a method for spatiotemporal trajectory tracking of industrial assets based on multi-source heterogeneous positioning data fusion.

[0194] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0195] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0196] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0197] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for spatiotemporal trajectory tracking of industrial assets based on multi-source heterogeneous positioning data fusion, characterized in that, include: By collecting multi-source heterogeneous asset positioning data through edge nodes deployed on industrial assets, the geometric features of the terrain reflection surface in the signal propagation path are inferred from the asset positioning data based on the changes in the location of the industrial assets, and the polarization direction of the LoRa and satellite communication link is selected based on the geometric features. The polarization direction includes the vertical polarization direction and the horizontal polarization direction. The edge nodes are divided into federated learning groups according to geographical regions. Within the federated learning group, the asset location data is transmitted through the polarization direction. Based on the transmission results, spatiotemporal feature segments of industrial asset trajectories are generated, and the spatiotemporal feature segments are broadcast across groups to adjacent federated learning groups through the LoRa and satellite communication links. Based on the spatiotemporal feature segments broadcast in the adjacent federated learning groups, and combined with the signal coverage blind zone information corresponding to the polarization direction, the fracture region of the cross-group industrial asset trajectory is topologically extended, and a continuous trajectory prediction path is generated based on the extended fracture region. Based on the signal attenuation differences in different terrain regions under the polarization direction, the continuous trajectory prediction paths of each federated learning group are integrated, and a spatiotemporal trajectory tracking strategy is generated based on the integrated continuous trajectory prediction paths.

2. The method according to claim 1, characterized in that, The process of dividing the edge nodes into federated learning groups according to geographical regions, transmitting asset location data within the federated learning groups via the polarization direction, generating spatiotemporal feature segments of industrial asset trajectories based on the transmission results, and broadcasting the spatiotemporal feature segments across groups to adjacent federated learning groups via the LoRa and satellite communication link includes: Based on the geometric features corresponding to the polarization direction, the edge nodes are divided into federated learning groups that cover the continuous region of the same reflective surface. The boundary of the federated learning group is determined by the straight-line penetration range of the signal propagation path under the polarization direction. Within the federated learning group, the transmission parameters of the asset location data are directionally matched based on the polarization direction. According to the matching results, edge nodes that form a stable transmission relationship with the polarization direction are selected. The asset location data is then divided into multiple data segments in descending order of transmission stability through the edge nodes. Calculate the time interval and position offset between the multiple data segments, combine the parts that are continuous in time and whose position offset is less than a preset distance threshold into a local trajectory sequence, and generate industrial asset trajectory feature segments containing spatiotemporal correlation based on the movement direction and speed change patterns of the local trajectory sequence. The industrial asset trajectory feature segments are sorted according to the signal coverage priority corresponding to the polarization direction, and the sorted feature segments are broadcast to adjacent federated learning groups in order of coverage range from near to far via the LoRa and satellite communication links.

3. The method according to claim 1, characterized in that, The geometric features of the terrain reflective surface in the signal propagation path are inferred from the asset location data based on changes in industrial asset location. The polarization direction of the LoRa-satellite communication link is then selected based on these geometric features. The polarization direction includes both vertical and horizontal polarization directions. The edge nodes synchronously collect both type I and type II positioning data of the industrial assets. The type I positioning data includes the continuous position coordinates of the industrial assets, and the type II positioning data includes the movement direction and speed changes of the industrial assets. The movement trajectory direction of the industrial asset is determined based on the continuous position coordinates. An initial straight line is drawn along the movement trajectory direction. The initial straight line is then corrected for curvature based on the movement direction and speed changes. A signal propagation path matching the terrain occlusion distribution is generated based on the corrected initial straight line. The height difference of the terrain reflective surfaces and the density distribution of obstacles are measured along the signal propagation path. Reflective surfaces with a height difference greater than the upper limit of the difference are marked as strong reflection areas, and reflective surfaces with an obstacle density distribution greater than a preset density are marked as multipath interference areas. The geometric boundary features of the strong reflection areas and the multipath interference areas are extracted. Based on the tilt angle and distribution range in the geometric boundary features, the vertical polarization direction of the LoRa communication link is selected to weaken the reflection interference of the tilt angle, and the horizontal polarization direction of the satellite communication link is selected to penetrate the distribution range.

4. The method according to claim 1, characterized in that, The step of topologically extending the fractured regions of cross-group industrial asset trajectories based on the spatiotemporal feature segments broadcast in the adjacent federated learning groups, combined with the signal coverage blind zone information corresponding to the polarization direction, and generating continuous trajectory prediction paths based on the extended fractured regions includes: Based on the signal coverage blind zone information corresponding to the polarization direction, the starting point and ending point of the fracture region of the cross-group industrial asset trajectory are determined, and the starting point and ending point are the position coordinates of the first and last valid trajectory points in the adjacent federated learning group, respectively. From the spatiotemporal feature segments broadcast by the adjacent federated learning groups, trajectory segments that match the starting point and the ending point are selected. The movement direction of the trajectory segments is superimposed with the shape of the signal coverage blind zone corresponding to the polarization direction to form a candidate extension path. The curvature of the candidate extension path is dynamically adjusted according to the signal reflection intensity distribution in the fracture area, and the adjusted curvature is used as a constraint to generate an arc-shaped extension path that fits the boundary of the signal coverage blind zone. The arc-shaped extension path and the starting point and ending point are smoothly connected to form a continuous arc-shaped extension path. After deleting the part of the continuous arc-shaped extension path that does not overlap with the signal coverage area of ​​the polarization direction, a continuous trajectory prediction path is generated.

5. The method according to claim 2, characterized in that, The generation of industrial asset trajectory feature segments containing spatiotemporal correlations based on the movement direction and speed variation patterns of the local trajectory sequence includes: Calculate the difference in the movement direction of adjacent trajectory points in the local trajectory sequence, and mark the continuous trajectory segments whose difference in movement direction is less than the upper limit of the difference as directionally stable segments; Extract a set of trajectory points with consistent velocity change trends from the stable directional segment, and divide the set of trajectory points into acceleration and deceleration segments according to the direction of increase or decrease of the velocity change trend. The parts of the acceleration and deceleration segments that have the same direction of movement and continuous speed trend are combined into basic feature units. Based on the spatiotemporal distribution density of the basic feature units, the boundaries of adjacent basic feature units are seamlessly spliced ​​together to form an industrial asset trajectory feature segment containing spatiotemporal correlation.

6. The method according to claim 3, characterized in that, The step of selecting the vertical polarization direction of the LoRa communication link to reduce reflection interference from the tilt angle based on the tilt angle and distribution range in the geometric boundary features, and selecting the horizontal polarization direction of the satellite communication link to penetrate the distribution range, includes: Based on the direction and magnitude of the tilt angle in the geometric boundary features, the vertical polarization direction of the LoRa communication link is selected, and the vertical polarization direction is orthogonal to the tilt angle. The horizontal polarization direction of the satellite communication link is determined based on the continuous distribution length of obstacles within the distribution range of the geometric boundary features, wherein the horizontal polarization direction is parallel to the main extension direction of the distribution range; Based on the change in tilt angle during the movement of industrial assets, the angle offset of the vertical polarization direction is corrected so that the orthogonal relationship is updated synchronously with the tilt state of the terrain reflector surface. Based on the density change of obstacles in the distribution range, the coverage width of the horizontal polarization direction is adjusted so that the horizontal polarization direction penetrates the distribution range.

7. The method according to claim 1, characterized in that, The step of integrating the continuous trajectory prediction paths of each federated learning group based on the signal attenuation differences in different terrain regions under the polarization direction, and generating a spatiotemporal trajectory tracking strategy based on the integrated continuous trajectory prediction paths, includes: According to the terrain reflector type corresponding to the polarization direction, the continuous trajectory prediction paths of each federated learning group are divided into flat terrain paths and complex terrain paths, and the signal attenuation of the flat terrain path is lower than that of the complex terrain path. The sections with the smallest signal strength attenuation in the flat terrain path and the sections with the largest signal reflection intensity fluctuations in the complex terrain path are extracted and marked as stable communication areas and dynamic compensation areas, respectively. The trajectory prediction paths in the stable communication zone are arranged in ascending order of signal attenuation, while the trajectory prediction paths in the dynamic compensation zone are arranged in descending order of signal reflection intensity fluctuation. A trajectory sequence with priority superposition is formed based on the results of the two arrangements. Based on the spatial overlap between the stable communication zone and the dynamic compensation zone in the trajectory sequence, the continuous trajectory prediction paths of each federated learning group are integrated to generate dynamic tracking rules covering mixed terrain areas.

8. A spatiotemporal trajectory tracking system for industrial assets based on multi-source heterogeneous positioning data fusion, characterized in that, include: The acquisition module is used to acquire multi-source heterogeneous asset positioning data through edge nodes deployed on industrial assets, infer the signal propagation path based on the changes in the location of industrial assets in the asset positioning data, extract the geometric features of the terrain reflection surface in the signal propagation path, and select the polarization direction of the LoRa and satellite communication link according to the geometric features. The transmission module is used to divide the edge nodes into federated learning groups according to geographical regions, transmit the asset location data within the federated learning group through the polarization direction, generate spatiotemporal feature segments of industrial asset trajectories locally based on the transmission results, and broadcast the spatiotemporal feature segments across groups to adjacent federated learning groups through the LoRa and satellite communication link. The extension module is used to extend the topology of the fracture region of the cross-group industrial asset trajectory based on the spatiotemporal feature segments broadcast in the adjacent federated learning groups and the signal coverage blind zone information corresponding to the polarization direction, and generate a continuous trajectory prediction path based on the extended fracture region. The verification module is used to dynamically adapt the signal attenuation characteristics of the polarization direction under extreme weather conditions to the update cycle of the continuous trajectory prediction path, and verify the spatiotemporal continuity of the polarization direction switching and the update cycle. The generation module is used to integrate the continuous trajectory prediction paths of each federated learning group based on the signal attenuation differences of different terrain regions under the polarization direction, and generate a spatiotemporal trajectory tracking strategy that integrates multi-terrain dynamic communication features based on the integrated continuous trajectory prediction paths.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement the spatiotemporal trajectory tracking method for industrial assets based on multi-source heterogeneous positioning data fusion as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The system contains a computer program that, when executed by a computer, implements a spatiotemporal trajectory tracking method for industrial assets based on the fusion of multi-source heterogeneous positioning data as described in any one of claims 1 to 7.

Images