Intelligent three-dimensional positioning method and system based on Bluetooth beacon and LoRa ad hoc network

Abstract

The application provides a kind of intelligent three-dimensional positioning method and system based on Bluetooth beacon and LoRa ad hoc network, belongs to the field of building construction safety.The technical scheme is: S1, establish the three-dimensional positioning constraint model of tunnel construction area and complete node system configuration;S2, positioning terminal collects and pre-processes multi-source observation data to generate unified observation package;S3, based on LoRa ad hoc network, unified observation package is cooperated, converged and reliably transmitted;S4, execute semantic constraint multi-source fusion three-dimensional positioning solution;S5, execute three-dimensional ambiguity resolution and output positioning result and closed loop update.The beneficial effects of the application are: the application improves alignment accuracy through ad hoc network time synchronization and windowed fusion, and guarantees data continuity through multi-hop reliable transmission and confidence retransmission;three-dimensional reachable domain and connectivity rules are used to resolve ambiguity and reduce cross-zone and cross-height misjudgment;confidence is output, and constraints are calibrated and dynamically updated online, so that positioning is long-term stable and reliable.

Description

Technical Field

[0001] This invention relates to the field of construction safety, and in particular to an intelligent three-dimensional positioning method and system based on Bluetooth beacons and LoRa self-organizing networks. Background Technology

[0002] In recent years, personnel positioning technologies for GPS-denied scenarios such as building construction, tunnel excavation, and underground utility tunnels have evolved rapidly: from early RFID / infrared-based area identification to the fusion of Wi-Fi fingerprinting, UWB time of arrival, Bluetooth beacon RSSI ranging, and inertial odometry; simultaneously, low-power wide-area communication technologies such as LoRa have expanded from star-shaped access to ad hoc networks and edge aggregation, forming an integrated "sensing-communication-computing" solution. Academia and industry have gradually introduced methods such as network time synchronization, windowed feature extraction, probabilistic graphical models, and particle filtering to achieve cross-source data fusion and continuous 3D trajectory estimation, continuously optimizing deployment scale, power consumption constraints, and engineering maintainability in complex underground spaces.

[0003] However, existing solutions still struggle to simultaneously meet the requirements of "continuous, interpretable, and spatially semantically consistent" three-dimensional positioning under conditions of strong multipath interference, obstruction, and structural repetition in construction tunnels. First, Bluetooth RSSI is significantly affected by reflections from metal devices and occlusion by crowds, and the fingerprint / ranging model is prone to drift, resulting in multi-peak posterior and cross-height ambiguity in the same mileage segment, and lacks an ambiguity resolution mechanism consistent with construction semantics (area identification and allowed migration rules); Secondly, LoRa positioning systems often use star topologies or rely on fixed gateways for backhaul. The relay and backhaul links are prone to forming single-point bottlenecks. Data packet loss and latency jitter can disrupt cross-source window alignment, thereby weakening the stability of multi-source fusion. Third, most implementations lack unified quality metrics and reliable transmission control, making it difficult to adaptively trigger acknowledgments / retransmissions or redundant forwarding under low-confidence observation conditions to ensure continuous delivery of positioning data. Fourth, the constraint priors often remain at the level of static maps or coarse-grained restricted areas, lacking a mechanism to inject the three-dimensional reachable domain and regional connectivity into the inference in a computable form and to dynamically update the constraint model under construction events or abnormal consistency conditions. Therefore, when the opening and closing of passages, temporary fences and shaft traffic change frequently, the positioning results are prone to problems such as trajectory breakage, cross-regional jumps or uninterpretable confidence. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent 3D positioning method and system based on Bluetooth beacons and LoRa self-organizing networks. This method improves alignment accuracy by integrating self-organizing network time synchronization and windowing, ensures data continuity by combining multi-hop reliable transmission and confidence retransmission, resolves ambiguity and reduces cross-region and cross-height misjudgments by using 3D reachability domains and connectivity rules, and outputs confidence scores with online calibration and dynamic constraint updates, thereby enabling long-term stable and reliable positioning.

[0005] This invention is achieved through the following measures: A smart 3D positioning method based on Bluetooth beacons and LoRa ad hoc networks, characterized by the following steps: S1. Establish a three-dimensional positioning constraint model of the tunnel construction area and complete the node system configuration, specifically including: establishing a local three-dimensional coordinate system for the tunnel and defining the area identifier ZonelD; constructing a three-dimensional reachable domain based on the tunnel spatial structure. and regional connectivity graph The edges of the connected region graph carry migration rules; Bluetooth beacons are deployed at key locations in the tunnel and the Bluetooth beacon ID, known spatial location, and Zone ID of each Bluetooth beacon are stored; LoRa nodes are deployed to form a LoRa ad hoc network and the LoRa node ID, node role, and known spatial location of each LoRa node are stored. S2. The positioning terminal collects and preprocesses multi-source observation data to generate a unified observation package, specifically including: collecting Bluetooth scan data, LoRa measurement data, and inertial / gait increment data, wherein the multi-source observation data consists of the Bluetooth scan data, the LoRa measurement data, and the inertial / gait increment data; establishing a network time reference through LoRa ad hoc network time synchronization messages and estimating the deviation between the positioning terminal's local clock and the network time to obtain a correction time; performing time alignment and aggregation on the Bluetooth scan data, the LoRa measurement data, and the inertial / gait increment data based on the correction time according to a preset window length, and extracting Bluetooth features from the Bluetooth scan data respectively. Extracting LoRa features from the LoRa measurement data Inertial / gait increments are formed from the inertial / gait increment data. Generate a unified observation package The unified observation package includes at least the positioning terminal identifier (TagID), the window representing the correction time, and Bluetooth characteristics. LoRa characteristics Inertia / gait increment And quality indicators; and based on the quality indicators, generate a pre-confidence level for communication reliability control. ; S3. Based on the LoRa self-organizing network, the unified observation packets are collaboratively aggregated and reliably transmitted, specifically including: each LoRa node calculates the routing cost based on link quality, hop count, and latency, and selects a multi-hop forwarding path; relay nodes or edge aggregation nodes statistically aggregate and compress the unified observation packets to generate a minimum positioning data packet for positioning and perform differential coding and / or event-triggered coding; according to preset reliability conditions, the minimum positioning data packet is acknowledged, retransmitted, and / or forwarded with multi-path redundancy to ensure that positioning data is continuously delivered to the positioning calculation end; S4. Perform multi-source fusion 3D localization solution with semantic constraints, specifically including: defining the localization state. A motion prior model is established based on inertia / gait increments to predict the state; Bluetooth observation likelihood functions are established based on Bluetooth features. LoRa observation likelihood function is established based on LoRa features. The three-dimensional reachable region Connectivity graph of the region Introducing constraint priors to generate constraint terms Calculate the Bluetooth and LoRa residuals, and adaptively adjust the weight coefficients of the Bluetooth and LoRa observation likelihood functions based on the residuals. Then, based on the motion prior model, the Bluetooth and LoRa observation likelihood functions, and the constraint terms, obtain the 3D position estimate using constrained particle filtering or factor graph optimization. ; S5. Perform 3D ambiguity resolution and output the localization result and loop closure update, specifically including: mapping candidate locations to region identifiers (ZoneID) and updating them according to the region connectivity graph. The migration rule eliminates or penalizes candidates that cross regions or altitudes to resolve ambiguities; location reliability is calculated based on posterior distribution concentration, observation residual consistency, and constraint violation rate. The output includes the positioning terminal identifier (TagID), calibration time, and 3D position estimate. Location reliability The positioning results are compared with the ZoneID; when preset stable or abnormal conditions are met, online updates are performed on the Bluetooth propagation parameters and / or LoRa measurement bias parameters, and the three-dimensional reachability domain is triggered based on construction events or abnormal consistency. and / or the region connectivity graph edge set Dynamic updates.

[0006] The invention also has the following specific features: The local three-dimensional coordinate system of the tunnel mentioned in step S1 adopts mileage coordinates. Cross-sectional transverse coordinates With vertical coordinates express; The ZoneID is composed of at least mileage segment code, cross-section interval code, and work area type code; The migration rules include at least the following: vertical migration is allowed only when the ZoneID of the candidate location is located in a preset communication channel area, shaft opening area, or stair section area; otherwise, vertical crossing is prohibited.

[0007] The three-dimensional reachable region described in step S1 Represented using a voxel grid set or a corridor polyhedron set; And the constraint terms mentioned in step S4 It at least includes a candidate state to the three-dimensional reachable domain. The soft penalty term determined by the minimum distance to the boundary.

[0008] The correction time mentioned in step S2 is obtained by the LoRa node periodically broadcasting a time synchronization message containing a network timestamp; The positioning terminal estimates the clock skew based on at least two time synchronization messages. and / or drift items, and convert the local timestamp to the corrected time. This allows for unified time alignment of Bluetooth scan data, LoRa measurement data, and inertial / gait increment data.

[0009] Bluetooth features mentioned in step S2 At least include the mean RSSI value, the variance of the RSSI value, and the preset number of Bluetooth beacons with the highest strength for each beacon within the window. A set of Bluetooth beacons; Median filtering and / or Huber suppression are applied to the Received Signal Strength Indicator (RSSI) sequence within the window to suppress multipath spike anomalies, and Bluetooth beacon observations with fewer reads than a threshold or variance higher than a threshold are removed.

[0010] The LoRa features described in step S2 One of the following: a) a time-of-flight-based ranging set And introduce ranging offset parameters for each anchor point. ; or b) a set of time differences based on arrival time differences ; Furthermore, the LoRa feature further includes the measurement variance derived from the signal-to-noise ratio and / or quality metric, which is used to weight the LoRa observation likelihood function in step S4.

[0011] The routing cost function mentioned in step S3 is:

[0012] in: The value of routing; hop is the hop count weighting coefficient; hop is the hop count. The link packet loss rate is the weighting coefficient; the loss is the link packet loss rate. "delay" represents the link delay weighting coefficient; "delay" represents the link delay. Furthermore, when the minimum positioning data packet corresponds to the pre-confidence level When the value is below a preset threshold, acknowledgment and adaptive retransmission are enabled for the minimum positioning data packet, and the number of retransmissions is related to... Positively correlated, or multi-path redundant forwarding is used for the minimum positioning data packet; in: Pre-confidence level; For window indexing.

[0013] When using constrained particle filtering to solve in step S4, it includes: propagating particles based on the motion prior model; and based on the Bluetooth observation likelihood function. The LoRa observation likelihood function and the aforementioned constraint terms Calculate and normalize the particle weights; Calculate the effective number of particles:

[0014] When the effective number of particles Resampling is triggered when the value is below a preset threshold; and a 3D position estimate is obtained based on the particle posterior distribution. ; in: It is a conditional probability function; This is a Bluetooth feature; LoRa characteristics Positioning status; For constraint terms; The effective number of particles; For the first The first window Normalized weights of individual particles; For window indexing.

[0015] The location confidence level described in step S5 The calculation is based at least on the posterior distribution concentration, observation residual consistency, and constraint violation rate, where the observation residual consistency is composed of the normalized sum of squares of the Bluetooth and LoRa residuals, and the constraint violation rate is calculated based on whether the candidate state falls within the three-dimensional reachable domain. The proportion outside; and when multiple consecutive windows meet the stability condition that the displacement increment is less than the threshold, the Bluetooth propagation parameters and / or LoRa ranging bias parameters are updated online by minimizing the observation residuals. When a change in the reachability domain or an abnormal consistency condition corresponding to a construction event is detected, the three-dimensional reachability domain is triggered. and / or the region connectivity graph edge set Dynamic updates.

[0016] A system employing the aforementioned intelligent 3D positioning method based on Bluetooth beacons and LoRa ad hoc networks, characterized in that it comprises: Constraint Modeling Module: Establishes a three-dimensional positioning constraint model of the tunnel construction area and completes the node system configuration; Observation preprocessing module: The positioning terminal collects and preprocesses multi-source observation data to generate a unified observation package; Self-organizing network transmission module: Based on LoRa self-organizing network, it performs collaborative aggregation and reliable transmission of the unified observation packets. Fusion localization module: performs multi-source fusion 3D localization solution with semantic constraints; Ambiguity closure module: Performs 3D ambiguity resolution and outputs localization results and closure update.

[0017] The beneficial effects of this invention are as follows: This invention improves alignment accuracy by integrating self-organizing network time synchronization and windowing, and ensures data continuity by combining multi-hop reliable transmission and confidence retransmission; it reduces misjudgment across regions and heights by resolving ambiguity with three-dimensional reachable domains and connectivity rules; and it outputs confidence and performs online calibration and dynamic update of constraints, making positioning stable and reliable in the long term. Attached Figure Description

[0018] Figure 1 The overall flowchart of the intelligent three-dimensional positioning method based on Bluetooth beacons and LoRa self-organizing networks provided in the embodiments of the present invention is shown below. Detailed Implementation

[0019] To clearly illustrate the technical features of this solution, the following detailed implementation method will be used to explain the solution.

[0020] Example 1 See Figure 1 A smart 3D positioning method based on Bluetooth beacons and LoRa ad hoc networks includes the following steps: S1. Establish a three-dimensional positioning constraint model of the tunnel construction area and complete the node system configuration, specifically including: establishing a local three-dimensional coordinate system for the tunnel and defining the area identifier ZonelD; constructing a three-dimensional reachable domain based on the tunnel spatial structure. and regional connectivity graph The edges of the region connectivity graph carry migration rules; Bluetooth beacons are deployed at key locations in the tunnel and the Bluetooth beacon ID, known spatial location, and Zone ID of each Bluetooth beacon are stored; LoRa nodes are deployed to form a LoRa ad hoc network and the LoRa node ID, node role, and known spatial location of each LoRa node are stored. In step S1, the local three-dimensional coordinate system of the tunnel adopts mileage coordinates. Cross-sectional transverse coordinates With vertical coordinates express; The ZoneID identifier consists of at least three parts: mileage segment code, cross-sectional interval code, and work area type code. The migration rules should include at least the following: vertical migration is allowed only when the ZoneID of the candidate location is located in the preset communication channel area, shaft opening area, or stair section area; otherwise, vertical crossing is prohibited.

[0021] Three-dimensional reachable region in step S1 Represented using a voxel grid set or a corridor polyhedron set; And the constraint terms in step S4 It should at least include a three-dimensional reachable domain based on candidate states. The soft penalty term determined by the minimum distance to the boundary.

[0022] Step S1 specifically includes: In tunnel excavation construction scenarios, due to the presence of long distances, curved sections and cross-sectional changes in the spatial form, and vertically connected structures such as connecting passage areas, shaft opening areas or staircase areas, as well as the fact that wireless signals are easily affected by the reflection of surrounding rock, the obstruction of mechanical equipment and multipath propagation, if only Bluetooth scanning data or LoRa measurement data is used for three-dimensional positioning, it is easy for candidate positions to fall into inaccessible spaces such as outside the lining, behind the partition wall or closed danger zone, and there may be jumps across areas or heights that do not conform to the actual construction passage.

[0023] Therefore, in this embodiment, a three-dimensional positioning constraint model of the tunnel construction area is first established in step S1, and the node system configuration is completed so that the subsequent positioning solution is carried out under clear spatial constraints and connectivity constraints.

[0024] (1) Establishment of local three-dimensional coordinate system in tunnel and Unified expression In this embodiment, the tunnel design axis or construction survey axis is preferably used as the tunnel centerline. Data sources, among which For centerline parameters.

[0025] If mileage marker data already exists, then with mileage coordinates Establish a corresponding relationship; If the centerline is given by discrete measurement points, then interpolation fitting is performed on the discrete measurement points to form a continuous line. And use arc length or mileage as a parameterized scale to ensure continuous representation of positions along the line. Based on Constructed in mileage coordinates cross-sectional orthogonal basis matrix , Used to project spatial positions onto the cross-sectional coordinate system, making the transverse coordinates of the cross-section... with vertical coordinates It has clear engineering significance; in curved sections or sections with changing cross-sections, Follow Continuous updates are performed to avoid discontinuities in subsequent ZoneID mapping and reachability determination caused by sudden coordinate changes. From the tunnel centerline at the mileage coordinates The tangential direction at the location and two mutually orthogonal directions within the cross-section constitute the coordinates, and vary with the mileage coordinates. Continuous updates are performed to maintain coordinate system consistency.

[0026] For any spatial position vector This embodiment obtains the result through "centerline projection + cross-sectional projection". To achieve a unified coordinate representation across the entire line:

[0027]

[0028] in: Mileage coordinates; The transverse coordinates of the cross section; Vertical coordinates; The spatial position vector to be transformed; For parameters The vector representing the position of the tunnel centerline; Centerline parameters; `arg min` is the L2 norm operator; `arg min` is the operator that takes the argument that minimizes the objective function. For mileage coordinates The orthogonal basis matrix of the cross section is defined at the location; It is the transpose operator; When the centerline parameter is taken as the same as the mileage coordinates The centerline position vector corresponding to the parameter values.

[0029] Using this unified coordinate system, subsequent ZoneID encoding and 3D reachability domains are implemented. The construction and storage of Bluetooth beacon and anchor node locations can all be completed under the same coordinate reference, thus avoiding boundary misalignment caused by "node locations in one coordinate system and reachable domains in another coordinate system". This misalignment is particularly sensitive in the narrow space of a tunnel: a small misalignment can cause candidate locations to fall into the inaccessible space, introducing unnecessary ambiguity.

[0030] (2) Scope of ZoneID Construction and Encoding Parameter Values In obtaining Subsequently, in this embodiment, the tunnel construction space is discretized into regional units with engineering semantics, and a regional identifier ZoneID is assigned to each regional unit.

[0031] ZoneID consists of at least mileage segmentation coding, cross-sectional interval coding, and work area type coding, so that the same location has both geometric segmentation attributes and work semantic attributes, providing a direct carrier for subsequent migration permission rules.

[0032] ZoneID can be expressed as: ZoneID

[0033] Where: ZoneID is the region identifier; Mileage segmentation coding; For cross-sectional interval coding; Code the work area type; Mileage coordinates; The transverse coordinates of the cross section; The length of each mileage segment; The width of the cross-sectional section; This represents the minimum boundary value of the transverse coordinate of the cross section; This is the floor operator.

[0034] in, and Not selected arbitrarily: The optimal approach is to match the granularity of construction management segments, the spacing of Bluetooth beacon deployments, and the typical obstruction scale within the tunnel to ensure that the mileage segment coding is observable for "movement along the route". The optimal configuration should match the effective bandwidth of the cross section and the width occupied by the equipment to ensure that the cross section interval encoding has the ability to discriminate "internal offset"; The cross-sectional boundaries determine the starting point for ensuring consistent transverse coding across different cross-sectional morphologies. If If the value is too large, different work points will be assigned to the same mileage segment encoding, weakening the ability of subsequent migration rules to constrain spatial semantics; if... If the value is too small, people will frequently cross segments during normal movement, causing the ZoneID to change rapidly and increasing the burden of ambiguity resolution.

[0035] (3) Three-dimensional reachable domain Construction process and two representation methods This embodiment constructs a three-dimensional reachable domain based on the tunnel's spatial structure. , This refers to a collection of spaces that people can enter, stay in, and pass through. During construction, it is based on the inner contour of the tunnel lining, the boundaries of isolation facilities, the boundaries of enclosed areas, and the delineation of hazardous zones. The space outside the lining, the solid space of the surrounding rock, and the enclosed hazardous zone are eliminated to obtain an accessible space consistent with the actual construction and passage.

[0036] Method 1: Voxel raster set representation During implementation, first determine the coverage area (covering the tunnel segments corresponding to each mileage segment code), set the voxel resolution, and generate the coordinate vector of the voxel grid center point. Then, based on the inner contour and closed boundary, reachability is determined for each voxel and a reachability marker variable is assigned. ,get:

[0037] in: It is a three-dimensional reachable region; For the first The coordinate vector of the center point of the individual pixel raster; For voxel indexing; A voxel-reachable marker variable; This indicates that the corresponding voxel is located in the reachable space.

[0038] Method 2: Representation of a corridor using a polyhedral set During implementation, the tunnel passage space is segmented and enveloped according to mileage segment coding, forming a corridor polyhedron within each segment. And by taking the union of all corridor polyhedra, we obtain the reachable region:

[0039] in: U represents a three-dimensional reachable region; U is the set union operator. The number of polyhedra in the corridor; For polyhedral indexing; For the first A collection of multifaceted spaces in a corridor.

[0040] pass The explicit construction of the modeling space restricts the candidate localization space to the reachable set during the modeling phase, effectively removing "unreachable areas" from the feasible domain. This removal is particularly relevant to tunnel scenarios because tunnels are long and narrow with clear boundaries, making geometric reachability constraints highly reliable and able to withstand short-term anomalies from wireless observations.

[0041] (4) Region connectivity graph Scope and Allowed Migration Rules In ZoneID and After construction is complete, this embodiment constructs a region connectivity graph. .

[0042] Node set It consists of region units corresponding to each region identifier (ZoneID); edge set Used to express the traversability between regions. Edge set It contains at least two types of edges: horizontal connectivity and vertical migration. It also allows migration rules to be carried as edge attributes to constrain vertical crossings without changing the existence of horizontal connectivity edges.

[0043] This embodiment determines the adjacency determination variable based on spatial adjacency or boundary contact relationship. ,when At that time, node With nodes There are walkable connections between them. Adjacency determination variables. The boundary contact relationship or passage connection relationship of the area unit corresponding to the ZoneID in the tunnel spatial structure is determined. This indicates that there is a passable connection between the two area units.

[0044] Secondly, vertical migration is allowed only when the ZoneID of the candidate location is located in the preset communication channel area, shaft opening area, or stair section area, and this migration rule is bound to the corresponding vertical migration edge.

[0045] The vertical migration permission function can be represented as follows: ZoneID

[0046] in: ZoneID This is the vertical migration permission function; ZoneID is the region identifier. A set of region identifiers that allow vertical migration.

[0047] This design solidifies the construction access fact that "vertical connectivity only occurs at specific structures" into the edge attribute of the regional connectivity graph. Even if wireless observations produce height anomalies in the short term, migration rules can be used to constrain cross-height candidates that do not conform to the access fact, thereby reducing the risk of misjudgment of height jumps caused by multipath.

[0048] (5) Bluetooth beacon and LoRa ad hoc network node deployment, coordinate mapping and home storage process After completing the three-dimensional constraint model, this embodiment deploys Bluetooth beacons at key locations in the tunnel. These key locations preferably include mileage segment boundaries, cross-sectional change points, tunnel intersections, and work area entrances to improve the distinguishability of Bluetooth scanning data between different ZoneIDs.

[0049] Each Bluetooth beacon stores a Bluetooth beacon ID, its known spatial location, and its zone ID (ZoneID). To ensure that the "known spatial location" is consistent with the aforementioned constraint model, this embodiment first converts the engineering or measurement coordinates of the Bluetooth beacon into the tunnel's local three-dimensional coordinate system. Then, based on the ZoneID encoding rules, the ZoneID identifier is calculated and written into the storage field, thereby achieving a consistent mapping of "coordinates-zone-constraint".

[0050] Simultaneously, LoRa nodes are deployed to form a LoRa self-organizing network. All LoRa nodes store their LoRa node identifier (NodeID) and node role; anchor nodes further store their known spatial locations. Similar to Bluetooth beacons, the known spatial locations of anchor nodes are preferably defined using (…). ) or can be mapped to The anchor node is stored in the form of a variable and its zone identifier (ZoneID) is determined according to the ZoneID encoding rules, thereby ensuring the spatial location of the anchor node and its three-dimensional reachability. and regional connectivity graph They are under the same coordinate system and regional semantic benchmark.

[0051] Through the above process of "coordinate mapping - region attribution - field storage", it is ensured that subsequent steps retrieve Bluetooth beacon location, anchor node location, and 3D reachability domain. With regional connectivity graph It does not produce ambiguity.

[0052] In summary, step S1 establishes a local three-dimensional coordinate system for the tunnel. Constructing ZoneID, constructing 3D reachable domains (Represented by a set of voxel grids or a set of corridor polyhedra), construct a region connectivity graph. It also carries migration rules on the side and completes the deployment and storage configuration of Bluetooth beacons and LoRa self-organizing network nodes, forming a computable constraint basis for the complex tunnel environment. It brings the facts of spatial accessibility and passage connectivity to the modeling stage, so that the subsequent positioning solution can be carried out within a reasonable feasible domain, thus having higher stability and interpretability.

[0053] S2. The positioning terminal collects and preprocesses multi-source observation data to generate a unified observation package, specifically including: collecting Bluetooth scan data, LoRa measurement data, and inertial / gait increment data, wherein the multi-source observation data consists of Bluetooth scan data, LoRa measurement data, and inertial / gait increment data; establishing a network time reference through LoRa self-organizing network time synchronization messages and estimating the deviation between the positioning terminal's local clock and the network time to obtain a correction time; according to a preset window length, performing time alignment and aggregation on the Bluetooth scan data, LoRa measurement data, and inertial / gait increment data based on the correction time, and extracting Bluetooth features from the Bluetooth scan data respectively. Extracting LoRa features from LoRa measurement data Forming inertial / gait increments from inertial / gait increment data Generate a unified observation package The unified observation package should include at least the positioning terminal identifier (TagID), the window representing the correction time, and the Bluetooth signature. LoRa characteristics Inertia / gait increment And quality metrics; and based on these quality metrics, generate pre-confidence levels for communication reliability control. ; In step S2, the calibration time is obtained by the LoRa node periodically broadcasting a time synchronization message containing a network timestamp. The positioning terminal estimates the clock skew based on at least two time synchronization messages. and / or drift items, and convert the local timestamp to the corrected time. This allows for unified time alignment of Bluetooth scan data, LoRa measurement data, and inertial / gait increment data.

[0054] Bluetooth features in step S2 At least include the mean RSSI value, the variance of the RSSI value, and the preset number of Bluetooth beacons with the highest strength for each beacon within the window. A set of Bluetooth beacons; Median filtering and / or Huber suppression are applied to the Received Signal Strength Indicator (RSSI) sequence within the window to suppress multipath spike anomalies, and Bluetooth beacon observations with fewer reads than a threshold or variance higher than a threshold are removed.

[0055] LoRa features in step S2 One of the following: a) a time-of-flight-based ranging set And introduce ranging offset parameters for each anchor point. ; or b) a set of time differences based on arrival time differences ; Furthermore, the LoRa features further include measurement variance derived from the signal-to-noise ratio and / or quality metrics, which is used to weight the LoRa observation likelihood function in step S4.

[0056] Step S2 specifically includes: After completing step S1, the local three-dimensional coordinate system of the tunnel Zone ID, 3D reachability and regional connectivity graph It has been established, and the Bluetooth beacon (including Bluetooth beacon ID, known spatial location, and zone ID) and LoRa ad hoc network nodes (including LoRa node ID, node role, and known spatial location of anchor node) have been deployed and their fields stored.

[0057] Step S2, based on the above constraint model and node system configuration, involves the positioning terminal collecting and preprocessing multi-source observation data to generate a unified observation package. The multi-source observation data consists of Bluetooth scan data, LoRa measurement data, and inertial / gait incremental data; the positioning terminal establishes a network time reference through the time synchronization message of the LoRa ad hoc network and estimates the deviation between the positioning terminal's local clock and the network time to obtain the correction time. The positioning terminal operates according to the preset window length. Based on the calibration time, Bluetooth scan data, LoRa measurement data, and inertial / gait increment data are time-aligned and aggregated to extract Bluetooth features. LoRa characteristics And form inertia / gait increment Generate a unified observation package And based on quality indicators, generate pre-confidence levels for communication reliability control. .

[0058] Preset window length The timing of the LoRa node's broadcast synchronization message, the Bluetooth scanning cycle of the positioning terminal, and the gait cycle of the person are matched to ensure that a stable aggregation result of Bluetooth scanning data and LoRa measurement data is formed within each window; a preset number of... Matching the LoRa self-organizing network bandwidth budget and minimum positioning data packet load constraints to control the information content and communication overhead of the unified observation packet; the reading count threshold and variance threshold are determined by the tunnel site wireless environment calibration, and are used for easily removed low sampling support observation and strong multipath unstable observation.

[0059] (1) Multi-source observation data acquisition and field unification The positioning terminal performs three types of observation acquisitions in each acquisition cycle: 1) Bluetooth Scanning Data: The positioning terminal scans the Bluetooth beacons deployed in the tunnel and records the BeaconID, Received Signal Strength Indicator (RSSI), and local time of each successful scan. The data includes fields such as the number of reads; and the BeaconID is associated with the known spatial location and ZoneID of the beacon stored in step S1.

[0060] 2) LoRa Measurement Data: The positioning terminal interacts with the anchor node through the LoRa ad hoc network, parses the LoRa measurement data to obtain the raw distance measurement or time difference of arrival, and records the anchor node identifier AnchorID and the corresponding local timestamp. Signal-to-noise ratio and quality indicators; LoRa measurement data in the first... Each window strictly forms either a "range set based on flight time" or a "time difference set based on arrival time difference".

[0061] 3) Inertial / Gait Incremental Data: The positioning terminal collects the output of the inertial device and combines it with the gait detection results to form inertial / gait incremental data, and records the corresponding local time. .

[0062] (2) Establish a network time reference and obtain the corrected time based on the LoRa self-organizing network time synchronization message. LoRa nodes periodically broadcast time synchronization messages containing network timestamps. The positioning terminal receives at least two time synchronization messages and estimates the clock offset based on the time pair of the two time synchronization messages. and / or drift terms And convert the local timestamp to the corrected time. .

[0063] The positioning mid-range adopts a "two-stage calibration aperture": first, drift correction is performed to obtain... Then, the clock offset is added to obtain The deterministic calculation process is as follows:

[0064]

[0065] in: This represents the drift of the local clock relative to the network time. The network timestamp for the first time synchronization message; For the network timestamp of the second time synchronization message; The local timestamp corresponding to the first time synchronization message received by the positioning terminal; The local timestamp corresponding to the second time synchronization message received by the positioning terminal; This is the local timestamp after drift correction; This is the clock deviation term; And for the correction time; This is the local timestamp corresponding to any observation sample.

[0066] This time correction mechanism aligns Bluetooth scan data, LoRa measurement data, and inertial / gait increment data on a unified time axis, eliminating mis-window fusion and window segmentation caused by cross-source time inconsistencies.

[0067] (3) Perform time alignment and aggregation according to the preset window length and determine the window to represent the correction time. Positioning terminal sets preset window length To correct time Windowed alignment is performed on the three types of observations using a single benchmark: for any observation sample, according to... Calculate window index The Bluetooth scan data, LoRa measurement data, and inertial / gait increment data falling into the same window are aggregated separately, and the window represents the correction time. Used to standardize the time labeling of observation packets:

[0068] in: For window indexing; This is the floor operator; For time correction; The reference time for the start of the window; The preset window length; For the first Each window represents a correction time; For the first The number of samples participating in the time-representation calculation within each window; For the first The first window The calibration time for each sample.

[0069] (4) Bluetooth features Extraction: Deterministic definitions, statistical calculations, and removal of median filtering and Huber suppression. In the Within a window, the positioning terminal aggregates RSSI sequences belonging to the same BeaconID into a group.

[0070] The positioning terminal first performs median filtering on each group of RSSI sequences to obtain... Then execute Huber suppression to obtain :

[0071]

[0072] in: The Received Signal Strength Indicator (RSSI) value is the output of the median filter. For the median operator; The set of RSSI samples that participate in median filtering; The Received Signal Strength Indicator (RSSI) value for Huber's suppressed output; For trimming operators; The Huber suppression threshold; For the first The first window The first Bluetooth beacon Received Signal Strength Indication (RSSI) value; For window indexing; For Bluetooth beacon indexing; To read the sequence number.

[0073] Positioning terminal based on Calculate the RSSI mean and RSSI variance, and count the number of reads; the RSSI variance is denoted as... :

[0074]

[0075] in: For the first The first window The average Received Signal Strength Indicator (RSSI) value of each Bluetooth beacon; For the first The first window The variance of the Received Signal Strength Indicator (RSSI) value for each Bluetooth beacon; For the first The first window Number of reads for each Bluetooth beacon; The Received Signal Strength Indicator (RSSI) value for Huber's suppressed output; For window indexing; For Bluetooth beacon indexing; To read the sequence number.

[0076] The positioning terminal reads fewer times than the threshold or Bluetooth beacon observations exceeding a threshold are removed, and then... Sort by strength from largest to smallest, and select the preset number with the highest strength. A set of Bluetooth beacons, and the RSSI mean, RSSI variance, and preset number are used. A set of Bluetooth beacons together constitutes the Bluetooth feature. .

[0077] (5) LoRa features Extraction: Distance set or time difference set, bias compensation and measurement variance In the Within each window, the positioning terminal extracts LoRa features from the LoRa measurement data. .

[0078] Strictly form one of the following: a) Time-of-flight-based ranging set And introduce ranging offset parameters for each anchor point. ;or b) Time difference set based on arrival time difference .

[0079] The positioning terminal parses the raw ranging data from the LoRa measurement data. or arrival time difference and indexed by window Aggregates into a range set or a time difference set.

[0080] For the ranging set, the positioning terminal performs offset compensation to obtain:

[0081] in: For the first Anchor nodes within a window The corresponding offset-compensated distance measurement value; For the first Anchor nodes within a window The corresponding original distance measurement value; For anchor nodes The corresponding ranging offset parameters; For window indexing; For anchor node indexing.

[0082] The positioning terminal further derives the signal-to-noise ratio and quality indicators as measurement variance and writes them into the system. The measurement variance is denoted as The benchmark variance is denoted as :

[0083] in: For the first Needle node within a window The corresponding measurement variance; The standard variance; For the first Needle node within a window The corresponding signal-to-noise ratio; For the first Anchor nodes within a window Corresponding quality indicators; For window indexing; For anchor node indexing.

[0084] (6) Inertia / gait increment Formation and Connection The positioning terminal aggregates the output of the inertial device with the gait detection results within a window to form an inertial / gait increment. and will Represented in the local three-dimensional coordinate system of the tunnel Below, the inertial / gait increment and ZoneID segmentation logic and subsequent... Constraints have consistent coordinate semantics.

[0085] (7) Unified observation package Generation and Pre-set Confidence calculate After completing alignment and feature extraction, the positioning terminal in the first... Each window generates a unified observation package. , It should at least include the positioning terminal identifier TagID and the window representing the calibration time. Bluetooth features LoRa characteristics Inertia / gait increment And quality indicators.

[0086] Quality indicators are derived from Bluetooth quality indicators. LoRa quality indicators With inertia / gait quality index Composition, in which: The number of valid Bluetooth beacons, the rejection ratio, and The statistic is determined; The number of valid LoRa measurements and The statistic is determined, and Depend on and Decide; It is determined by the continuity and missing data of inertial / gait increments.

[0087] The positioning terminal integrates the three quality indicators according to their weights to obtain a comprehensive quality index. And generate a pre-defined confidence level through monotonic mapping. :

[0088]

[0089] in: For the first The overall quality indicators of each window; This is the Bluetooth quality weighting coefficient; For the first Bluetooth quality metrics for each window; This refers to the LoRa quality weighting coefficient; For the first LoRa quality metrics for each window; Inertia / gait mass weighting coefficient; For the first Inertia / gait quality metrics for each window; Pre-confidence level; It is an exponential function; The mapping slope parameter; This is the mapping threshold parameter; For window indexing.

[0090] Through the above process, step S2 achieves unified time benchmark alignment, unified window aggregation, unified feature representation, and unified quality characterization of multi-source observation data, and encapsulates the results into a unified observation package. and pre-confidence This provides consistent, quantifiable, and reproducible observation inputs for communication reliability control in subsequent step S3 and multi-source fusion three-dimensional positioning solution in step S4.

[0091] S3. Based on the LoRa self-organizing network, the unified observation packets are collaboratively aggregated and reliably transmitted. Specifically, this includes: each LoRa node calculates the routing cost based on link quality, hop count, and latency, and selects a multi-hop forwarding path; relay nodes or edge aggregation nodes statistically aggregate and compress the unified observation packets to generate the minimum positioning data packet for positioning and perform differential coding and / or event-triggered coding; according to preset reliability conditions, the minimum positioning data packet is acknowledged, retransmitted, and / or forwarded with multi-path redundancy to ensure that positioning data is continuously delivered to the positioning calculation end. The routing cost function in step S3 is:

[0092] in: The value of routing; hop is the hop count weighting coefficient; hop is the hop count. The link packet loss rate is the weighting coefficient; the loss is the link packet loss rate. "delay" represents the link delay weighting coefficient; "delay" represents the link delay. Furthermore, when the minimum positioning data packet corresponds to the pre-confidence level When the value is below a preset threshold, acknowledgment and adaptive retransmission are enabled for the smallest location data packet, and the number of retransmissions is related to... Positively correlated, or use multi-path redundancy forwarding for the smallest location data packet; in: Pre-confidence level; For window indexing.

[0093] Step S3 specifically includes: After completing the configuration of the three-dimensional positioning constraint model and node system of the tunnel construction area in step S1, the LoRa self-organizing network has formed a multi-hop communication foundation; In step S2, the positioning terminal has been aligned with the calibration time. After completing window alignment and aggregation, the result contains TagID and window representation of the correction time. Bluetooth features LoRa characteristics Inertia / gait increment Quality indicators are mapped to obtain the pre-defined confidence level. Unified observation package .

[0094] Step S3, based on this, utilizes LoRa ad hoc network to perform collaborative aggregation and reliable transmission of unified observation packets: on the one hand, it resists link fluctuations through multi-hop path selection driven by routing costs; on the other hand, it generates minimum positioning data packets through statistical aggregation and compression and performs differential coding and / or event-triggered coding to reduce communication load. At the same time, it performs acknowledgment, retransmission and / or multi-path redundant forwarding on the minimum positioning data packets according to preset reliability conditions, so that positioning data is continuously delivered to the positioning calculation end and the input data stream is maintained according to the window organization.

[0095] (1) Route cost calculation, link state statistics and multi-hop forwarding path selection Each LoRa node periodically maintains the neighbor link state, which consists of link packet loss, link delay, and hop count to the location computing terminal. This state is updated using a sliding time window to ensure that routing selection always reflects the latest tunnel wireless environment. The link packet loss is obtained from the sequence number continuity and acknowledgment statistics of the minimum location data packet: the ratio of failed acknowledgments to sent packets is calculated within the sliding window, or the sequence number gap ratio is used as the packet loss rate. Link delay is measured under network time reference: the difference between the sending correction time and receiving correction time for each forwarding record is recorded by the node, and the average value is taken within the sliding window as the link delay. The hop count is updated by the destination distance identifier maintained by the node: the location computing terminal or edge aggregation node periodically publishes hop count announcements, and each node takes the minimum value from the neighbor announcements plus one as its own destination hop count, updating it in real time as the topology changes.

[0096] Each time a unified observation packet or minimum location packet needs to be forwarded, the node calculates the routing cost based on the candidate next hop and selects the next hop with the lowest routing cost, thus forming an end-to-end multi-hop forwarding path.

[0097] The routing cost function takes the following form:

[0098] in: The value of routing; hop is the hop count weighting coefficient; hop is the hop count. The link packet loss rate is the weighting coefficient; the loss is the link packet loss rate. `delay` is the link delay weighting coefficient; `delay` is the link delay. The weighting coefficients adopt fixed and executable setting rules: when the construction site is in a time-priority state, the weighting coefficients are increased. and reduce This causes routing to favor low-latency paths; when in a reliability-first state, it improves... and reduce This causes routing to favor paths with low packet loss; when under energy-constrained and congestion-limited conditions, it improves... This causes routing to favor low-hop-count paths in order to reduce forwarding times and duty cycle.

[0099] The above rules enable consistent routing strategy adjustments to be achieved through weight switching under different construction tasks using the same cost function, thus avoiding the failure of a single strategy under conditions of tunnel link fluctuations and multi-hop congestion.

[0100] (2) Statistical aggregation, deduplication and merging, compression and minimum location data packet generation of relay nodes or edge aggregation nodes Unified observation package After ascending along a multi-hop path to a relay node or edge convergence node, the node performs statistical convergence and compression on the observations. Convergence is strictly represented by a window representing the correction time. Using the positioning terminal identifier TagID as the entity key, a definite deduplication key is formed:

[0101] in: This is the deduplication key; TagID is the location terminal identifier. The window represents the calibration time; For window indexing; This is an ordered pair construction operator.

[0102] When the same deduplication key corresponds to multiple arriving packets (introduced by retransmission or multi-path redundancy), the node generates the final version used for positioning according to a fixed merging rule: first, the field completeness is compared, and the one with higher field completeness is given priority; When fields have the same completeness, the one with the later arrival time takes precedence; if conflicts still exist with the same field, then the one with the later arrival time is retained. A lower-confidence version is used as the final version to ensure that low-confidence observations are prioritized and not overwritten during subsequent reliable transmission and fusion positioning stages.

[0103] Subsequently, the node retains only the fields necessary for positioning calculation and generates a minimum positioning data packet. The minimum positioning data packet consistently includes: the positioning terminal identifier (TagID) and the window representing the calibration time. Bluetooth features LoRa characteristics Inertia / gait increment With pre-confidence .

[0104] The minimum location data packet uses a fixed field arrangement order and a compact encoding method: the packet header consists of TagID and... ; Control field is and quality indicators; Bluetooth features Press Top - The Bluetooth beacon set stores the BeaconID sequence and its corresponding value in sequence. and Sequence; LoRa features Storing distance measurement sets sequentially according to AnchorID sequence. or time difference set And synchronously store the corresponding measurement variance Inertia / gait increment Stored at the end of the packet, this order ensures that the localization calculation end can recover window observations according to fixed resolution rules without relying on additional context.

[0105] (3) Differential coding and event-triggered coding: Differential vector construction caliber and triggering criteria To reduce the duty cycle and multi-hop forwarding load in LoRa ad hoc networks, relay nodes or edge aggregation nodes perform differential coding and / or event-triggered coding on the minimum positioning data packet. Differential coding uses the minimum positioning data packets of adjacent windows with the same TagID as a reference, encoding only the change in the current window relative to the previous window.

[0106] Differential coding applies only to the smallest localization data packet, containing numeric fields that are concatenated in a fixed order to form a vector of numeric fields. Numeric field vectors must consist of at least one Bluetooth features and Ranging values ​​in sequence and LoRa features or time difference Series and measurement variance Sequence, and inertia / gait increments The components are assembled in a predetermined order; discrete identifiers such as BeaconID and AnchorID are not included. Its changes are handled separately through event-triggered coding.

[0107] The core operation of differential coding is defined as follows:

[0108] in: For the first The differential coding amount of each window; For the first A vector of numeric fields for each window; For the first A vector of numeric fields for each window; For window indexing.

[0109] The event triggering code uses a fixed triggering criterion: if and only if Top- Changes in the Bluetooth beacon set, the index set of the ranging set or the time difference set, or the differential coding amount The transmission is triggered when the norm exceeds a preset threshold; transmission is suppressed when the differential coding amount is zero and the index set remains unchanged, thereby mapping the "observation stability phase" to low-frequency communication and the "observation change phase" to high-confidence information transmission.

[0110] (4) The preset confidence level output by performing confirmation, retransmission and / or multipath redundancy forwarding steps S2 according to preset reliability conditions. This is directly applied to reliable transmission control in step S3. For each minimum location data packet, the network decides whether to enable acknowledgment and adaptive retransmission, or to use multi-path redundant forwarding, based on preset reliability conditions. The acknowledgment granularity is fixed at hop-by-hop: each hop forwarding waits for an acknowledgment message from the next hop; successful acknowledgment completes the delivery of the current hop, while failure to acknowledgment initiates the retransmission process. This hop-by-hop acknowledgment mechanism decomposes end-to-end uncertainty into link-by-link controllable uncertainty, enabling the tunnel multi-hop network to maintain global delivery continuity even when local link fluctuations occur.

[0111] When the minimum positioning data packet corresponds to the pre-confidence When the value is below a preset threshold, the node enables acknowledgment and adaptive retransmission for the smallest location data packet, and the number of retransmissions is related to... Maintain a positive correlation; When both conditions are met When the loss of the selected path is lower than the preset threshold and the loss of the selected path is higher than the preset packet loss threshold or the delay is higher than the preset delay threshold, the node enables multi-path redundancy forwarding for the minimum location data packet, and sends it up in parallel along two or more different multi-hop forwarding paths to resist the instantaneous failure of a single path.

[0112] The adaptive retransmission count is calculated using the following deterministic method:

[0113] in: For the first The minimum number of retransmissions required for a given window to locate a data packet; To select the smaller value operator; This is the maximum number of retransmissions. This is the floor operator; This is the retransmission scaling factor; Pre-confidence level; For window indexing.

[0114] Through "routing cost-driven multi-hop path selection + ( Statistical aggregation and deduplication merging with ) as the key + minimum location data packet generation with fixed field set + differential encoding and / or event-triggered encoding + with With "hop-by-hop confirmation, adaptive retransmission and multi-path redundant forwarding as the core", step S3 establishes a measurable, adjustable and reproducible control chain at the communication level to address the uncertainties of the tunnel environment: routing cost enables the data flow to avoid high packet loss and high latency links, convergence compression and differential / event coding make the multi-hop network load controllable, and reliable transmission control driven by pre-confidence ensures that critical window observations are continuously delivered to the positioning calculation end even under link fluctuations, thereby providing a stable, continuous and time-aligned input data flow for the fusion positioning solution in subsequent steps.

[0115] S4. Perform multi-source fusion 3D localization solution with semantic constraints, specifically including: defining the localization state. A motion prior model is established based on inertia / gait increments to predict the state; Bluetooth observation likelihood functions are established based on Bluetooth features. LoRa observation likelihood function is established based on LoRa features. ; Three-dimensional reachable region With regional connectivity graph Introducing constraint priors to generate constraint terms The Bluetooth and LoRa residuals are calculated, and the weight coefficients of the Bluetooth and LoRa observation likelihood functions are adaptively adjusted based on the residuals. Then, based on the motion prior model, Bluetooth and LoRa observation likelihood functions, and constraint terms, constrained particle filtering or factor graph optimization is used to obtain the 3D position estimate. .

[0116] When using constrained particle filtering to solve step S4, it includes: propagating particles based on the motion prior model; and observing the likelihood function based on Bluetooth. LoRa observation likelihood function and constraint terms Calculate and normalize the particle weights; Calculate the effective number of particles:

[0117] When the number of effective particles Resampling is triggered when the value is below a preset threshold; and a 3D position estimate is obtained based on the particle posterior distribution. ; in: It is a conditional probability function; This is a Bluetooth feature; LoRa characteristics Positioning status; For constraint terms; The effective number of particles; For the first The first window Normalized weights of individual particles; For window indexing.

[0118] Step S4 specifically includes: After completing the construction of the three-dimensional positioning constraint model of the tunnel construction area in step S1, the local three-dimensional coordinate system of the tunnel has been solidified. Zone ID, 3D reachability and regional connectivity graph It also includes rules that allow migration; In step S2, the positioning terminal completes the time alignment and aggregation of multi-source observation data to generate a unified observation package. It includes at least TagID and a window representing the correction time. Bluetooth features LoRa characteristics Inertia / gait increment and pre-confidence In step S3, the unified observation packets are continuously delivered to the positioning calculation terminal in the form of minimum positioning data packets after collaborative aggregation and reliable transmission.

[0119] This step involves using the window index on the location calculation end. The organization decodes and reassembles the minimum localization data packets, performs multi-source fusion 3D localization solution with semantic constraints, and stably outputs 3D position estimates. .

[0120] (1) Positioning status definition and input organization Positioning calculation terminal defines positioning status For the local three-dimensional coordinate system of the tunnel The three-dimensional position state below, The three components correspond to the mileage coordinates, the transverse horizontal coordinates, and the vertical coordinates, respectively.

[0121] The location computing terminal uses TagID and To reassemble the minimum location data packets into window observation entries, it is strictly guaranteed that each window corresponds to only one set. And perform consistency processing on missing fields: if a window is missing or If the remaining observations are retained, the estimation is completed based on the motion priors and constraints. If duplicate arrival packets exist in the same window, the deduplication and merging results from step S3 are used instead of repeated merging.

[0122] (2) Motion prior model and particle propagation (including random perturbation) The positioning calculation terminal will output the inertial / gait increment from step S2. As a driving force for motion, a priori motion model is constructed to predict particle propagation. Inertia / gait increment. In this embodiment, it is a three-dimensional incremental vector, corresponding to respectively Displacement increments in three axes. To ensure the particle filter has an executable diffusion mechanism, the positioning computation end superimposes the process perturbation vector onto each particle during propagation, so that the particle ensemble maintains the necessary state diversity even when tunnel multipath and occlusion cause short-term observation degradation. The first window The propagation of individual particles follows the core relationship:

[0123] in: For the first The first window The positioning status of each particle; For the first The first window The positioning status of each particle; For the first Inertia / gait increments for each window; For the first The first window The process perturbation vector of each particle; For window indexing; For particle indexing.

[0124] The magnitude of the process disturbance vector is determined by the quality information in the minimum positioning data packet and... Co-regulation: When Reduce or LoRa measurement variance When the overall size increases, the positioning calculation end increases the perturbation amplitude to expand the search space; when When the amplitude of the disturbance is high and the two types of observations are consistent, the positioning calculation end converges to improve the stability of the estimation.

[0125] (3) Bluetooth and LoRa observation likelihood construction, residual calculation and weight adaptation The positioning computing end constructs Bluetooth observation likelihood functions respectively. LoRa observation likelihood function Its construction depends on the "known spatial location" of S1 and the "observation statistics / variance caliber" of S2.

[0126] Bluetooth observation likelihood construction: for Top Middle The Bluetooth beacon set, the positioning computing terminal calculates the status based on the known spatial location associated with each BeaconID. The geometric distance to the beacon is calculated, and the expected RSSI is obtained using a calibrated log-distance fading model; mean RSSI in Subtract the expected RSSI to form the Bluetooth residual, and use the RSSI variance Normalization is performed to obtain a window-level Bluetooth residual consistency metric; the positioning calculation end monotonically maps this consistency metric to... This results in candidate states with larger RSSI variance and larger residuals obtaining smaller likelihood values.

[0127] LoRa observation likelihood construction: when For the range set At that time, the positioning calculation terminal calculates the geometric ranging prediction based on the known spatial position of the anchor node associated with AnchorID, and the ranging residual is... The difference between the geometric prediction and the variance is measured. Normalization; when Time difference set At that time, the positioning calculation terminal calculates the position based on the two anchor points and their states. The predicted time difference is obtained by dividing the geometric distance difference by the propagation speed. This is then subtracted from the observed time difference to form the time difference residual. This residual is normalized using the dual variance obtained by combining the variances of the corresponding two anchor points. Thus, under the TDOA branch, a window-level consistency measure is also obtained and monotonically mapped to... .

[0128] Residual-driven adaptive weighting: The localization calculation terminal simultaneously calculates the tooth monitoring residual consistency metric and the LoRa residual consistency metric, and adaptively adjusts the weight coefficients of the two types of observation likelihoods accordingly. and .

[0129] Weight adjustment follows a fixed rule: observations with higher residual consistency (smaller normalized residuals) receive greater weight; weights tend to balance when the consistency between the two is close; when... When the weights are reduced, the positioning computing end simultaneously reduces the upper limit of the weights of the two types of observations and increases the influence of the constraint terms, thereby explicitly transforming the behavior of "decreased communication reliability" into "fusion becoming more dependent on priors and constraints".

[0130] (4) Semantic constraint prior Deterministic generation and hard constraint injection The positioning calculation end will have a three-dimensional reachable domain. With regional connectivity graph Introducing constraint priors to generate constraint terms .

[0131] Its calculation process is deterministic and reproducible: 1) Reachability domain feasibility determination: for any candidate state (particle) To determine whether it has fallen into ;when When using a voxel raster set representation, a voxel index lookup table is used for determination; when When using a corridor polyhedron set representation, the determination is made using points within the polyhedron. 2) Feasibility determination of area relocation: The positioning calculation terminal determines the feasibility of relocation based on the ZoneID encoding rules in S1 (mileage segment encoding, cross-sectional interval encoding, work area type encoding). Perform interval determination to obtain the current candidate state's region. and the previous window Construct transfer pairs; if If the migration rules are met (vertical migration is only allowed when the ZoneID of the candidate location is located in the connecting passage area, shaft entrance area, or stair section area; otherwise, vertical crossing is prohibited), then the migration is feasible; otherwise, the migration is not feasible.

[0132] 3) Hard constraint injection method: When the candidate state does not meet the reachability domain or migration rule, the constraint term of the particle is directly set to 0, and the unnormalized weight of the particle is reset to zero in the weight update. In this way, the tunnel geometric boundary and semantic migration rule are injected into the posterior inference in a "hard suppression" manner to avoid impossible trajectories such as wall crossing, layer crossing, and boundary crossing.

[0133] (5) Constrained particle filter solution: weight update, normalization, effective particle count and resampling, position estimation output The positioning calculation uses constrained particle filtering. For the first... Each window positions the computational endpoint for the state of each particle. Calculate Bluetooth observation likelihood LoRa observational likelihood and constraint terms and according to weighting coefficients By weighting and combining two similar factors, the unnormalized weights of the particles are obtained and then normalized.

[0134] in: For the first The first window Normalized weights of individual particles; For Bluetooth observation likelihood function; This is a Bluetooth feature; For the first The first window Individual particle states; For the first Bluetooth observation weighting coefficients for each window; For LoRa observation likelihood function; LoRa characteristics; For the first LoRa observation weighting coefficients for each window; For constraint terms; The number of particles; For window indexing; For particle indexing; For particle indexing.

[0135] The positioning calculation unit further calculates the effective number of particles to determine the degree of particle degradation:

[0136] in: The effective number of particles; The number of particles; For the first The first window Normalized weights of individual particles; For window indexing; For particle indexing.

[0137] when When the value is below a preset threshold, resampling is triggered. After resampling, the number of particles remains unchanged and particle diversity is restored. When the constraint term causes a large number of particles to have zero weights, the localization calculation end first removes all zero particles and supplements the propagated particles before resampling to ensure that the normalized denominator is non-zero and the posterior can be calculated.

[0138] Finally, the localization calculation terminal outputs based on the particle posterior distribution. and will Index by window It generates continuous trajectory output for real-time positioning and safety monitoring of tunnel construction personnel.

[0139] (6) Closed-loop implementation of factor graph optimization branch (same as particle filter) When using factor graph optimization, the location calculation terminal uses the location status of each window. As graph nodes, based on motion priors (by...) The generated adjacent state constraints) are used as motion factors, with Bluetooth observation likelihood (by...) The resulting observation consistency constraint) is used as the Bluetooth factor, with the LoRa observation likelihood (generated by...) The resulting observation consistency constraint) is used as the LoRa factor, and with and and allow migration rule generation As a constraint factor.

[0140] The location calculation terminal accumulates the negative log-likelihoods of all factors and uses iterative optimization to obtain the result that minimizes the overall cost. It maintains the same residual definition, variance normalization method, and weight adaptive rule as particle filtering, thus ensuring that the output results of the two solution methods are consistent and interpretable under the same input data stream.

[0141] Through the above process, step S4 performs probabilistic fusion of the multi-source observations generated in step S2 and continuously delivered by step S3 under a unified window structure, and injects the geometric reachability and semantic transfer rules of step S1 into the inference in a hard constraint manner, so as to stably obtain a three-dimensional position estimate that satisfies spatial semantic constraints even under tunnel multipath, occlusion and link fluctuation conditions. This provides a continuous sequence of positioning results for subsequent positioning output and safety monitoring.

[0142] S5. Perform 3D ambiguity resolution and output localization results and loop closure update, specifically including: mapping candidate locations to zone identifiers (ZoneID) and updating based on the region connectivity graph. The migration rule eliminates or penalizes candidates that cross regions or altitudes to resolve ambiguities; location reliability is calculated based on posterior distribution concentration, observation residual consistency, and constraint violation rate. The output includes the positioning terminal identifier (TagID), calibration time, and 3D position estimate. Location reliability The positioning results are compared with the ZoneID; when preset stable or abnormal conditions are met, online updates are performed on the Bluetooth propagation parameters and / or LoRa measurement bias parameters, and the 3D reachability domain is triggered based on construction events or abnormal consistency. and / or region connectivity graph edge set Dynamic updates.

[0143] Location reliability in step S5 The calculation is based at least on the posterior distribution concentration, observation residual consistency, and constraint violation rate. The observation residual consistency is composed of the normalized sum of squares of the Bluetooth and LoRa residuals, and the constraint violation rate is calculated based on whether the candidate state falls within the three-dimensional reachable domain. The proportion outside; and when multiple consecutive windows meet the stability condition that the displacement increment is less than the threshold, the Bluetooth propagation parameters and / or LoRa ranging bias parameters are updated online by minimizing the observation residuals. When a change in the reachability domain or an abnormal consistency condition corresponding to a construction event is detected, the three-dimensional reachability domain is triggered. and / or region connectivity graph edge set Dynamic updates.

[0144] Step S5 specifically includes: Step S5 performs 3D ambiguity resolution on the posterior results output by S4 and outputs the localization results and loop closure update: the candidate locations are mapped to the Zone ID and updated according to the region connectivity graph. The migration rule eliminates or penalizes candidates that cross regions or altitudes to resolve ambiguities; location reliability is calculated based on posterior distribution concentration, observation residual consistency, and constraint violation rate. The output includes the positioning terminal identifier (TagID), calibration time, and 3D position estimate. Location reliability The positioning results are compared with the ZoneID; when preset stable or abnormal conditions are met, online updates are performed on the Bluetooth propagation parameters and / or LoRa measurement bias parameters, and the 3D reachability domain is triggered based on construction events or abnormal consistency. and or region connectivity graph edge set The dynamic updates ensure that the positioning output remains continuous and reliable over a long period of time, even under changes in the construction environment.

[0145] (1) Three-dimensional ambiguity resolution: ZoneID mapping and allow migration rule elimination / penalty The localization computation unit takes the posterior distribution obtained in step S4 as input and performs spatial semantic consistency filtering on the posterior candidate states. For each candidate state... The positioning calculation terminal calculates the ZoneID according to the encoding rules (mileage segment encoding, cross-section interval encoding, and work area type encoding) in step S1. Map to the Zone ID and construct a cross-zone migration pair with the Zone ID confirmed in the previous window. .

[0146] The positioning calculation terminal strictly follows the regional connectivity map. edge set And allow migration rules to perform ambiguity resolution, and ambiguity resolution adopts a determination rule of "removal priority, penalty as a last resort": When the cross-regional migration pair corresponding to the candidate state does not belong to the edge set If a candidate state simultaneously satisfies the conditions that the mileage segment codes are not adjacent or the cross-sectional interval codes are not adjacent, the positioning calculation end will directly remove the candidate state. When the cross-regional migration pair corresponding to the candidate state does not belong to the edge set However, when a candidate state satisfies the conditions of adjacent mileage segment codes and adjacent cross-sectional interval codes, the positioning calculation end does not remove the candidate state but performs a penalty process, that is, the posterior weight of the candidate state is reduced by a fixed proportion, so that it may only be retained when other candidates are also inconsistent. When a candidate state involves vertical crossing, the positioning calculation terminal only allows vertical migration if the ZoneID of the candidate state is located in the preset connecting passage area, shaft opening area, or stair section area; otherwise, the candidate state is directly eliminated.

[0147] After the elimination or penalty is completed, the localization calculation terminal renormalizes the weights of the remaining candidate states and takes the state cluster corresponding to the largest posterior weight peak after renormalization as the final candidate peak. The localization calculation terminal uses the weighted estimate of this candidate peak as the 3D position estimate for this window. and output The corresponding ZoneID forms a location output that is consistent with the spatial location and semantic region.

[0148] This processing addresses the posterior multi-peak phenomenon caused by long tunnel corridors and structural repetition. Without changing the fusion framework of step S4, it uses regional connectivity graphs and allowable migration rules to quickly suppress pseudo-peaks across regions and heights, thus avoiding impossible cross-layer jumps and cross-regional instantaneous shifts in the positioning trajectory.

[0149] (2) Location reliability Calculation: Concentration Residual consistency With constraint violation rate

[0150] After resolving ambiguities, the location computing unit calculates the location reliability according to the established procedure. It is based on at least three types of quantities: posterior distribution concentration. Consistency of observation residuals With constraint violation rate .

[0151] 1) Posterior distribution concentration Fixed-caliber calculation: The localization calculation terminal uses the maximum value of the particle posterior normalized weights as the concentration index.

[0152] This indicator increases monotonically as it converges from multi-peaked to uni-peaked in the posterior, and can directly reflect the degree of ambiguity.

[0153] 2) Consistency of observation residuals Using a fixed-caliber calculation: The positioning calculation end adds the window-level Bluetooth residual consistency value used for observation likelihood calculation in step S4 to the window-level LoRa residual consistency value to obtain the result. ,Right now It consists of the normalized sum of squares of the Bluetooth and LoRa residuals, and the consistency of the two residuals follows the principle of "summing the squared residuals after normalizing them according to their corresponding variances". This index increases monotonically when multipath spikes, occlusion, or accumulated measurement bias cause deviations between observations and geometric predictions.

[0154] 3) Constraint violation rate Defined as a candidate state falling within a two-dimensional reachable region The proportion outside. The positioning calculation terminal checks each particle in the particle set to see if it falls into the range. And calculate the violation rate:

[0155] in: For the first Constraint violation rate of each window; The number of particles; This is a reachable domain indicator function; For the first The first window Individual particle states; For window indexing.

[0156] The location calculation unit fuses the above three types of quantities according to a defined monotonic mapping relationship to obtain the location reliability. Furthermore, the fusion coefficient is set as a fixed configuration parameter, so that the confidence level can be reproducibly calculated under different working conditions.

[0157] Location reliability takes the following form:

[0158] in: For the first Location reliability of a single window; This represents the concentration integration coefficient. For the first The posterior distribution concentration of each window; The residual fusion coefficient; For the first Consistency of observation residuals for each window; The violation rate fusion coefficient; For the first Constraint violation rate of each window.

[0159] This confidence level construction maps "degree of ambiguity, observation consistency, and constraint feasibility" into a comparable benchmark: Distinguish between unimodal convergence and multimodal coexistence. Inconsistencies caused by quantization multipath / occlusion / bias Explicitly suppressing out-of-bounds candidates, the three factors together ensure that even spurious peaks with short-term high likelihood cannot obtain high confidence, thus enabling the localization output to have an interpretable confidence metric.

[0160] (3) Location result output: The fields are fixed and consistent with the time semantics. The positioning calculation terminal outputs positioning result records in each window. The record fields are fixed and include: positioning terminal identifier (TagID) and calibration time. 3D position estimation Location reliability With ZoneID. Calibration time uses steps. The given window represents the correction time. Ensure that the output time is strictly consistent with the S2 window organization, S3 convergence key, and S4 inference window; the zone identifier (ZoneID) is determined by... The encoding rules in step S1 are used to ensure that the output semantic region is consistent with the spatial location.

[0161] (4) Online update: Stability condition determination and minimizing observation residuals to update Bluetooth propagation parameters and / or LoRa measurement bias parameters When multiple consecutive windows meet the stability condition that the displacement increment is below a threshold, the positioning calculation terminal enters online update mode. The stability condition uses a fixed criterion: the positioning calculation terminal calculates the displacement increment of consecutive windows. When continuous Each window satisfies When the condition for stability is met, then... Configure parameters for the number of consecutive windows. Configure parameters for the displacement increment threshold. The stability condition means that the actual displacement is approximately zero. At this time, the observation residual is mainly caused by the drift of wireless propagation parameters and the drift of measurement bias, making the online update of "minimizing the observation residual" identifiable and convergent.

[0162] Online updates have clearly defined object boundaries: For the Bluetooth side, the positioning calculation terminal uses the Bluetooth propagation parameters used in the Bluetooth observation likelihood in step S4 as the set of parameters to be updated. It uses the Bluetooth normalized squared residuals within the stable window set as the objective function to perform a recursive update that minimizes the observation residuals, and immediately applies the updated Bluetooth propagation parameters to subsequent windows. calculate; For the LoRa side, the positioning calculation terminal uses the LoRa ranging offset parameters introduced in step S2. As a parameter to be updated, and only for the AnchorIDs that actually appear in the stable window set. Perform a recursive update, with the objective function being the LoRa normalized squared residuals within the stable window set, and then update the result. Immediately used in subsequent windows calculate.

[0163] This online update incorporates propagation parameter drift caused by slow environmental changes and ranging offset caused by equipment temperature drift / installation deviations into the model parameters in real time, preventing long-term accumulation of residuals. Continued rise and The output continues to decrease, thus ensuring long-term stability of the positioning output.

[0164] (5) Dynamic update: Triggered by consistency of construction events or anomalies AND / or edge set renew When a change in the reachability domain or an abnormal consistency condition corresponding to a construction event is detected, the positioning calculation terminal triggers an analysis of the three-dimensional reachability domain. and / or region connectivity graph edge set Dynamic updates are implemented to keep the constraint model consistent with the actual traffic conditions on site.

[0165] 1) Construction event triggering mechanism: The positioning calculation terminal receives the construction event identifier and parses its impact range and type. The construction event is identified as an event that causes changes in the passage boundary or connectivity relationship.

[0166] Location calculation terminal updates according to event type :when When represented as a voxel raster set, perform reachability / unreachability marker updates on the affected voxels; when When representing a set of corridor polyhedra, update the boundary parameters of the corresponding corridor polyhedra.

[0167] 2) The abnormal consistency triggering mechanism adopts a fixed judgment criterion: when consecutive Each window simultaneously satisfies and Monotonically increasing, or continuously increasing Each window satisfies At that time, the positioning calculation terminal determines that the constraint model deviates from the on-site traffic status and triggers a dynamic update, in which Configure parameters for the number of abnormal windows. Configure parameters for the confidence threshold. Configure parameters for the violation rate threshold.

[0168] 3) Edge set Update rule: When the localization results repeatedly show a certain ZoneID migration pair under high confidence conditions, and this migration pair is not in In the middle, the positioning calculation end adds the migration pair to the edge set. When a certain edge causes a large number of candidate states to be allowed to migrate within a continuous window, but the violation rate continues to increase in the reachability domain determination, the positioning calculation terminal removes that edge from the range. Remove or disable the connection to make the connectivity consistent with the open / closed status of the on-site passage.

[0169] Through the closed-loop mechanism of "ambiguity resolution - confidence quantification - online update of stable conditions - dynamic update triggered by construction / anomalies", step S5 transforms the posterior inference result of step S4 into a final positioning output that satisfies spatial semantic constraints and has interpretable confidence, and integrates Bluetooth propagation parameters, LoRa measurement bias, and... and edge set Maintaining consistency under changing construction environments ensures that the positioning results, as the final output, remain stable, continuous, and reliable over the long term.

[0170] Example 2 A system employing an intelligent 3D positioning method based on Bluetooth beacons and LoRa ad hoc networks, characterized in that it includes: Constraint Modeling Module: Establishes a three-dimensional positioning constraint model of the tunnel construction area and completes the node system configuration; Observation preprocessing module: The positioning terminal collects and preprocesses multi-source observation data to generate a unified observation package; Self-organizing network transmission module: Based on LoRa self-organizing network, it performs collaborative aggregation and reliable transmission of unified observation packets. Fusion localization module: performs multi-source fusion 3D localization solution with semantic constraints; Ambiguity closure module: Performs 3D ambiguity resolution and outputs localization results and closure update.

[0171] The technical features of this invention not described can be implemented by or using existing technology, and will not be repeated here. Of course, the above description is not a limitation of this invention, and this invention is not limited to the examples above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of this invention should also be within the protection scope of this invention.

Claims

1. A smart 3D positioning method based on Bluetooth beacons and LoRa ad hoc networks, characterized in that, Includes the following steps: S1. Establish a three-dimensional positioning constraint model of the tunnel construction area and complete the node system configuration, specifically including: establishing a local three-dimensional coordinate system for the tunnel and defining the area identifier ZonelD; constructing a three-dimensional reachable domain based on the tunnel spatial structure. and regional connectivity graph The edges of the connected region graph carry migration rules; Bluetooth beacons are deployed at key locations in the tunnel and the Bluetooth beacon ID, known spatial location, and Zone ID of each Bluetooth beacon are stored; LoRa nodes are deployed to form a LoRa ad hoc network and the LoRa node ID, node role, and known spatial location of each LoRa node are stored. S2. The positioning terminal collects and preprocesses multi-source observation data to generate a unified observation package, specifically including: collecting Bluetooth scan data, LoRa measurement data, and inertial / gait increment data, wherein the multi-source observation data consists of the Bluetooth scan data, the LoRa measurement data, and the inertial / gait increment data; establishing a network time reference through LoRa ad hoc network time synchronization messages and estimating the deviation between the positioning terminal's local clock and the network time to obtain a correction time; performing time alignment and aggregation on the Bluetooth scan data, the LoRa measurement data, and the inertial / gait increment data based on the correction time according to a preset window length, and extracting Bluetooth features from the Bluetooth scan data respectively. Extracting LoRa features from the LoRa measurement data Inertial / gait increments are formed from the inertial / gait increment data. Generate a unified observation package The unified observation package includes at least the positioning terminal identifier (TagID), the window representing the correction time, and Bluetooth characteristics. LoRa characteristics Inertia / gait increment And quality indicators; and based on the quality indicators, generate a pre-confidence level for communication reliability control. ; S3. Based on the LoRa self-organizing network, the unified observation packets are collaboratively aggregated and reliably transmitted, specifically including: each LoRa node calculates the routing cost based on link quality, hop count, and latency, and selects a multi-hop forwarding path; relay nodes or edge aggregation nodes statistically aggregate and compress the unified observation packets to generate a minimum positioning data packet for positioning and perform differential coding and / or event-triggered coding; according to preset reliability conditions, the minimum positioning data packet is acknowledged, retransmitted, and / or forwarded with multi-path redundancy to ensure that positioning data is continuously delivered to the positioning calculation end; S4. Perform multi-source fusion 3D localization solution with semantic constraints, specifically including: defining the localization state. A motion prior model is established based on inertia / gait increments to predict the state; Bluetooth observation likelihood functions are established based on Bluetooth features. LoRa observation likelihood function is established based on LoRa features. The three-dimensional reachable region Connectivity graph of the region Introducing constraint priors to generate constraint terms Calculate the Bluetooth and LoRa residuals, and adaptively adjust the weight coefficients of the Bluetooth and LoRa observation likelihood functions based on the residuals. Then, based on the motion prior model, the Bluetooth and LoRa observation likelihood functions, and the constraint terms, obtain the 3D position estimate using constrained particle filtering or factor graph optimization. ; S5. Perform 3D ambiguity resolution and output the localization result and loop closure update, specifically including: mapping candidate locations to region identifiers (ZoneID) and updating them according to the region connectivity graph. The migration rule eliminates or penalizes candidates that cross regions or altitudes to resolve ambiguities; location reliability is calculated based on posterior distribution concentration, observation residual consistency, and constraint violation rate. The output includes the positioning terminal identifier (TagID), calibration time, and 3D position estimate. Location reliability The positioning results are compared with the ZoneID; when preset stable or abnormal conditions are met, online updates are performed on the Bluetooth propagation parameters and / or LoRa measurement bias parameters, and the three-dimensional reachability domain is triggered based on construction events or abnormal consistency. and / or the region connectivity graph edge set Dynamic updates.

2. The intelligent three-dimensional positioning method based on Bluetooth beacons and LoRa ad hoc networks according to claim 1, characterized in that, The local three-dimensional coordinate system of the tunnel mentioned in step S1 adopts mileage coordinates. Cross-sectional transverse coordinates With vertical coordinates express; The ZoneID is composed of at least mileage segment code, cross-section interval code, and work area type code; The migration rules include at least the following: vertical migration is allowed only when the ZoneID of the candidate location is located in a preset communication channel area, shaft opening area, or stair section area; otherwise, vertical crossing is prohibited.

3. The intelligent three-dimensional positioning method based on Bluetooth beacons and LoRa ad hoc networks according to claim 2, characterized in that, The three-dimensional reachable region described in step S1 Represented using a voxel grid set or a corridor polyhedron set; And the constraint terms mentioned in step S4 It at least includes a candidate state to the three-dimensional reachable domain. The soft penalty term determined by the minimum distance to the boundary.

4. The intelligent three-dimensional positioning method based on Bluetooth beacons and LoRa ad hoc networks according to claim 1, characterized in that, The correction time mentioned in step S2 is obtained by the LoRa node periodically broadcasting a time synchronization message containing a network timestamp; The positioning terminal estimates the clock skew based on at least two time synchronization messages. and / or drift items, and convert the local timestamp to the corrected time. This allows for unified time alignment of Bluetooth scan data, LoRa measurement data, and inertial / gait increment data.

5. The intelligent three-dimensional positioning method based on Bluetooth beacons and LoRa ad hoc networks according to claim 1, characterized in that, Bluetooth features mentioned in step S2 At least include the mean RSSI value, the variance of the RSSI value, and the preset number of Bluetooth beacons with the highest strength for each beacon within the window. A set of Bluetooth beacons; Median filtering and / or Huber suppression are applied to the Received Signal Strength Indicator (RSSI) sequence within the window to suppress multipath spike anomalies, and Bluetooth beacon observations with fewer reads than a threshold or variance higher than a threshold are removed.

6. The intelligent three-dimensional positioning method based on Bluetooth beacons and LoRa ad hoc networks according to claim 1, characterized in that, The LoRa features described in step S2 One of the following: a) a time-of-flight-based ranging set And introduce ranging offset parameters for each anchor point. ; or b) a set of time differences based on arrival time differences ; Furthermore, the LoRa feature further includes the measurement variance derived from the signal-to-noise ratio and / or quality metric, which is used to weight the LoRa observation likelihood function in step S4.

7. The intelligent three-dimensional positioning method based on Bluetooth beacons and LoRa ad hoc networks according to claim 1, characterized in that, The routing cost function mentioned in step S3 is: in: The value of routing; hop is the hop count weighting coefficient; hop is the hop count. The link packet loss rate is the weighting coefficient; the loss is the link packet loss rate. "delay" represents the link delay weighting coefficient; "delay" represents the link delay. Furthermore, when the minimum positioning data packet corresponds to the pre-confidence level When the value is below a preset threshold, acknowledgment and adaptive retransmission are enabled for the minimum positioning data packet, and the number of retransmissions is related to... Positively correlated, or multi-path redundant forwarding is used for the minimum positioning data packet; in: Pre-confidence level; For window indexing.

8. The intelligent three-dimensional positioning method based on Bluetooth beacons and LoRa ad hoc networks according to claim 1, characterized in that, When using constrained particle filtering to solve in step S4, it includes: propagating particles based on the motion prior model; and based on the Bluetooth observation likelihood function. The LoRa observation likelihood function and the aforementioned constraint terms Calculate and normalize the particle weights; Calculate the effective number of particles: When the effective number of particles Resampling is triggered when the value is below a preset threshold; and a 3D position estimate is obtained based on the particle posterior distribution. ; in: It is a conditional probability function; This is a Bluetooth feature; LoRa characteristics Positioning status; For constraint terms; The effective number of particles; For the first The first window Normalized weights of individual particles; For window indexing.

9. The intelligent three-dimensional positioning method based on Bluetooth beacons and LoRa ad hoc networks according to claim 1, characterized in that, The location confidence level described in step S5 The calculation is based at least on the posterior distribution concentration, observation residual consistency, and constraint violation rate, where the observation residual consistency is composed of the normalized sum of squares of the Bluetooth and LoRa residuals, and the constraint violation rate is calculated based on whether the candidate state falls within the three-dimensional reachable domain. The proportion outside; and when multiple consecutive windows meet the stability condition that the displacement increment is below the threshold, the Bluetooth propagation parameters and / or LoRa ranging bias parameters are updated online by minimizing the observation residuals. When a change in the reachability domain or an abnormal consistency condition corresponding to a construction event is detected, the three-dimensional reachability domain is triggered. and / or the region connectivity graph edge set Dynamic updates.

10. A system employing the intelligent three-dimensional positioning method based on Bluetooth beacons and LoRa ad hoc networks as described in any one of claims 1-9, characterized in that, include: Constraint Modeling Module: Establishes a three-dimensional positioning constraint model of the tunnel construction area and completes the node system configuration; Observation preprocessing module: The positioning terminal collects and preprocesses multi-source observation data to generate a unified observation package; Self-organizing network transmission module: Based on LoRa self-organizing network, it performs collaborative aggregation and reliable transmission of the unified observation packets. Fusion localization module: performs multi-source fusion 3D localization solution with semantic constraints; Ambiguity closure module: Performs 3D ambiguity resolution and outputs localization results and closure update.

Images