A downhole multi-scene adaptive positioning method and device

Abstract

The present application relates to a kind of downhole multi-scene adaptive positioning method and device, belong to positioning registration technical field.The present application constructs downhole typical scene set, configures corresponding scene adaptive constraint weight for different downhole scenes, and constructs mapping relationship;Mobile equipment extracts downhole scene feature during moving in downhole;Using downhole multi-scene identification model, the current scene is identified according to the scene feature vector;Mobile equipment adaptively matches scene adaptive constraint weight according to the scene label of current time, to adjust positioning constraint weight adaptively;Registration residual and cross-scene continuity residual influenced by scene adaptive constraint weight are unified to optimization target, and cross-scene adaptive joint optimization model is obtained;By solving cross-scene adaptive joint optimization model, the continuous pose sequence of mobile equipment in downhole environment is obtained.The present application realizes accurate smooth positioning by pose constraint adjustment and cross-scene continuity residual.

Description

Technical Field

[0001] This invention relates to the field of adaptive positioning technology for multiple scenarios in underground mines, and specifically to an adaptive positioning method and device for multiple scenarios in underground mines. Background Technology

[0002] With the development of intelligent underground transportation, unmanned inspection, mining robots, and autonomous operation technologies in underground spaces, mobile equipment in mines faces higher demands for continuous and stable positioning capabilities. Unlike open environments on the surface, underground environments typically cannot receive reliable satellite navigation signals and are characterized by enclosed spaces, insufficient lighting, dust interference, repetitive structures, and severe obstruction. Therefore, fusion positioning methods based on sensors such as lidar, inertial measurement units, and odometers have become an important means for unmanned equipment in underground mines to achieve autonomous navigation.

[0003] Existing positioning and mapping methods can achieve good positioning results in single-structure environments. For example, in regular straight tunnels, inertial navigation units can assist lidar in positioning; in curved areas, changes in environmental curvature can provide certain directional constraints; and in intersection areas, branching structures can provide richer geometric features. However, the real-world underground working environment is not a single scene, but a continuous combination of various scenes such as straight tunnels, curves, intersections, and wide tunnels. When mobile equipment switches between different scenes, the environmental geometry, observable constraint directions, point cloud matching stability, and sensor constraint reliability all change significantly.

[0004] The existing technology has the following main shortcomings: First, traditional positioning methods typically use fixed registration parameters and uniform residual weights, making it difficult to adapt to the different constraints in various underground scenarios. For example, the geometric constraints along the direction of the straight tunnel are relatively weak, while the structural constraints are more complex in intersection scenarios. If the same constraint weights are used, the positioning results may become unstable in some scenarios.

[0005] Second, most methods lack explicit scene recognition and scene switching modeling. When a robot moves from a straight aisle to a curve, intersection, or wide aisle, the point cloud distribution and matching residuals will change abruptly. If the system cannot recognize the current scene and adjust the positioning constraints in time, pose jumps, trajectory drifts, or short-term positioning interruptions are likely to occur.

[0006] Third, existing fusion localization methods often focus more on the registration accuracy of a single frame, while insufficiently considering the constraints on trajectory continuity across scenes. At scene boundaries, the local map structure changes significantly, and relying solely on the current frame's point cloud registration may lead to pose estimation being affected by local mismatches, thereby disrupting the overall trajectory continuity.

[0007] Fourth, the spatial scale varies significantly in areas such as wide tunnels and intersections underground, and the local geometric boundaries are unstable. Traditional registration methods based on fixed neighborhoods or fixed weights are prone to constraint imbalance. Especially in wide tunnel areas, the observable range of point clouds expands, but the effective structural features are dispersed, which can easily lead to local matching fluctuations.

[0008] Therefore, there is an urgent need for an adaptive positioning method for multiple scenarios in underground mines, which can identify the current scenario category, adaptively adjust the positioning residual weights according to the geometric structure and constraint characteristics of different scenarios, and introduce continuity constraints during scenario switching, thereby improving the stability, continuity and robustness of the positioning system in satellite-free underground environments. Summary of the Invention

[0009] In order to solve the above-mentioned technical problems, or at least partially solve the above-mentioned technical problems, and in view of the above-mentioned shortcomings, the present invention provides a downhole multi-scenario adaptive positioning method and device.

[0010] In a first aspect, the present invention provides a downhole multi-scenario adaptive positioning method, comprising: A set of typical downhole scenarios is constructed, and corresponding scene adaptive constraint weights are configured for different downhole scenarios. A mapping relationship between scenarios and scene adaptive constraint weights is constructed, wherein the scene adaptive constraint weights are set according to the characteristics of the impact of the scenario on positioning accuracy. During the movement of the mobile device underground, underground scene features are extracted; the underground scene feature vector includes tunnel width features, tunnel curvature features, scene branch structure features, and observability features; The current scene is identified based on the scene feature vector using an underground multi-scene recognition model. Mobile devices adaptively match scene adaptive constraint weights based on the scene labels at the current moment, so as to adaptively adjust the positioning constraint weights. The point cloud registration residuals and cross-scene continuity residuals affected by scene adaptive constraint weights are unified into the optimization objective to obtain the cross-scene adaptive joint optimization model; by solving the cross-scene adaptive joint optimization model, the continuous pose sequence of the mobile equipment in the underground environment is obtained.

[0011] Furthermore, a set of typical downhole scenarios is pre-built, corresponding scene adaptive constraint weights are configured for different downhole scenarios, and a mapping relationship between scenarios and scene adaptive constraint weights is constructed: ; in, This represents the mapping function from scene to scene adaptive constraint weights. This represents the scene at time k, and the adaptive constraint weights for the scene at time k. Includes: point cloud registration constraint weights Inertial constraint weight and odometer constraint weights .

[0012] Furthermore, the mapping function from scene to scene adaptive constraint weights is constructed into a lookup table form: ; To adapt to different scenarios and improve positioning accuracy, the constraint characteristics of the scene adaptive constraint weights for each scenario are set as follows: Adaptive constraint weights for straight lane scenarios Reduce the dominance of point cloud registration on axial displacement, increase the weights of inertial navigation units and odometry, and suppress cumulative drift along the tunnel direction; adaptive constraint weights for curved tunnel scenarios. Increase the point cloud registration weights while maintaining constraints on the inertial navigation unit and odometry to improve the stability of attitude change estimation; adaptive constraint weights for intersection scenarios. Enhanced point cloud registration constraints utilize multi-directional geometric information of point clouds to complete registration; adaptive constraint weights for wide-lane scenes. The weights of the inertial navigation unit and wheel speed odometer are increased to suppress pose fluctuations caused by mismatches in local point clouds.

[0013] Furthermore, the tunnel width feature is defined using the lateral boundary distance of the point cloud; The curvature characteristics of the tunnel are based on historical trajectories or local tunnel centerline curvature. The scene branch structure features are defined by the number of local point clouds occupying grids or the number of traffic directions. The observability features Calculated from the effective point cloud quantity and spatial distribution uniformity: ; in: This indicates the number of valid points participating in localization in the current frame's point cloud; This indicates the total number of point clouds in the current frame; It indicates the degree of dispersion of the spatial distribution of point clouds.

[0014] Furthermore, the scene classification function uses the scene feature vector at the current moment. The input is used to identify the current underground scene category using preset rules based on scene features.

[0015] Furthermore, the feature vector of a given scene is calculated using a scoring function corresponding to the scene category. In the case that the current scene belongs to category The probability of: ; in: Represents the feature vector in a given scene In the case that the current scene belongs to category The probability of; This represents the scoring function corresponding to the scene category, and the current scene feature vector. The higher the degree of matching with a certain scene, the stronger the current scene feature vector. The higher the score of the scoring function corresponding to the scene category, the higher the score. S represents the set of typical underground scenes, and the final scene label is determined by the highest probability.

[0016] Furthermore, by unifying the registration residuals affected by the scene adaptation constraint weights and the cross-scene continuity residuals into the optimization objective, a cross-scene adaptive joint optimization model is obtained: ; in: Let represent the registration residual vector at time k; To be based on scene tags Adaptive constraint weights for adaptive matching scenarios; This indicates a smooth scene transition constraint. This indicates the scene switching weight.

[0017] Furthermore, the cross-scene adaptive joint optimization model controls the continuity of cross-scene localization by limiting the tuning amount of the system state through cross-scene continuity residuals. The calculation process of the cross-scene continuity residuals includes: The system state is related to the motion state of the mobile device. A propagation model of the system state is pre-constructed based on the kinematic model of the mobile device. , in: This indicates the predicted current system state; Represents the state propagation function; This represents the system state at the previous moment. This indicates the motion input provided by the inertial measurement unit or odometer; During operation underground, mobile equipment constructs the system state using multi-source observation data and utilizes the system state propagation model based on the system state at the previous moment. and the predicted current system state ; Then, subtracting the predicted current system state from the current optimized system state yields the cross-scenario continuity residual: ; in, Indicates the continuity residual across scenarios; Indicates the current optimization state; Indicates the change from the system state at the previous moment. and motion input The predicted current state.

[0018] Furthermore, the system state includes the mobile device's position, attitude, speed, acceleration bias, and gyroscope bias. Alternatively, the system status of the mobile device may include the mobile device's location and its orientation rotation matrix. Secondly, the present invention provides an underground multi-scenario adaptive positioning device, comprising: at least one processing unit, the processing unit being connected to a storage unit via a bus unit, the storage unit storing a computer program, and the processing unit implementing the underground multi-scenario adaptive positioning method by running the computer program stored in the storage unit.

[0019] Thirdly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed, implements the downhole multi-scenario adaptive positioning method.

[0020] The technical solutions provided in the embodiments of the present invention have the following advantages compared with the prior art: First, this invention addresses the problem of continuous positioning in multiple underground scenarios. It establishes a set of typical scenarios, including straight tunnels, curved tunnels, intersections, and wide tunnels, and identifies the current environment category through scene features. This allows it to adapt to complex changes in underground scene structures and achieve accurate scene recognition.

[0021] Second, based on the geometric constraint characteristics of different scenarios, the present invention adaptively adjusts the weights of point cloud registration constraints, inertial navigation unit constraints, and odometry constraints to improve positioning accuracy in scenarios involving multiple scene switching.

[0022] Third, the cross-scene adaptive joint optimization model of this invention introduces cross-scene continuous residuals, introduces state continuity constraints when switching scenes, and limits the amount of change in the state of the optimization system by combining scene switching weights and cross-scene continuous residuals to suppress pose jumps, so that the mobile equipment maintains smooth trajectory and stable positioning when switching between straight lanes, curved lanes, intersections and wide lanes. Attached Figure Description

[0023] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

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

[0025] Figure 1 A flowchart of a downhole multi-scenario adaptive positioning method provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating the construction of a cross-scenario adaptive joint optimization model provided in an embodiment of the present invention; Figure 3 An architecture diagram of downhole multi-scenario adaptive positioning provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of an underground multi-scenario adaptive positioning device provided in an embodiment of the present invention. Detailed Implementation

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

[0027] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0028] Example 1 This invention provides an adaptive localization method for multiple scenarios in underground mines. It constructs a set of typical underground scenarios, extracts point cloud geometric features, motion features, and historical pose features to achieve real-time identification of typical underground scenarios, adaptively adjusts the weights of point cloud registration constraints, inertial constraints, and odometry constraints according to the identified scenario categories, and outputs smooth, continuous, and stable localization results by solving a cross-scenario continuous optimization model that considers scenario continuity.

[0029] like Figure 1 and Figure 3 As shown, the specific process of the method includes: A set of typical downhole scenarios is constructed, and corresponding scene adaptive constraint weights are configured for different downhole scenarios. A mapping relationship between scenarios and scene adaptive constraint weights is constructed, where the scene adaptive constraint weights are set according to the characteristics of the impact of scenarios on positioning accuracy.

[0030] First, a set of typical underground scenarios is constructed, covering straight tunnels, curved tunnels, intersections, and wide tunnels. Different underground scenarios have different positioning constraint characteristics. To increase positioning accuracy, this application configures corresponding scene adaptive constraint weights for different underground scenarios.

[0031] The scene adaptation constraint weights at time k include: point cloud registration constraint weights. Used to adjust the impact of point cloud geometric matching on the localization results; inertial constraint weights Used to adjust the effects of inertial prediction and short-term motion constraints; odometry constraint weights This is used to adjust the impact of the odometry on pose estimation. The scene adaptive constraint weights at time k are expressed as: .

[0032] During real-time positioning, it is necessary to select scene adaptive constraint weights based on the current scene. Specifically, this application constructs a mapping relationship between the scene and the scene adaptive constraint weights: ; in, This represents the mapping function from scene to scene adaptive constraint weights. Let k represent the scene at time k. In one example, the mapping function from scene to scene adaptive constraint weights is written in lookup table form: .

[0033] To adapt to the impact of different scenarios on positioning and improve positioning accuracy, the constraint characteristics of the scene adaptive constraint weights for each scenario are set as follows: In straight tunnels, the geometric constraints of the point cloud along the tunnel direction are relatively weak, and the scene adaptive constraint weights for straight tunnels are... To reduce the dominance of point cloud registration on axial displacement, the weights of the inertial navigation unit and odometry are increased to suppress cumulative drift along the tunnel direction. In curves, where trajectory direction changes significantly, the point cloud geometry can provide certain directional constraints, and adaptive constraint weights are implemented for curved tunnel scenarios. To improve the stability of attitude change estimation, the point cloud registration weights are increased while maintaining constraints on the inertial navigation unit and wheel speed odometry. At intersections, multi-directional walls, branches, and boundary structures provide richer geometric constraints, and the intersection scene adaptive constraint weights are implemented. Enhanced point cloud registration constraints utilize multi-directional geometric information of the point cloud for registration. In wide aisles, the number of point cloud features increases, and axial constraints are strengthened compared to straight aisles. However, effective structural features are dispersed, and local matching may be unstable. Adaptive constraint weights are implemented for wide aisle scenarios. The weights of the inertial navigation unit and wheel speed odometer are increased to suppress pose fluctuations caused by mismatches in local point clouds.

[0034] During the movement of the mobile device underground, features of the underground scene are extracted.

[0035] Based on the current frame point cloud, local map, and historical trajectory, the mobile device constructs the underground scene feature vector at the current moment. The underground scene feature vector at time k includes tunnel width features. tunnel curvature characteristics Scene branch structure features and observability features The extracted features of the underground scene are then represented as: .

[0036] Lane width characteristics Using the horizontal boundary distance of the point cloud: ,in, This represents the estimated distance from the current location of the mobile device to the left lane boundary; This represents the estimated distance from the current location of the mobile device to the right lane boundary. tunnel curvature characteristics The curvature of the centerline of a historical tunnel or a local tunnel is used. Taking the historical trajectory as an example: ; in, These represent the positions at three consecutive time points: k-2, k-1, and k, respectively. This is to smooth out terms and prevent extremely small positive numbers with a denominator of zero.

[0037] Scene branch structure features It is obtained from the number of grid cells or the number of traffic directions occupied by local point clouds.

[0038] Taking the number of traffic directions as an example, ; in, This indicates the number of passable branches detected near the current location. When this value is large, the current scene is more likely to be an intersection.

[0039] Observability features Calculated from the effective point cloud quantity and spatial distribution uniformity: ; in: This indicates the number of valid points participating in localization in the current frame's point cloud; This indicates the total number of point clouds in the current frame; It indicates the degree of dispersion of the spatial distribution of point clouds.

[0040] Mobile devices utilize an underground multi-scene recognition model to identify the current scene based on the scene feature vector: ; in: Represents a scene classification function; This represents the identified scene category. In one example, the scene classification function uses the scene feature vector at the current moment. As input, a preset rule based on scene features is used to determine the current underground scene category. For example, when the tunnel width feature is used... Less than the set width threshold, branch structure features Weaker than set value, curvature The change is less than the set curvature threshold, and the observable characteristics are... When the distribution is uniform, it can be identified as a straight tunnel; when the curvature characteristics are uniform... When the curve is large, it can be identified as a bend; when a branch is passable... The number is large and the observable characteristics are significant. When the distribution is scattered, it can be identified as an intersection; when the width characteristics of the alleyway... Significantly increased and observable features When the distribution is relatively scattered, it can be identified as a wide tunnel.

[0041] To improve the scene recognition stability of the downhole multi-scene recognition model, a scoring function is further introduced to calculate the feature vector of a given scene. In the case that the current scene belongs to category The probability of: ; in: Represents the feature vector in a given scene In this case, the current scene belongs to the category The probability of; This represents the scoring function corresponding to the scene category, and the current scene feature vector. The higher the degree of matching with a certain scene, the stronger the current scene feature vector. The higher the score of the scoring function corresponding to the scene category, the higher the score. For example, the more branches, the higher the score of the intersection scene; the greater the curvature, the higher the score of the curve scene; S represents the set of typical underground scenes, and the final scene label is determined by the maximum probability. The example scoring function is based on the similarity of scene feature vectors.

[0042] Mobile devices based on the scene label at the current time k Adaptive matching of scene adaptive constraint weights, in order to adaptively adjust the positioning constraint weights.

[0043] To provide more accurate positioning, this application configures different scene-adaptive constraint weights for different scenarios. Since mobile devices move continuously, pose estimation jumps occur during weight switching. To avoid pose estimation jumps during scene transitions, a cross-scene adaptive joint optimization model is constructed. Cross-scene continuity residuals are introduced to limit the tuning amount of the optimization system state, achieving continuity in cross-scene positioning. Figure 2 As shown, the specific items include: The system state is related to the motion state of the mobile device. A propagation model of the system state is pre-constructed based on the kinematic model of the mobile device. , in: This indicates the predicted current system state; Represents the state propagation function; This represents the system state at the previous moment. This indicates the motion input provided by the inertial measurement unit or odometer.

[0044] In one example, the mobile device acquires real-time multi-source observation data through its own lidar, inertial measurement unit, and odometry to obtain the real-time system state of the mobile device. The system state at time k includes the position of the mobile device. Mobile device posture Mobile device speed Zero bias acceleration of mobile devices Zero bias of gyroscope in mobile devices Then the system state at time k is: . In one example, the mobile device acquires real-time multi-source observation data through its own LiDAR and odometry to obtain the real-time system state of the mobile device. The system state at time k includes the location of the mobile device. Mobile device attitude rotation matrix Then the system state at time k is: . During operation underground, mobile equipment constructs the system state using multi-source observation data and utilizes the system state propagation model based on the system state at the previous moment. and the predicted current system state .

[0045] Then, subtracting the predicted current system state from the current optimized system state yields the cross-scenario continuity residual: ; in, Indicates the continuity residual across scenarios; Indicates the current optimization state; Indicates the change from the system state at the previous moment. and motion input The predicted current state.

[0046] By unifying the registration residuals affected by scene-adaptive constraint weights and the cross-scene continuity residuals into the optimization objective, a cross-scene adaptive joint optimization model is obtained: ; in: This represents the registration residual vector at time k, which includes the point cloud registration residual, the inertial navigation registration residual, and the odometry registration residual. To be based on scene tags Adaptive constraint weights for adaptive matching scenarios; This indicates a smooth scene transition constraint. This indicates the scene switching weight. The cross-scene adaptive joint optimization model can automatically adjust the importance of different constraint terms according to the current downhole scene, and enhance the continuity constraint at scene switching points, thereby suppressing pose jumps and trajectory interruptions.

[0047] By solving the cross-scenario adaptive joint optimization model, the continuous pose sequence of the mobile equipment in the underground environment is obtained: ; in: This represents the optimal pose sequence from the initial time to the current time. This represents the optimal pose at the current time k. Ultimately, it achieves smooth switching and trajectory continuity between scenarios such as straight tunnels, curves, and intersections, and enables stable positioning output in underground environments without satellite support.

[0048] Example 2 like Figure 4 As shown, this embodiment of the invention provides a downhole multi-scene adaptive positioning device, including: at least one processing unit, the processing unit being connected to a storage unit via a bus unit, the storage unit serving as a computer-readable storage medium, which can be used to store software programs, computer-executable programs, and modules, such as the software program, computer-executable program, and module corresponding to the downhole multi-scene adaptive positioning method in this embodiment of the invention. The processing unit implements the aforementioned downhole multi-scene adaptive positioning method by running the software program, computer-executable program, and module stored in the storage unit.

[0049] Of course, the computer program stored in the storage unit of the downhole multi-scenario adaptive positioning device provided in the embodiments of the present invention is not limited to the method operation described above, but can also execute related operations in the downhole multi-scenario adaptive positioning method provided in any embodiment of the present invention.

[0050] In the embodiments provided by this invention, it should be understood that the disclosed structures and methods can be implemented in other ways. For example, the structural embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, structures, or units, and may be electrical, mechanical, or other forms.

[0051] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0052] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0053] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A multi-scenario adaptive positioning method for underground drilling, characterized in that, include: Construct a set of typical downhole scenarios and configure corresponding scenario adaptive constraint weights for different downhole scenarios; And construct a mapping relationship from scene to scene adaptive constraint weight, where the scene adaptive constraint weight is set according to the characteristics of the scene's impact on positioning accuracy; During the movement of the mobile device underground, features of the underground scene are extracted; The feature vector of the underground scene includes tunnel width features, tunnel curvature features, scene branch structure features, and observability features; The current scene is identified based on the scene feature vector using an underground multi-scene recognition model. Mobile devices adaptively match scene adaptive constraint weights based on the scene labels at the current moment, so as to adaptively adjust the positioning constraint weights. By unifying the point cloud registration residuals affected by scene adaptive constraint weights and the cross-scene continuity residuals into the optimization objective, a cross-scene adaptive joint optimization model is obtained: ； in: Let represent the registration residual vector at time k; These are the adaptive constraint weights for the scene, which are obtained through adaptive matching based on the scene labels. This indicates a smooth scene transition constraint. Indicates the scene switching weight. This represents the scenario at time k. Indicates the continuity residual across scenarios; By solving the cross-scenario adaptive joint optimization model, a continuous pose sequence of mobile equipment in the underground environment is obtained.

2. The downhole multi-scenario adaptive positioning method according to claim 1, characterized in that, A set of typical downhole scenarios is pre-built, and corresponding scene adaptive constraint weights are configured for different downhole scenarios. A mapping relationship between scenarios and scene adaptive constraint weights is also constructed. ； in, The function representing the mapping from scene to scene adaptive constraint weights, where k represents the scene's adaptive constraint weights. Includes: point cloud registration constraint weights Inertial constraint weight and odometer constraint weights .

3. The downhole multi-scenario adaptive positioning method according to claim 2, characterized in that, The mapping function from scene to scene adaptive constraint weights is constructed into a lookup table form: ； To adapt to different scenarios and improve positioning accuracy, the constraint characteristics of the scene adaptive constraint weights for each scenario are set as follows: Adaptive constraint weights for straight lane scenarios Reduce the dominance of point cloud registration on axial displacement, increase the weights of inertial navigation units and odometry, and suppress cumulative drift along the tunnel direction; adaptive constraint weights for curved tunnel scenarios. Increase the point cloud registration weights while maintaining constraints on the inertial navigation unit and odometry to improve the stability of attitude change estimation; adaptive constraint weights for intersection scenarios. Enhanced point cloud registration constraints utilize multi-directional geometric information of point clouds to complete registration; adaptive constraint weights for wide-lane scenes. The weights of the inertial navigation unit and wheel speed odometer are increased to suppress pose fluctuations caused by mismatches in local point clouds.

4. The downhole multi-scenario adaptive positioning method according to claim 1, characterized in that, The tunnel width feature is defined by the lateral boundary distance of the point cloud. The curvature characteristics of the tunnel are based on historical trajectories or local tunnel centerline curvature. The scene branch structure features are defined by the number of local point clouds occupying grids or the number of traffic directions. The observability features Calculated from the effective point cloud quantity and spatial distribution uniformity: ； in: This indicates the number of valid points participating in localization in the current frame's point cloud; This indicates the total number of point clouds in the current frame; It indicates the degree of dispersion of the spatial distribution of point clouds.

5. The downhole multi-scenario adaptive positioning method according to claim 1, characterized in that, The scene classification function uses the scene feature vector at the current moment. The input is used to identify the current underground scene category using preset rules based on scene features.

6. The downhole multi-scenario adaptive positioning method according to claim 5, characterized in that, Calculate the feature vector of a given scene using a scoring function corresponding to the scene category. The probability that the current scene belongs to the following category: ； in: Represents the feature vector in a given scene The probability that the current scene belongs to a certain category; This represents the scoring function corresponding to the scene category, and the current scene feature vector. The higher the degree of matching with a certain scene, the stronger the current scene feature vector. The higher the score of the scoring function corresponding to the scene category, the higher the score. S represents the set of typical underground scenes, and the final scene label is determined by the highest probability.

7. The downhole multi-scenario adaptive positioning method according to claim 1, characterized in that, The cross-scene adaptive joint optimization model controls the continuity of cross-scene localization by limiting the tuning amount of the system state through cross-scene continuity residuals. The calculation process of the cross-scene continuity residuals includes: The system state is related to the motion state of the mobile device. A propagation model of the system state is pre-constructed based on the kinematic model of the mobile device. ， in: This indicates the predicted current system state; Represents the state propagation function; This represents the system state at the previous moment. This indicates the motion input provided by the inertial measurement unit or odometer; During operation underground, mobile equipment constructs the system state using multi-source observation data and utilizes the system state propagation model based on the system state at the previous moment. and the predicted current system state ; Then, subtracting the predicted current system state from the current optimized system state yields the cross-scenario continuity residual: ; in, Indicates the current optimization state; Indicates the change from the system state at the previous moment. and motion input The predicted current state.

8. The downhole multi-scenario adaptive positioning method according to claim 1, characterized in that, System status includes the mobile device's position, attitude, speed, acceleration bias, and gyroscope bias. Alternatively, the system status of the mobile device may include the mobile device's location and its orientation rotation matrix.

9. A multi-scenario adaptive positioning device for underground wells, comprising: At least one processing unit is provided, wherein the processing unit is connected to a storage unit via a bus unit, wherein the storage unit stores a computer program, and the processing unit implements the downhole multi-scenario adaptive positioning method according to any one of claims 1-8 by running the computer program stored in the storage unit.

Images