Intelligent charging system for tunnel excavation

Abstract

The application discloses an intelligent charging system and method for tunnel excavation, and belongs to the field of mechanical hands and control systems thereof.The system comprises environment sensing and data acquisition, spatial topology planning, master-slave visual cooperative guidance, pose compensation and alignment execution, intelligent pushing control, safety interlocking and installation, and archive recording modules.The application constructs a blast hole topology relationship network through global scanning and plans an optimal sequence, combines master-slave visual cooperative guidance to process the movement of the mechanical arm and pre-aiming in parallel, in the execution phase, uses a flexible guide sleeve to realize tolerance alignment, and introduces a vibration gap-seeking self-adaptive pushing mechanism based on real-time force feedback.The application can effectively overcome problems such as insufficient light and dust interference in a poor tunnel environment, realize rapid, accurate and safe automatic charging, and improve the efficiency of full-face charging operation, the single success rate and the autonomy of equipment operation.

Description

Technical Field

[0001] This invention relates to the field of robotic arms and their control systems, and in particular to an intelligent explosive loading system for tunnel excavation. Background Technology

[0002] Tunnel excavation is a crucial step in infrastructure construction, with drill-and-blast being a commonly used method. In the blasting process, the quality of the explosive charge directly affects the blasting effect and construction safety. To improve operational efficiency, utilizing industrial robots or robotic arms to replace manual labor in tasks such as borehole positioning, explosive charge delivery, and detonation charge installation has become an important development direction in the field of tunnel engineering automation.

[0003] Among related technologies, Chinese invention patent CN111664762A discloses an automatic loading control system and method for a mixed explosives cart based on borehole positioning parameters, including a data acquisition module, a calculation module, and a loading positioning module. This technology uses the data acquisition module to acquire the three-dimensional coordinates and depth of the borehole during drilling; calculates the explosive charge amount corresponding to each borehole; and positions the explosives cart each time the loading personnel move the loading hose to a borehole, matching the coordinates with the acquired coordinates, thereby controlling the explosives cart to load the corresponding explosive charge into the borehole.

[0004] Regarding the aforementioned technologies, the inventors believe that although this solution reduces the technical requirements for personnel, it still has the following drawbacks: First, its explosive loading positioning relies on the manual movement of the hose to the borehole by the loading personnel, resulting in limited automation and low efficiency during large-scale full-face operations. Second, the technology employs a rigid pushing and matching method, which requires extremely high regularity of the blast holes. When faced with irregular blast holes or rock debris interference within the tunnel, the explosive loading hose is prone to jamming or positioning deviation, leading to operation interruption. Furthermore, it lacks effective environmental awareness and fault self-recovery capabilities, making it unsuitable for long-term stable use under complex and variable tunnel conditions. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides an intelligent charging system for tunnel excavation. It employs a technical solution that combines spatial topology planning, master-slave vision collaborative guidance, flexible tolerance alignment, and intelligent push control, enabling the robotic arm to perform efficient, precise, and safe automated charging operations.

[0006] To achieve the above objectives, this application adopts the following technical solution:

[0007] In a first aspect, an intelligent charging system for tunnel excavation is provided, comprising: an environmental perception and data acquisition module for acquiring a global depth image of the tunnel blasting face collected by a global vision device, and acquiring borehole design data of the face; a spatial topology planning module for generating a borehole spatial topology network containing all borehole nodes and their spatial relationships based on the global depth image and borehole design data, and generating sequential operation instructions to guide the robotic arm's movements based on the borehole spatial topology network; and a master-slave vision collaborative guidance module for driving the robotic arm to move toward the target borehole according to the sequential operation instructions, and simultaneously initiating a master-slave vision collaborative guidance process, wherein the master-slave vision collaborative guidance process is used to fuse first positioning data acquired by the global vision device and second positioning data acquired by an end-effector mounted on the end of the robotic arm to generate an end-effector positioning data. The system includes: a dynamic attitude adjustment command; a posture compensation and alignment execution module, used to respond to the end-effector's dynamic attitude adjustment command, controlling the actuator at the end of the robotic arm to move towards the borehole, and completing the tolerance alignment of the propellant tube by contacting the borehole with the flexible guide sleeve at the front end of the actuator; an intelligent push control module, used to send a vibration gap-finding push command after the tolerance alignment is completed, receive thrust and velocity feedback data monitored in real time by the internal sensors of the tube feeder, and generate an adaptive push control command or a constant speed push control command to control the tube feeder to complete the push operation; a safety interlock and installation module, used to acquire the propellant tube in-situ confirmation signal reflecting the propellant tube's in-situ status and the detonation charge in-situ confirmation signal reflecting the detonation charge's material supply status, and control the robotic arm to perform the detonation charge installation operation based on the safety interlock logic; and a file recording module, used to record and upload the loading process file including the thrust and velocity feedback data.

[0008] Based on the above technical solutions, the intelligent charging system for tunnel excavation provided in this application adopts a combination of spatial topology planning, master-slave vision collaborative guidance, flexible tolerance alignment and intelligent push control, which can control the robotic arm to achieve efficient, accurate and safe automated charging operations.

[0009] In conjunction with the first aspect above, in one possible implementation, the spatial topology planning module includes: an initial pose extraction unit, used to process the global depth image using a target detection algorithm to identify the initial pose set of all visible boreholes within the field of view; a spatial matching and completion unit, used to spatially register the initial pose set with the borehole design data using a registration algorithm, and apply a transformation matrix to correct the borehole design data, constructing a borehole spatial topology network containing the true poses of all target boreholes; a movement cost calculation unit, used to calculate the movement cost between any two borehole nodes in the borehole spatial topology network, the movement cost being weighted by the spatial distance between the two nodes and the expected recognition confidence of the target borehole node by the end vision device; and a job instruction planning unit, used to apply a path search algorithm to solve for the path that traverses all borehole nodes and has the minimum total movement cost based on the movement cost, and generate a serialized job instruction.

[0010] In conjunction with the first aspect described above, in one possible implementation, the master-slave vision collaborative guidance module includes: a macroscopic offset tracking unit, used to continuously track the spatial position of the robotic arm end effector relative to the current target borehole area through the global vision device, to obtain a coarse spatial offset as first positioning data; a local fine scanning unit, used to perform a three-dimensional scan of the current target borehole through the end effector vision device during the movement of the robotic arm, to obtain fine point cloud data of the borehole as second positioning data; a data fusion estimation unit, used to apply a Kalman filter algorithm to fuse the coarse spatial offset with the fine point cloud data to estimate the real-time relative pose of the actuator relative to the target borehole opening; and a trajectory dynamic correction unit, used to compare the real-time relative pose with a preset optimal observation pose to calculate the pose error, and generate speed commands to drive the movement of each joint of the robotic arm as end effector pose dynamic adjustment commands.

[0011] In conjunction with the first aspect above, in one possible implementation, the pose compensation and alignment execution module includes: a pose error compensation unit, used to drive the end effector of the robotic arm to perform translation and rotation compensation according to the end effector pose dynamic adjustment command, so that the central axis of the actuator is aligned with the orifice normal vector of the target borehole; a flexible contact feedback unit, used to determine whether a reliable flexible sealing connection is established between the actuator and the borehole by monitoring the compression deformation state of the flexible guide sleeve during the process of the actuator approaching the borehole; and a tolerance alignment verification unit, used to confirm that the passage for the propellant tube to enter the borehole is connected within a preset allowable radial deviation range based on the physical limiting space of the flexible guide sleeve.

[0012] In conjunction with the first aspect above, in one possible implementation, the intelligent push control module includes: a vibration excitation control unit, used to control the tube feeder to perform high-frequency micro-vibration with a set vibration frequency and amplitude; an initial low-speed push unit, used to drive the tube feeder to advance the propellant tube forward at a set initial low speed under the high-frequency micro-vibration state; and a friction reduction and clearance-finding guidance unit, used to guide the propellant tube to advance along the path of least resistance in the initial section of the borehole through the friction reduction effect generated by the coupling effect of the high-frequency micro-vibration and the initial low speed.

[0013] In conjunction with the first aspect above, in one possible implementation, the safety interlock and installation module includes: a precondition verification unit, used to verify whether, before generating the control command for grabbing the detonating charge, both a charge tube in-place confirmation signal reflecting the placement of the charge tube inside the actuator and a detonating charge in-place confirmation signal reflecting sufficient material at the target workstation of the detonating charge supply device are simultaneously received; an instruction authorization control unit, used to authorize the execution of the control command for grabbing the detonating charge only when the verification results of the precondition verification unit are both valid; and a grabbing status confirmation unit, used to verify the grabbing success feedback signal from the actuator gripper after executing the control command for grabbing the detonating charge, and determine whether to continue executing the subsequent detonating charge installation operation.

[0014] In conjunction with the first aspect above, in one possible implementation, the intelligent push control module further includes: a jamming state determination unit, used to determine that jamming has occurred and trigger an obstacle encounter response strategy when the thrust and velocity feedback data meet the preset jamming conditions of abnormally increased thrust and abnormally decreased velocity; an adaptive instruction generation unit, used to generate an adaptive push control instruction including a retraction instruction, a vibration mode adjustment instruction, and a secondary propulsion instruction in response to the obstacle encounter response strategy; and a strategy iteration execution unit, used to control the tube feeder to execute the retraction instruction to relieve stress, and to execute a secondary propulsion instruction to attempt to overcome the obstacle after adjusting the vibration frequency or amplitude.

[0015] In conjunction with the first aspect described above, in one possible implementation, the system further includes a vision self-maintenance module: a real-time confidence calculation unit, used to calculate in real time the continuous recognition confidence of the target bore output by the end vision device; a contamination event determination unit, used to determine that a lens contamination trigger event has occurred when the continuous recognition confidence is lower than a set threshold and meets preset degradation characteristics, provided that the ambient lighting conditions have been optimized; and a self-cleaning execution unit, used to automatically control the cleaning mechanism installed on the protective cover of the end vision device to perform cleaning actions in response to the lens contamination trigger event.

[0016] In conjunction with the first aspect described above, in one possible implementation, the visual self-maintenance module further includes: a grayscale feature analysis unit, used to acquire the grayscale histogram distribution features of the current image acquired by the end vision device; a supplementary light signal generation unit, used to generate a dynamic supplementary light control signal based on the difference between the grayscale histogram distribution features and a preset ideal contrast range; and a brightness closed-loop adjustment unit, used to adjust the brightness of the supplementary light lamp according to the dynamic supplementary light control signal to make the contrast of the current image approach the target range, so as to prioritize improving the confidence of continuous recognition by improving illumination.

[0017] Secondly, an intelligent charging method for tunnel excavation is provided, comprising: acquiring a global depth image of the tunnel blasting face collected by a global vision device, and acquiring borehole design data of the face; generating a borehole spatial topology network containing all borehole nodes and their spatial relationships based on the global depth image and borehole design data, and planning and generating sequential operation instructions to guide the robotic arm's movements according to the borehole spatial topology network; driving the robotic arm to move towards the target borehole according to the sequential operation instructions, and simultaneously initiating a master-slave vision collaborative guidance process, wherein the master-slave vision collaborative guidance process is used to fuse first positioning data acquired by the global vision device and second positioning data acquired by an end vision device mounted on the end of the robotic arm, generating... The system generates a dynamic end-effector orientation adjustment command; responding to the dynamic end-effector orientation adjustment command, it controls the actuator at the end of the robotic arm to move towards the borehole, and completes the tolerance alignment of the propellant tube by contacting the borehole through the flexible guide sleeve at the front end of the actuator; after the tolerance alignment is completed, it sends a vibration gap-finding push command, receives thrust and velocity feedback data monitored in real time by the internal sensors of the tube feeder, and generates an adaptive push control command or a constant speed push control command to control the tube feeder to complete the push operation; it acquires a propellant tube in-situ confirmation signal reflecting the propellant tube's in-situ status and an detonation charge in-situ confirmation signal reflecting the detonation charge's material supply status, and controls the robotic arm to perform the detonation charge installation operation based on safety interlock logic; it records and uploads the loading process archive, including the thrust and velocity feedback data.

[0018] Compared with the prior art, the present invention has the following advantages:

[0019] This invention constructs a master-slave collaborative guidance architecture that includes global vision and end-effector vision, and combines it with a flexible-guided end effector, enabling the robotic arm system to perceive and adapt to the actual positional deviation of the blast holes on the tunnel working face in real time. This eliminates the dependence on high-precision pre-programmed paths and improves the positioning accuracy and operational adaptability of the robotic arm in unstructured environments.

[0020] This invention proposes a global path planning method based on spatial topology network and movement cost calculation, so that the working path of the robotic arm is no longer a fixed linear sequence, but an optimal traversal sequence dynamically planned according to the movement distance and visual recognition confidence, which shortens the overall operation cycle time and improves the working efficiency of the robotic arm.

[0021] This invention integrates an intelligent push control strategy based on force and speed feedback with a strict safety interlock logic. The former effectively solves the problem of the propellant tube getting stuck in irregular blast holes through vibration gap finding and adaptive obstacle detection and escape, while the latter ensures the absolute safety of handling high-risk materials such as detonating charges through multiple precondition checks, thereby improving the reliability and safety of robotic arm operations.

[0022] This invention adds a visual self-maintenance module, which can proactively diagnose and handle the problem of decreased recognition ability caused by lens contamination. By prioritizing intelligent supplemental lighting optimization and performing automatic cleaning when necessary, it ensures the long-term stable operation of the core perception system, enhances the autonomy and environmental tolerance of the entire robotic arm system, and reduces the frequency of manual maintenance.

[0023] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[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, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A flowchart illustrating an intelligent charging method for tunnel excavation provided in this application embodiment;

[0026] Figure 2A structural architecture diagram of an intelligent charging system for tunnel excavation provided in this application embodiment;

[0027] Figure 3 This is an iterative convergence curve of the path planning algorithm provided in the embodiments of this application.

[0028] Figure 4 This is a distribution diagram showing the influence of vibration parameters on the equivalent friction coefficient provided in the embodiments of this application. Detailed Implementation

[0029] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0030] The intelligent charging system for tunnel boring provided in this application embodiment can be applied to, for example... Figure 1 In the intelligent charging method for tunnel excavation shown, such as Figure 1 As shown, the method includes:

[0031] Acquire global depth images of the tunnel blasting face from a global vision device, and obtain the borehole design data for the face.

[0032] Based on the global depth image and borehole design data, a borehole spatial topology network containing all borehole nodes and the spatial relationships between nodes is generated, and a sequenced operation instruction to guide the robotic arm's movements is planned and generated according to the borehole spatial topology network.

[0033] The robotic arm is driven to move toward the target bore according to the serialized operation instructions, and the master-slave vision collaborative guidance process is started simultaneously. The master-slave vision collaborative guidance process is used to fuse the first positioning data obtained by the global vision device and the second positioning data obtained by the end vision device mounted on the end of the robotic arm to generate the end pose dynamic adjustment instruction.

[0034] In response to the end-position dynamic adjustment command, the actuator at the end of the robotic arm is controlled to move toward the borehole, and the tolerance alignment of the propellant tube is completed by the flexible guide sleeve at the front end of the actuator contacting the borehole.

[0035] After the tolerance alignment is completed, a vibration gap-finding push command is sent, and the thrust and speed feedback data monitored and fed back in real time by the internal sensors of the tube feeder are received. An adaptive push control command or a constant speed push control command is generated to control the tube feeder to complete the push operation.

[0036] The system acquires a tube-in-place confirmation signal reflecting the tube's position and an initiation charge-in-place confirmation signal reflecting the supply status of the initiation charge, and controls the robotic arm to perform the initiation charge installation operation based on safety interlock logic.

[0037] Record and upload the loading process archive, which includes the thrust and velocity feedback data.

[0038] like Figure 2 As shown in the figure, this application provides an intelligent charging system for tunnel boring, including:

[0039] The environmental perception and data acquisition module is used to acquire global depth images of the tunnel blasting working face collected by the global vision device, and to acquire the blast hole design data of the working face.

[0040] The spatial topology planning module is used to generate a spatial topology network of boreholes that includes all borehole nodes and the spatial relationships between nodes based on the global depth image and borehole design data, and to plan and generate sequential operation instructions to guide the robotic arm's movements based on the borehole spatial topology network.

[0041] The master-slave vision collaborative guidance module is used to drive the robotic arm to move toward the target gun hole according to the serialized operation instructions, and simultaneously start the master-slave vision collaborative guidance process. The master-slave vision collaborative guidance process is used to fuse the first positioning data obtained by the global vision device and the second positioning data obtained by the end vision device mounted on the end of the robotic arm to generate the end pose dynamic adjustment instruction.

[0042] The pose compensation and alignment execution module is used to respond to the end pose dynamic adjustment command, control the actuator at the end of the robotic arm to move toward the borehole, and complete the tolerance alignment of the propellant tube by contacting the borehole with the flexible guide sleeve at the front end of the actuator.

[0043] The intelligent push control module is used to send a vibration gap-finding push command after the tolerance alignment is completed, receive the thrust and speed feedback data monitored and fed back in real time by the internal sensors of the tube feeder, and generate an adaptive push control command or a constant speed push control command to control the tube feeder to complete the push operation.

[0044] The safety interlock and installation module is used to acquire the in-situ confirmation signal of the explosive tube reflecting the in-situ status and the in-situ confirmation signal of the detonating explosive charge reflecting the supply status of the detonating explosive charge, and to control the robotic arm to perform the installation operation of the detonating explosive charge based on the safety interlock logic.

[0045] The file recording module is used to record and upload the loading process file, which includes the thrust and velocity feedback data.

[0046] It should be noted that by integrating global 3D vision with the design blueprint, a realistic spatial topology network of blast holes on the blasting face was constructed, and based on this, the most efficient global operation sequence was planned. During execution, a master-slave vision collaboration approach was adopted, using global vision for macroscopic path guidance and robotic arm end-effector vision for local fine-grained pose correction, achieving dynamic and precise approximation of the target blast hole. In the final docking stage, the flexible guide sleeve at the end of the robotic arm made flexible physical contact with the blast hole opening, utilizing its own guidance to complete the final tolerance alignment. During the propellant delivery, vibration gap-finding technology was used to overcome the initial stage of resistance, and adaptive control was achieved through real-time force and velocity feedback to cope with complex conditions inside the hole. In the critical detonator installation stage, strict safety interlocking logic was introduced, and multiple status signals were used for confirmation to ensure absolute operational safety. Key data from the entire operation process were recorded and archived, providing a basis for subsequent analysis and traceability.

[0047] In one possible implementation of the embodiments of this application, combined with Figure 2 The spatial topology planning module includes:

[0048] The initial pose extraction unit is used to process the global depth image using a target detection algorithm to identify the initial pose set of all visible boreholes within the field of view.

[0049] The spatial matching and completion unit is used to spatially register the initial pose set with the borehole design data through a registration algorithm, and to apply a transformation matrix to correct the borehole design data, thereby constructing a borehole spatial topology network containing the true poses of all target boreholes.

[0050] The movement cost calculation unit is used to calculate the movement cost between any two borehole nodes in the borehole spatial topology network. The movement cost is composed of the spatial distance between the two nodes and the expected recognition confidence of the target borehole node by the end vision device.

[0051] The operation instruction planning unit is used to apply a path search algorithm to find the path that traverses all borehole nodes and minimizes the total movement cost based on the movement cost, and generate serialized operation instructions.

[0052] In some implementations, the initial pose extraction unit quickly identifies all clearly visible boreholes from complex field images. It calls a deep learning-based object detection algorithm, such as YOLOv7 or Mask R-CNN, to process the global depth image provided by the environment perception and data acquisition module. This algorithm, trained on a large number of borehole image samples, can accurately identify the 2D contours of the boreholes in the image and, combined with depth data, calculates the 3D spatial coordinates of the center point of each borehole and the normal vector of its orifice plane, thus forming an initial pose set. To ensure extraction accuracy, the pose data of a borehole is only included in the initial pose set if the algorithm's output recognition confidence score is higher than a preset threshold, such as 0.9. The spatial matching and completion unit uses a few identified boreholes as a reference to align and correct the entire borehole design blueprint, ultimately generating a spatial topology network of boreholes containing the true poses of all boreholes. This unit employs an iterative nearest-point registration algorithm to register the 3D point cloud in the initial pose set with the corresponding points in the borehole design data. The algorithm iteratively calculates the optimal 3D spatial transformation matrix, which precisely describes the overall translational and rotational deviations of the actual borehole group relative to its theoretical design position. This unit applies this transformation matrix to all borehole design data to correct the pose of each designed borehole, including those not initially identified due to obstruction or poor lighting. This ultimately constructs a complete borehole spatial topology network highly consistent with the actual site conditions. This network is a graph structure, where each node represents a borehole with a precise 3D pose. The movement cost calculation unit processes the constructed borehole spatial topology network, quantifying the overall cost of the robotic arm moving from one borehole to another, providing a basis for subsequent path optimization decisions. For any two borehole nodes in the network... and This unit calculates the cost of movement between them, which is not simply a matter of spatial distance, but rather a weighted average of distance cost and visual guidance risk cost. (Movement Cost) Calculated using the following formula:

[0053] ;

[0054] in, This represents the final cost of the move. It is the dimensionless value of the spatial Euclidean distance between the two borehole nodes after normalization, reflecting the time and energy cost of movement. The expected identification risk of the target borehole node is calculated as follows: ,in This refers to the expected recognition confidence level of the end-view vision device for a target blast hole. This confidence level is pre-assessed based on factors such as the expected imaging angle of the target blast hole and whether there are obstructions. A blast hole that is easier to identify has a higher risk value. Lower. and This is a weighting coefficient, an empirical parameter obtained during the debugging phase based on the robotic arm's motion performance and the on-site visual environment calibration, used to balance the importance of movement efficiency and visual guidance success rate. This formula ensures a unified quantitative comparison of different physical quantities. The operation instruction planning unit solves for the optimal path that traverses all borehole nodes and minimizes the total movement cost, generating the final serialized operation instructions. This problem is essentially a traveling salesman problem. Considering computational efficiency, this unit applies heuristic path search algorithms such as ant colony optimization or simulated annealing. The algorithm takes the cost matrix generated by the movement cost calculation unit as input, simulates the path exploration process of multiple generations of "ants" in the borehole spatial topology network, and gradually converges to a traversal path with a total movement cost close to the global optimum through the positive feedback mechanism of pheromones. The output of the algorithm is a clear borehole visit sequence, such as [bombhole 7, borehole 21, borehole 15, ...]. This sequence is the serialized operation instruction that guides the robotic arm to perform automated loading operations and is sent to the master-slave visual collaborative guidance module to start the physical operation process. Figure 3 As shown, through iterative calculation, the total movement cost of the entire cross section can be converged to the global optimal interval in a very short time, thus verifying the role of spatial topology planning network in improving operational efficiency.

[0055] For example, in a tunnel loading operation, the initial pose extraction unit processes the global depth image using the YOLO algorithm, identifying 80 visible boreholes with a confidence level higher than 0.9 from 120 designed boreholes and forming a pose set; the spatial matching and completion unit uses the ICP algorithm to register this set with the design data and calculates the translation amount. The spatial transformation matrix is ​​used to correct the poses of all 120 boreholes and construct a complete spatial topology network of boreholes. When preparing to start from the currently completed borehole... Move to the blast hole to be operated When calculating, substitute the values ​​into the formula, assuming the normalized distance between the two holes. The value is 0.5, and the end-view vision device is used to locate the blast hole. Expected identification confidence A value of 0.8 indicates that risk has been identified. Set weight coefficients and Then the movement cost of this path The task planning unit compares this cost with the costs of other candidate paths and uses an ant colony algorithm to find the serialized task instruction with the minimum total movement cost within 15 seconds, such as determining the priority to access the blast hole. Rather than the more costly gun ports This guides the robotic arm to begin the physical operation process.

[0056] In one possible implementation, combining Figure 2 The master-slave visual collaborative guidance module includes:

[0057] The macroscopic offset tracking unit is used to continuously track the spatial position of the robotic arm end relative to the current target borehole group area through the global vision device, so as to obtain a rough spatial offset as the first positioning data.

[0058] A local fine scanning unit is used to perform a three-dimensional scan of the current target borehole through the end vision device during the movement of the robotic arm, so as to obtain fine point cloud data of the borehole as second positioning data.

[0059] The data fusion estimation unit is used to apply the Kalman filter algorithm to fuse the coarse spatial offset with the fine point cloud data in order to estimate the real-time relative pose of the actuator with respect to the target borehole.

[0060] The trajectory dynamic correction unit is used to compare the real-time relative pose with the preset optimal observation pose to calculate the pose error, and generate speed commands to drive the movement of each joint of the robotic arm as end-effector pose dynamic adjustment commands.

[0061] In some implementations, the macroscopic offset tracking unit provides a continuous and stable global position reference during the large-scale movement of the robotic arm. A global vision device mounted on the equipment platform continuously monitors the cooperative target fixed to the end effector of the robotic arm at a low frame rate, such as 5 to 10 times per second. The target center is located using image recognition algorithms, and combined with depth information, a coarse spatial offset between the current position of the robotic arm's end effector and the center of the target borehole area in the sequential operation command is calculated. This three-dimensional offset vector serves as the first positioning data, with an accuracy typically at the centimeter level. Its main function is to overcome the cumulative errors and low-frequency vibrations of the robotic arm itself, ensuring that the robotic arm can reach the approximate area of ​​the target borehole. When the robotic arm's end effector is guided into a preset distance from the target borehole, such as 0.5 meters, based on the first positioning data, a local fine scanning unit acquires high-precision three-dimensional geometric information of the target borehole opening and its surrounding area. The end effector vision device mounted on the end effector of the robotic arm, typically a structured light 3D camera or a laser contour scanner, performs one or more rapid scans of the target borehole. The scanning process generates fine point cloud data containing thousands to tens of thousands of data points, accurately depicting the actual shape of the borehole, its center position, and the direction of the normal vector of the borehole plane. This high-density point cloud serves as the second localization data, with local measurement accuracy reaching sub-millimeter level, providing crucial geometric basis for subsequent accurate pose estimation. The data fusion estimation unit begins processing the two-channel vision data. It optimally fuses macroscopic, low-frequency coarse localization data with local, high-precision fine localization data to generate a smooth and accurate real-time relative pose estimate of the actuator. A Kalman filter algorithm is employed as the core. The filter's state vector is defined as the six-degree-of-freedom relative pose of the actuator relative to the target borehole, i.e., three-dimensional position and three-dimensional attitude. In the prediction step of the filtering, the kinematic model of the robotic arm and the state at the previous moment are used to predict the current pose; in the update step, the coarse spatial offset output by the macroscopic offset tracking unit and the borehole center pose calculated by the local fine scanning unit are used as measurements to correct the predicted state. In this way, the algorithm effectively filters out various measurement noises and model uncertainties, outputting a high-fidelity estimate of the real-time relative pose between the actuator and the borehole. The trajectory dynamic correction unit generates the final control command based on the fused pose estimate. It calculates the velocity command required to drive the robotic arm to eliminate the current pose error, forming a dynamic closed loop. This unit compares the real-time relative pose output by the data fusion estimation unit with the preset optimal observation pose. The optimal observation pose is an ideal spatial pose defined for the end-effector, which allows for a clear and complete observation of the borehole while facilitating the next actuator insertion. The deviation between the two constitutes the pose error. Accordingly, the unit generates velocity commands to drive the movement of each joint of the robotic arm through a proportional controller, serving as the end-effector pose dynamic adjustment command. It can be calculated using the following formula:

[0062] ;

[0063] in, It outputs a six-dimensional velocity command, which includes the linear velocity and angular velocity components that the end effector needs to execute. It is the six-degree-of-freedom pose error vector between the current real-time relative pose and the optimal observation pose. It is a 6x6 gain matrix, whose diagonal elements are calibrated to adjust the system's response speed and stability; its physical unit is Hertz. This end-effector pose dynamic adjustment command is directly sent to the robotic arm's underlying controller, driving the robotic arm to smoothly correct its trajectory until the pose error converges to a very small threshold range, completing the cooperative guidance process.

[0064] For example, as the robotic arm moves towards the target borehole cluster, the macroscopic offset tracking unit monitors the end-effector at a frequency of 5Hz using a global vision device, and calculates the first positioning data between the current end-effector and the center of the target area, i.e., the coarse spatial offset. Centimeters; once the robotic arm enters the preset range, the local fine scanning unit activates the end-effector structured light camera to perform a 3D scan, acquiring sub-millimeter-level fine point cloud data as the second positioning data; the data fusion estimation unit then uses a Kalman filter algorithm to fuse the two data streams to calculate the real-time relative pose of the actuator relative to the borehole. for Centimeters; the trajectory dynamic correction unit compares the real-time relative pose with the preset optimal observation attitude. The pose error vector is calculated by comparing the centimeters. for centimeters, if the gain matrix is ​​set If the diagonal element of the position gain is 0.5 Hz, then the speed command for correcting the trajectory of the robotic arm can be calculated according to the formula. for At a speed of centimeters per second, the actuator is guided to converge smoothly and precisely to the ideal working position.

[0065] In one possible implementation, combining Figure 2 The pose compensation and alignment execution module includes:

[0066] The pose error compensation unit is used to drive the end of the robotic arm to perform translation and rotation compensation according to the end pose dynamic adjustment command, so that the central axis of the actuator is aligned with the orifice normal vector of the target gun hole;

[0067] The flexible contact feedback unit is used to determine whether a reliable flexible sealing connection has been established between the actuator and the borehole by monitoring the compression deformation state of the flexible guide sleeve during the process of the actuator approaching the borehole.

[0068] The tolerance alignment verification unit is used to confirm that the passage for the propellant tube to enter the borehole is connected within a preset allowable radial deviation range, based on the physical limiting space of the flexible guide sleeve.

[0069] In some implementations, the pose error compensation unit responds to end-effector pose dynamic adjustment commands received from the master-slave vision collaborative guidance module. It performs final fine-tuning to align the actuator with the target borehole axis. The received velocity commands are parsed into precise motion control for the six joints of the robotic arm, driving the end-effector to perform millimeter- or sub-millimeter-level translational compensation and sub-angular-level rotational compensation. This ensures the angle between the actuator's central axis and the borehole normal vector calculated by the end-effector vision device is less than a preset minimum threshold, such as 0.5 degrees, while simultaneously ensuring the lateral deviation between the actuator center and the borehole center is less than 2 millimeters, thus achieving high-precision pose alignment. After pose alignment, the flexible contact feedback unit confirms a stable flexible connection between the actuator and the borehole wall through physical contact. After pose error compensation, the controller instructs the actuator to approach the borehole along its own axis at a very low preset speed, such as 5 mm / s. The flexible guide sleeve installed at the front end of the actuator is a conical or bowl-shaped component made of highly elastic polyurethane or rubber. When the guide sleeve contacts the uneven rock wall of the borehole, it undergoes compressive deformation. The contact state is determined by real-time monitoring of the force sensor readings connected to the actuator or by directly measuring the sensor signal of the flexible guide sleeve's displacement. When the force sensor reading exceeds a set contact force threshold, such as 50 Newtons, or the compressive deformation of the flexible guide sleeve reaches a preset value, such as 5 to 10 millimeters, it is determined that a reliable flexible seal connection has been formed between the actuator and the borehole, and a signal is immediately sent to stop the actuator's forward movement. The tolerance alignment verification unit performs the final path confirmation, confirming that the path for the propellant to enter the borehole is open based on the physical characteristics of the flexible guide sleeve, without the need for additional sensors. The function of this unit is logical; its verification is based on the fact that the inner cavity of the flexible guide sleeve itself is a mechanical guiding structure. Its tapered inner wall can still smoothly guide the propellant into the borehole even when there is a certain radial deviation between the actuator center and the borehole center. This allowable radial deviation range is determined by the geometry of the flexible guide sleeve. For example, for a 45 mm diameter borehole, if the inner diameter of the guide sleeve is narrowed from 60 mm to 42 mm, it can theoretically compensate for a radial alignment deviation of up to 7.5 mm. Therefore, once the flexible contact feedback unit confirms reliable contact, it means that the relative position of the actuator and the borehole has fallen within this physical limit space. Based on this, the tolerance alignment verification unit confirms that the path is connected and sends an alignment completion signal. This signal will serve as a trigger condition to activate the intelligent push control module and begin the next step of the propellant tube push operation.

[0070] For example, the pose error compensation unit responds to commands to drive the robotic arm to perform sub-millimeter-level fine adjustments, bringing the angle between the actuator's central axis and the borehole normal vector to 0.3 degrees; the flexible contact feedback unit controls the actuator to approach the borehole at a speed of 5 mm / s. When the polyurethane flexible guide sleeve installed at the front end contacts the rock wall and the force sensor detects a contact force of 50 Newtons, it determines that a flexible sealing connection has been established and stops moving; the tolerance alignment verification unit performs physical verification based on the geometry of the guide sleeve. Assuming the borehole diameter is 45 mm and the inner diameter of the guide sleeve's flare is from... Narrowed to millimeters Millimeters, according to the formula, its theoretically permissible maximum radial deviation compensation amount can be calculated as follows: If the residual deviation of visual positioning is 6 mm at this time, since it is less than the physical limit range of 9 mm, the verification unit confirms that the passage for the propellant tube to enter the blast hole has been connected and triggers the subsequent push command.

[0071] In one possible implementation, combining Figure 2 The intelligent push control module includes:

[0072] The vibration excitation control unit is used to control the tube feeder to perform high-frequency micro-vibration at a set vibration frequency and amplitude;

[0073] An initial low-speed pushing unit is used to drive the tube feeder to advance the drug tube forward at a set initial low speed under the high-frequency micro-vibration state.

[0074] The friction-reducing and gap-finding guiding unit is used to guide the propellant tube along the path of least resistance in the initial section of the borehole through the friction-reducing effect generated by the coupling effect of the high-frequency micro-vibration and the initial low speed.

[0075] In some implementations, the vibration excitation control unit sends precise drive signals to an electromagnetic or piezoelectric vibrator integrated within the tube feeder, controlling it to perform high-frequency micro-vibration at a set frequency and amplitude. Depending on the hardness and roughness of the rock wall, this vibration frequency is typically set between 50 and 200 Hz, while the vibration amplitude is controlled between 0.1 and 0.5 mm. This small, rapid reciprocating motion effectively breaks the static friction between the outer wall of the propellant tube and the inner wall of the borehole, placing them in a quasi-suspended contact state. Simultaneously with the vibrator starting and stabilizing its vibration output, the initial low-speed pushing unit outputs a low-speed control command to the tube feeder's push motor, driving the propellant tube forward at a set initial low speed. This speed value is experimentally calibrated and is generally set to 20% to 40% of the normal pushing speed, for example, 50 to 150 mm / s. Maintaining a low-speed advance allows the propellant tube sufficient time to respond to the gap-finding effect of the vibration, avoiding rigid collisions and jamming caused by excessive speed. The friction-reducing clearance-finding guiding unit embodies the physical effect of the coupling of the two aforementioned actions, achieving passive adaptive guidance of the propellant tube. When the tip of the propellant tube encounters a protrusion or gravel in the initial section of the borehole during low-speed advancement, continuous high-frequency micro-vibrations prevent it from rigidly jamming. Instead, the vibration induces a slight lateral or rotational displacement at the tip of the propellant tube, continuously probing the surrounding clearances. Since the propellant tube tends to move in the direction of least force, this vibration-coupled low-speed advancement mode essentially guides the propellant tube along the path of least resistance, cleverly bypassing obstacles. This vibration clearance-finding pushing process typically operates for the first 200 to 300 millimeters of the propellant tube entering the borehole. After successfully passing through this initial section, it switches to a constant-speed or adaptive pushing control mode to complete the subsequent full-range pushing operation. Figure 4 As shown, the physical mechanism by which the static friction between the drug tube and the orifice wall is transformed into dynamic equivalent friction by adjusting the high-frequency micro-vibration parameters is intuitively demonstrated, proving the significant contribution of the friction reduction effect to reducing the initial pushing resistance.

[0076] For example, after receiving the tolerance alignment completion signal, the vibration excitation control unit starts the vibrator integrated in the tube feeder and sets the vibration frequency. 150 Hz, amplitude To generate high-frequency micro-vibrations of 0.3 mm; subsequently, the initial low-speed pushing unit drives the drug tube to... The system is set to advance at a low speed of millimeters per second; this speed is only the normal pushing speed. 25% of millimeters per second; the friction-reducing and clearance-finding guiding unit utilizes the friction-reducing effect generated by this coupling to change the contact force between the front end of the drug tube and the orifice wall from static friction to intermittent dynamic friction, according to the equivalent friction coefficient formula. ,in The initial coefficient of friction is 0.5. As an environmental correction factor, if calculated as follows If the value is reduced to 0.15, the equivalent resistance required for the drug delivery tube to advance will be... Significantly reduced; under this state, when the propellant tip enters the initial section 300 mm before the borehole, it can bypass the rock debris barrier with a diameter of about 2 mm by a small lateral displacement, thus smoothly completing the initial insertion and transitioning to the subsequent pushing mode along the path of least resistance.

[0077] In one possible implementation, combining Figure 2 The safety interlock and installation module includes:

[0078] The precondition verification unit is used to verify whether, before generating the control command to grab the detonating charge, it simultaneously receives a tube in-place confirmation signal reflecting that the tube in the actuator is in place, and a detonating charge in-place confirmation signal reflecting that the material at the target station of the detonating charge supply device is sufficient.

[0079] The instruction authorization control unit is used to authorize the execution of the control instruction to grab the detonating charge only when the verification results of the precondition verification unit are all valid;

[0080] The grabbing status confirmation unit is used to verify the grabbing success feedback signal from the actuator gripper after executing the control command for grabbing the detonating charge, and to determine whether to continue the subsequent detonating charge installation operation.

[0081] In some implementations, the precondition verification unit performs rigorous dual verification of materials and status before the system generates any intention to grab the detonating charge. After the ordinary propellant tube in a borehole is pushed in, two independent binary status signals are monitored in parallel. The first is a propellant tube presence confirmation signal, emitted by a sensor installed at the end of the actuator or tube feeder, such as a microswitch or photoelectric sensor. This signal becomes valid after the propellant tube has been pushed to a predetermined depth, confirming that the main charge is in place. The second is a detonating charge presence confirmation signal, originating from a sensor on the detonating charge supply device corresponding to the grabbing station, such as a vision sensor or weight sensor, used to confirm that there is a grabbable detonating charge at that station and that its status is normal. A strict logical AND operation is performed; only when both confirmation signals are simultaneously valid is the precondition considered satisfied. The command authorization control unit makes a decision based on the precondition verification results. This unit acts as an inviolable "command gate," releasing the execution authority for high-risk actions only under absolutely safe conditions. The control command to grasp the detonating charge is authorized and issued to the robotic arm controller only upon receiving a "condition satisfied" flag from the precondition verification unit. This command contains the complete motion sequence of the robotic arm moving to the target position of the detonating charge supply device and performing the gripper closing action. If the precondition verification fails, such as when the charge tube is not in place or the detonating charge is missing, the unit will lock the grasping command and report a waiting or abnormal status to the main control system, thereby preventing the robotic arm from performing an invalid or dangerous grasping action at the source. After the grasping command is executed, the grasping status confirmation unit confirms whether the physical action of grasping has indeed been successfully completed as expected, preventing the continuation of subsequent dangerous steps when it is "believed" that a grasp has been made but has not actually been made. Feedback signals from the actuator grippers are monitored. This successful grasping feedback signal is usually a composite signal, which may include a force sensor reading at the gripper fingertip reaching a preset gripping force range, such as 2 to 5 Newtons, and an encoder reading from the gripper drive motor indicating that the gripper has closed to the correct opening to hold the detonating charge. Only when a clear success signal is received within a preset time window, such as 1.5 seconds, will the unit determine that the grab was successful and unlock the subsequent control of the robotic arm to move the actuator carrying the detonating charge to the borehole and perform the installation. If the feedback fails or the timeout occurs, the current task will be immediately terminated, and an alarm or retry procedure will be triggered.

[0082] For example, when the precondition verification unit receives a logic level 1 confirmation signal from the end sensor of the tube feeder indicating that the explosive tube is in place, and a confirmation signal from the weight sensor of the detonating charge supply unit indicating that the detonating charge is in place and there is sufficient material, the safety interlock logic determination result is a logical AND operation. The triggering condition is met; subsequently, the instruction authorization control unit releases the permission, driving the robotic arm to perform the grasping action. The grasping status confirmation unit performs a physical closed-loop verification of the actuator gripper. Assuming the standard diameter of the detonating charge is 40 mm, the preset gripper closure success determination opening range is... Millimeters, and the fingertip force sensor has a preset clamping force threshold. for If the encoder then feeds back the gripper opening... millimeter and force sensor readings Newton determines that the grab is successful and authorizes the robotic arm to perform the subsequent detonation charge installation operation because all parameters fall within the preset range. Otherwise, if the above feedback is not received within 1.5 seconds, the task will be immediately terminated and an alarm will be triggered.

[0083] In one possible implementation, combining Figure 2 The intelligent push control module further includes:

[0084] The jamming state determination unit is used to determine that jamming has occurred and trigger the jamming response strategy when the thrust and velocity feedback data meet the preset jamming conditions of abnormal thrust increase and abnormal velocity decrease.

[0085] An adaptive instruction generation unit is used to generate an adaptive push control instruction that includes a pullback instruction, a vibration mode adjustment instruction, and a secondary propulsion instruction in response to the obstruction response strategy.

[0086] The strategy iteration execution unit is used to control the pipe feeder to execute the pullback command to relieve stress, and after adjusting the vibration frequency or amplitude, execute a secondary propulsion command to attempt to overcome the obstacle.

[0087] In some implementations, the jamming status determination unit performs real-time health monitoring of the tube feeder's operating status and accurately identifies jamming events. It continuously receives and analyzes thrust and velocity feedback data from internal sensors of the tube feeder. A preset jamming condition is defined as follows: when the thrust feedback data rapidly increases to more than 150% of the normal stable thrust reference or exceeds an absolute threshold, such as 400 Newtons, within a short period (e.g., 200 milliseconds), and simultaneously the velocity feedback data significantly decreases to less than 20% of the commanded velocity, jamming is determined to have occurred. This composite condition determination logic effectively eliminates instantaneous data jumps caused by normal friction fluctuations. Once the condition is met, the unit immediately triggers the obstruction response strategy. Once the obstruction response strategy is triggered, the adaptive command generation unit quickly generates a set of structured adaptive push control commands based on preset escape logic. This set of commands is not a single command but a sequence of actions containing multiple steps. First, a clear retraction command is generated, instructing the delivery tube to retract a small distance, such as 50 to 100 millimeters. Next, a vibration mode adjustment command is generated to change vibration parameters, such as reducing the vibration frequency from 150 Hz to a more impactful 80 Hz, or increasing the vibration amplitude by 30%. Finally, a secondary propulsion command is generated, typically accompanied by a lower velocity setting than the initial push. This complete command sequence is sent as a whole to the strategy iteration execution unit. The strategy iteration execution unit is responsible for translating macroscopic strategy commands into precise physical actions. It strictly controls the delivery tube to perform the obstacle avoidance operation according to the command sequence and manages the retry logic. First, the delivery tube is controlled to execute the retraction command, the physical intention of which is to immediately relieve stress concentration between the propellant tube and the borehole wall, creating space for subsequent adjustments. After retraction, the vibration mode adjustment command is immediately executed, changing the vibrator's operating state. After a brief adjustment delay, such as 0.5 seconds, the secondary propulsion command is executed, controlling the delivery tube to attempt to push forward again in the new vibration mode, utilizing the altered dynamic characteristics to overcome or bypass the obstacle. If this attempt triggers jamming again, the "pullback-parameter adjustment-pullback" cycle can be repeated three times, based on a preset number of iterations, with different combinations of vibration parameters applied each time. Only if all attempts fail will it be considered a hard jam, at which point operation will be stopped, an anomaly reported, and manual intervention requested.

[0088] For example, during the constant-speed pushing phase, the jamming state determination unit monitors the thrust reference of the tube feeder in real time. The instantaneous thrust is 100N and is fed back within 200 milliseconds. Soaring to 260N, which meets the requirements. The proportional threshold, and instantaneous speed From command speed Descending to When the threshold of less than 20% is met, a jamming condition is detected, and an obstruction response strategy is triggered. The adaptive command generation unit then generates an action sequence containing an 80mm retraction command, a vibration mode adjustment command to change the vibration frequency from 150Hz to 80Hz, and a secondary propulsion command. The strategy iteration execution unit first drives the tube feeder to retract 80mm to relieve stress concentration. After switching to the 80Hz impact vibration mode, it attempts to propel the tube again at a low speed of 0.2m / s. If the equivalent thrust required to overcome the jamming point is calculated... If the force is less than the system output limit, the medicine tube will successfully break through the obstacle and return to the constant speed pushing mode.

[0089] In one possible implementation, the system further includes a visual self-maintenance module:

[0090] A real-time confidence calculation unit is used to calculate in real time the continuous recognition confidence of the target borehole output by the end vision device;

[0091] The contamination event determination unit is used to determine that a lens contamination trigger event has occurred when the confidence level of continuous identification is lower than a set threshold and the preset degradation characteristics are met, provided that the ambient lighting conditions have been optimized.

[0092] The self-cleaning actuator is used to automatically control the cleaning mechanism installed on the end vision device protective cover to perform cleaning actions in response to the lens contamination triggering event.

[0093] In some implementations, continuous monitoring by a real-time confidence calculation unit is relied upon to quantitatively evaluate the reliability of the current vision system in identifying target boreholes in real time. In each iteration of the master-slave vision collaborative guidance process, the end-user vision device processes image or point cloud data and attempts to locate the target borehole. Its built-in recognition algorithm, such as a deep learning model or point cloud registration algorithm, synchronously outputs a confidence score reflecting the success rate of this recognition. A continuous confidence score sequence is captured and recorded; this score is typically a floating-point number between 0 and 1, providing basic data for subsequent contamination determination. The contamination event determination unit analyzes this continuous confidence data stream. It accurately distinguishes between systemic performance degradation caused by lens contamination and normal recognition fluctuations caused by other factors, such as sudden changes in illumination or temporary target occlusion. The judgment logic of this unit is based on composite conditions. It confirms that the system has prioritized illumination optimization, ruling out the possibility of recognition difficulties due to insufficient illumination. Under this premise, if the continuous recognition confidence score output by the real-time confidence score calculation unit remains below the set warning threshold (e.g., 0.7) for a period of time, such as more than 2 consecutive seconds, and the confidence score sequence exhibits a monotonically decreasing or low-level oscillation characteristic, a lens contamination trigger event is determined to have occurred. This trend- and duration-based judgment avoids misjudgment based on a single recognition failure, enhancing diagnostic accuracy. Once the lens contamination trigger event is confirmed, the self-cleaning execution unit performs specific physical cleaning actions to quickly restore the lens's optical performance. An execution command is immediately sent to the cleaning mechanism mounted on the end-vision device's protective cover. This cleaning mechanism typically integrates a micro-nozzle and a flexible scraper. The execution process involves the nozzle spraying a small amount of high-pressure clean gas or cleaning fluid to disperse or dissolve the attached contaminants; after a delay of approximately 200 milliseconds, the flexible scraper, driven by a micro-motor, scrapes across the lens surface once or twice, removing residual dirt and liquid. The entire cleaning action typically takes less than 2 seconds. After the cleaning action is completed, the robotic arm is re-aligned with the target borehole, and the confidence level is re-evaluated by the confidence level calculation unit in real time to confirm whether the cleaning effect meets the standard, thus completing a complete self-maintenance closed loop.

[0094] For example, during the master-slave vision collaborative guidance process, the confidence calculation unit identifies the blast hole through the end vision device and captures a set of 10 consecutive frames of confidence score sequence. After obtaining the sequence, the pollution event determination unit confirms that the current supplementary lighting is already in a high-brightness operating state, i.e., the lighting has been optimized, and then calculates the average confidence level of the sequence. ;because If the image quality falls below the warning threshold of 0.7 and exhibits a monotonically decreasing degradation characteristic, a lens contamination trigger event is determined to have occurred. The self-cleaning execution unit then responds and activates the cleaning mechanism on the protective cover. After controlling the nozzle to spray high-pressure gas, it drives the flexible scraper to scrape away dust contaminants from the lens surface in one cycle. The entire action takes about 1.5 seconds. After cleaning is completed, the confidence level is reassessed and rises to 0.92.

[0095] In one possible implementation, the visual self-maintenance module further includes:

[0096] The grayscale feature analysis unit is used to obtain the grayscale histogram distribution features of the current image acquired by the end vision device;

[0097] The supplementary light signal generation unit is used to generate a dynamic supplementary light control signal based on the difference between the grayscale histogram distribution characteristics and the preset ideal contrast range;

[0098] The brightness closed-loop adjustment unit is used to adjust the brightness of the fill light according to the dynamic fill light control signal so that the contrast of the current image approaches the target range, so as to improve the confidence of continuous recognition by prioritizing the improvement of illumination.

[0099] In some implementations, this step aims to proactively improve imaging conditions to fundamentally enhance the stability and accuracy of visual recognition. Before attributing a decrease in recognition confidence to lens contamination, intelligent adjustment of auxiliary illumination is prioritized to eliminate interference from insufficient or overexposed lighting, thereby reducing unnecessary cleaning actions and ensuring the vision system always operates under optimal optical conditions. The grayscale feature analysis subunit performs a rapid, quantitative evaluation of the quality of the currently captured image. Real-time image frames are acquired from the end-view vision device and converted into 8-bit grayscale images. A grayscale histogram of this image is calculated; this is a statistical chart showing the distribution of the number of pixels at each grayscale level from darkest to brightest, i.e., 0 to 255. By analyzing the mean, standard deviation, and cumulative percentage of pixels in the extremely dark or bright regions at both ends of the histogram, it is possible to accurately determine whether the image is generally too dark, too bright, or lacks contrast. Based on the above analysis results, the supplementary lighting signal generation subunit immediately intervenes. Precise adjustment instructions are generated based on the difference between the current image and the ideal image. This subunit internally stores a preset ideal contrast range. This range is not defined by a fixed value, but rather by a set of statistical characteristics describing a well-developed image. For example, the histogram distribution should be roughly uniform, and the sum of pixels clustered near 0 and 255 should be less than 5% of the total pixels. The grayscale histogram distribution characteristics of the current image are compared to this ideal range. If the image is too dark, meaning a large number of pixels are concentrated in low grayscale areas, a positive dynamic fill light control signal is generated; if the image is overexposed, a negative signal is generated. The magnitude of the signal can be proportional to the degree of deviation from the ideal range, thus achieving proportional control. The brightness closed-loop adjustment subunit is responsible for performing physical adjustments and completing the entire optimization loop. It forms a feedback loop, converting the control signal into actual changes in illumination intensity and iterating continuously until the target is reached. Based on the received dynamic fill light control signal, the output brightness of the fill light, typically a pulse-width modulation (PWM) signal supplied to an LED ring light array deployed around the lens, is changed by adjusting the duty cycle of the PWM signal. After brightness adjustment, a new frame image is immediately acquired, and the grayscale feature analysis subunit is restarted for evaluation. This closed-loop process of "analysis-decision-adjustment" iterates at an extremely high frequency, such as 15 to 30 times per second, until the contrast features of the current image enter the preset target range, or the resulting continuous improvement in recognition confidence and its stabilization at a normal level. Only when the brightness has been adjusted to the optimal level, but the recognition confidence remains low, will the problem be attributed to lens contamination, and the self-cleaning execution unit will be authorized to intervene.

[0100] For example, in the visual self-maintenance process, the grayscale feature analysis subunit acquires a frame of image from the end vision device and calculates its grayscale histogram mean. With a measurement range of only 65 (0-255), the displayed image is generally dark; the supplementary light signal generation subunit then compares this feature with the mean target of a preset ideal contrast range. Compare and calculate the mean deviation. Based on this deviation, the brightness closed-loop adjustment subunit generates a dynamic supplementary lighting control signal through a PID control algorithm, assuming the current PWM duty cycle of the supplementary light... It is 40%, according to the adjustment formula. Set the proportional gain Calculate the new duty cycle Immediately increase the brightness of the LED lights and acquire new images to quickly bring the image contrast close to the target range. If the confidence level of continuous recognition rises from 0.65 and stabilizes at 0.95, it is determined that the lighting conditions have been optimized and the false start of the self-cleaning actuator has been successfully suppressed.

[0101] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0102] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. An intelligent charging system for tunnel excavation, characterized in that, The system includes: The environmental perception and data acquisition module is used to acquire global depth images of the tunnel blasting working face collected by the global vision device, and to acquire the blast hole design data of the working face. The spatial topology planning module is used to generate a spatial topology network of boreholes that includes all borehole nodes and the spatial relationships between nodes based on the global depth image and borehole design data, and to plan and generate sequential operation instructions to guide the robotic arm's movements based on the borehole spatial topology network. The master-slave vision collaborative guidance module is used to drive the robotic arm to move toward the target gun hole according to the serialized operation instructions, and simultaneously start the master-slave vision collaborative guidance process. The master-slave vision collaborative guidance process is used to fuse the first positioning data obtained by the global vision device and the second positioning data obtained by the end vision device mounted on the end of the robotic arm to generate the end pose dynamic adjustment instruction. The pose compensation and alignment execution module is used to respond to the end pose dynamic adjustment command, control the actuator at the end of the robotic arm to move toward the borehole, and complete the tolerance alignment of the propellant tube by contacting the borehole with the flexible guide sleeve at the front end of the actuator. The intelligent push control module is used to send a vibration gap-finding push command after the tolerance alignment is completed, receive the thrust and speed feedback data monitored and fed back in real time by the internal sensors of the tube feeder, and generate an adaptive push control command or a constant speed push control command to control the tube feeder to complete the push operation. The safety interlock and installation module is used to acquire the in-situ confirmation signal of the explosive tube reflecting the in-situ status and the in-situ confirmation signal of the detonating explosive charge reflecting the supply status of the detonating explosive charge, and to control the robotic arm to perform the installation operation of the detonating explosive charge based on the safety interlock logic. The file recording module is used to record and upload the loading process file, which includes the thrust and velocity feedback data.

2. The intelligent charging system for tunnel excavation according to claim 1, characterized in that, The spatial topology planning module includes: The initial pose extraction unit is used to process the global depth image using a target detection algorithm to identify the initial pose set of all visible boreholes within the field of view. The spatial matching and completion unit is used to spatially register the initial pose set with the borehole design data through a registration algorithm, and to apply a transformation matrix to correct the borehole design data, thereby constructing a borehole spatial topology network containing the true poses of all target boreholes. The movement cost calculation unit is used to calculate the movement cost between any two borehole nodes in the borehole spatial topology network. The movement cost is composed of the spatial distance between the two nodes and the expected recognition confidence of the target borehole node by the end vision device. The operation instruction planning unit is used to apply a path search algorithm to find the path that traverses all borehole nodes and minimizes the total movement cost based on the movement cost, and generate serialized operation instructions.

3. The intelligent charging system for tunnel excavation according to claim 1, characterized in that, The master-slave visual collaborative guidance module includes: The macroscopic offset tracking unit is used to continuously track the spatial position of the robotic arm end relative to the current target borehole group area through the global vision device, so as to obtain a rough spatial offset as the first positioning data. A local fine scanning unit is used to perform a three-dimensional scan of the current target borehole through the end vision device during the movement of the robotic arm, so as to obtain fine point cloud data of the borehole as second positioning data. The data fusion estimation unit is used to apply the Kalman filter algorithm to fuse the coarse spatial offset with the fine point cloud data in order to estimate the real-time relative pose of the actuator with respect to the target borehole. The trajectory dynamic correction unit is used to compare the real-time relative pose with the preset optimal observation pose to calculate the pose error, and generate speed commands to drive the movement of each joint of the robotic arm as end-effector pose dynamic adjustment commands.

4. The intelligent charging system for tunnel excavation according to claim 1, characterized in that, The pose compensation and alignment execution module includes: The pose error compensation unit is used to drive the end of the robotic arm to perform translation and rotation compensation according to the end pose dynamic adjustment command, so that the central axis of the actuator is aligned with the orifice normal vector of the target gun hole; The flexible contact feedback unit is used to determine whether a reliable flexible sealing connection has been established between the actuator and the borehole by monitoring the compression deformation state of the flexible guide sleeve during the process of the actuator approaching the borehole. The tolerance alignment verification unit is used to confirm that the passage for the propellant tube to enter the borehole is connected within a preset allowable radial deviation range, based on the physical limiting space of the flexible guide sleeve.

5. The intelligent charging system for tunnel excavation according to claim 1, characterized in that, The intelligent push control module includes: The vibration excitation control unit is used to control the tube feeder to perform high-frequency micro-vibration at a set vibration frequency and amplitude; An initial low-speed pushing unit is used to drive the tube feeder to advance the drug tube forward at a set initial low speed under the high-frequency micro-vibration state. The friction-reducing and gap-finding guiding unit is used to guide the propellant tube along the path of least resistance in the initial section of the borehole through the friction-reducing effect generated by the coupling effect of the high-frequency micro-vibration and the initial low speed.

6. The intelligent charging system for tunnel excavation according to claim 1, characterized in that, The safety interlock and installation module includes: The precondition verification unit is used to verify whether, before generating the control command to grab the detonating charge, it simultaneously receives a tube in-place confirmation signal reflecting that the tube in the actuator is in place, and a detonating charge in-place confirmation signal reflecting that the material at the target station of the detonating charge supply device is sufficient. The instruction authorization control unit is used to authorize the execution of the control instruction to grab the detonating charge only when the verification results of the precondition verification unit are all valid; The grabbing status confirmation unit is used to verify the grabbing success feedback signal from the actuator gripper after executing the control command for grabbing the detonating charge, and to determine whether to continue the subsequent detonating charge installation operation.

7. The intelligent charging system for tunnel excavation according to claim 5, characterized in that, The intelligent push control module also includes: The jamming state determination unit is used to determine that jamming has occurred and trigger the jamming response strategy when the thrust and velocity feedback data meet the preset jamming conditions of abnormal thrust increase and abnormal velocity decrease. An adaptive instruction generation unit is used to generate an adaptive push control instruction that includes a pullback instruction, a vibration mode adjustment instruction, and a secondary propulsion instruction in response to the obstruction response strategy. The strategy iteration execution unit is used to control the pipe feeder to execute the pullback command to relieve stress, and after adjusting the vibration frequency or amplitude, execute a secondary propulsion command to attempt to overcome the obstacle.

8. The intelligent charging system for tunnel excavation according to claim 1, characterized in that, The system also includes a visual self-maintenance module: A real-time confidence calculation unit is used to calculate in real time the continuous recognition confidence of the target borehole output by the end vision device; The contamination event determination unit is used to determine that a lens contamination trigger event has occurred when the confidence level of continuous identification is lower than a set threshold and the preset degradation characteristics are met, provided that the ambient lighting conditions have been optimized. The self-cleaning actuator is used to automatically control the cleaning mechanism installed on the end vision device protective cover to perform cleaning actions in response to the lens contamination triggering event.

9. The intelligent charging system for tunnel excavation according to claim 8, characterized in that, The visual self-maintenance module also includes: The grayscale feature analysis unit is used to obtain the grayscale histogram distribution features of the current image acquired by the end vision device; The supplementary light signal generation unit is used to generate a dynamic supplementary light control signal based on the difference between the grayscale histogram distribution characteristics and the preset ideal contrast range; The brightness closed-loop adjustment unit is used to adjust the brightness of the fill light according to the dynamic fill light control signal so that the contrast of the current image approaches the target range, so as to improve the confidence of continuous recognition by prioritizing the improvement of illumination.

10. An intelligent charging method for tunnel excavation, characterized in that, The method includes: Acquire global depth images of the tunnel blasting face from a global vision device, and obtain the borehole design data for the face. Based on the global depth image and borehole design data, a borehole spatial topology network containing all borehole nodes and the spatial relationships between nodes is generated, and a sequenced operation instruction to guide the robotic arm's movements is planned and generated according to the borehole spatial topology network. The robotic arm is driven to move toward the target bore according to the serialized operation instructions, and the master-slave vision collaborative guidance process is started simultaneously. The master-slave vision collaborative guidance process is used to fuse the first positioning data obtained by the global vision device and the second positioning data obtained by the end vision device mounted on the end of the robotic arm to generate the end pose dynamic adjustment instruction. In response to the end-position dynamic adjustment command, the actuator at the end of the robotic arm is controlled to move toward the borehole, and the tolerance alignment of the propellant tube is completed by the flexible guide sleeve at the front end of the actuator contacting the borehole. After the tolerance alignment is completed, a vibration gap-finding push command is sent, and the thrust and speed feedback data monitored and fed back in real time by the internal sensors of the tube feeder are received. An adaptive push control command or a constant speed push control command is generated to control the tube feeder to complete the push operation. The system acquires a tube-in-place confirmation signal reflecting the tube's position and an initiation charge-in-place confirmation signal reflecting the supply status of the initiation charge, and controls the robotic arm to perform the initiation charge installation operation based on safety interlock logic. Record and upload the loading process archive, which includes the thrust and velocity feedback data.

Images