A mine underground unmanned equipment intelligent control method and system

Abstract

The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine underground unmanned equipment intelligent control method and system, it is related to intelligent mine equipment automatic control technical field.The application discloses a kind of mine

Description

Technical Field

[0001] This invention relates to the field of automatic control technology for intelligent mining equipment, and in particular to an intelligent control method and system for unmanned underground mining equipment. Background Technology

[0002] Currently, unmanned equipment in underground mines (such as unmanned loaders, unmanned transport trucks, and inspection robots) mainly relies on a "high-precision mapping—high-precision positioning—path planning" technical framework for automated control. Typically, the mine production control center formulates task flows based on operational requirements and constructs a static 3D environmental map using high-precision sensors (such as LiDAR, visual SLAM, and inertial measurement units). This map is then combined with path planning algorithms (such as hybrid A*, Dijkstra's algorithm, and RRT) to generate the equipment's travel paths and operational plans, which are ultimately executed by a rule-based controller. While this technical solution is adaptable to structured, geometrically stable surface environments, it faces significant limitations in the underground mining environment.

[0003] Specifically, the underground working environment is characterized by its enclosed and narrow shape, high dust and humidity, and drastic changes in lighting. Furthermore, there is no GNSS signal coverage. Even with the integration of multi-source sensing methods such as UWB and visual SLAM, it remains difficult to achieve continuous and stable high-precision positioning under long-distance, multi-area operating conditions. Especially during operation, the tunnel structure frequently expands, deforms, or is disturbed, causing a mismatch between the existing map and the real environment, further exacerbating positioning drift and map failure. Therefore, existing control schemes relying on high-precision maps and positioning face challenges such as high-frequency updates, high maintenance costs, poor stability, and poor generalization, making it difficult to support long-term robust operation in underground mines. In addition, current systems mostly adopt a control method based on preset paths and rules, lacking sufficient perception-driven and intelligent decision-making capabilities, making it difficult to fully integrate with upper-level scheduling systems. At the operation execution level, critical tasks such as loading and unloading still require manual intervention or semi-automatic control, failing to achieve truly autonomous operation throughout the entire process.

[0004] In summary, existing unmanned mine control technologies still have significant shortcomings in the following aspects: First, they rely excessively on high-precision maps and positioning systems, resulting in low stability, high system deployment costs, and heavy maintenance burdens. Second, existing models designed for long tasks have complex structures, high coupling, large training overhead, and lack cross-scenario generalization capabilities, making it difficult to support efficient migration and deployment across multiple tasks and scenarios. Third, the methods based on preset paths and rules, along with the aforementioned problems, make it difficult to integrate with upper-level scheduling systems to achieve flexible scheduling.

[0005] Therefore, proposing an intelligent control method and system for unmanned equipment in underground mines to solve the difficulties existing in the prior art is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention provides an intelligent control method and system for unmanned equipment in underground mines. Through the division of work areas, decomposition of work tasks, decoupling of action models, reinforcement learning training and control logic map and event-driven control mechanism, it realizes stable, reliable and efficient control and scheduling of unmanned equipment operations under coarse positioning conditions, and has good robustness, scalability and generalization.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A method for intelligent control of unmanned equipment in underground mines includes the following steps:

[0009] S1. Construction of Control Logic Map: Construct a control logic map based on the centerline of the mine roadway;

[0010] S2. Subtask partitioning and model decoupling: Long-process job tasks are decomposed into multiple independent subtask units, and an independent end-to-end subtask control model is trained for each subtask and subtask running scenario.

[0011] S3. Central Task Control: When each task starts or an event is triggered, a task navigation map is generated based on the control logic map, the task being executed, and the current location;

[0012] S4. Equipment Perception and Execution: The front-end equipment triggers the corresponding sub-task control model based on the task navigation map, perception information and location information, and completes the autonomous control of the current stage of the operation task through real-time perception information.

[0013] Optionally, in the above method, the control logic map constructed in S1 is a topology map based on the centerline of the lane. The topology map contains several nodes, edges, and lengths. Among them, nodes represent turnouts or other key scene control points, edges represent lanes, and lengths are the measured length or design length of the lane.

[0014] Optionally, the task navigation map generated in S3 includes the running path and subtask boundaries, which are used for device operation navigation and subtask switching trigger logic.

[0015] Optionally, in the above method, a training scenario is constructed in the simulation environment in S4, and a reinforcement learning method is used to train the end-to-end sub-task control model. The input of the sub-task control model is the device's own perception information, and the output is continuous or discrete control commands. After training, it is deployed on the front-end device and fine-tuned according to the real scenario.

[0016] Optionally, the device-specific sensing information in the above method may include images, point clouds, localization, and device operating conditions.

[0017] An intelligent control system for unmanned equipment in underground mines, and an intelligent control method for unmanned equipment in underground mines applying any of the above, comprising a central control module, a sensing module, a control execution module, and a communication module connected in sequence;

[0018] The central control module is used to generate a task navigation map based on the control logic map and the tasks being executed, and then send the task navigation map to the front-end device.

[0019] The perception module, including the positioning submodule and the scene recognition submodule, is used to correct positioning errors or provide position redundancy by identifying key scene locations.

[0020] The control execution module is used to deploy multiple sub-task control models and call and execute corresponding actions based on perception information and task navigation map;

[0021] The communication module is used for status synchronization and information exchange between various modules and devices.

[0022] Optionally, the perception module of the above system can achieve positioning based on SLAM, UWB or wheel odometry, identify key scenes, correct the errors in positioning based on SLAM, UWB or wheel odometry by using control points on the task navigation map, or use the identified location as the location information when positioning fails completely.

[0023] The aforementioned system may include, optionally, key control points such as ore outlets, turnouts, and ore unloading points.

[0024] Optionally, the communication module in the above system is used to transmit location information, status feedback, and task navigation map.

[0025] As can be seen from the above technical solution, compared with the prior art, the present invention provides an intelligent control method and system for unmanned equipment in underground mines, which has the following beneficial effects:

[0026] (1) This invention constructs a control logic map and task navigation map on the central side, and implements a sub-task control mechanism based on this, thus building a closed-loop scheduling control framework of "central side task planning - device side behavior switching - device side model execution - device side information feedback", enabling the system to respond to the scheduling control of the upper-level scheduling system in a timely manner.

[0027] (2) Without relying on high-precision maps and high-precision positioning, this invention achieves stable, reliable and efficient control of unmanned equipment operations under coarse positioning conditions through operation area division, operation task decomposition, action model decoupling, reinforcement learning training and control logic map and event-driven control mechanism. It has good robustness, scalability and generalization.

[0028] (3) This invention effectively decouples the action model by dividing the complex downhole task scenario into multiple simple scenarios with a single structure, which is conducive to reinforcement learning training. At the same time, it supports rapid adaptation to changing scenarios and flexible configuration of work processes. It has significant engineering deployability and system expansion potential, and can be widely applied to unmanned operation tasks in various downhole scenarios. It has good robustness, adaptability and generalization ability. Attached Figure Description

[0029] 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 only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0030] Figure 1 A flowchart of an intelligent control method for unmanned equipment in underground mines provided by the present invention;

[0031] Figure 2 This invention provides a structural block diagram of an intelligent control system for unmanned equipment in underground mines.

[0032] Figure 3 A schematic diagram of the control logic map for downhole unmanned equipment provided by the present invention;

[0033] Figure 4 A schematic diagram of the mission navigation map for unmanned equipment in wells provided by the present invention;

[0034] Figure 5 This invention provides a schematic diagram of intelligent operation of unmanned equipment in underground mines.

[0035] Figure 6 A schematic diagram of the composition of the unmanned underground equipment provided by the present invention;

[0036] Figure 7 This is a schematic diagram of the intelligent control process for unmanned downhole equipment provided in an embodiment of the present invention;

[0037] Among them, 301-Control Logic Map, 302-Control Point, 401-Sub-map Path, 402-Travel Task Area, 403-Branch Decision Area, 404-Loading Operation Area, 501-Downhole Intelligent Equipment, 502-Sensing Information, and 503-Task Navigation Map. Detailed Implementation

[0038] 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, and 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.

[0039] Reference Figure 1 As shown, this invention discloses an intelligent control method for unmanned equipment in underground mines, comprising the following steps:

[0040] S1. Construction of Control Logic Map: Construct a control logic map based on the centerline of the mine roadway;

[0041] S2. Subtask partitioning and model decoupling: Long-process job tasks are decomposed into multiple independent subtask units, and an independent end-to-end subtask control model is trained for each subtask and subtask running scenario.

[0042] S3. Central Task Control: When each task starts or an event is triggered, a task navigation map is generated based on the control logic map, the task being executed, and the current location;

[0043] S4. Equipment Perception and Execution: The front-end equipment triggers the corresponding sub-task control model based on the task navigation map, perception information and location information, and completes the autonomous control of the current stage of the operation task through real-time perception information.

[0044] Furthermore, the control logic map constructed in S1 is a topology map based on the center line of the lane. The topology map contains several nodes, edges, and lengths. Among them, nodes represent turnouts or other key scene control points, edges represent lanes, and lengths are the measured length or design length of the lane.

[0045] Furthermore, the task navigation map generated in S3 includes the running path and subtask boundaries, which are used for device operation navigation and subtask switching trigger logic.

[0046] Furthermore, in S4, training scenarios are constructed in a simulation environment, and reinforcement learning methods are used to train end-to-end sub-task control models. The input of the sub-task control model is the device's own perception information, and the output is continuous or discrete control commands. After training, the model is deployed to the front-end device and fine-tuned according to the real scenario.

[0047] Furthermore, the device's sensing information includes images, point clouds, positioning, and device operating conditions.

[0048] In a specific embodiment, such as Figure 7As shown, it includes upper-level central equipment and front-end equipment. The upper-level central equipment is responsible for the planning of production tasks and equipment scheduling management in the mining area. Internally, it includes production task generation, control logic map construction, real-time equipment information, and task navigation map. The front-end intelligent equipment is deployed in the underground environment and has a certain degree of perception and operation control capabilities.

[0049] (1) Production task generation: Based on the mining operation requirements, the ore extraction task process is set up, and the operation sequence and stage objectives of each piece of equipment are clarified. Real-time equipment information is used to continuously receive the positioning information, operation status and scene perception results uploaded by each front-end equipment, and combined with the control logic map, to determine the current path node and task stage of the equipment, forming a global perception of the spatial distribution of equipment and the progress of task execution on the central side.

[0050] A control logic map is constructed based on the centerline of the mine tunnels. The tunnel structure is abstractly modeled to generate a topology map (control logic map) based on the tunnel centerline. Unlike urban road navigation maps, underground tunnels are inherently confined spaces, with vehicle travel areas strictly defined by physical boundaries. The topology map, formed by the tunnel centerlines, is expressed using a graph structure, containing several nodes and edges. Nodes represent key locations (such as junctions, sub-task areas, etc.), while edges represent feasible pathways between nodes. Length, as an edge attribute, represents the tunnel length, supporting path planning and dynamic scheduling. This structure ensures the map remains lightweight while maintaining complete path connectivity and control capabilities. More importantly, by employing a topology-centric abstraction method, the scheduling logic map construction does not rely on high-precision geometric models, but only on the tunnel centerline and task semantic information. When the mining area structure expands (such as tunnel advancement) or undergoes local deformation, only the node and region definitions need to be updated to complete map reconstruction and scheduling adaptation. It has good reconfigurability, scalability and generalization ability, and is suitable for underground operation environments with frequent scene deformation.

[0051] During actual operation, the central control module plans the execution path and forms a topology subgraph in real time from the global control logic map based on the equipment type, task objectives, and current status. Based on this subgraph, and considering operational requirements, the module rapidly divides the path segments into regions according to the principles of "structural simplicity" and "semantic consistency." This division includes task regions such as "walking zone," "avoidance zone," "loading zone," and "unloading zone," each region being bound to a corresponding sub-task control model. Finally, the subgraph structure and region division together form a complete "task navigation map," guiding the current equipment's path execution and model scheduling within the current task cycle. This task subgraph only contains the region information required for the current equipment's operation, ensuring scheduling accuracy during system operation while significantly reducing the computational burden and data loading pressure of the online system.

[0052] (2) Subtask division and model decoupling: The long-process job task is decomposed into multiple independent subtask units, and an independent end-to-end subtask control model is trained for each subtask and subtask running scenario.

[0053] Specifically, as an example, for shoveling and transportation tasks, the task sequence of equipment is formulated according to the functional definition of each shaft and tunnel project in the mining area, and the target area and operation logic of each stage such as shoveling, transportation, and unloading are clearly defined.

[0054] Long-process operations are divided into several sub-task units with well-defined functions and limited state spaces, and each sub-task corresponds to an independent sub-task control model.

[0055] Each sub-task control model is trained end-to-end using reinforcement learning, enabling it to make autonomous decisions for similar equipment in specific operational scenarios. This avoids problems such as overfitting, policy failure, and insufficient generalization that easily occur when a unified large model handles cross-scenario and cross-state tasks, structurally improving the modularity and adaptability of the control system.

[0056] (3) Central task control: When each task is started or an event occurs, a task navigation map is generated based on the control logic map, the task being executed, and the current location;

[0057] Specifically, the central control module, based on the control logic map, device location information, and operational status feedback, determines the current task stage of the device in real time. Combining the task objective and device type, it extracts the corresponding topological sub-graph from the global map, divides the area, and generates a task navigation map containing path structure and task semantics. This navigation map contains the path information required for the device's current task, the functional division of the area, and the sub-model activation rules, and is sent to the device via the communication link. The device then autonomously parses the task stage and loads the corresponding sub-model to execute operations. This mechanism enables the central control module to distribute task control logic in a structured manner, reducing communication load and improving scheduling efficiency and system robustness.

[0058] (4) Equipment perception and execution: The front-end equipment triggers the corresponding sub-task control model based on the task navigation map, perception information and location information to complete the autonomous control of the current stage of the operation task.

[0059] Specifically, a sub-task control model trained by reinforcement learning is deployed. The corresponding sub-task control model is activated based on the current task navigation map and the location of the scene to perform action reasoning and control execution, thereby realizing specific operation tasks such as loading, walking, and unloading.

[0060] During the perception process, models based on YOLO V12 and GAnet can be integrated to detect key structures or objects in the work area (such as intersections, avoidance chambers, loading points, unloading platforms, beacon points, etc.), thereby achieving type judgment and coarse semantic region localization of the current work scene. This visual recognition process not only improves the sub-model's ability to understand environmental semantics, but can also be used to compensate for and correct errors caused by localization, achieving a fusion enhancement of visual perception and motion estimation.

[0061] The perception results can be represented using the Frenet coordinate system based on the centerline of the topological logic map. The current pose of the device is mapped to a certain arc segment through the projection of the nearest point. The longitudinal position is denoted as s, and the lateral offset d is ideally set as d=0, which means that the device is regarded as always moving along the centerline.

[0062] In terms of control execution, the equipment has a locally deployed sub-model library, which integrates multiple end-to-end sub-task control models for front-end devices and trained using reinforcement learning. These models correspond to typical operational tasks such as walking, loading, and unloading. After receiving the task navigation map from the central control unit, the front-end device switches to the corresponding sub-task control model based on its own task area and performs real-time action reasoning based on local perception information (such as environmental point clouds, images, and coarse localization results) to autonomously complete the current stage of operation control.

[0063] (5) Training of the sub-task control model, the training method mainly includes the following two stages:

[0064] 1. Simulation Training

[0065] In the initial training phase, high-fidelity simulation platforms (such as IsaacLab and Gazebo) are used to construct typical operating environments for pre-training control models of different sub-tasks. Based on the semantic features and operational requirements of various tasks, corresponding training scenarios (such as driving through narrow passages, turning, obstacle avoidance, loading / unloading operations, and ore unloading operations) are designed, and reasonable dynamic constraints and interaction rules are set in conjunction with equipment physical parameters and kinematic models. Model training employs reinforcement learning methods (such as PPO and DDPG), using reward functions to guide the strategy in learning objectives such as task completion efficiency, path stability, and safety.

[0066] To address the operational requirements under coarse positioning conditions in this invention, all training scenarios undergo appropriate boundary expansion to cover the positioning error range that may occur in actual deployment. This also simulates various postures and initial states of the device entering the area under different positioning deviations, enabling the training model to possess a certain degree of fault tolerance and adaptability. It should be noted that, to maintain the principle of scene structural uniformity during the training process of the sub-task control model, this expansion range is not arbitrarily increased, but rather controlled within an acceptable accuracy range based on the upper limit of the error of the adopted coarse positioning method. Specifically, the expansion boundary only covers the maximum error band that the positioning system may generate in actual operation, ensuring that the model possesses the necessary robustness and generalization ability without compromising the semantic and structural uniformity of the scene. This design effectively avoids problems such as unclear model learning objectives, enhanced scene interference, and decreased policy generalization ability caused by excessively blurred region boundaries, thereby ensuring the stability and transferability of training results.

[0067] Meanwhile, to enhance the generalization ability and cross-scenario adaptability of the sub-task control model in actual deployment, a unified training framework for diverse semantic environments was constructed during the training process, addressing the control requirements of similar devices in different task areas. This invention does not rely on unique modeling of a specific physical space, but rather, based on the consistency of task semantics, abstracts and extracts the functional structural features (such as path morphology, boundary constraints, etc.) of various work scenarios. In simulation, multiple scenario sample sets with similar structures but diverse environments are constructed for training the same sub-task control model.

[0068] This method introduces scene variability and layout perturbation during policy training, enabling the model to learn robust control policies. When facing the same type of task objectives, it can still achieve reliable behavioral outputs under different topological structures, diverse perceptual inputs, or the presence of positioning error perturbations, thereby significantly enhancing the model's generalization ability and robustness when deployed in mining environments.

[0069] 2. Model Deployment

[0070] After training, the model needs to be transferred to an actual mining area for testing and optimization to ensure that it adapts to the real environment and can perform tasks stably.

[0071] First, multiple rounds of testing were conducted in a simulation environment to verify the accuracy of subtask switching and to evaluate the stability and performance of the subtask control model under different operating conditions. By comparing the data differences between the simulation and the real environment, the normalization method of the model input was adjusted to reduce domain offset issues.

[0072] During the trial operation phase in the mining area, the trained sub-task control model was deployed to the underground shovel and unloading equipment. Actual operational data, including navigation accuracy and operational efficiency, were collected, and the model was evaluated. Since there may be mismatches between the real and simulation environments, offline reinforcement learning (Offline RL) was used to fine-tune the model to better suit the mining conditions and improve its robustness.

[0073] In another specific embodiment, the following provides an example of a scheduling logic map and a description of a typical job process:

[0074] like Figure 3 As shown, a typical control logic map includes several key structural elements, mainly composed of the control logic map 301 and control points 302. The tunnel center topology 301 describes the direction, length, and spatial connectivity between the main paths in the mining area, forming the global basis for task scheduling and path planning. Control points 302 mark key locations involved in task execution, such as junctions, loading areas, and unloading areas, serving as important references for generating task navigation maps and dividing work areas.

[0075] like Figure 4 As shown, a task navigation map is dynamically generated by the central control module based on the control logic map 301, combined with the current equipment type and the production tasks it needs to perform. The sub-map path 401 extracted from the control logic map is used to define the operating range within the current task cycle; 402 is the corresponding walking task area, 403 is the junction decision area, and 404 is the loading operation area. These task areas are semantically divided by the central control module based on equipment status and task flow, and bound to the corresponding sub-task control model to guide the equipment in executing appropriate control strategies at each stage.

[0076] Combination Figure 5 As shown below, an example of the application of a scheduling logic map in an actual work process is given.

[0077] At time T0, the downhole intelligent device 501 uploads its current location, sensing information 502, and equipment operating status to the central control module, and receives a task navigation map 503 generated by the central control module based on its equipment model and current production task. This navigation map clearly plans the path structure, sub-task area division, and corresponding control model of the equipment within the current task cycle. Guided by the task navigation map 503, the equipment starts executing the "walking model" from time T0, and approaches the fork decision area 403 at time T1. According to the model calling rules embedded in the task navigation map, it selects and switches to the "fork path decision model" from the local model library, completes the left turn operation, correctly enters the designated channel, and realizes path selection and action control.

[0078] like Figure 6 As shown, to implement the method of this embodiment of the invention, this embodiment also provides an intelligent control device for downhole unmanned equipment, comprising: an upper-level central device and a front-end device, wherein each module includes at least one device power supply, a computing unit, a storage unit, and a wireless network transceiver unit. The various components in the intelligent control device for downhole unmanned equipment are coupled together through a bus system. It can be understood that the bus system is used to realize the connection and communication between these components. In addition to a data bus, the bus system also includes a power bus, a control bus, and a status signal bus.

[0079] The upper-level central device-side control method disclosed in this invention is implemented by a computing unit, which includes multiple processing modules: an information transmission and reception module, a control map construction module, a task database generation module, and a task scheduling module. The software modules and the database are located in a storage unit. The computing unit reads information from the storage unit and, in conjunction with its hardware, completes the upper-level central device-side control method.

[0080] The front-end device-side control method disclosed in this invention is implemented by a computing unit, which includes multiple processing modules: an information transmission and reception module, a scene perception module, a control model selection module, a sub-task control model, and an action execution module. The software modules and the database are located in a storage unit. The computing unit reads information from the storage unit and, in conjunction with its hardware, completes the front-end device-side control method.

[0081] and Figure 1 Corresponding to the method described above, this embodiment of the invention also provides an intelligent control system for unmanned equipment in underground mines, used for... Figure 1 The specific implementation of the method is shown in the following structural diagram. Figure 2 As shown, it includes a central control module, a sensing module, a control execution module, and a communication module connected in sequence;

[0082] The central control module is used to generate a task navigation map based on the control logic map and the tasks being executed, and then send the task navigation map to the front-end device.

[0083] The perception module, including the positioning submodule and the scene recognition submodule, is used to correct positioning errors or provide position redundancy by identifying key scene locations.

[0084] The control execution module is used to deploy multiple sub-task control models and call and execute corresponding actions based on perception information and task navigation map;

[0085] The communication module is used for status synchronization and information exchange between various modules and devices.

[0086] Specifically, the front-end intelligent devices are deployed in the underground environment and have a certain degree of perception and operation control capabilities.

[0087] Furthermore, the perception module achieves positioning based on SLAM, UWB, or wheeled odometers, and identifies key scenes. It corrects the errors in positioning based on SLAM, UWB, or wheeled odometers by using control points on the task navigation map, or uses the identified location as the location information when positioning completely fails.

[0088] Specifically, the perception module is mainly used to identify the current operating environment and task area of ​​the equipment, providing necessary scene semantic information for the action decisions of the equipment's terminal task control model. As an independent perception system, it relies on multi-source sensors to perform real-time perception and processing of the environmental state. The perception module consists of three parts: a positioning submodule, a scene recognition submodule, and a sub-task execution perception module.

[0089] The positioning submodule, for example, employs a lightweight, low-cost positioning solution based on SLAM or a fusion of UWB and wheeled odometers. This solution does not rely on a high-precision map positioning system; its positioning error, as long as it is within the maximum error band mentioned above, is sufficient to meet the area judgment and task switching requirements of this invention. Coarse positioning information is used to assist in determining the device's location node and its corresponding task area, serving as an important reference for subsequent action model invocation.

[0090] Scene Recognition Submodule: Based on visual perception models (such as YOLO V12, GAnet) or other lightweight visual recognition networks, the scene recognition submodule identifies the scene type and key areas of the current environment of the device. Its main functions include: first, extracting semantic features of the current work scene to support the identification and determination of the task area; second, as a redundancy mechanism, periodically correcting the location information obtained based on positioning.

[0091] Subtask execution perception: as input to the subtask control model, it includes images, point clouds, localization, and equipment operating conditions.

[0092] Furthermore, key control points in key scenarios include ore outlets, turnouts, and ore unloading points.

[0093] Furthermore, the communication module is used to transmit location information, status feedback, and task navigation maps.

[0094] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0095] The above description of the disclosed embodiments enables those skilled in the art to make or use 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 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 disclosed herein.

Claims

1. An intelligent control method for unmanned equipment in underground mines, characterized in that, Includes the following steps: S1. Construction of Control Logic Map: Construct a control logic map based on the centerline of the mine roadway; S2. Subtask partitioning and model decoupling: Long-process job tasks are decomposed into multiple independent subtask units, and an independent end-to-end subtask control model is trained for each subtask and subtask running scenario. S3. Central Task Control: When each task starts or an event is triggered, a task navigation map is generated based on the control logic map, the task being executed, and the current location; S4. Equipment Perception and Execution: The front-end equipment triggers the corresponding sub-task control model based on the task navigation map, perception information and location information, and completes the autonomous control of the current stage of the operation task through real-time perception information; The control logic map constructed in S1 is a topology map based on the center line of the lane. The topology map contains several nodes, edges and lengths. Among them, nodes represent turnouts or other key scene control points, edges represent lanes, and lengths are the measured length or design length of the lane. The task navigation map generated in S3 includes the running path and subtask boundaries, which are used for device operation navigation and subtask switching trigger logic.

2. The intelligent control method for unmanned underground equipment in a mine according to claim 1, characterized in that, In S4, training scenarios are built in a simulation environment, and reinforcement learning is used to train end-to-end sub-task control models. The input of the sub-task control model is the device's own perception information, and the output is continuous or discrete control commands. After training, the model is deployed to the front-end device and fine-tuned according to the real scenario.

3. The intelligent control method for unmanned underground equipment in a mine according to claim 2, characterized in that, The equipment's sensing information includes images, point clouds, positioning, and equipment operating conditions.

4. An intelligent control system for unmanned equipment in underground mines, characterized in that, The intelligent control method for unmanned equipment in underground mines according to any one of claims 1-3 includes a central control module, a sensing module, a control execution module and a communication module connected in sequence. The central control module is used to generate a task navigation map based on the control logic map and the tasks being executed, and then send the task navigation map to the front-end device. The perception module, including the positioning submodule and the scene recognition submodule, is used to correct positioning errors or provide position redundancy by identifying key scene locations. The control execution module is used to deploy multiple sub-task control models and call and execute corresponding actions based on perception information and task navigation map; The communication module is used for status synchronization and information exchange between various modules and devices.

5. The intelligent control system for unmanned underground equipment in a mine according to claim 4, characterized in that, The perception module achieves positioning based on SLAM, UWB, or wheeled odometers, and identifies key scenes. It corrects the errors in positioning based on SLAM, UWB, or wheeled odometers by using control points on the task navigation map, or uses the identified location as the location information when positioning completely fails.

6. The intelligent control system for unmanned underground equipment in mines according to claim 4, characterized in that, Key control points in key scenarios include ore outlets, turnouts, and ore unloading points.

7. The intelligent control system for unmanned underground equipment in mines according to claim 4, characterized in that, The communication module is used to transmit location information, status feedback, and task navigation maps.

Images