Underground coal mine robot inspection system and management and control method thereof

Abstract

The invention relates to the technical field of coal mine safety and intelligent inspection, and discloses a coal mine underground robot inspection system, which comprises an environment sensing module used for collecting three-dimensional environment data of a coal mine tunnel, including obstacle position, tunnel form and dynamic change information; the path optimization module is used for optimizing the inspection path based on the three-dimensional environment data and generating an optimal path with optimized path length and curvature, the invention further provides a control method of the underground coal mine robot inspection system, and the control method comprises the following steps: environment sensing: collecting the three-dimensional environment data of a coal mine tunnel through the environment sensing module; and path optimization: generating an optimal path with path length and curvature optimization based on the environment data through a path optimization module. Through multi-sensor sensing, path optimization, reinforcement learning decision and intelligent energy consumption management, high efficiency, intelligence, endurance optimization and task execution accuracy of coal mine underground robot inspection are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of coal mine safety and intelligent inspection, and in particular to a coal mine underground robot inspection system and a control method thereof. Background Art

[0002] The production environment of coal mines is complex and the risks are extremely high. Inspection work, as an important part of coal mine safety production, has long relied on manual work. This traditional method has problems such as low efficiency, great safety hazards, and limited data collection accuracy. In order to improve the safety of coal mine production and the efficiency of inspection tasks, robotics technology has gradually been applied to the underground environment of coal mines and has become an important means to solve this problem. Robots reduce the need for personnel to go down to work through automated inspections, and provide a scientific basis for coal mine management through real-time perception and data processing.

[0003] In the existing technology, the underground robot inspection system of coal mines has achieved certain results. Some technical solutions use lidar or cameras to achieve obstacle detection and environmental modeling, which can meet the path planning requirements in simple tunnels. Some solutions use fixed-rule obstacle avoidance methods to improve the robot's task completion rate in a static environment. At the same time, some technical solutions can issue a warning before the power is exhausted through the energy monitoring module, and complete the inspection and return in combination with the preset path. The application of these technologies reduces the safety risks of manual inspections and improves the operating efficiency in some coal mine scenarios.

[0004] However, the existing technology still has some shortcomings: first, the environmental perception capability is limited. A single sensor solution cannot fully obtain complex information such as dynamic obstacles, tunnel details and gas concentration, resulting in insufficient basic data for path planning; second, path planning is mostly based on fixed rules or static environment assumptions and cannot be adjusted dynamically. Robots running in complex tunnels are prone to vibration or path failure problems; third, the dynamic decision-making ability is weak. The existing system has a lag in responding to areas with dense obstacles or sudden changes in scenes, affecting inspection efficiency and safety; fourth, energy consumption management lacks intelligence, power output and navigation direction are not optimized according to task requirements, and endurance is difficult to guarantee; fifth, the execution module has a single function, poor obstacle avoidance support for dynamic environments, and insufficient task data feedback. Summary of the invention

[0005] In view of the shortcomings of the existing technology, the present invention provides a coal mine underground robot inspection system and a control method thereof, which solves the problems of insufficient environmental perception, inflexible path planning, weak dynamic decision-making ability, and imperfect energy consumption management and task execution in the existing technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A coal mine underground robot inspection system, comprising:

[0007] Environmental perception module, used to collect three-dimensional environmental data of coal mine tunnels, including obstacle locations, tunnel morphology and dynamic change information;

[0008] The path optimization module is used to optimize the inspection path based on the three-dimensional environmental data and generate the optimal path with optimized path length and curvature;

[0009] A reinforcement learning decision module is used to adjust path planning and power output based on real-time perception of dynamic environmental information;

[0010] Energy consumption management module, used to monitor the robot's power status and dynamically allocate power output and navigation direction;

[0011] The execution module is used to receive control instructions from the path optimization module and the reinforcement learning decision module to drive the robot to complete the inspection task.

[0012] Preferably, the environment perception module includes:

[0013] LiDAR, used to collect information on the location and shape of obstacles in coal mine tunnels;

[0014] Infrared cameras to sense dynamic changes in the lanes;

[0015] Gas detectors are used to monitor the concentration of harmful gases and environmental conditions in the tunnel.

[0016] Preferably, the path optimization module includes:

[0017] The path modeling unit based on Finsler geometry is used to calculate the environmental cost and curvature cost of the inspection path and generate the path optimization target;

[0018] The curvature optimization unit is used to smooth the curvature of the path to ensure path continuity and smooth movement;

[0019] The global path generation unit is used to generate the optimal path with comprehensive optimization of path length and curvature.

[0020] Preferably, the reinforcement learning decision module includes:

[0021] A state space construction unit, used to construct a dynamic state space including the robot's position, velocity, acceleration, power state, and obstacle distance;

[0022] Motion control unit, used to control power output and navigation direction adjustment;

[0023] A reward calculation unit, used to calculate the reward value according to the path length, curvature cost, energy consumption cost and obstacle avoidance requirements;

[0024] The policy update unit is used to update the decision policy based on the ProximalPolicyOptimization algorithm.

[0025] Preferably, the energy consumption management module includes:

[0026] The power monitoring unit is used to monitor the remaining power of the robot battery in real time;

[0027] Energy distribution unit, used to dynamically allocate power output and navigation energy consumption according to mission requirements and path status;

[0028] The early warning and return unit is used to issue an early warning and plan the return route when the battery power is lower than the preset threshold.

[0029] Preferably, the execution module includes:

[0030] Motion control unit, used to control the robot's movement and navigation on the optimal path;

[0031] Obstacle avoidance unit, used to adjust the path to avoid obstacles by sensing environmental data in real time;

[0032] The data feedback unit is used to record the robot path, energy consumption and environmental status during the inspection process, and upload the data to the system background.

[0033] The present invention also provides a control method for a coal mine underground robot inspection system, comprising the following steps:

[0034] Environmental perception: Collecting three-dimensional environmental data of coal mine tunnels through the environmental perception module;

[0035] Path optimization, using the path optimization module to generate the optimal path with optimized path length and curvature based on environmental data;

[0036] Real-time decision-making, adjusting path planning and power output through reinforcement learning decision-making module combined with dynamic environment information;

[0037] Energy consumption management: The energy consumption management module monitors the power status in real time, dynamically allocates power output and navigation direction according to the route and mission requirements, and issues an early warning and plans the return route when the power is lower than the preset threshold.

[0038] Task execution, receiving the optimized path and power instructions through the execution module, drives the robot along the optimal path to complete the inspection task, and records the task data during the inspection process for feedback and analysis.

[0039] Preferably, the environmental perception includes:

[0040] Collect obstacle location and shape information;

[0041] Perceive dynamic changes in the lane;

[0042] Monitor the gas concentration and environmental conditions in the tunnel.

[0043] Preferably, the path optimization includes:

[0044] Model the path based on Finsler geometry and calculate the path length and environmental cost;

[0045] Smoothing the path based on the curvature optimization method to reduce the curvature change of the path;

[0046] Generate the optimal path with comprehensive optimization of path length and curvature according to the optimization target.

[0047] Preferably, the real-time decision-making includes:

[0048] Construct a dynamic state space, including position, velocity, acceleration, power state, and obstacle distance;

[0049] Calculate path length cost, curvature cost, energy cost and obstacle avoidance reward based on the dynamic state space;

[0050] Adjust power output and navigation direction based on reinforcement learning algorithms to dynamically optimize path planning.

[0051] The present invention provides a coal mine underground robot inspection system and a control method thereof. It has the following beneficial effects:

[0052] 1. The present invention uses the technical solution of reinforcement learning to adjust path planning and power output in real time, thus realizing intelligent obstacle avoidance and path optimization in a dynamic environment. Compared with the prior art, it overcomes the defects of traditional path planning that relies on fixed rules and is inflexible in response to dynamically changing scenarios, and significantly improves the adaptability of robots in the complex environment of coal mines.

[0053] 2. The present invention effectively extends the robot's battery life by using dynamic energy consumption allocation and low-battery warning technology. In the low-battery state, the shortest return path can be planned first to ensure the safe completion of the task. Compared with the passive monitoring method of power consumption in the prior art, this technology realizes more reasonable energy scheduling and solves the problem of inspection interruption caused by insufficient battery life.

[0054] 3. The present invention combines Finsler geometric modeling and curvature optimization to generate an optimal path with low path cost and smooth curvature, ensuring the stability of the robot's operation. Unlike the existing technology that only aims at the shortest path, this solution eliminates the vibration and unstable operation of the robot caused by the uneven path, and performs better in complex lane environments.

[0055] 4. The execution module converts the optimized path and power instructions into specific motion behaviors, has precise motion control and real-time obstacle avoidance capabilities, and records inspection data for subsequent analysis. Compared with the single-function execution system in the existing technology, this module significantly improves the integrity and reliability of the inspection task through comprehensive data feedback and efficient task execution. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 is a system structure diagram of the present invention;

[0057] Figure 2 This is a module architecture diagram of the environment perception of the present invention;

[0058] Figure 3 A module architecture diagram for path optimization of the present invention;

[0059] Figure 4 The module architecture diagram of the reinforcement learning decision of the present invention;

[0060] Figure 5 This is a module architecture diagram of the energy consumption management of the present invention;

[0061] Figure 6 A module architecture diagram for the implementation of the present invention;

[0062] Figure 7 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0063] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] Please refer to the attached Figure 1 -Attached Figure 6 The embodiment of the present invention provides a coal mine underground robot inspection system, comprising:

[0065] Environmental perception module, used to collect three-dimensional environmental data of coal mine tunnels, including obstacle locations, tunnel morphology and dynamic change information;

[0066] The environmental perception module includes a variety of sensing devices such as laser radar, infrared camera and gas detector. These devices work in coordination to collect three-dimensional environmental data in coal mine tunnels in real time.

[0067] LiDAR is mainly used to collect the shape and location information of obstacles in the tunnel. Specifically:

[0068] LiDAR obtains the distance between the obstacle and the robot by emitting laser pulses and measuring the return time.

[0069] In one possible implementation, the point cloud data generated by the LiDAR can be used to construct a 3D model of the roadway. The model is represented in the form of a matrix, where each element corresponds to a spatial coordinate in the point cloud data:

[0070] M={(x i ,y i , z i )|i=1,2,...,N}

[0071] Where: M: 3D point cloud dataset; x i ,y i , z i : The three-dimensional coordinates of the i-th point in the point cloud data; N: The total number of points in the point cloud data.

[0072] Specifically, x i ,y i , z i The value of is calculated by the following formula:

[0073] x i =r i cos(θ i ), y i =r i sin(θ i ), z i =h i

[0074] Where: r i : The straight-line distance from the i-th point to the laser radar, in meters (m), measured by the laser radar; θ i : The scanning angle of the laser radar, in radians (rad), in the range of [0, 2π]; h i : The height of the lidar relative to the reference surface, in meters (m).

[0075] Through this formula, each point in the point cloud data can be restored to a three-dimensional coordinate through distance and angle, thus forming a three-dimensional model of the roadway.

[0076] Infrared cameras are used to capture dynamic changes in the tunnel, especially in low-light or completely dark environments. Specifically:

[0077] Infrared cameras can sense thermal radiation characteristics and generate infrared images to identify the location and movement direction of dynamic obstacles.

[0078] In some embodiments, the infrared image extracts contour information through an edge detection algorithm and marks dynamic obstacles as specific areas:

[0079] R={(x,y)|T(x,y)>T threshold ]

[0080] Where: R: the two-dimensional pixel set of the dynamic obstacle area; (x, y): the pixel coordinates in the image; T(x, y) is the temperature value in degrees Celsius (℃); T threshold : Temperature threshold of the dynamic obstacle area, in degrees Celsius (℃).

[0081] Temperature threshold T threshold Dynamically set according to the ambient temperature in the tunnel and the thermal radiation characteristics of dynamic obstacles, for example, set to 1.5 times the ambient temperature to avoid misjudgment;

[0082] Gas detectors are used to monitor the concentration of harmful gases and environmental conditions in the tunnel. Specifically:

[0083] The gas detector detects the concentration of gases such as methane, oxygen, carbon monoxide, and carbon dioxide through built-in sensors.

[0084] In one possible implementation, the gas detector generates a gas concentration matrix by periodic sampling:

[0085] G={(g1, g2, ..., g m )|m=1,2,...,M]

[0086] Where: G: gas concentration matrix; g m : The concentration value of the mth gas, in ppm (parts per million); M: The total number of monitored gas types.

[0087] As an option, the gas detector can be combined with lidar and infrared cameras to assist in determining the dangerousness of certain dynamic obstacles by analyzing changes in gas concentration.

[0088] In one possible implementation, the data output by the environment perception module is integrated through the data fusion unit:

[0089] Point cloud data is used to construct a three-dimensional model of the roadway;

[0090] Infrared images are used to mark areas of dynamic obstacles;

[0091] The gas concentration matrix is ​​used to evaluate the safety status of the tunnel environment.

[0092] The integrated environmental data are represented in the form of a multi-dimensional matrix and serve as the input of the subsequent path optimization module.

[0093] Through the above implementation method, the environmental perception module can comprehensively and accurately obtain the three-dimensional environmental data of the coal mine tunnel, providing reliable basic data support for the operation of subsequent modules.

[0094] The path optimization module is used to optimize the inspection path based on the three-dimensional environmental data and generate the optimal path with optimized path length and curvature;

[0095] The core function of the path optimization module is to calculate and generate the optimal inspection path based on the three-dimensional environmental data provided by the environmental perception module. This module combines differential geometry methods and curvature optimization strategies to comprehensively consider the length, smoothness and environmental cost of the path, thereby providing an efficient path solution suitable for complex coal mine environments. The results of path optimization directly affect the efficiency and safety of the robot inspection task.

[0096] Generally speaking, the coal mine tunnel environment has the characteristics of complex paths, random obstacle distribution and dynamic changes. In order to deal with these problems, this module adopts a combination of Finsler geometry and curvature optimization to generate inspection paths that meet specific optimization goals through a series of calculations and modeling.

[0097] In this embodiment, the specific implementation of the path optimization module is as follows:

[0098] The path optimization module includes a path modeling unit based on Finsler geometry, a curvature optimization unit, and a global path generation unit. These units work together to build a complete path optimization system.

[0099] This module first uses Finsler geometry to model the path and calculates the comprehensive cost of the path, including environmental cost and curvature cost. Specifically:

[0100] Finsler geometry defines the path length via the local metric formula for the path:

[0101]

[0102] in: The comprehensive cost of the path; Finsler metric function represents the local environmental cost and curvature cost of the path.

[0103] The Finsler metric function is defined as follows:

[0104]

[0105] Where: g ij (x): weight matrix of path optimization; The velocity component of the path; x: the position coordinate of the path point in three-dimensional space.

[0106] In general, the weight matrix g ij The specific form of (x) consists of the following two parts:

[0107] Fixed environment weight, which represents the basic cost of the area where the waypoint is located;

[0108] Dynamically adjust weights and optimize based on obstacle distribution and roadway complexity.

[0109] The smoothness of the path directly affects the stability and energy consumption of the robot. In order to reduce the curvature change of the path, this module further optimizes the curvature of the path. Specifically:

[0110] The path curvature K(x) is defined by the following formula:

[0111]

[0112] Where: K(x): the curvature of the path point; The rate of change of the path curve; The velocity vector of the path.

[0113] In one possible implementation, curvature optimization is achieved by minimizing the curvature integral:

[0114] J curvature =∫0 T βK(x)dt

[0115] Among them: J curvature : Curvature optimization target value; β: Curvature optimization weight factor, used to balance the curvature cost and path length.

[0116] Based on path modeling and curvature optimization, this module combines the length and curvature of the path to generate the global optimal path. Specifically:

[0117] The global path optimization objective function is:

[0118]

[0119] Among them: J path : Optimization target value of the global path; Environmental cost of the path; K(x): curvature cost of the path.

[0120] As an option, the results of path optimization can be stored in the form of a multi-dimensional matrix, with the coordinates, speed and curvature of each path point fully recorded for subsequent module calls.

[0121] The output of the path optimization module includes the point set and curvature data of the global optimal path. These data are passed to the reinforcement learning decision module as the basic data for dynamically adjusting the path planning. At the same time, the path optimization results directly affect the task allocation strategy of the energy consumption management module, thereby ensuring the efficiency and reliability of the overall operation of the system.

[0122] A reinforcement learning decision-making module is used to adjust path planning and power output based on real-time perception of dynamic environmental information;

[0123] The reinforcement learning decision module combines the global path output of the path optimization module and the dynamic data input of the environment perception module to adjust the robot's inspection path and power output in real time. In the complex underground environment of coal mines, dynamic changes are unpredictable, so it is necessary to rely on reinforcement learning algorithms to achieve optimal decisions in a dynamic environment. The goal of the module is to enable the robot to adapt to environmental changes in real time while optimizing path planning, power output, and energy consumption management.

[0124] Generally speaking, the distribution of obstacles, changes in gas concentration, and the complexity of the lanes in a dynamic environment will significantly affect the safety and efficiency of robot inspections. The reinforcement learning decision module ensures the autonomy and intelligence of the robot inspection process by constructing a dynamic state space, calculating reward values ​​in real time, and updating strategies.

[0125] In this embodiment, the specific implementation of the reinforcement learning decision module is as follows:

[0126] The reinforcement learning decision module first dynamically models the system state and constructs a state space to describe the robot's current operating state and environmental information. Specifically:

[0127] The state space is defined as:

[0128] s t =[x t ,y t , z t , v t , a t , E t , d obs，t ]

[0129] Where: s t : The state vector of the robot at time t; x t ,y t , z t : The three-dimensional coordinates of the robot's current position, in meters (m); v t : The current speed of the robot, in meters per second (m / s); a t : The current acceleration of the robot, in meters per second squared (m / s 2 );E t: The current battery status of the robot, in percentage (%); d obs，t : The distance between the robot and the nearest obstacle, in meters (m).

[0130] As an option, the dimension of the state space can be expanded according to specific application scenarios. For example, when there are specific hazardous areas in the tunnel, area identifiers can be added to enhance the state modeling capability.

[0131] In reinforcement learning, the action space describes the set of actions that a robot can perform. Specifically:

[0132] The action space is defined as:

[0133] a t =[u 1，t ,u 2，t ]

[0134] Among them: a t : the action performed by the robot at time t; u 1，t : Power output adjustment value, dimensionless unit; u 2，t : Navigation direction adjustment value, in radians (rad).

[0135] The power output adjustment value is used to control the robot's forward speed and acceleration, and the navigation direction adjustment value is used to control the robot's path direction when avoiding obstacles.

[0136] The reward function is a key part of the reinforcement learning decision module, defining the reward value the robot receives after performing a certain action. Specifically:

[0137] The reward function is defined as:

[0138]

[0139] Where: R(s t , a t ): The robot is in state s t Next, perform action a t Reward value; α1, α2, α3, α4: reward weight factors, all dimensionless; The environmental cost of the path; K(x): the curvature cost of the path; P(u 1，t ,u 2，t ): energy cost of the action; d obs，t : The distance between the robot and the obstacle.

[0140] Specifically, the environmental cost and curvature cost reflect the complexity and smoothness of the path respectively, the energy consumption cost is used to constrain the rationality of power output and navigation adjustment, and the obstacle distance is used to motivate the robot to maintain safe operation.

[0141] The reinforcement learning decision module updates the control strategy through the Proximal Policy Optimization (PPO) algorithm to ensure that the robot can gradually learn the optimal behavior. The specific implementation is as follows:

[0142] The policy function is defined as:

[0143] π θ (a t |s t )

[0144] Where: π θ : Policy function, represented by parameter θ; a t : In state s t Next, select the action.

[0145] The policy update targets are:

[0146]

[0147] in: Importance sampling ratio; A t : advantage function, used to measure the degree of improvement of the current action relative to the strategy; ∈: clipping threshold, limiting the update range of the strategy. The output of the reinforcement learning decision module includes the optimized action instructions and the strategy update results. These instructions are directly sent to the execution module to drive the robot to complete the inspection task. At the same time, the module iteratively updates the strategy, making the robot more adaptable in subsequent tasks.

[0148] Energy consumption management module, used to monitor the robot's power status and dynamically allocate power output and navigation direction;

[0149] The main function of the energy management module is to monitor the robot's power status in real time, dynamically allocate power output and navigation direction according to the inspection path and task requirements, and issue an early warning and plan the return path when the power is low. Through the reasonable allocation and management of energy consumption, the module can effectively extend the robot's battery life during inspection tasks and ensure the smooth completion of the task.

[0150] Generally speaking, the endurance of robots in underground coal mine environments is crucial to inspection efficiency and task completion. To this end, the present invention achieves efficient energy allocation through an energy consumption management module under the collaboration of a path optimization module and a reinforcement learning decision module, thereby reducing the risk of task interruption.

[0151] In this embodiment, the specific implementation of the energy consumption management module is as follows:

[0152] The energy consumption management module first monitors the robot's power status in real time. Specifically:

[0153] The state of charge is described by the following formula:

[0154] E(t)=E init -∫0 t P(u 1，τ ,u 2，τ )dτ

[0155] Where: E(t): the remaining power of the robot at time t, in percentage (%); E init : Initial power, in percentage (%); P(u 1，τ ,u 2，τ ): Energy consumption rate of the robot in time τ, in power (W).

[0156] As an option, the energy consumption rate P(u 1，τ ,u 2，τ ) is determined by the comprehensive energy consumption of power output and navigation direction, and can be calculated by the following formula:

[0157]

[0158] Where: c1, c2: energy consumption coefficients for power output and navigation direction adjustment, dimensionless; u 1，τ : Power output adjustment value at time τ, dimensionless; u 2，τ : Navigation direction adjustment value at time τ, dimensionless.

[0159] Based on energy consumption monitoring, the energy consumption management module dynamically allocates energy consumption between power output and navigation direction according to path complexity and task requirements. Specifically:

[0160] The energy consumption allocation aims to minimize the total energy consumption. The optimization formula is as follows:

[0161] min∫0 T P(u 1，t ,u 2，t )dt

[0162] Where: T: the total time of the inspection task, in seconds (s).

[0163] As a possible implementation method, the energy consumption management module combines the path cost information output by the path optimization module to dynamically adjust the power output value and navigation direction value according to parameters such as path flatness and obstacle distribution. Through such adjustments, energy waste on complex paths can be significantly reduced.

[0164] When the battery status approaches the critical value, the energy consumption management module triggers a low battery warning and prioritizes planning the return route.

[0165] Power warning threshold E thresholdThrough experimental settings, for example, when the remaining power is less than 20%, the system will sound an alarm.

[0166] The return path planning takes the shortest distance as the optimization goal, and combines dynamic obstacle information to avoid obstacles. The formula for optimizing the path is:

[0167]

[0168] Among them: J return : Optimization target value of the return path, dimensionless; T return : The time length of the return path, in seconds (s).

[0169] As a possible implementation method, dynamic obstacle avoidance on the return path is assisted by a reinforcement learning decision-making module to ensure that the robot can still safely return to its initial position when the battery is low.

[0170] The output of the energy management module includes power status information, real-time energy allocation plan and return path instructions. These output data are passed to the execution module to drive the robot to complete the inspection task, and interact with the reinforcement learning decision module to provide support for dynamic adjustment.

[0171] The execution module is used to receive control instructions from the path optimization module and the reinforcement learning decision module to drive the robot to complete the inspection task;

[0172] The execution module is the terminal control unit of the underground coal mine robot inspection system of the present invention, which is used to receive instructions from the path optimization module, the reinforcement learning decision module and the energy consumption management module, and complete the inspection task through actual operation. The main functions of the execution module include motion control, obstacle avoidance and task data feedback. Its role is to convert the optimization results of the front-end module into specific motion behaviors, and record and feedback the data during the inspection process.

[0173] Generally speaking, the complexity of the underground environment of coal mines requires robots to have high-precision control capabilities and real-time response capabilities. This module ensures the efficiency and safety of inspection tasks in a dynamic environment by coordinating multiple execution units.

[0174] In this embodiment, the specific implementation of the execution module is as follows:

[0175] The motion control unit in the execution module is responsible for the precise movement of the robot along the path. Specifically:

[0176] The motion control unit receives the optimal path point set (x t ,y t , z t ) and convert it into a continuous motion trajectory.

[0177] In one possible implementation, the motion trajectory is calculated by the following formula:

[0178]

[0179] in: The velocity vector of the trajectory, in meters per second (m / s); γ(t): the continuous trajectory composed of the path point set, in meters (m); t: time, in seconds (s); T: the total task time, in seconds (s).

[0180] Specifically, the motion control unit ensures that the robot always moves along the optimal trajectory by adjusting the power output and navigation direction in real time.

[0181] As an option, the unit can combine dynamic adjustment of speed and acceleration to optimize motion smoothness and avoid severe vibrations that could affect inspection tasks.

[0182] The obstacle avoidance unit in the execution module is responsible for the robot's obstacle avoidance operations in a dynamic environment. Specifically:

[0183] Obstacle avoidance is based on the navigation direction adjustment value (u 2，t ), and dynamically update it in combination with the real-time data of the environmental perception module.

[0184] In one possible implementation, the obstacle avoidance path is generated by the following optimization formula:

[0185]

[0186] Among them: J avoid : Optimization target of obstacle avoidance path, dimensionless; d obs，t : The distance between the robot and the nearest obstacle, in meters (m); β: Curvature weight factor, dimensionless; K(x): Path curvature cost, in inverse meters (m -1 ).

[0187] Generally, the obstacle avoidance unit prioritizes the shortest path that avoids dynamic obstacles while ensuring path smoothness.

[0188] The task data feedback unit is used to record the path data, energy consumption data and environmental status information during the inspection process. These data are uploaded to the system background in real time through the wireless communication module for subsequent analysis and optimization.

[0189] As an option, the task data can be stored in a multi-dimensional matrix form, for example:

[0190] D={(t i , x i ,y i, z i , E i , d obs，i )|i=1,2,...,N}

[0191] Where: D: task data matrix; t i : the i-th time point, in seconds (s); x i ,y i , z i : The position coordinates of the robot at the i-th time point, in meters (m); E i : The battery status of the robot at the i-th time point, in percentage (%); d obs，i : The distance between the robot and the nearest obstacle at the i-th time point, in meters (m).

[0192] Specifically, these data can be used to analyze the robot inspection efficiency, path optimization effect and energy consumption allocation rationality.

[0193] The execution module receives the optimal path from the path optimization module, the real-time control instructions from the reinforcement learning decision module, and the power allocation results from the energy consumption management module to complete the actual operation of the inspection task. The output of the module includes environmental data collected during the inspection process, path execution records, and energy consumption feedback data.

[0194] The execution module can efficiently complete the specific operations of the robot inspection task while ensuring the task safety and the integrity of data feedback in a dynamic environment.

[0195] The control method of a coal mine underground robot inspection system described below and the coal mine underground robot inspection system described above can be referred to each other.

[0196] Please refer to the attached Figure 7 The present invention also provides a method for controlling a coal mine underground robot inspection system, comprising the following steps:

[0197] Environmental perception: Collecting three-dimensional environmental data of coal mine tunnels through the environmental perception module;

[0198] Path optimization, using the path optimization module to generate the optimal path with optimized path length and curvature based on environmental data;

[0199] Real-time decision-making, adjusting path planning and power output through reinforcement learning decision-making module combined with dynamic environment information;

[0200] Energy consumption management: The energy consumption management module monitors the power status in real time, dynamically allocates power output and navigation direction according to the route and mission requirements, and issues an early warning and plans the return route when the power is lower than the preset threshold.

[0201] Task execution, receiving the optimized path and power instructions through the execution module, drives the robot along the optimal path to complete the inspection task, and records the task data during the inspection process for feedback and analysis.

[0202] The method of this embodiment can be used to execute the above system embodiment, and its principles and technical effects are similar, which will not be repeated here.

[0203] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A coal mine underground robot inspection system, characterized in that: include: Environmental perception module, used to collect three-dimensional environmental data of coal mine tunnels, including obstacle locations, tunnel morphology and dynamic change information; The path optimization module is used to optimize the inspection path based on the three-dimensional environmental data and generate the optimal path with optimized path length and curvature; A reinforcement learning decision module is used to adjust path planning and power output based on real-time perception of dynamic environmental information; Energy consumption management module, used to monitor the robot's power status and dynamically allocate power output and navigation direction; The execution module is used to receive control instructions from the path optimization module and the reinforcement learning decision module to drive the robot to complete the inspection task.

2. A coal mine underground robot inspection system according to claim 1, characterized in that: The environment perception module comprises: LiDAR, used to collect information on the location and shape of obstacles in coal mine tunnels; Infrared cameras to sense dynamic changes in the lanes; Gas detectors are used to monitor the concentration of harmful gases and environmental conditions in the tunnel.

3. The coal mine underground robot inspection system according to claim 1, characterized in that: The path optimization module includes: The path modeling unit based on Finsler geometry is used to calculate the environmental cost and curvature cost of the inspection path and generate the path optimization target; The curvature optimization unit is used to smooth the curvature of the path to ensure path continuity and smooth movement; The global path generation unit is used to generate the optimal path with comprehensive optimization of path length and curvature.

4. The coal mine underground robot inspection system according to claim 1, characterized in that: The reinforcement learning decision module includes: A state space construction unit, used to construct a dynamic state space including the robot's position, velocity, acceleration, power state, and obstacle distance; Motion control unit, used to control power output and navigation direction adjustment; A reward calculation unit, used to calculate the reward value according to the path length, curvature cost, energy consumption cost and obstacle avoidance requirements; The policy update unit is used to update the decision policy based on the ProximalPolicyOptimization algorithm.

5. The coal mine underground robot inspection system according to claim 1, characterized in that: The energy consumption management module comprises: The power monitoring unit is used to monitor the remaining power of the robot battery in real time; Energy distribution unit, used to dynamically allocate power output and navigation energy consumption according to mission requirements and path status; The early warning and return unit is used to issue an early warning and plan the return route when the battery power is lower than the preset threshold.

6. The coal mine underground robot inspection system according to claim 1, characterized in that: The execution module includes: Motion control unit, used to control the robot's movement and navigation on the optimal path; Obstacle avoidance unit, used to adjust the path to avoid obstacles by sensing environmental data in real time; The data feedback unit is used to record the robot path, energy consumption and environmental status during the inspection process, and upload the data to the system background.

7. A method for controlling a coal mine underground robot inspection system, characterized in that: Using a coal mine underground robot inspection system according to any one of claims 1 to 6, comprising the following steps: Environmental perception: Collecting three-dimensional environmental data of coal mine tunnels through the environmental perception module; Path optimization, using the path optimization module to generate the optimal path with optimized path length and curvature based on environmental data; Real-time decision-making, adjusting path planning and power output through reinforcement learning decision-making module combined with dynamic environment information; Energy consumption management: The energy consumption management module monitors the power status in real time, dynamically allocates power output and navigation direction according to the route and mission requirements, and issues an early warning and plans the return route when the power is lower than the preset threshold. Task execution, receiving the optimized path and power instructions through the execution module, drives the robot along the optimal path to complete the inspection task, and records the task data during the inspection process for feedback and analysis.

8. A method for controlling a coal mine underground robot inspection system according to claim 7, characterized in that: The environmental perception includes: Collect obstacle location and shape information; Perceive dynamic changes in the lane; Monitor the gas concentration and environmental conditions in the tunnel.

9. A method for controlling a coal mine underground robot inspection system according to claim 7, characterized in that: The path optimization includes: Model the path based on Finsler geometry and calculate the path length and environmental cost; Smoothing the path based on the curvature optimization method to reduce the curvature change of the path; Generate the optimal path with comprehensive optimization of path length and curvature according to the optimization target.

10. A method for controlling a coal mine underground robot inspection system according to claim 7, characterized in that: The real-time decision making includes: Construct a dynamic state space, including position, velocity, acceleration, power state, and obstacle distance; Calculate path length cost, curvature cost, energy cost and obstacle avoidance reward based on the dynamic state space; Adjust power output and navigation direction based on reinforcement learning algorithms to dynamically optimize path planning.

Images