Robot system for coal bunker cleaning

Abstract

The invention discloses a robot system for coal bunker cleaning, and belongs to the technical field of coal bunker cleaning, a cooperative control module comprises a global task decomposition unit and a local path planning unit, and the global task decomposition unit is used for obtaining coal bunker overall cleaning requirements and caking coal distribution information collected by an environment sensing module; a coal bunker cleaning area is divided into a plurality of sub-task areas, sub-tasks are distributed to all cleaning robots, and the local path planning unit plans paths from the current position to all cleaning points in the sub-task areas for all the cleaning robots on the basis of an improved Frenet algorithm and a cubic spline interpolation algorithm in combination with obstacle position information. According to the invention, environmental data are collected through multi-sensor fusion, and caking coal distribution and obstacle positions can still be accurately identified even in a coal bunker environment with high dust and blocked sight. Meanwhile, the improved Frenet algorithm can automatically generate a dynamic guide line, does not need to depend on a fixed map, and is adaptive to the unstructured environment of the coal bunker.

Description

Technical Field

[0001] This invention belongs to the field of coal bunker cleaning technology, and specifically relates to a robot system for coal bunker cleaning. Background Technology

[0002] In the coal production and storage process, coal bunkers, as critical transit facilities, are prone to coal buildup on their inner walls and bottom after prolonged use. This coal buildup not only reduces the effective volume of the bunker and decreases coal conveying efficiency, but also can lead to safety accidents such as blockages and collapses due to excessive thickness. Traditional coal bunker cleaning methods are mainly manual, requiring workers to enter the dusty, confined space of the bunker, which poses risks of carbon monoxide poisoning and falls from heights, and use tools such as high-pressure water guns and shovels to clean it. This method is not only labor-intensive and inefficient, but also poses extremely high safety hazards.

[0003] With the development of automation technology, some companies have tried using single robots for coal bunker cleaning, but existing technologies still have many shortcomings: First, the interior of a coal bunker is a typical unstructured environment with no fixed travel path, and the positions of obstacles change dynamically. Existing robots mostly rely on preset paths or single sensors, making it difficult to adapt to environmental changes in real time, and prone to path deviations or collisions. Second, the cleaning efficiency of a single robot is limited, making it difficult to meet the rapid cleaning needs of large coal bunkers. When multiple robots work together, there is a lack of effective task allocation and path conflict coordination mechanisms, often resulting in "repeated cleaning" or "cleaning blind spots," leading to serious resource waste. Third, the evaluation of cleaning effect relies on manual post-cleaning inspections, making it impossible to determine in real time whether the cleaned area meets the standards. If the cleaning is not thorough, a second entry into the coal bunker is required for re-cleaning, increasing operating costs and risks. Fourth, path planning does not consider the robot's energy consumption characteristics. Some paths have excessive curvature and travel distances, resulting in high energy consumption, shortening the robot's endurance, and affecting the continuity of cleaning tasks.

[0004] Furthermore, while existing technologies have incorporated Frenet and cubic spline interpolation algorithms for path planning, they haven't been adapted to the unstructured environment of coal bunkers. For example, they lack dynamic guide line generation mechanisms and still rely on fixed map data. Simultaneously, sensor data fusion is low, and camera accuracy is easily affected by dust obstruction, making it difficult to meet the perception needs in complex environments when used alone. Therefore, there is an urgent need for a coal bunker cleaning robot system that can adapt to the complex environment of coal bunkers, achieve efficient multi-robot collaboration, provide real-time evaluation of cleaning effectiveness, and optimize energy consumption. Summary of the Invention

[0005] The purpose of this invention is to provide a robotic system for cleaning coal bunkers to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a robot system for cleaning coal bunkers, comprising multiple cleaning robots, an environmental perception module, a collaborative control module, a cleaning effect evaluation module, and an energy consumption optimization module. The environmental perception module is mounted on the cleaning robots and uses cameras, infrared imagers, lidar, and ultrasonic sensors to collect information on the distribution of sticky coal, the location of obstacles, and images and infrared data of the cleaned areas in real time.

[0007] The collaborative control module includes a global task decomposition unit and a local path planning unit. The global task decomposition unit is used to obtain the overall cleaning requirements of the coal bunker and the coal distribution information collected by the environmental perception module, divide the coal bunker cleaning area into multiple sub-task areas, and assign the sub-tasks to each cleaning robot. The local path planning unit is based on the improved Frenet algorithm and cubic spline interpolation algorithm, combined with obstacle location information, to plan the path from the current position to each cleaning point in the sub-task area for each cleaning robot. The local path planning unit also has a conflict detection sub-unit, which is used to detect in real time whether there is a cross or overlapping conflict between the paths of multiple robots. If a conflict exists, the path or travel speed of the relevant robots is adjusted.

[0008] The cleaning effect evaluation module receives images and infrared data of the cleaned area collected by the environmental perception module, and judges the cleaning effect of the cleaned area through image recognition technology. If the effect is not met, a secondary cleaning instruction is generated and sent to the collaborative control module. The energy consumption optimization module is used to establish an energy consumption model of the cleaning robot, calculate the energy consumption value of each path through the energy consumption model, select the path with the lowest energy consumption as the optimal path, and feed it back to the local path planning unit. The cleaning robot receives the path instructions sent by the collaborative control module, moves along the planned path and performs cleaning operations, and simultaneously feeds back its own position, movement status and cleaning operation status to the collaborative control module in real time.

[0009] The preferred improvement of the Frenet algorithm lies in the automatic generation of dynamic guide lines, taking into account the unstructured environment of the coal bunker where there are no fixed guide lines. The generation process of the dynamic guide lines is as follows:

[0010] First, based on the coordinates of the cleaning points within the sub-task area, a clustering algorithm is used to group the cleaning points into multiple cleaning point clusters. Then, the center coordinates of each cleaning point cluster are calculated, and the center coordinates are connected sequentially to form an initial guide line. Finally, based on the obstacle location information collected by the environmental perception module, the initial guide line is smoothly adjusted to avoid obstacles, resulting in a dynamic guide line.

[0011] Preferably, the path generation process based on the improved Frenet algorithm includes: constructing a Frenet coordinate system based on the dynamic guide line, converting the position of the cleaning robot into Frenet coordinates (s, d), where s is the arc length along the guide line, d is the lateral distance from the guide line, and a lateral path and a longitudinal path are generated by fitting a fifth-order polynomial, where the lateral path is a ds curve, the longitudinal path is a vs curve, and v is the traveling speed of the cleaning robot.

[0012] Preferably, the cubic spline interpolation algorithm is used to interpolate the path points generated by the local path planning unit. The specific process is as follows: the path points are divided into multiple intervals, and a cubic polynomial is constructed in each interval so that the first and second derivatives of the cubic polynomials of adjacent intervals are continuous at the connection point, thereby generating a smooth path without sharp turns.

[0013] Preferably, the conflict detection process of the conflict detection subunit includes: acquiring the current position, travel speed and planned path of each cleaning robot in real time; predicting the position trajectory of each robot within a preset time period based on the planned path of each robot; and determining that there is a path conflict if the distance between the position trajectories of any two robots at the same time point is less than a preset safe distance.

[0014] Preferably, the evaluation process of the cleaning effect evaluation module includes: converting the image of the cleaned area captured by the camera into a grayscale image, extracting the contour information of the area through an edge detection algorithm, calculating the contour area ratio, wherein the contour area ratio reflects the distribution range of coal residue, and using infrared data collected by an infrared imager to analyze the temperature distribution of the cleaned area, obtaining the area of ​​the coal residue area by segmenting through a temperature threshold, and calculating the coal residue rate by combining the contour area ratio and the coal residue area. If the coal residue rate is greater than a preset threshold, it is determined that the cleaning has not met the standard, and a secondary cleaning instruction is generated.

[0015] Preferably, the energy consumption optimization module fits the relationship between robot energy consumption and path parameters through experiments, and uses a multiple linear regression algorithm to fit an energy consumption model. The expression of the energy consumption model is E=K1*L+K2*C+K3*F, where E is the energy consumption value, L is the travel distance, C is the average curvature of the path, F is the robot load coefficient, and K1, K2 and K3 are preset weight coefficients obtained by fitting experimental data.

[0016] Preferably, it also includes a communication module, which adopts a wireless mesh network and consists of a mobile node carried by the robot and a fixed node in the coal bunker. Communication time slots are allocated through TDMA technology, and a retransmission mechanism is used for critical instructions.

[0017] Preferably, the cleaning robot includes a walking mechanism, a cleaning mechanism, a positioning module, and an emergency handling module. The walking mechanism adopts a tracked structure, the cleaning mechanism consists of a high-pressure water gun and a rotating brush, the positioning module adopts a laser SLAM radar and an inertial navigation module, and achieves positioning through the ROS cartographer algorithm. The emergency handling module is developed based on an Arduino microcontroller. When an obstacle is detected or the remaining power is ≤10%, an emergency signal is sent to the collaborative control module.

[0018] Preferably, the collaborative control module uses an industrial-grade computer.

[0019] Compared with the prior art, the beneficial effects of the present invention are:

[0020] 1. This invention uses multi-sensor fusion to collect environmental data, enabling accurate identification of coal distribution and obstacle locations even in dusty, obstructed coal bunker environments. Simultaneously, the improved Frenet algorithm automatically generates dynamic guide lines, eliminating the need for fixed maps and adapting to the unstructured environment of coal bunkers. This solves the problem of traditional robot path planning relying on fixed routes, making it suitable for the complex and ever-changing environment within coal bunkers.

[0021] 2. This invention achieves reasonable task allocation through a global task decomposition unit, combined with a conflict detection and resolution mechanism, to avoid multi-robot path conflicts, reduce repetitive cleaning and resource waste, and significantly improve overall cleaning efficiency.

[0022] 3. This invention uses image and infrared data fusion to quantitatively calculate the residual coal sticking rate and determine in real time whether the cleaning meets the standards. No manual post-inspection is required. The secondary cleaning command accurately locates the non-compliant areas, avoiding repeated cleaning and effectively reducing operating costs. Attached Figure Description

[0023] Figure 1 This is a system diagram of the present invention;

[0024] Figure 2 This is a flowchart of the present invention;

[0025] Figure 3 This is a flowchart of the environmental perception module of the present invention;

[0026] Figure 4 This is a flowchart of the cleaning effect evaluation module of the present invention;

[0027] Figure 5 This is a flowchart of the energy consumption optimization module of the present invention;

[0028] Figure 6 This is a flowchart illustrating the workflow of a single cleaning robot according to the present invention. Detailed Implementation

[0029] 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.

[0030] Please see Figure 1 , Figure 2 and Figure 3 This invention provides a robot system for cleaning coal bunkers, including multiple cleaning robots, an environmental perception module, a collaborative control module, a cleaning effect evaluation module, and an energy consumption optimization module. The environmental perception module, mounted on the cleaning robots, uses cameras, infrared imagers, lidar, and ultrasonic sensors to collect real-time information on the distribution of sticky coal, the location of obstacles, and images and infrared data of the cleaned areas within the coal bunker. The collaborative control module employs an industrial-grade computer. It includes a global task decomposition unit and a local path planning unit. The global task decomposition unit obtains the overall cleaning requirements of the coal bunker and the sticky coal distribution information collected by the environmental perception module, dividing the coal bunker cleaning area into multiple sub-task areas and assigning the sub-tasks to each cleaning robot. The local path planning unit, based on an improved Frenet algorithm and a cubic spline interpolation algorithm, combined with obstacle location information, plans a path for each cleaning robot from its current position to each cleaning point within the sub-task area. The local path planning unit also includes a conflict detection sub-unit, used to detect in real-time whether there are intersections or overlaps between the paths of multiple robots. If a conflict exists, the path or travel speed of the relevant robots is adjusted. The improvement of the Frenet algorithm lies in its automatic generation of dynamic guide lines, taking into account the unstructured environment of coal bunkers without fixed guide lines. The generation process of the dynamic guide lines is as follows:

[0031] First, based on the coordinates of the cleaning points within the sub-task area, a clustering algorithm is used to group the cleaning points into multiple cleaning point clusters. Then, the center coordinates of each cleaning point cluster are calculated, and the center coordinates are connected sequentially to form an initial guide line. Finally, based on the obstacle location information collected by the environmental perception module, the initial guide line is smoothly adjusted to avoid obstacles, resulting in a dynamic guide line.

[0032] The path generation process based on the improved Frenet algorithm includes: constructing a Frenet coordinate system on the basis of dynamic guide lines, converting the position of the cleaning robot into Frenet coordinates (s, d), where s is the arc length along the guide line and d is the lateral distance from the guide line. The lateral path and the longitudinal path are generated by fitting a fifth-order polynomial, where the lateral path is the ds curve and the longitudinal path is the vs curve, and v is the traveling speed of the cleaning robot.

[0033] The cubic spline interpolation algorithm is used to interpolate path points generated by local path planning units. The specific process is as follows: the path points are divided into multiple intervals, and a cubic polynomial is constructed in each interval so that the first and second derivatives of the cubic polynomials of adjacent intervals are continuous at the connection point, thus generating a smooth path without sharp turns.

[0034] The conflict detection process of the conflict detection subunit includes: real-time acquisition of the current position, speed and planned path of each cleaning robot; prediction of the position trajectory of each robot within a preset time period based on the planned path of each robot; and determination of path conflict if the distance between the position trajectories of any two robots at the same time point is less than the preset safe distance.

[0035] Further, system initialization and environmental data acquisition: The system was deployed inside the circular coal bunker of Linhuan Coal Mine. First, all cleaning robots and modules were activated. The environmental perception module used a Hikvision DS-2CD8A20F-EI3 camera, an FLIRA65 infrared imager, an RPLIDARRA3 lidar, and an HC-SR04 ultrasonic sensor to collect data at a frequency of 10 frames per second. After hardware-triggered synchronization, the data was sent to the collaborative control module. The environmental perception module then collected information on the distribution of sticky coal, showing three main sticky coal areas on the inner wall of the coal bunker, with areas of 30m², 25m², and 20m², and a thickness of 2–8cm. Obstacle information showed two coal gangue obstacles at the bottom of the coal bunker, measuring 1m × 0.8m × 0.5m and 0.6m × 0.5m × 0.4m respectively. Meanwhile, the three cleaning robots reported their initial positions: to the left, right, and middle of the coal bunker entrance, respectively. All three robots had 100% remaining battery power and reported the status of their cleaning mechanisms.

[0036] Global Task Decomposition and Subtask Allocation: The global task decomposition unit of the collaborative control module acquires the 3D laser scanning model of the coal bunker and the coal adhesion distribution data from the environmental perception module. The DBSCAN clustering algorithm is used to divide the coal bunker cleaning area into three subtask areas, each corresponding to one cleaning robot. During task allocation, a model is established with the goal of minimizing the total cleaning time. Input parameters include: distance from robot 1 to subtask area 1: 3m; battery level: 100%; water pressure: 25MPa. Distance from robot 2 to subtask area 2: 4m; battery level: 100%; water pressure: 24MPa. Distance from robot 3 to subtask area 3: 2.5m; battery level: 100%; water pressure: 26MPa. The Hungarian algorithm is used to solve for the allocation scheme: robot 1 is responsible for subtask area 1, robot 2 for subtask area 2, and robot 3 for subtask area 3. Subtask instructions are sent to each robot via the communication module.

[0037] Local Path Planning and Energy Consumption Optimization: Taking Robot No. 1 as an example, the local path planning unit generates dynamic guide lines based on an improved Frenet algorithm. First, using the K-means algorithm with a cluster size of 3, the cleaning points within subtask area 1 are divided into 3 clusters. The coordinates of the cluster centers ((2m, 3m), (5m, 7m), (8m, 4m)) are calculated and connected to form the initial guide line. Then, the initial guide line is adjusted based on obstacle location information to avoid the bottom coal gangue, resulting in a dynamic guide line with a total length of 12m. In the Frenet coordinate system, the lateral path (ds curve) and longitudinal path (vs curve) are fitted using a fifth-order polynomial. The maximum acceleration of the lateral path is 0.4m / s², and the speed of the longitudinal path is adjusted according to the curvature (curvature 0.3m). -1 (The speed is 0.6 m / s). Then, a cubic spline interpolation algorithm is used to interpolate the path points at 0.5 m intervals to generate a smooth path. The energy optimization module sets the path parameters as follows: L=12m, C=0.3m. -1 Substituting F=1.3 into the energy consumption model, E=0.8×12+15×0.3+200×1.3=0.96+4.5+260=265.46Wh. Simultaneously, the energy consumption values ​​of the other two candidate paths are calculated as 278.2Wh and 285.6Wh respectively. The current path is selected as the optimal path and fed back to the local path planning unit. The collaborative control module sends the optimal path command to each robot. Robot 1 moves along the planned path and activates the cleaning mechanism: for 5-8cm thick sticky coal, it first activates the rotating brush for pretreatment, then turns on the high-pressure water gun for rinsing. At the same time, the robot reports its position, speed, and cleaning status to the collaborative control module in real time.

[0038] Conflict Detection and Robot Operation Execution: The conflict detection subunit acquires the positions, speeds, and planned paths of the three robots in real time. Based on Kalman filtering, it predicts the trajectory for the next 5 seconds and detects that robots 2 and 3 will have a conflict in the middle of the coal bunker with a trajectory distance of 0.8m after 1.5 seconds. Because robot 2 has a higher task priority than robot 3, the conflict detection subunit adjusts the path of robot 3 by shifting it laterally by 0.4m. After recalculating the trajectory, the distance between the two robots is ≥1.5m, and the conflict is eliminated.

[0039] Cleaning Effect Evaluation and Secondary Cleaning: After robot 1 completes the cleaning of sub-task area 1, the environmental perception module collects images and infrared data of the area. After image preprocessing, the area of ​​the sticky coal outline is extracted as S1 = 1.2㎡, the cleaned area is S2 = 45㎡, and R1 = 1.2 / 45 ≈ 2.67%. The infrared data segmentation yields the residual sticky coal area as S3 = 0.8㎡, and R2 = 0.8 / 45 ≈ 1.78%. The residual sticky coal rate is calculated as R = 0.6 × 2.67% + 0.4 × 1.78% ≈ 1.6% + 0.71% = 2.31% ≤ 5%, indicating that the cleaning meets the standard, and a "cleaning completed" signal is sent to the collaborative control module. After robot 2 completes the cleaning of sub-task area 2, the residual sticky coal rate is R = 6.8% > 5%. The cleaning effect evaluation module generates a secondary cleaning instruction. The coordinates of the unqualified area are (7m, 6m) ~ (9m, 8m). The collaborative control module schedules robot 2 to return to this area and adjusts the high-pressure water gun pressure for secondary cleaning.

[0040] Task Completion and System Shutdown: After all three robots have completed their cleaning tasks, the collaborative control module summarizes the cleaning data and sends it to the remote monitoring terminal. Each robot returns to its docking point at the coal bunker entrance, shuts down its cleaning mechanism, and the system automatically shuts down, completing the coal bunker cleaning operation.

[0041] Please see Figure 5 This invention provides a robot system for cleaning coal bunkers. The energy consumption optimization module is used to establish an energy consumption model for the cleaning robot. The energy consumption model is used to calculate the energy consumption value of each path, select the path with the lowest energy consumption as the optimal path, and feed it back to the local path planning unit. The energy consumption optimization module fits the relationship between robot energy consumption and path parameters through experiments, and uses a multiple linear regression algorithm to fit the energy consumption model. The expression of the energy consumption model is E=K1*L+K2*C+K3*F, where E is the energy consumption value, L is the travel distance, C is the average curvature of the path, F is the robot load coefficient, and K1, K2 and K3 are preset weight coefficients obtained by fitting experimental data.

[0042] Furthermore, during the experiment, 10 paths with different parameters were selected, and each path was tested 10 times, with the energy consumption values ​​recorded and averaged. Linear regression fitting was performed using Python's scikit-learn library to obtain the coefficients under typical operating conditions: K1=0.8, indicating that the travel distance has the most direct impact on energy consumption; K2=15, indicating that the greater the curvature, the higher the turning energy consumption; K3=200, indicating that the load coefficient is positively correlated with the high-pressure water gun power and has a significant impact. During path selection, the local path planning unit generates 3-5 candidate paths, and the energy consumption optimization module calculates the E value for each path, selecting the path with the smallest E value. If the difference in E values ​​among multiple paths is ≤5%, the path with the smaller curvature is prioritized, balancing energy consumption and travel stability.

[0043] Please see Figure 4This invention provides a robot system for cleaning coal bunkers. A cleaning effect evaluation module receives images and infrared data of the cleaned area collected by an environmental perception module, and uses image recognition technology to determine the cleaning effect of the cleaned area. If the effect is not satisfactory, a secondary cleaning instruction is generated and sent to the collaborative control module. The evaluation process of the cleaning effect evaluation module includes: converting the cleaned area image collected by the camera into a grayscale image; extracting the contour information within the area using an edge detection algorithm; calculating the contour area ratio, where the contour area ratio reflects the distribution range of coal residue; analyzing the temperature distribution of the cleaned area using infrared data collected by an infrared imager; obtaining the area of ​​the coal residue area by segmenting it using a temperature threshold; and calculating the coal residue rate by combining the contour area ratio and the coal residue area. If the coal residue rate is greater than a preset threshold, the cleaning is deemed unsatisfactory, and a secondary cleaning instruction is generated.

[0044] Furthermore, the high-resolution image is converted into an 8-bit grayscale image, and a 5×5 Gaussian filter is used to remove dust noise. Adaptive threshold segmentation is used to distinguish the foreground from the background. Then, the Canny algorithm is used to extract the foreground contour, and the total area of ​​the contour S1 and the total area of ​​the cleaned area S2 are calculated to obtain the contour area ratio R1 = S1 / S2.

[0045] Infrared data processing: The temperature data collected by the infrared imager ranges from -20 to 150℃. Based on the coal bunker environment, the coal-sticking temperature threshold is calibrated, and the area of ​​residual coal sticking, S3, is segmented. The area ratio R2 = S3 / S2 is calculated. The coal sticking rate R = 0.6 × R1 + 0.4 × R2, with a preset threshold of 5%. When R > 5%, a secondary cleaning instruction is generated, containing the coordinates of the substandard area and recommended cleaning parameters.

[0046] This invention provides a robot system for cleaning coal bunkers, which also includes a communication module. The communication module adopts a wireless mesh network and consists of a mobile node mounted on the robot and a fixed node in the coal bunker. Communication time slots are allocated through TDMA technology, and a retransmission mechanism is used for critical instructions.

[0047] Furthermore, mobile nodes utilize the 2.4GHz frequency band with a communication rate ≥150Mbps. Fixed nodes are deployed every 50m along the top and side walls of the coal bunker, forming a mesh topology to ensure no signal blind spots. During TDMA time slot allocation, each node's time slot length is set to 10ms to avoid data transmission conflicts. Critical instructions such as task allocation, path adjustment, and emergency signals employ CRC checksums. If the receiver fails the checksum, the transmitter retransmits within 100ms, with a retransmission count ≤3 times, ensuring reliable instruction transmission. Communication latency is ≤80ms, meeting the system's real-time requirements.

[0048] Please see Figure 6This invention provides a robot system for cleaning coal bunkers. The cleaning robot receives path instructions from a collaborative control module, travels along a planned path, and performs cleaning operations. Simultaneously, it provides real-time feedback on its own position, travel status, and cleaning operation status to the collaborative control module. The cleaning robot includes a walking mechanism, a cleaning mechanism, a positioning module, and an emergency handling module. The walking mechanism adopts a tracked structure, the cleaning mechanism consists of a high-pressure water gun and a rotating brush, the positioning module uses a laser SLAM radar and an inertial navigation module, and achieves positioning through the ROS cartographer algorithm, and the emergency handling module is developed based on an Arduino microcontroller. When an obstacle is detected or the remaining battery power is ≤10%, an emergency signal is sent to the collaborative control module.

[0049] Furthermore, the walking mechanism features anti-slip grooves on the track surface to adapt to the uneven ground inside the coal bunker. Steering is achieved via a differential motor, with a minimum turning radius of ≤0.5m, meeting the needs of working in confined spaces. The cleaning mechanism utilizes a fan-shaped nozzle design with a 60° spray angle to improve cleaning coverage. The rotating brush bristles are made of wear-resistant nylon, capable of cleaning stubborn coal particles ≤10cm thick. The contact pressure between the brush and the bunker wall is adaptively adjusted via a spring to prevent damage to the bunker wall.

[0050] Positioning module: The laser SLAM radar has a scanning frequency of 10Hz, an effective detection range of 0.1~80m, and an accuracy of ±30mm. The inertial navigation module has an acceleration measurement range of ±16g and an angular velocity measurement range of ±2000dps. The two are fused through the ROS cartographer algorithm, with a positioning error of ≤0.5m. Even in environments with a dust concentration of ≥50mg / m³, it can still maintain stable positioning.

[0051] Emergency Response Module: This module collects obstacle distance data from the LiDAR in real time. When the distance is ≤0.5m, it triggers an obstacle avoidance procedure. The battery sensor monitors the remaining battery power in real time; when it is ≤10%, it sends a low battery signal to the collaborative control module. If no support command is received within one minute, the robot automatically returns to the charging dock along its original path. Furthermore, the emergency response module also sends a fault signal when the cleaning mechanism experiences abnormal water pressure or the brushes become stuck, ensuring the robot's safety.

[0052] The working principle and usage process of this invention are as follows: all cleaning robots and modules are started. The environmental perception module begins to collect data on the distribution of sticky coal and the location of obstacles in the coal bunker and sends it to the collaborative control module. The cleaning robots then report their initial position, remaining power, and cleaning mechanism status to the collaborative control module.

[0053] The global task decomposition unit of the collaborative control module divides the sub-task areas based on the 3D model of the coal bunker and environmental perception data. It uses the Hungarian algorithm to assign the sub-tasks to each cleaning robot. The local path planning unit generates dynamic guide lines based on the improved Frenet algorithm and generates multiple candidate paths by combining the cubic spline interpolation algorithm. The energy consumption optimization module calculates the energy consumption value of each candidate path and selects the optimal path. The conflict detection sub-unit detects whether there are conflicts in the paths of multiple robots. If there are, the paths are adjusted.

[0054] Once the optimal path is determined, the collaborative control module sends the optimal path instruction to the cleaning robot. The cleaning robot moves along the planned path, starts the cleaning mechanism to perform cleaning operations, and at the same time provides real-time feedback on the movement status and cleaning status to the collaborative control module.

[0055] After the cleaning robot completes cleaning a sub-task area, the environmental perception module collects images and infrared data of that area and sends them to the cleaning effect evaluation module. The cleaning effect evaluation module calculates the coal residue rate. If the rate meets the standard, it notifies the collaborative control module; otherwise, it generates a secondary cleaning instruction, which the collaborative control module then dispatches the cleaning robot to perform a secondary cleaning. After all sub-task areas have been cleaned to the required standard, the collaborative control module summarizes the cleaning data and sends it to the monitoring terminal. The cleaning robot then returns to its designated docking point, shuts down its cleaning mechanism, and the system completes the cleaning task and automatically shuts down.

[0056] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to the above embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A robotic system for cleaning coal bunkers, characterized in that, It includes multiple cleaning robots, an environmental perception module, a collaborative control module, a cleaning effect evaluation module, and an energy consumption optimization module. The environmental perception module is mounted on the cleaning robot and uses cameras, infrared imagers, lidar, and ultrasonic sensors to collect real-time information on the distribution of sticky coal in the coal bunker, the location of obstacles, and images and infrared data of the cleaned area. The collaborative control module includes a global task decomposition unit and a local path planning unit. The global task decomposition unit is used to obtain the overall cleaning requirements of the coal bunker and the coal distribution information collected by the environmental perception module, divide the coal bunker cleaning area into multiple sub-task areas, and assign the sub-tasks to each cleaning robot. The local path planning unit is based on the improved Frenet algorithm and cubic spline interpolation algorithm, combined with obstacle location information, to plan the path from the current position to each cleaning point in the sub-task area for each cleaning robot. The local path planning unit also has a conflict detection sub-unit, which is used to detect in real time whether there is a cross or overlapping conflict between the paths of multiple robots. If a conflict exists, the path or travel speed of the relevant robots is adjusted. The cleaning effect evaluation module receives images and infrared data of the cleaned area collected by the environmental perception module, and judges the cleaning effect of the cleaned area through image recognition technology. If the effect is not met, a secondary cleaning instruction is generated and sent to the collaborative control module. The energy consumption optimization module is used to establish an energy consumption model of the cleaning robot, calculate the energy consumption value of each path through the energy consumption model, select the path with the lowest energy consumption as the optimal path, and feed it back to the local path planning unit. The cleaning robot receives the path instructions sent by the collaborative control module, moves along the planned path and performs cleaning operations, and simultaneously feeds back its own position, movement status and cleaning operation status to the collaborative control module in real time.

2. The robot system for cleaning coal bunkers according to claim 1, characterized in that, The improvement of the Frenet algorithm lies in its automatic generation of dynamic guide lines, taking into account the unstructured environment of coal bunkers without fixed guide lines. The generation process of the dynamic guide lines is as follows: First, based on the coordinates of the cleaning points within the sub-task area, a clustering algorithm is used to group the cleaning points into multiple cleaning point clusters. Then, the center coordinates of each cleaning point cluster are calculated, and the center coordinates are connected sequentially to form an initial guide line. Finally, based on the obstacle location information collected by the environmental perception module, the initial guide line is smoothly adjusted to avoid obstacles, resulting in a dynamic guide line.

3. The robot system for cleaning coal bunkers according to claim 2, characterized in that, The path generation process based on the improved Frenet algorithm includes: constructing a Frenet coordinate system on the basis of dynamic guide lines, converting the position of the cleaning robot into Frenet coordinates (s, d), where s is the arc length along the guide line and d is the lateral distance from the guide line. The lateral path and the longitudinal path are generated by fitting a fifth-order polynomial, where the lateral path is the ds curve and the longitudinal path is the vs curve, and v is the traveling speed of the cleaning robot.

4. The robot system for cleaning coal bunkers according to claim 1, characterized in that, The cubic spline interpolation algorithm is used to interpolate the path points generated by the local path planning unit. The specific process is as follows: the path points are divided into multiple intervals, and a cubic polynomial is constructed in each interval so that the first and second derivatives of the cubic polynomials of adjacent intervals are continuous at the connection point, thereby generating a smooth path without sharp turns.

5. The robot system for cleaning coal bunkers according to claim 1, characterized in that, The conflict detection process of the conflict detection subunit includes: acquiring the current position, speed and planned path of each cleaning robot in real time; predicting the position trajectory of each robot within a preset time period based on the planned path of each robot; and determining that there is a path conflict if the distance between the position trajectories of any two robots at the same time point is less than a preset safe distance.

6. The robot system for cleaning coal bunkers according to claim 1, characterized in that, The evaluation process of the cleaning effect evaluation module includes: converting the image of the cleaned area captured by the camera into a grayscale image; extracting the contour information of the area through an edge detection algorithm; calculating the contour area ratio, where the contour area ratio reflects the distribution range of coal residue; analyzing the temperature distribution of the cleaned area using infrared data collected by an infrared imager; obtaining the area of ​​the coal residue area by segmenting through a temperature threshold; and calculating the coal residue rate by combining the contour area ratio and the coal residue area. If the coal residue rate is greater than a preset threshold, the cleaning is deemed unsatisfactory, and a secondary cleaning instruction is generated.

7. The robot system for cleaning coal bunkers according to claim 1, characterized in that, The energy consumption optimization module fits the relationship between robot energy consumption and path parameters through experiments, and uses a multiple linear regression algorithm to fit the energy consumption model. The expression of the energy consumption model is E=K1*L+K2*C+K3*F, where E is the energy consumption value, L is the travel distance, C is the average curvature of the path, F is the robot load coefficient, and K1, K2 and K3 are preset weight coefficients obtained by fitting experimental data.

8. The robot system for cleaning coal bunkers according to claim 1, characterized in that, It also includes a communication module, which uses a wireless mesh network and consists of a mobile node mounted on the robot and a fixed node in the coal bunker. Communication time slots are allocated through TDMA technology, and a retransmission mechanism is used for critical instructions.

9. A robot system for cleaning coal bunkers according to claim 1, characterized in that, The cleaning robot includes a walking mechanism, a cleaning mechanism, a positioning module, and an emergency response module. The walking mechanism adopts a tracked structure, the cleaning mechanism consists of a high-pressure water gun and a rotating brush, the positioning module uses a laser SLAM radar and an inertial navigation module, and achieves positioning through the ROS cartographer algorithm, and the emergency response module is developed based on an Arduino microcontroller. When an obstacle is detected or the remaining power is ≤10%, an emergency signal is sent to the collaborative control module.

10. A robot system for cleaning coal bunkers according to claim 1, characterized in that, The collaborative control module uses an industrial-grade computer.

Images