Multi-machine cooperative efficient leveling system based on robot group

Abstract

The invention relates to the technical field of leveling control, and particularly discloses a multi-machine cooperative efficient leveling system based on a robot group, which comprises initial three-dimensional grain surface map construction, dynamic task initial allocation and operation path planning. The bottleneck of single-machine operation is broken through in a multi-machine cooperation mode, dynamic task allocation and path planning enable the robot to accurately adapt to grain surface differences, intensive operation in a high convex area and efficient traversal in a flat area, and a fine leveling process is combined, so that the time consumed for leveling the granary is remarkably shortened, the efficiency is improved by several times compared with that of manual work, the operation period is greatly shortened, and the rapid operation requirement of a large-scale granary is met; high-precision point cloud acquisition of initial mapping, laser radar dynamic scanning in operation, IMU attitude stable control and error strict control of fine leveling links are realized, grain surface flatness is guaranteed in the whole process, highly convex areas are accurately identified and processed, obstacles are avoided, a foundation is built for subsequent links of grain storage, and the problems of low efficiency and long period of traditional leveling are solved.

Description

Technical Field

[0001] The present invention relates to the field of liquidation control technology, and in particular to a multi-machine collaborative and efficient liquidation system based on a robot swarm. Background Art

[0002] After the grain is stored in the warehouse, it needs to be leveled in a timely manner to ensure subsequent standardized management and the application of new grain storage technologies. In the grain storage operation system, leveling operations are a key link in ensuring the storage stability of the grain pile and laying the foundation for subsequent ventilation, temperature measurement and other processes.

[0003] The traditional closing mode relies on manual operation and has significant drawbacks: on the one hand, manual closing is labor-intensive and extremely inefficient. Faced with large-scale grain warehouses, the operation cycle is long and it is difficult to ensure quality uniformity; on the other hand, single-machine automated closing equipment is limited by rigid path planning and lagging environmental perception. When faced with dynamic changes in the grain surface, it is unable to flexibly adjust the operation strategy, and is prone to problems such as missed closing and repeated operations. In addition, there is a lack of effective scheduling mechanism for multi-machine collaboration, and path conflicts are prone to occur between robots, further reducing operational efficiency. Summary of the Invention

[0004] In view of the above situation, in order to overcome the defects of the existing technology, the present invention provides a multi-machine collaborative and efficient closing system based on a robot swarm to solve the above-mentioned technical defects.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a multi-machine collaborative efficient liquidation system based on a robot swarm, comprising:

[0006] Step 1: System initialization: Utilizing the command and dispatch system deployed on the silo roof, which integrates a lidar, starlight camera, and central processing unit, it performs initialization operations after startup, controls the start-up of each spiral-driven warehouse-closing robot, initializes the Unity3D digital twin engine, loads the pre-built silo 3D model, and establishes a data interface with the command and dispatch system.

[0007] Step 2: Constructing an initial 3D grain surface map: The command and dispatch system uses a LiDAR to conduct a full-scale scan of the grain surface in the warehouse, collecting grain surface point cloud data to construct an initial 3D grain surface map and assess its surface flatness. Simultaneously, the terrain prediction algorithm uses a bulk grain flow dynamics model, combining grain pile shape parameters and grain particle characteristics to model grain surface morphology changes in real time.

[0008] Step 3: Initial dynamic task allocation: This involves the dynamic task scheduling module of the system. Based on the grain conditions and robot status in different areas of the warehouse, it calculates the task urgency coefficient for each area. Based on the grain surface point cloud data collected by the LiDAR, it calculates the point cloud data variance to quantify the regional terrain complexity and generate a terrain complexity factor. Finally, to achieve the optimal match between the robot and the work area, an improved Hungarian algorithm is used to construct a cost function C(i, j) as the evaluation basis for task allocation.

[0009] Step 4: Operation path planning: When the robot moves from the initial position to the target operation area, it uses A * The robot uses the Dijkstra algorithm or the Dijkstra algorithm to plan the path in combination with the three-dimensional grain surface map. After entering the operation area, the robot performs the closing operation along a zigzag or zigzag path.

[0010] Step 5: Execute the leveling operation: Each leveling robot drives along the path planned by the command and dispatch system to perform grain leveling operations. During the operation, the laser radar dynamically scans the grain surface in the working area. Each robot in the warehouse broadcasts its own position coordinates, driving speed vector, and working area boundary information through the wireless communication module at a preset period. After receiving the multi-machine status data, the command and dispatch system uses a collision prediction algorithm to predict collision risks based on the Euclidean distance and velocity vector angle between the robots.

[0011] Step 6: Dynamic task reallocation trigger: The monitoring and dispatching system continuously monitors the task urgency coefficient of each operation area. After responding to the trigger signal, the dynamic task scheduling module dispatches robots that are idle or underloaded from adjacent operation areas to support areas with high task urgency or faulty robots.

[0012] Step 7: Intelligent Monitoring and Exception Handling: The intelligent monitoring platform uses the Unity3D engine to build a digital twin model of the warehouse, rendering the grain surface shape, robot operation trajectory, and obstacle location information in real time. The platform also deploys a machine learning prediction model to analyze robot operation data in real time and predict faults in advance.

[0013] Step 8. Acceptance of the warehouse closing effect: The control system uses the point cloud data of the entire warehouse grain surface collected by the lidar and uses the three-dimensional data analysis algorithm to calculate the flatness of the entire warehouse grain surface. The intelligent monitoring platform integrates the data of the entire warehouse closing operation process and generates a standardized operation report.

[0014] Furthermore, in step one, the data interface with the command and dispatch system is used to receive point cloud data and robot status information transmitted by the command and dispatch system in real time, laying a data foundation for the monitoring platform to visualize the operation scene in the warehouse and the operation status of the robot. The point cloud data reflects the three-dimensional shape of the grain surface, and the robot status information includes but is not limited to position, power, and operation mode.

[0015] Furthermore, in step 2, the method for evaluating the flatness of the grain surface is: based on the point cloud data after denoising, the height variance of the grain surface is calculated to determine the initial grain surface flatness error, the flatness error threshold is set to ±5 cm, and the high convex areas exceeding the threshold are identified through the region growing clustering algorithm of the three-dimensional point cloud.

[0016] Furthermore, in the step three, the task urgency coefficient of each area is calculated by multiplying the difference between the actual height of the grain pile in each area and the target leveling height, the dust concentration in each area, and the remaining power of each robot by the corresponding weight coefficient, and finally obtaining the task urgency coefficient of each area; wherein, the dust concentration in each area is a value determined by normalizing the sensor collection value with the maximum dust concentration in the warehouse.

[0017] Furthermore, in step six, while the task redistribution is being executed, the command system controls the laser radar to perform high-frequency scanning of the affected work area, collect the latest grain surface point cloud data, and update the three-dimensional grain surface map.

[0018] Furthermore, it also includes:

[0019] Step 9. System reset and data archiving: After the closing operation is completed, each robot receives instructions from the command system, autonomously plans its return path, and drives to the designated charging area. The intelligent monitoring platform standardizes the closing operation data and stores it on the cloud server via an encrypted transmission protocol.

[0020] Beneficial effects of the present invention:

[0021] 1. The multi-machine collaborative mode breaks the bottleneck of single-machine operation. Dynamic task allocation and path planning allow robots to accurately adapt to differences in grain surface, perform intensive operations in high-convex areas, and efficiently traverse flat areas. Combined with the precise leveling process, the time spent on leveling is significantly shortened, and efficiency is several times higher than manual operation. The operation cycle is greatly compressed, adapting to the rapid operation needs of large-scale grain warehouses, and solving the problems of low efficiency and long cycle of traditional leveling.

[0022] 2. From high-precision point cloud collection for initial mapping, to dynamic laser radar scanning and IMU attitude stabilization control during operation, to strict error control in the leveling process, the entire process ensures the flatness of the grain surface, accurately identifies and processes high-convex areas, avoids obstacles, lays a solid foundation for subsequent grain storage links, and solves the problems of low efficiency and long cycle time in traditional leveling.

[0023] 3. Digital twin monitoring enables visualization and remote intervention within the warehouse. Fault prediction models provide early warning and preventive maintenance. Data archiving drives algorithm iteration, giving the system self-optimization capabilities. Multi-machine obstacle avoidance and collaboration prevent operational conflicts. Dynamic reallocation responds to changes in grain level and equipment failures, enabling the system to intelligently adapt to complex storage environments. This addresses the lack of intelligence and ability to respond to emergencies in traditional warehouse closing. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The present invention will be further described below with reference to the accompanying drawings.

[0025] Figure 1 This is a schematic diagram of the principle architecture of a multi-machine collaborative efficient liquidation system based on a robot swarm according to an embodiment of the present invention;

[0026] Figure 2 The present invention is a flowchart of the steps for implementing a multi-machine collaborative efficient liquidation system based on a robot swarm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] Below, the exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0028] Example 1

[0029] See also Figure 1 and Figure 2 As shown in the figure, the multi-machine collaborative efficient liquidation system based on robot swarm includes:

[0030] Step 1: System initialization: Utilize the command and dispatch system deployed on the silo roof, which integrates the lidar, starlight camera, and central processing unit. After startup, perform initialization operations to control the startup of each spiral-driven warehouse-closing robot, initialize the Unity3D digital twin engine, load the pre-built silo 3D model, and establish a data interface with the command and dispatch system.

[0031] Specifically, the laser radar model equipped with the control system is XYZ-100, with a scanning angle of 360° and a vertical resolution of 0.5°. It actively scans the grain surface at a frequency of 10Hz to obtain grain surface point cloud data. The point cloud data contains the three-dimensional coordinates (x, y, z) information of each scanning point, where the z value represents the height of the grain surface. The scanning coverage area is a circular area with a diameter of 20m, and the scanning parameters can be adjusted according to the actual size of the granary. Each warehouse-leveling robot is equipped with an IMU attitude sensor, model XYZ-IMU200, with an accuracy of ±0.1°. During the robot's operation, it collects real-time motion acceleration data of grain particles generated by the spiral wheel operation, including acceleration components in the x, y, and z directions. The sampling frequency is synchronized with the laser radar and is 10Hz.

[0032] It should be noted that by self-checking the working status of the lidar scanning frequency, communication module and dynamic task scheduling module, it is ensured that each functional module can operate normally, providing basic support for subsequent grain surface scanning, task scheduling and other operations; at the same time, the spiral wheel drive system, IMU attitude sensor, and multi-sensor fusion obstacle avoidance module are self-checked in turn to verify the integrity of hardware equipment and functions. Among them, the multi-sensor fusion obstacle avoidance module covers lidar, millimeter-wave radar, and infrared sensor; a network connection is established with the command and control system through the WiFi6 communication protocol to realize the establishment of a data interaction channel between the robot and the command and control system, ensuring the reception of subsequent operation instructions and the return of status information.

[0033] Specifically, the data interface with the command and dispatch system is used to receive point cloud data and robot status information transmitted by the command and dispatch system in real time, laying a data foundation for the monitoring platform to visualize the operation scene in the warehouse and the operation status of the robot. The point cloud data reflects the three-dimensional shape of the grain surface, and the robot status information includes but is not limited to position, power, and operation mode.

[0034] Step 2: Construct an initial three-dimensional grain surface map: The laser radar carried by the command and dispatch system is used to conduct a full-scale scanning operation on the grain surface in the warehouse to collect grain surface point cloud data. The collected raw point cloud data is denoised by the Kalman filter algorithm. After removing the noise interference, an initial three-dimensional grain surface map is constructed and the flatness of the grain surface is evaluated. At the same time, the bulk grain flow dynamics model is called through the terrain prediction algorithm unit. Combined with the shape parameters of the grain pile and the characteristics of the grain particles, the morphological changes of the grain surface are modeled in real time.

[0035] Specifically, when making basic assumptions in the bulk grain flow dynamics model, the grain particles are regarded as discrete rigid spheres, and the wear and deformation of the particles after long-term operation are ignored. The silo wall is assumed to be a rigid boundary, and the influence of the silo wall deformation on the grain flow is not considered. The adhesion between grain particles is ignored, and only the contact force and friction between particles are considered.

[0036] The method for evaluating grain surface flatness is as follows: based on the denoised point cloud data, the height variance of the grain surface is calculated to determine the initial grain surface flatness error. The flatness error threshold is set to ±5 cm, and the high-convex areas exceeding this threshold are identified through the region growing clustering algorithm of the three-dimensional point cloud. Specifically, grain piles with a height greater than 1.2 m are identified to clarify the location and characteristics of the initial uneven areas on the grain surface, providing a basis for the subsequent flattening task planning.

[0037] The specific manifestations of predicting the possible morphological changes of the grain surface within 0.5 hours are: for areas with slopes greater than 25°, analyzing the sliding trend of bulk grain, predicting the dynamic changes of the grain surface in advance, assisting in subsequent operation path planning and task adjustment, and improving the adaptability and foresight of closing operations.

[0038] It should be noted that during the map construction process, the location and height of the grain pile are accurately marked, and the coordinates of obstacles such as temperature measuring cables are identified and marked, clearly presenting the initial terrain of the grain surface in the warehouse and the distribution of obstacles; the shape parameters of the grain pile include but are not limited to the volume and bottom shape of the grain pile, and the characteristics of the grain particles include but are not limited to particle size, stacking density and internal friction angle.

[0039] Step 3: Initial dynamic task allocation: This refers to the dynamic task scheduling module of the system. It calculates the task urgency coefficient of each area based on the grain conditions and robot status in different areas of the warehouse. Based on the grain surface point cloud data collected by the lidar, it calculates the point cloud data variance to quantify the regional terrain complexity and generate a terrain complexity factor. Finally, to achieve the optimal match between the robot and the working area, the improved Hungarian algorithm is used to construct the cost function C(i,j) as the evaluation basis for task allocation.

[0040] The task urgency coefficient of each area is calculated as follows: the difference between the actual height of the grain pile in each area and the target leveling height, the dust concentration in each area, and the remaining power of each robot are multiplied by the corresponding weight coefficient, and finally the task urgency coefficient of each area is obtained; among which, the dust concentration in each area is determined by normalizing the sensor collection value with the maximum dust concentration in the warehouse.

[0041] It should be noted that the value range of the terrain complexity factor is set to 0.5-1.2. Specifically, in areas where the grain pile surface is undulating and the point cloud data variance is large, such as the edge of the grain pile and the height mutation, the terrain complexity factor is close to 1.2; in areas where the grain surface is relatively flat and the point cloud data variance is small, the terrain complexity factor is close to 0.5. The terrain complexity factor is used in the cost function of subsequent task allocation, reflecting the impact of terrain on the difficulty and efficiency of robot operations.

[0042] It should be noted that the formula of the cost function C(i,j) is defined as C(i,j)=D(i,j)×α j +T(i,j)×β, where D(i,j) is the straight-line distance from the i-th robot to the j-th working area, α j It is expressed as the terrain complexity factor of the j-th operating area, T(i,j) is expressed as the estimated operating time of the i-th robot in the j-th operating area, which is pre-calculated based on factors such as the grain pile volume and the robot's operating efficiency, and β is expressed as the task priority coefficient, which is dynamically adjusted according to the task urgency coefficient. The larger the task urgency coefficient, the higher the task priority coefficient.

[0043] Specifically, when scheduling six robots to work in the three high-convex areas identified in the warehouse, the algorithm uses a cost function as the basis, comprehensively considering distance, terrain, operation time and task priority, and divides the six robots into three groups, with two robots in each group responsible for one high-convex area, ensuring balanced robot load and improving overall closing operation efficiency.

[0044] Step 4: Operation path planning: When the robot moves from the initial position to the target operation area, it uses A * The robot uses the Dijkstra algorithm or the Dijkstra algorithm to plan the path in combination with the three-dimensional grain surface map. After entering the operation area, it performs the closing operation according to the zigzag or zigzag path.

[0045] It should be noted that during the planning process, obstacles in the warehouse are identified and avoided, and a 0.8m obstacle buffer is set to ensure that the minimum distance between the robot's driving path and obstacles is not less than the buffer, avoiding collision risks and ensuring operational safety.

[0046] The control system dynamically adjusts the traversal density according to the height distribution of the grain surface: in high-convex areas, the traversal interval is reduced to 0.5m, increasing the frequency of operation coverage to ensure the leveling effect of the grain surface in high-convex areas; in flat areas, the traversal interval is set to 1m, while ensuring the quality of the flattening, optimizing the operation efficiency and reducing invalid operation paths.

[0047] Furthermore, the terrain prediction unit is based on the bulk grain flow dynamics model and real-time grain surface data. By adopting an improved discrete element model to simulate the bulk grain flow, combined with the interactive mechanics of the spiral drive wheel, it generates a grain surface change trend map for the next hour in advance. If it is predicted that the grain pile in a certain area will undergo morphological changes such as slippage, the command and control system will automatically trigger the path pre-adjustment mechanism, re-plan the robot's operation path, avoid potential dangerous areas, ensure operation continuity and robot safety, and improve the system's adaptability to dynamic grain surface environments.

[0048] Step 5. Execution of the leveling operation: Each leveling robot travels according to the path planned by the command and dispatch system. Its spiral wheel drive system integrates driving and leveling functions, and can carry out grain leveling operations simultaneously. During the operation, the laser radar dynamically scans the grain surface in the working area. Each robot in the warehouse broadcasts its own position coordinates, driving speed vector and working area boundary information through the wireless communication module at a preset period. After receiving the multi-machine status data, the command and dispatch system adopts a collision prediction algorithm to predict the collision risk within time t based on the Euclidean distance and velocity vector angle between the robots. When the rough leveling operation is completed, that is, the grain surface flatness error is detected to be less than 10cm, the fine leveling operation process is triggered, and the robot automatically mounts the fine leveling mechanism to carry out secondary fine leveling operations on the grain surface.

[0049] It should be noted that the scanning frequency of the laser radar for dynamic scanning of the grain surface in the working area is adaptively adjusted within the range of 5-15Hz according to the complexity of the grain surface, and the grain surface morphology data is obtained in real time. At the same time, the robot continuously monitors its own posture through the IMU attitude sensor. When it detects that the slope of the grain surface in the working area is greater than 45°, it automatically switches to low-speed anti-skid mode. By adjusting the speed difference of the left and right spiral wheels, the robot's posture is ensured to be stable when working on inclined grain surfaces, avoiding risks such as slipping and overturning, and ensuring the continuity and quality of the closing operation.

[0050] Specifically, if it is determined that there is a possibility of a collision, the command and control system sends instructions to the relevant robots, causing them to actively adjust their operating parameters, such as slowing down by 30% to reduce kinetic energy, or turning 15° to change the driving path. By dynamically adjusting to avoid path conflicts, it can achieve safe obstacle avoidance during collaborative operation of multiple machines and ensure the smoothness of parallel operation of multiple robots.

[0051] It should be noted that during the precision leveling operation, the lidar and IMU sensor work together to monitor the flatness of the grain surface in real time, ensuring that the final grain surface flatness error is ≤5cm, meeting the high standards for grain surface flatness in grain storage, and laying a good foundation for subsequent grain storage, ventilation, temperature measurement and other operations.

[0052] Step 6. Dynamic task reallocation trigger: The monitoring and command system continuously monitors the task urgency coefficient of each operation area. After the dynamic task scheduling module responds to the trigger signal, it dispatches robots in idle or low-load states from adjacent operation areas to support areas with high task urgency or faulty robots. While task reallocation is being executed, the command system controls the lidar to perform high-frequency scanning of the affected operation areas, collect the latest grain surface point cloud data, and update the three-dimensional grain surface map.

[0053] It should be noted that when the K value change rate in a certain operating area exceeds 20% or a robot malfunctions, the dynamic task reallocation mechanism is triggered to adjust the operating tasks in a timely manner to ensure the efficiency and progress of the closing operation.

[0054] Specifically, the improved Hungarian algorithm is re-called to calculate the task allocation cost function C(i,j), and the idle robots are reasonably allocated to the working area with the highest K value, reducing the loss of working efficiency caused by task reallocation and maintaining the overall efficient advancement of the closing operation in the warehouse.

[0055] It should be noted that the terrain prediction unit recalculates the grain surface change trend based on the updated grain surface data, adjusts the subsequent operation path planning, and makes the robot's operation path adapt to the new grain surface shape to ensure operation accuracy and efficiency.

[0056] Step 7. Intelligent Monitoring and Exception Handling: The intelligent monitoring platform uses the Unity3D engine to build a digital twin model of the warehouse, rendering the grain surface shape, robot operation trajectory, and obstacle location information in real time. By deploying a machine learning prediction model on the platform, it conducts real-time analysis of robot operation data and predicts faults in advance.

[0057] Furthermore, if an emergency occurs in the warehouse, such as a robot being buried in a grain pile, the operator can quickly switch to remote control mode through the intelligent monitoring platform. Based on 1.4G / 2.4G penetrating communication technology, the operator can directly send control instructions to the trapped robot to control the robot to perform escape actions. At the same time, the command and dispatch system automatically suspends the operation of other robots in the area to avoid secondary accidents caused by the operation of other robots, ensure the safety of personnel and equipment, and enhance the system's ability to respond to emergencies.

[0058] Specifically, the platform supports remote access of VR devices. Operators can use VR devices to immersively view the working scenes in the warehouse and manually mark key areas of concern (such as high dust concentration areas, dust concentration> 100mg / m 3 After receiving the tag information, the command and dispatch system automatically adjusts the operating parameters of the robots in the corresponding area, such as reducing the robot's driving speed by 50% to reduce dust during operation, thereby ensuring a safe working environment and the life of the robot equipment.

[0059] It should be noted that by building a fault feature library, potential robot failures can be predicted 72 hours in advance. When a failure is predicted, the platform automatically generates a maintenance work order and pushes it to the management personnel, clarifying the fault type, impact level and recommended maintenance time, to achieve preventive maintenance and reduce the impact of robot failure and downtime on liquidation operations.

[0060] Step 8. Acceptance of the warehouse closing effect: The control system uses the point cloud data of the entire warehouse grain surface collected by the lidar and uses the three-dimensional data analysis algorithm to calculate the flatness of the entire warehouse grain surface. The intelligent monitoring platform integrates the data of the entire warehouse closing operation process and generates a standardized operation report.

[0061] Specifically, the system iterates through the height values ​​of each coordinate point in the point cloud data and calculates the deviation between the grain surface height and the target leveling height. When the flatness error of the entire grain warehouse surface is ≤5cm and there are no grain piles with a height greater than 0.5m, the leveling operation is deemed to have met the quality standard and is complete. If the flatness error exceeds the threshold or there are still high and convex grain piles, the command and dispatch system automatically triggers the task reallocation mechanism, returning to the dynamic task allocation stage of the operation process, redividing the work area and dispatching robots to level the substandard areas until the leveling quality requirements are met.

[0062] It's important to note that the report covers: Operation Time: This records the total duration from system startup to operation completion, accurate to the second. For example, a 5,000-ton grain silo closing operation took 8.7 hours, clearly demonstrating operational efficiency; Robot Energy Consumption: This measures the power or fuel consumption of each robot during operation, analyzes the energy consumption distribution of multiple robots, and provides data support for energy optimization; Grain Surface Flatness Curve: This plots a curve with time as the horizontal axis and grain surface flatness error as the vertical axis, visually demonstrating the improvement trend in grain surface flatness during the closing operation; and Fault Records: This records the type, occurrence time, duration, and troubleshooting measures of faults encountered during robot operations, providing a reference for equipment maintenance and fault prevention. The operation report allows for a quantitative assessment of the efficiency and quality of the closing operation.

[0063] Step 9. System reset and data archiving: After the closing operation is completed, each robot receives instructions from the command system, autonomously plans its return path, and drives to the designated area. The intelligent monitoring platform standardizes the closing operation data and stores it on the cloud server via an encrypted transmission protocol.

[0064] Specifically, after each robot arrives at the charging area, it automatically docks with the charging equipment and enters the charging or standby state. At the same time, the command and control system controls the lidar to adjust the working mode, reducing the scanning frequency to 1Hz during non-operating periods, reducing equipment energy consumption, reducing unnecessary scanning of the warehouse environment, extending the service life of the lidar, and maintaining basic monitoring capabilities of the warehouse environment to quickly respond to the next operation needs.

[0065] It should be noted that the data of the warehouse closing operation include but are not limited to the initial and final point cloud maps of the grain surface, robot operation trajectory data, fault logs, etc. The warehouse closing operation data is used as a historical operation sample for subsequent operation strategy optimization: Task allocation algorithm optimization: Analyze the multi-machine collaborative operation data, adjust the weights of the terrain complexity factor and the task priority coefficient in the dynamic task allocation algorithm, for example, according to the operation data of different grain types and warehouse types, optimize the cost function C(i,j) to improve the rationality of task allocation and operation efficiency; Fault prediction model optimization: Based on the fault log data, enrich the training samples of the machine learning fault prediction model, optimize the model parameters, improve the accuracy and advance time of fault prediction, and achieve more accurate preventive maintenance; Path planning algorithm optimization: Combine the robot operation trajectory data to analyze the path planning effect under different grain surface environments, optimize A * The heuristic function and weight setting of the algorithm or Dijkstra algorithm can improve the efficiency of path planning and obstacle avoidance capabilities.

[0066] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The size of the coefficient is a specific value obtained by quantifying each parameter. Regarding the size of the coefficient, as long as it does not affect the proportional relationship between the parameter and the quantified value, it is acceptable.

[0067] In addition, it will be understood by those skilled in the art that various aspects of the present invention may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of the present invention may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of the present invention may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0068] Unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology and should not be interpreted in an idealized or highly formal sense, unless expressly defined as such herein.

[0069] The above is an illustration of the present invention and should not be considered as limiting thereof. Although several exemplary embodiments of the present invention have been described, it will be readily understood by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present invention. Therefore, all such modifications are intended to be included within the scope of the present invention as defined by the claims. It should be understood that the above is an illustration of the present invention and should not be considered as being limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present invention is defined by the claims and their equivalents.

Claims

1. A multi-machine collaborative and efficient liquidation system based on a robot swarm, characterized by: include: Step 1: System initialization: Utilizing the command and dispatch system deployed on the silo roof, which integrates a lidar, starlight camera, and central processing unit, it performs initialization operations after startup, controls the start-up of each spiral-driven warehouse-closing robot, initializes the Unity3D digital twin engine, loads the pre-built silo 3D model, and establishes a data interface with the command and dispatch system. Step 2: Constructing an initial 3D grain surface map: The command and dispatch system uses a LiDAR to conduct a full-scale scan of the grain surface in the warehouse, collecting grain surface point cloud data to construct an initial 3D grain surface map and assess its surface flatness. Simultaneously, the terrain prediction algorithm uses a bulk grain flow dynamics model, combining grain pile shape parameters and grain particle characteristics to model grain surface morphology changes in real time. Step 3: Initial dynamic task allocation: This involves the dynamic task scheduling module of the system. Based on the grain conditions and robot status in different areas of the warehouse, it calculates the task urgency coefficient for each area. Based on the grain surface point cloud data collected by the LiDAR, it calculates the point cloud data variance to quantify the regional terrain complexity and generate a terrain complexity factor. Using the improved Hungarian algorithm, it constructs a cost function C(i, j) as the evaluation basis for task allocation. Step 4: Operation path planning: When the robot moves from the initial position to the target operation area, it uses A * The robot uses the Dijkstra algorithm or the Dijkstra algorithm to plan the path in combination with the three-dimensional grain surface map. After entering the operation area, the robot performs the closing operation along a zigzag or zigzag path. Step 5: Execute the leveling operation: Each leveling robot drives along the path planned by the command and dispatch system to perform grain leveling operations. During the operation, the laser radar dynamically scans the grain surface in the working area. Each robot in the warehouse broadcasts its own position coordinates, driving speed vector, and working area boundary information through the wireless communication module at a preset period. After receiving the multi-machine status data, the command and dispatch system uses a collision prediction algorithm to predict collision risks based on the Euclidean distance and velocity vector angle between the robots. Step 6: Dynamic task reallocation trigger: The monitoring and dispatching system continuously monitors the task urgency coefficient of each operation area. After responding to the trigger signal, the dynamic task scheduling module dispatches robots that are idle or underloaded from adjacent operation areas to support areas with high task urgency or faulty robots. Step 7: Intelligent Monitoring and Exception Handling: The intelligent monitoring platform uses the Unity3D engine to build a digital twin model of the warehouse, rendering the grain surface shape, robot operation trajectory, and obstacle location information in real time. The platform also deploys a machine learning prediction model to analyze robot operation data in real time and predict faults in advance. Step 8. Acceptance of the warehouse closing effect: The control system uses the point cloud data of the entire warehouse grain surface collected by the lidar and uses the three-dimensional data analysis algorithm to calculate the flatness of the entire warehouse grain surface. The intelligent monitoring platform integrates the data of the entire warehouse closing operation process and generates a standardized operation report.

2. The multi-machine collaborative efficient liquidation system based on robot swarm according to claim 1 is characterized in that: In step one, the data interface with the command and dispatch system is used to receive point cloud data and robot status information transmitted by the command and dispatch system in real time, laying a data foundation for the monitoring platform to visualize the operation scene in the warehouse and the operation status of the robot. The point cloud data reflects the three-dimensional shape of the grain surface, and the robot status information includes but is not limited to position, power, and operation mode.

3. The multi-machine collaborative efficient liquidation system based on robot swarm according to claim 1 is characterized in that: In step 2, the method for evaluating the flatness of the grain surface is as follows: based on the point cloud data after denoising, the height variance of the grain surface is calculated to determine the initial grain surface flatness error, the flatness error threshold is set to ±5 cm, and the high convex areas exceeding the threshold are identified through the region growing clustering algorithm of the three-dimensional point cloud.

4. The multi-machine collaborative efficient liquidation system based on robot swarm according to claim 1 is characterized in that: In step three, the task urgency coefficient of each area is calculated as follows: the difference between the actual height of the grain pile in each area and the target leveling height, the dust concentration in each area, and the remaining power of each robot are multiplied by the corresponding weight coefficient, and finally the task urgency coefficient of each area is obtained; wherein, the dust concentration in each area is a value determined by normalizing the sensor collection value with the maximum dust concentration in the warehouse.

5. The multi-machine collaborative efficient liquidation system based on robot swarm according to claim 1 is characterized in that: In step six, while the task redistribution is being executed, the command and dispatch system controls the laser radar to perform high-frequency scanning of the affected work area, collect the latest grain surface point cloud data, and update the three-dimensional grain surface map.

6. The multi-machine collaborative efficient liquidation system based on robot swarm according to claim 1 is characterized in that: Also includes: Step 9. System reset and data archiving: After the closing operation is completed, each robot receives instructions from the command system, autonomously plans its return path, and drives to the designated charging area. The intelligent monitoring platform standardizes the closing operation data and stores it on the cloud server via an encrypted transmission protocol.

Images