An automatic retort loading method and system based on an AGV collaborative robot

Abstract

The application discloses an automatic steaming method and system based on AGV collaborative robots, relates to the technical field of liquor solid-state brewing equipment automation, and realizes real-time sensing of steaming pot opening states through point clouds and thermal imaging, accurate identification of steam dissipation areas, formation of operation constraints, perfect matching of core process requirements of steam detection and steaming, continuous operation beat guarantee through multi-AGV collaborative scheduling, realization of uniform spreading of fermented grains through ballistic solution and two-arm collaborative adjustment, online parameter updating and cyclic material supplementing, further improvement of spreading uniformity, and smooth steam passage in the steaming pot. The automatic steaming operation is completed in an automatic manner, labor intensity is greatly reduced, distillation conditions are stabilized, and the problems of complicated equipment, high land occupation and height requirement, many fault points and high maintenance cost of existing automatic steaming schemes are solved. Without the configuration of a complicated material feeding and conveying system, the fermented grain transfer link is greatly reduced, material leakage and sticking problems are avoided from the root, the production clean production demand is met, and the process flexibility of the production layout is improved.

Description

Technical Field

[0001] This application relates to the field of automation technology for solid-state brewing equipment for baijiu (Chinese liquor), specifically to an automatic steaming method and system based on AGV collaborative robots. Background Technology

[0002] In the solid-state brewing process of baijiu, loading the still is a crucial step connecting the loading of raw materials with distillation. Its core technological requirements are typically manifested in "exploring steam before loading and spreading evenly," meaning that under the constraints of steam distribution and the steam flow within the still, the mash is spread loosely, evenly, and thinly into the still to ensure smooth steam flow, stable distillation, and consistent liquor quality. With the upgrading of brewing production towards digitalization and intelligence, the automation, standardization, and controllability of the loading process have become important directions for improving efficiency, reducing labor intensity, and stabilizing product quality.

[0003] Existing automated loading solutions mostly employ a combination of fixed loading robots and supporting feeding systems to achieve continuous feeding. The feeding system typically consists of multiple devices such as vertical conveyors, chain conveyors or chain plate conveyors, tilting feeders, and feeding hoppers, to transport the mash from the stockpile to the still opening for loading. While this approach can improve feeding efficiency to some extent, it places high demands on site layout, space, and equipment integration, resulting in significant system complexity and high maintenance and modification costs. Furthermore, the long conveying process and multiple receiving and transfer stages are prone to leakage, material adhesion, and cleaning difficulties, posing challenges to clean production and stable operation.

[0004] Existing solutions generally suffer from numerous problems, including a large number of devices, multiple points of failure, high requirements for floor space and height, high investment and maintenance costs, and difficulty in avoiding material leakage and adhesion during conveying and transfer, as well as challenges in achieving cleanliness and process flexibility. Furthermore, traditional solutions lack real-time sensing of the steam state at the steamer opening. Therefore, how to achieve continuous and automatic steaming based on the steam state and thickness at the steamer opening without relying on a complex feeding and conveying system, while ensuring that the steaming process satisfies steam detection and avoidance as well as gentle and even spreading, is a pressing technical problem that needs to be solved in this field. Summary of the Invention

[0005] In order to solve the above-mentioned technical problems, this application proposes the following technical solution: In a first aspect, embodiments of this application provide an automatic steamer loading method based on an AGV collaborative robot, including: Establish and unify the world coordinate system, complete the unification and calibration of the coordinate systems of AGV, AGV dual-cooperative robotic arm, AGV hopper, sensor and still, and ensure that the AGV hopper discharge port and release point are mapped to the world coordinate system; Point cloud and thermal imaging were collected above the steamer opening, the steamer opening plane was fitted and the steamer coordinate system was refined, and the steam escape area and boundary were identified to form no-pour and slow-pour constraints. By combining the loading stage with the steam boundary to generate a target thickness field, and by incurring costs such as travel, material picking, and scattering time as well as the risk of steam overflow, multiple AGV task assignments are completed to ensure continuous cycle time. The AGV picks up and positions the material, generates the landing point according to the target thickness, solves the scattering parameters based on the trajectory, and adjusts the tilt angle and shaking of the scoop in coordination with both arms to achieve light and even spreading, and updates the shaking parameters online. After spreading, the thickness field and steam boundary are updated. Based on the height and coverage, the cycle is replenished or terminated, and the cleaning is completed and the crew is removed.

[0006] In one possible implementation, establishing and unifying the world coordinate system, and completing the unification and calibration of the coordinate systems of the AGV, the AGV dual-cooperative robotic arm, the AGV hopper, the sensor, and the still, ensuring that the AGV hopper outlet and release point are mapped to the world coordinate system, includes: Establish and unify the world coordinate system, AGV coordinate system, robotic arm base coordinate system, scoop coordinate system, sensor coordinate system, and still coordinate system; The installation relationship between the AGV, the dual collaborative robotic arm, and the robot's handheld scoop is calibrated to obtain a coordinate transformation chain for mapping points in the scoop coordinate system to the world coordinate system, thereby unifying the scoop discharge point, scoop release point, and their posture to the world coordinate system during operation.

[0007] In one possible implementation, the step of acquiring point cloud and thermal imaging above the steamer opening, fitting the steamer opening plane and refining the steamer coordinate system, identifying steam escape areas and boundaries to form no-dumping and slow-dumping constraints includes: The drive arm swings the visual detection device above the mouth of the target still, collects point cloud and / or distance information of the still mouth area and forms a point set; The point set is used to fit the steamer opening plane to obtain the unit normal vector and plane offset of the steamer opening plane, thereby determining the geometric center of the steamer opening and updating the steamer coordinate system. At the same time, a set of loading grid points is generated on the steamer opening plane. , in: Indicates to and Optimize to minimize the objective function in the subsequent steps. The unit normal vector of the steamer opening plane. For planar offset, The number of points involved in the fitting. For the first point cloud One point, This indicates that the following constraints are satisfied. Let be the norm of a vector; A thermal imaging temperature field of the steamer opening area is collected. Based on the thermal imaging temperature field, a steam significance score is calculated and the steam escape area is extracted. Then, a steam boundary curve is obtained to characterize the prohibited and slow-to-add regions within the steamer opening.

[0008] In one possible implementation, the thermal imaging temperature field of the steam steamer opening area is collected, a steam significance score is calculated based on the thermal imaging temperature field, and the steam escape region is extracted to obtain a steam boundary curve, which is used to characterize the prohibited and slow-release regions within the steam steamer opening, including: in: For steam confidence level, Represents the horizontal pixel coordinates of the image. Represents the pixel coordinates in the vertical direction of the image. As the weight of the temperature term, For the temperature of thermal imaging pixels, The average temperature. For temperature standard deviation, For gradient term weights, The temperature gradient is used to characterize changes at the vapor edge. The Sigmoid function normalizes the scores to a range of 0-1. This is a steam escape zone. This is the threshold for the steam region.

[0009] In one possible implementation, the process of generating a target thickness field by combining the loading stage with the steam boundary, and at the cost of travel, material handling, and spillage time, as well as the risk of steam spillage, to complete multi-AGV task assignment to ensure continuous cycle time includes: Based on the loading stage and the steam boundary curve, a target material thickness field is generated for the loading grid point set, causing the target material thickness to decrease near the steam boundary and vary with the loading stage, including: in: For the target thickness field, The first plane of the steamer opening Grid points, Based on the thickness of the base material, The thickness increment scale varies with the loading process. This is a function representing the steam loading stage, gradually increasing from the initial steam spill to the stable filling stage. This is the steaming stage. To reduce the feed intensity in the steam zone, the closer the distance, the greater the feed reduction. It is an exponential function. The distance from the grid point to the steam boundary. For steam boundary, For parameters affecting the range of influence; The system obtains the current location and task status of multiple AGVs, their corresponding estimated travel time, estimated material collection time, and estimated spilling time, and also obtains the steam overflow risk of the target still at the current moment. The estimated travel time, estimated material collection time, estimated spillage time, and steam overflow risk are constructed as task allocation costs and assigned for optimization to determine the target stills and arrival sequence of each AGV, thereby reducing the still loading time and suppressing steam overflow.

[0010] In one possible implementation, the step of constructing the estimated travel time, estimated material collection time, estimated spillage time, and steam overflow risk as task allocation costs and performing assignment optimization to determine the target still and arrival sequence of each AGV, thereby reducing the still loading idle time and suppressing steam overflow, includes: in: Here is the cost matrix. For AGV number index, For the target index, This represents the travel time of the AGV from its current location to the still. For the first The loading time of one AGV. For the first The time required for one-time scattering operation of a single steamer. As the spillover risk penalty coefficient, For the first Each steamer at a moment The steam spillover risk function, Indicates all possible assignment schemes In the process, we seek the solution that minimizes the objective function. This represents the total cost of the assignment plan. In a given assignment In the case of the first Each AGV corresponds to the assigned target The cost of time, For the number of AGVs, To give the first The target index assigned to the AGV.

[0011] In one possible implementation, the AGV picks up and positions the material, generates a landing point based on the target thickness, solves for the scattering parameters based on the ballistics, and coordinates the tilt angle and shaking of the scoop with both arms to achieve gentle and even spreading. The shaking parameters are updated online, including: Under the assigned and optimized task, the target AGV is driven to move to the mash pile and the dual cooperative robotic arms perform the material picking action to load the mash into the robot's handheld winnowing basket; The target AGV is then driven to move to the target still and complete the positioning and docking process. During the positioning and docking process, closed-loop trajectory tracking control is performed on the AGV to reduce docking errors. in: To track errors, This indicates the current pose and position of the AGV. For reference trajectory state, To control the output speed, As the reference velocity vector, This is the feedback gain matrix; Based on the target material thickness field and real-time material thickness information, the material shortage area is determined and the target landing point set is generated. The release point and target landing point of the winnowing basket are determined in the world coordinate system. Under the given flight time condition, the equivalent initial velocity vector of the material discharge is solved according to the ballistic relationship. Then, the dual cooperative robotic arms are controlled to coordinately adjust the tilt angle and back and forth shaking action of the winnowing basket to throw the mash into the spatial area corresponding to the target landing point set. Meanwhile, based on the relationship between the dispersion width and the scoop tilt angle, the amplitude of the shaking angle and the shaking frequency, a cost function is constructed with the uniformity of the landing point density distribution, the deviation between the actual thickness field and the target thickness field, and the penalty for the steam escape area as indicators. The scoop tilt angle, shaking angle amplitude, and shaking frequency are updated online based on the cost function to achieve light and even spreading and reduce the probability of covering areas where steam escapes.

[0012] In one possible implementation, the process of determining the under-material area and generating a target landing point set based on the target material thickness field and real-time material thickness information, determining the scoop release point and target landing point in the world coordinate system, and solving the equivalent initial velocity vector of the discharged material according to the ballistic relationship under a given flight time condition, thereby controlling the dual cooperative robotic arms to coordinately adjust the tilt angle and back-and-forth shaking motion of the scoop to scatter the mash into the spatial area corresponding to the target landing point set, includes: in: The position of the target landing point in the world coordinate system. This represents the position of the sieve's discharge port in the world coordinate system. The equivalent initial velocity vector for material discharge. For release duration, This is the acceleration due to gravity.

[0013] In one possible implementation, the cost function, constructed based on the relationship between the dispersion width and the scoop tilt angle, the shaking angle amplitude, and the shaking frequency, uses the uniformity of the landing point density distribution, the deviation between the actual thickness field and the target thickness field, and the penalty for the steam escape area as indicators, including: in; The width of the spread per unit of spray. The angle of inclination of the winnowing basket. The amplitude of the jitter angle. The frequency of the jitter. The calibrator coefficient for the tilt angle of the winnowing basket. This is the calibration coefficient for the jitter angle amplitude. This is the jitter frequency calibration coefficient. For learning rate, Cost function The gradient of the control parameters, The coefficient of variation of density distribution; the smaller the value, the more uniform the distribution. For the density distribution of landing points, The root mean square error of thickness deviation characterizes thinness and flatness. For the actual thickness of the stockpile, This is a penalty for areas where steam escapes; the closer to an area with strong steam coverage, the greater the penalty. The weights are the coefficients of variation of the density distribution. The root mean square error weight for thickness deviation The penalty weight for the steam escape region.

[0014] In one possible implementation, the process of updating the thickness field and vapor boundary after scattering, and determining whether to replenish or terminate the cycle based on height and coverage criteria, to complete the cleaning process and evacuate, includes: After each spill, the visual detection device is driven to collect point cloud and / or distance information of the steam pot opening area again to update the actual stockpile thickness field, and to collect thermal imaging temperature field again to update the steam escape area and steam boundary curve. The loading process is judged based on the loading completion criteria, which include at least: the maximum value of the actual material thickness field reaches a preset maximum material height threshold, and the proportion of grid points in the loading grid point set that satisfy the condition that the actual material thickness is not less than the target material thickness minus the allowable error band reaches a preset coverage threshold. in: This represents the maximum value of the actual thickness field. This indicates that for all grid points Take the maximum. The actual thickness of the material at the grid points The value at that location, The maximum allowable stacking height threshold, For the number of grid points, For indicator functions, For the target paving thickness at the grid points The value at that location, For thickness tolerance zone, The coverage threshold is the minimum required level. If the loading completion criterion is not met, update the set of material shortage area and target landing point and the process cannot continue; When the loading completion criterion is met, a clean-up and finishing action is performed to shake off / recycle the residual material in the hopper and drive the AGV to leave the target wine still station.

[0015] Secondly, embodiments of this application provide an automated steamer loading system based on an AGV collaborative robot, comprising: The coordinate unification and calibration module is used to establish and unify the world coordinate system, complete the unification and calibration of the coordinate systems of AGV, AGV dual cooperative robotic arm, AGV hopper, sensor and still, and ensure that the AGV hopper discharge port and release point are mapped to the world coordinate system. The steam pot mouth modeling and steam detection identification module is used to collect point cloud and thermal imaging above the steam pot mouth, fit the steam pot mouth plane and refine the steam pot coordinate system, identify the steam escape area and boundary to form no-pour and slow-pour constraints; The planning and scheduling module is used to generate a target thickness field by combining the loading stage and the steam boundary, and to complete the assignment of multiple AGV tasks at the cost of travel, material picking, scattering time and steam overflow risk to ensure continuous cycle time. The execution control module is used for AGV to pick up materials and position and stop, generate landing points according to the target thickness, solve the scattering parameters based on the ballistics, and coordinate the two arms to adjust the tilt angle and shaking of the scoop to achieve light and even spreading, and update the shaking parameters online. The status closing module is used to update the thickness field and steam boundary after spreading. It determines whether to replenish or stop the cycle based on the height and coverage criteria, completes the cleaning closing, and then withdraws.

[0016] In this embodiment, the state of the still opening is perceived in real time through point cloud and thermal imaging, accurately identifying the steam escaping area and forming operational constraints, perfectly matching the core process requirements of steam exploration and still loading. Multi-AGV collaborative scheduling ensures continuous operation, and ballistic solution combined with dual-arm coordinated adjustment achieves light and even spreading of the mash. Online parameter updates and cyclical replenishment further improve spreading uniformity, ensuring smooth steam flow within the still. The entire still loading operation is automated, significantly reducing manual labor intensity, stabilizing distillation conditions, effectively improving the consistency of liquor quality, and balancing still loading accuracy and production efficiency. It effectively solves the problems of existing automatic still loading solutions, such as complex equipment, high footprint and height requirements, numerous failure points, and high maintenance costs. It eliminates the need for complex feeding and conveying systems, significantly reducing mash transfer links, preventing leakage and sticking problems at the source, meeting the requirements of clean brewing production, and improving the process flexibility of the production layout. Attached Figure Description

[0017] Figure 1 A flowchart illustrating an automatic steamer loading method based on an AGV collaborative robot, provided in an embodiment of this application; Figure 2 This is a schematic diagram of the AGV structure provided in an embodiment of this application; Figure 3 This is a schematic diagram of an AGV working scenario provided in an embodiment of this application; Figure 4 This is a schematic diagram of an automatic steamer loading system based on an AGV collaborative robot, provided as an embodiment of this application. Detailed Implementation

[0018] The present solution will now be described in conjunction with the accompanying drawings and specific embodiments.

[0019] See Figure 1 The automatic steamer loading method based on AGV collaborative robots provided in this embodiment includes: S101, Establish and unify the world coordinate system, complete the unification and calibration of the coordinate systems of AGV, AGV dual-cooperative robotic arm, AGV hopper, sensor and still, and ensure that the AGV hopper discharge port and release point are mapped to the world coordinate system.

[0020] See Figure 2 This is a schematic diagram of the AGV (Automated Guided Vehicle) structure in this embodiment, including the AGV body 1, wireless receiver 2, dual collaborative robotic arms 3, and robotic handheld winnowing basket 4. Sensors (not shown) are also included. The wireless receiver 2 is controlled by the control system, and the robotic handheld winnowing basket 4 is connected to the collaborative robots on the left and right sides, according to the size of a winnowing basket used for manually loading food.

[0021] See Figure 3The vision system 5 consists of a thermal imaging camera, a laser rangefinder, and a 3D scanner. It is set up in pairs of stills and fixed to the rotating arm 6, allowing for the detection and positioning of any still. The control system comprises a steam detection and still-loading module and an AGV control module. The steam detection and still-loading module fully mimics the manual throwing action of the stills, achieving the requirements of light, loose, even, accurate, thin, and flat loading. Simultaneously, the AGV is controlled for movement and positioning through real-time positioning.

[0022] Before the intelligent still loading operation begins, system calibration and coordinate system are core prerequisites for ensuring the accuracy of subsequent operations. Since the AGV mobile platform, dual collaborative robotic arms, robot handheld sieves, and vision detection equipment each have their own independent local coordinate systems, the lack of a unified reference will lead to problems such as positioning deviations in the mash throwing and disordered coordination of mechanical movements, directly affecting the loading quality and efficiency.

[0023] Therefore, it is necessary to first establish and unify the world coordinate system, AGV coordinate system, robotic arm base coordinate system, scoop coordinate system, sensor coordinate system, and still coordinate system. Among them, the world coordinate system serves as a global reference benchmark, used to integrate the motion and positioning data of all equipment. The AGV coordinate system provides the positioning basis for the autonomous navigation of the mobile platform, the robotic arm base coordinate system guides the motion trajectory planning of the arm joints, the scoop coordinate system accurately describes the position and posture of the discharge port, the sensor coordinate system ensures the accurate conversion of visual detection data, and the still coordinate system focuses on the positioning and modeling of the target area.

[0024] To achieve effective correlation between various coordinate systems, a laser tracker combined with visual calibration technology is required to accurately calibrate the installation relationship between the AGV and the dual-cooperative robotic arm, and between the robotic arm and the robot's handheld dustpan. By collecting coordinate data of multiple sets of feature points, the transformation matrix between each coordinate system is calculated. (From world coordinate system to AGV coordinate system) (From AGV coordinate system to robotic arm base coordinate system) (From the robot arm base coordinate system to the scoop coordinate system), ultimately forming a complete coordinate transformation chain. This transformation chain can transform any point in the scoop coordinate system. (Such as discharge point, mash release point) precisely mapped to the world coordinate system. Their mapping relationship satisfies the formula: This calibration process ensures that the discharge position and release posture of the winnowing basket remain consistent with the relative position of the still during the operation, laying the foundation for accurate subsequent sprinkling.

[0025] S102 collects point cloud and thermal imaging above the steamer opening, fits the steamer opening plane and refines the steamer coordinate system, identifies the steam escape area and boundary to form no-pour and slow-pour constraints.

[0026] After system calibration is completed, the modeling of the still and the identification of steam boundaries are required. This is a crucial step in realizing personalized still-loading planning. The core requirement of the still-loading operation is to formulate a suitable material-laying strategy based on the actual structure and steam distribution of the still. Therefore, it is necessary to accurately obtain the geometric parameters of the still opening and the steam dissipation range.

[0027] During the still modeling stage, the system drives a rotating arm carrying a visual detection device consisting of a 3D LiDAR (or depth camera) to smoothly swing directly above the opening of the target still, ensuring that the detection range completely covers the opening area. The visual detection device will quickly collect point cloud data and distance information from the surface of the still opening, forming a point set containing thousands of feature points. .

[0028] To accurately determine the spatial position and shape of the steamer opening, the steamer opening plane needs to be fitted based on the aforementioned point set to obtain the unit normal vector and plane offset of the steamer opening plane. This determines the geometric center of the steamer opening and updates the steamer coordinate system. Simultaneously, a set of loading grid points is generated on the steamer opening plane. , in: Indicates to and Optimize to minimize the objective function in the subsequent steps. This is the unit normal vector of the steamer opening plane, used to represent the orientation of the plane. In this embodiment, The normal direction corresponding to the plane of the still opening can be used to define the z-axis direction of the still coordinate system. The plane offset represents the position of the plane relative to the origin. The number of points involved in the fitting, i.e., the number of effective points in the point cloud used to estimate the steamer opening plane. The larger the value and the more uniform the point distribution, the better the fitting stability generally is. For the first point cloud In this embodiment, there are several points. The three-dimensional point cloud originates from the steamer opening area or is reconstructed from distance data. This indicates that the following constraints are satisfied. Let be the norm of the vector.

[0029] By solving this optimization problem, the unit normal vector of the steamer opening plane can be obtained. Offset from the plane This allows for the calculation of the geometric center coordinates of the still opening, which is then used as the origin to update the still's dedicated coordinate system. Simultaneously, a set of loading grid points is generated on the still opening plane according to a uniform distribution principle, with each grid point serving as the basic unit for subsequent material layer thickness planning and placement positioning.

[0030] During the steam boundary identification stage, considering that the mash in the steam escaping area of ​​the still is prone to clumping due to high temperature, affecting the aeration and fermentation effect, it is necessary to accurately delineate the prohibited and slow-addition areas. The system will use the thermal imaging module on the visual detection device to collect the thermal imaging temperature field of the still opening area. To highlight the characteristics of the steam area, a steam significance score is calculated based on the thermal imaging temperature field, and the steam escaping area is extracted, thereby obtaining the steam boundary curve, which is used to characterize the prohibited and slow-addition areas within the still opening: in: For steam confidence level, Represents the horizontal pixel coordinates of the image. Represents the vertical pixel coordinates of the image, ( , This corresponds to a pixel in a frame of a thermal imaging image. The weighting of the temperature term adjusts for the influence of standardized temperature bias on the significance score. The larger the value, the higher the overall temperature. For thermal imaging pixels ( , The temperature at that location The average temperature is the temperature of all pixels within the current frame or a specified region (e.g., the ROI region). The average value is obtained and used as a reference for temperature normalization to offset changes in ambient temperature and thermal drift. The standard deviation of temperature is used to... Standardize the scales to make the scoring scales comparable under different working conditions. The gradient term weights adjust the influence of the temperature gradient magnitude on the significance score. The temperature gradient is used to characterize changes at the vapor edge. The Sigmoid function normalizes the scores to a range of 0-1. The vapor emission region is defined as the set of all pixel coordinates that satisfy the threshold condition. The threshold for the steam region. The larger the size, the stricter the criteria; the smaller the steam region, but the purer it is. A smaller value indicates a more lenient judgment, resulting in a larger vapor area but potentially more noise.

[0031] S103 combines the loading stage with the steam boundary to generate a target thickness field, and uses the time for driving, picking up, and scattering, as well as the risk of steam overflow, to complete the assignment of multiple AGV tasks to ensure continuous cycle time.

[0032] Based on the still geometry model and steam boundary curve, it is necessary to further develop a scientific target material thickness field and achieve efficient collaborative scheduling of multiple AGVs to balance the quality of still loading and operational efficiency.

[0033] The core idea of ​​the target thickness field planning is "adapting to local conditions and dynamically matching". The loading operation is divided into three stages: initial stage, middle stage and final stage. The material requirements of different stages are different: the initial stage requires thin layering to ensure bottom aeration, the middle stage requires even layering to ensure consistent fermentation, and the final stage requires moderate thickening to compact the mash.

[0034] Based on the loading stage and the steam boundary curve, a target material thickness field is generated for the loading grid point set, causing the target material thickness to decrease near the steam boundary and vary with the loading stage. Simultaneously, to prevent excessive material thickness from covering the steam passage in the steam boundary region, the material thickness needs to be reduced near the steam boundary. The target thickness field is finally determined using the following formula: in: For the target thickness field, at the grid points in the steamer opening plane. The desired thickness value is used to constrain thin, flat, and uniform spreading, and actively thins near the strong steam zone to meet the requirements of steam exploration and steam loading. The first plane of the steamer opening The grid points are typically generated in the still plane of the still coordinate system after geometric modeling of the still opening, and are used to represent the discretization of the thickness field. The base layer thickness indicates the default target thickness level when there is no staged thickening and it is far from the influence of the steam boundary, corresponding to the initial thin spreading or a certain base layer standard. It is a thickness increment scale that changes with the loading process, used to change the overall target thickness level as the loading stage progresses, so that the target thickness gradually changes from the initial stage to the later stage, such as from thin to thick or from the outside to the inside. This is a function for the steaming stage, with the steaming stage as input. The index outputs dimensionless stage coefficients, gradually increasing from initial steam spillage to stable filling. Initial stage. Smaller size, thinner target; mid-to-late stage: The height is increased to bring the target closer to the final loading height. During the loading stage, the parameters can be adjusted by the cumulative amount of material fed, time, thickness, field progress, or by manual settings. To reduce the feed intensity in the steam zone, the closer the distance, the greater the feed reduction. It is an exponential function. The distance from the grid point to the steam boundary. For steam boundary, This refers to the influence range parameter. In this embodiment... , for A point on the surface.

[0035] In terms of multi-AGV scheduling, to meet the continuous operation requirements of large-scale production, it is necessary to coordinate the task allocation and arrival sequence of multiple AGVs to avoid workstation congestion or waiting for idle time. The system will acquire information such as the current position, remaining power, and task status (idle / retrieving / traveling) of each AGV in real time, and calculate the estimated travel time from each AGV to each target still, the estimated retrieval time (the average retrieval time of the AGV, which is related to the moisture content of the mash and the height of the material pile), and the estimated spreading time (the single spreading time of the still, which is related to the target spreading amount). At the same time, by monitoring the steam overflow of the still in real time, a steam overflow risk function is constructed (the risk of steam overflow from the still; the higher the risk, the faster the spreading needs to be completed), and a weighting coefficient is introduced to adjust the proportion of risk factors in task allocation.

[0036] The estimated travel time, estimated material retrieval time, estimated spillage time, and steam overflow risk are constructed as task allocation costs and assigned for optimization to determine the target stills and arrival sequence for each AGV, thereby reducing the still loading idle time and suppressing steam overflow. in: Let be the cost matrix, representing the cost of the th... AGVs are assigned to serve the first The target distillation still or the first The overall cost corresponding to each workstation is used for assignment optimization. For AGV number index, For the target index, The travel time of the AGV from its current position to the still can be estimated by the AGV navigation module based on path length, speed limit, and congestion / obstacle avoidance status. For the first The loading time of one AGV is compared with the first The AGV's robotic arm action template, material pile position, material picking depth, posture, and other related information. For the first The time required for a single scattering operation in a steamer is usually related to the current material shortage level in that steamer, the number of scattering points, the number of scattering operations, and the strength of the steam avoidance strategy. This is the spillover risk penalty coefficient, used to adjust the impact of steam spillover risk on the total cost: The larger the steamer, the more the optimization will prioritize serving the steamers with higher steam risk in order to reduce idle time. The smaller the value, the more the optimization will favor the shortest pure time. For the first Each steamer at a moment The steam overflow risk function reflects the risk of steam escaping or overflowing if the steamer is not put on in time under the current operating conditions. It can be constructed from statistics of the thermal imaging steam significance score within the ROI at the steamer opening, such as: the proportion of steam area, the mean significance, and the intensity of boundary fluctuations. Indicates all possible assignment schemes In the process, we seek the solution that minimizes the objective function. This represents the total cost of the assignment plan. In a given assignment In the case of the first Each AGV corresponds to the assigned target The cost of time, For the number of AGVs, To give the first The target index assigned to the AGV.

[0037] S104: The AGV picks up and positions the material, generates the landing point according to the target thickness, solves the scattering parameters based on the trajectory, and adjusts the tilt angle and shaking of the scoop in coordination with both arms to achieve light and even spreading, and updates the shaking parameters online.

[0038] Once the scheduling plan is determined, the system will enter the execution phase of material collection, movement, and spreading. This phase requires precise control of equipment movements to ensure that the mash is spread according to the target thickness.

[0039] First, the target AGV will autonomously move to the fermentation mash pile area along a preset path according to the scheduling instructions. During the movement, it will use LiDAR and visual SLAM technology for real-time positioning and correction of travel deviations. Upon reaching the pile, the dual-arm collaborative robotic system will perform a coordinated material-retrieving action based on the pile height and the state of the fermentation mash: the force control sensor at the end of the robotic arm provides real-time feedback on the gripping force to ensure that the amount of material retrieved matches the target material thickness, avoiding excessive spillage or insufficient material that would affect work efficiency. Finally, the fermentation mash will be smoothly loaded into the robot's handheld sieve. After material retrieval, the AGV will restart and head to the target stilling station. As it approaches the stilling station, the system will activate closed-loop trajectory tracking control to further improve stopping accuracy and define the positional deviation. in: To track errors, This is the current pose and position state of the AGV. In this solution, It can be output in real time by AGV positioning system, laser SLAM, odometer, vision positioning, etc. The reference trajectory state represents the expected reference position and orientation that the AGV should reach at time t, with dimensions and... Consistent. Provided by the path planning or trajectory generation module, such as a reference trajectory to the stockpile / to the stop point beside the steamer. To control the output speed, As the reference velocity vector, This is the feedback gain matrix.

[0040] Based on the target material thickness field and real-time material thickness information, the insufficient material area is determined and a target landing point set is generated. The release point and target landing point of the winnowing basket are determined in the world coordinate system. Under a given flight time, the equivalent initial velocity vector of the discharged material is solved according to the ballistic relationship. Then, the two cooperative robotic arms are controlled to coordinately adjust the tilt angle and back-and-forth shaking motion of the winnowing basket, scattering the mash into the spatial area corresponding to the target landing point set. in: The target landing point is represented in the world coordinate system, indicating the spatial coordinates of the distillation mash particle bundle landing on the still opening plane or the target area inside the still at the end of its flight. It is typically a three-dimensional column vector. In this scheme, It is determined by the set of missing grid points or target landing points, and unified to the world coordinate system through coordinate transformation. The position of the hopper discharge port in the world coordinate system represents the spatial coordinates of the hopper discharge port at the instant the material leaves the hopper. In this embodiment, It is calculated from the scoop posture and end pose through a coordinate transformation chain. The equivalent initial velocity vector for material discharge represents the equivalent velocity of the material at the instant it leaves the hopper, and is used to characterize the scattering intensity and the offset of the landing point. Release duration indicates the time from the release point. Fly to landing point The time interval. Based on the initial velocity caused by gravity, the system controls the two collaborative robotic arms to adjust the tilt angle and back-and-forth shaking motion of the winnowing basket, so as to accurately throw the mash to the target landing area.

[0041] To further optimize the scattering effect, the system needs to dynamically adjust the scoop's motion parameters: based on the relationship between the scattering width and the scoop's tilt angle, the amplitude of the shaking angle, and the shaking frequency, a cost function is constructed with indicators such as the uniformity of the landing density distribution, the deviation between the actual thickness field and the target thickness field, and the penalty for the steam escape area. in; The spread width per unit is the equivalent width of the material covering the surface of the steamer formed by a single spread, and is used to measure the spread range of uniform distribution. The hopper inclination angle represents the angle of inclination of the hopper relative to the horizontal plane, which determines the material discharge flow rate and discharge direction component. The larger the inclination angle, the faster the discharge, the larger the ejection component, and the more the distribution pattern will change. The shaking angle amplitude value represents the angular amplitude of the winnowing basket when it makes periodic back-and-forth swinging or shaking during the scattering process. The frequency of the jitter. The calibrator coefficient for the tilt angle of the winnowing basket. This is the calibration coefficient for the jitter angle amplitude. This is the jitter frequency calibration coefficient. For learning rate, Cost function The gradient of the control parameters, The value is the coefficient of variation of the density distribution; the smaller the value, the more uniform the distribution. The drop density distribution represents the spatial distribution density of the material on the steamer opening plane or grid after being thrown, such as the estimated amount of material thrown per unit area of ​​each grid or the estimated number of times the material is thrown. The root mean square error of thickness deviation characterizes thinness and flatness. For the actual thickness of the stockpile, This is a penalty for areas where steam escapes; the closer to an area with strong steam coverage, the greater the penalty. The weights are the coefficients of variation of the density distribution. The root mean square error weight for thickness deviation The penalty weight for the steam escape region.

[0042] The scoop tilt angle, shaking angle amplitude, and shaking frequency are updated online based on the cost function to achieve light and even spreading and reduce the probability of covering areas where steam escapes.

[0043] S105, after spreading, updates the thickness field and steam boundary, and determines whether to replenish or stop the cycle based on height and coverage criteria, completes the cleaning and evacuation.

[0044] After each application of material, the system needs to enter the status update and completion determination phase. Through a closed-loop feedback mechanism, the loading quality is continuously optimized to ensure that the operation meets the preset standards. After each application, the visual detection device is driven to collect point cloud and / or distance information of the steam pot opening area again to update the actual material thickness field, and to collect thermal imaging temperature field again to update the steam escape area and steam boundary curve.

[0045] The loading process is judged based on the loading completion criteria, which include at least: the maximum value of the actual material thickness field reaches a preset maximum material height threshold, and the proportion of grid points in the loading grid point set that satisfy the condition that the actual material thickness is not less than the target material thickness minus the allowable error band reaches a preset coverage threshold. in: The maximum value of the actual thickness field is the thickness of the highest point of the material stack in the grid at the steamer opening, used to determine whether the upper limit of the steamer loading height has been reached or the height requirement has been met. This indicates that for all grid points Take the maximum. The actual thickness of the material at the grid points The values ​​at a given location are usually obtained from point cloud or 3D scan reconstruction. This is the maximum permissible stacking height threshold, used to determine whether the loading height meets process requirements, such as being close to the target loading amount or target material layer height. This value can be calculated from process experience, the steamer volume, or the target feed rate. For the number of grid points, For indicator functions, For the target paving thickness at the grid points The value at that location, For thickness tolerance zone, The coverage threshold is the target.

[0046] If the above completion criteria are not met, the system will return to step S103 to reschedule the AGV task based on the updated set of material shortage areas and target landing points, or directly return to step S104 to execute the next scattering operation, continuously optimizing material replenishment. If the completion criteria are met, it indicates that the loading operation has met the quality requirements, and the system will initiate a clean-up process: the dual collaborative robotic arms drive the winnowing basket to perform high-frequency shaking to shake off the remaining mash (avoiding material waste and equipment contamination). Subsequently, the AGV leaves the target fermentation still station according to the preset evacuation path and goes to the designated area to wait for the next round of operation instructions, thus completing the closed loop of the entire loading process.

[0047] Corresponding to the above embodiment of the automatic steamer loading method based on AGV collaborative robots, this application also provides an embodiment of an automatic steamer loading system based on AGV collaborative robots.

[0048] See Figure 4 The automated steamer loading system 20 based on AGV collaborative robots in this embodiment includes: The coordinate unification and calibration module 201 is used to establish and unify the world coordinate system, complete the unification and calibration of the coordinate systems of AGV, AGV dual cooperative robotic arm, AGV hopper, sensor and wine still, and ensure that the AGV hopper discharge port and release point are mapped to the world coordinate system.

[0049] The steam pot mouth modeling and steam detection identification module 202 is used to collect point cloud and thermal imaging above the steam pot mouth, fit the steam pot mouth plane and refine the steam pot coordinate system, identify the steam escape area and boundary to form no-pour and slow-pour constraints.

[0050] The planning and scheduling module 203 is used to generate a target thickness field by combining the loading stage and the steam boundary, and to complete the assignment of multiple AGV tasks at the cost of travel, material picking, scattering time and steam overflow risk to ensure continuous cycle time.

[0051] The execution control module 204 is used for AGV to pick up materials and position and stop, generate landing points according to the target thickness, solve the scattering parameters based on the trajectory, and coordinate the two arms to adjust the tilt angle and shaking of the scoop to achieve light and even spreading, and update the shaking parameters online.

[0052] The status closing module 205 is used to update the thickness field and steam boundary after spreading, and to replenish or end the cycle based on the height and coverage criteria, so as to complete the cleaning and closing and withdraw.

[0053] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, the simultaneous existence of A and B, or the existence of B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects have an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, and c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0054] The above description is merely a specific embodiment of this application. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the protection scope of this application. The protection scope of this application should be determined by the protection scope of the claims.

Claims

1. An automated steamer loading method based on AGV collaborative robots, characterized in that, include: Establish and unify the world coordinate system, complete the unification and calibration of the coordinate systems of AGV, AGV dual-cooperative robotic arm, AGV hopper, sensor and still, and ensure that the AGV hopper discharge port and release point are mapped to the world coordinate system; Point cloud and thermal imaging were collected above the steamer opening, the steamer opening plane was fitted and the steamer coordinate system was refined, and the steam escape area and boundary were identified to form no-pour and slow-pour constraints. By combining the loading stage with the steam boundary to generate a target thickness field, and by incurring costs such as travel, material picking, and scattering time as well as the risk of steam overflow, multiple AGV task assignments are completed to ensure continuous cycle time. The AGV picks up and positions the material, generates the landing point according to the target thickness, solves the scattering parameters based on the trajectory, and adjusts the tilt angle and shaking of the scoop in coordination with both arms to achieve light and even spreading, and updates the shaking parameters online. After spreading, the thickness field and steam boundary are updated. Based on the height and coverage, the cycle is replenished or terminated. After cleaning and evacuation, the process is completed.

2. The automatic steamer loading method based on AGV collaborative robots according to claim 1, characterized in that, The establishment and unification of the world coordinate system, completing the unification and calibration of the coordinate systems of AGV, AGV dual-cooperative robotic arm, AGV hopper, sensors, and still, ensuring that the AGV hopper discharge port and release point are mapped to the world coordinate system, includes: Establish and unify the world coordinate system, AGV coordinate system, robotic arm base coordinate system, scoop coordinate system, sensor coordinate system, and still coordinate system; The installation relationship between the AGV, the dual collaborative robotic arm, and the robot's handheld scoop is calibrated to obtain a coordinate transformation chain for mapping points in the scoop coordinate system to the world coordinate system, thereby unifying the scoop discharge point, scoop release point, and their posture to the world coordinate system during operation.

3. The automatic steamer loading method based on AGV collaborative robots according to claim 1, characterized in that, The process of acquiring point cloud and thermal imaging data above the still opening, fitting the still opening plane and refining the still coordinate system, identifying steam escape areas and boundaries to form no-docking and slow-docking constraints includes: The drive arm swings the visual detection device above the mouth of the target still, collects point cloud and / or distance information of the still mouth area and forms a point set; The point set is used to fit the steamer opening plane to obtain the unit normal vector and plane offset of the steamer opening plane, thereby determining the geometric center of the steamer opening and updating the steamer coordinate system. At the same time, a set of loading grid points is generated on the steamer opening plane. ， in: Indicates to and Optimize to minimize the objective function in the subsequent steps. The unit normal vector of the steamer opening plane. For planar offset, The number of points involved in the fitting. For the first point cloud One point, This indicates that the following constraints are satisfied. Let be the norm of a vector; A thermal imaging temperature field of the steamer opening area is collected. Based on the thermal imaging temperature field, a steam significance score is calculated and the steam escape area is extracted. Then, a steam boundary curve is obtained to characterize the prohibited and slow-to-add regions within the steamer opening.

4. The automatic steamer loading method based on AGV collaborative robots according to claim 3, characterized in that, The thermal imaging temperature field of the steamer opening area is collected. Based on the thermal imaging temperature field, a steam significance score is calculated and the steam escape area is extracted, thereby obtaining a steam boundary curve, which is used to characterize the prohibited and restricted feeding areas within the steamer opening, including: in: For steam confidence level, Represents the horizontal pixel coordinates of the image. Represents the pixel coordinates in the vertical direction of the image. As the weight of the temperature term, For the temperature of thermal imaging pixels, The average temperature. For temperature standard deviation, For gradient term weights, The temperature gradient is used to characterize changes at the vapor edge. The Sigmoid function normalizes the scores to a range of 0-1. This is a steam escape zone. This is the threshold for the steam region.

5. The automatic steamer loading method based on AGV collaborative robots according to claim 3 or 4, characterized in that, The method combines the loading stage with the steam boundary to generate a target thickness field, and, at the cost of travel, material retrieval, and spillage time, as well as the risk of steam spillage, completes multi-AGV task assignment to ensure continuous cycle time, including: Based on the loading stage and the steam boundary curve, a target material thickness field is generated for the loading grid point set, causing the target material thickness to decrease near the steam boundary and vary with the loading stage, including: in: For the target thickness field, The first plane of the steamer opening Grid points, Based on the thickness of the base material, The thickness increment scale varies with the loading process. This is a function representing the steam loading stage, which gradually increases from the initial steam spill to the stable filling stage. This is the steaming stage. To reduce the feed intensity in the steam zone, the closer the distance, the greater the feed reduction. It is an exponential function. The distance from the grid point to the steam boundary. For steam boundary, For parameters affecting the range of influence; The system obtains the current location and task status of multiple AGVs, their corresponding estimated travel time, estimated material collection time, and estimated spilling time, and also obtains the steam overflow risk of the target still at the current moment. The estimated travel time, estimated material collection time, estimated spillage time, and steam overflow risk are constructed as task allocation costs and assigned for optimization to determine the target stills and arrival sequence of each AGV, thereby reducing the still loading time and suppressing steam overflow.

6. The automatic steamer loading method based on AGV collaborative robots according to claim 5, characterized in that, The process of constructing the estimated travel time, estimated material collection time, estimated spillage time, and steam overflow risk into task allocation costs and optimizing assignment to determine the target stills and arrival sequence for each AGV, thereby reducing the stilling idle time and suppressing steam overflow, includes: in: Here is the cost matrix. For AGV number index, For the target index, This represents the travel time of the AGV from its current location to the still. For the first The loading time of one AGV. For the first The time required for one-time scattering operation of a single steamer. As the spillover risk penalty coefficient, For the first Each steamer at a moment The steam spillover risk function, Indicates all possible assignment schemes In the process, we seek the solution that minimizes the objective function. This represents the total cost of the assignment plan. In a given assignment In the case of the first Each AGV corresponds to the assigned target The cost of time, For the number of AGVs, To give the first The target index assigned to the AGV.

7. The automatic steamer loading method based on AGV collaborative robots according to claim 6, characterized in that, The AGV picks up and positions the material, generates landing points based on the target thickness, solves for the scattering parameters based on the ballistics, and coordinates the tilt angle and shaking of the scoop with both arms to achieve gentle and even spreading. The shaking parameters are updated online, including: Under the assigned and optimized task, the target AGV is driven to move to the mash pile and the dual cooperative robotic arms perform the material picking action to load the mash into the robot's handheld winnowing basket; The target AGV is then driven to move to the target still and complete the positioning and docking process. During the positioning and docking process, closed-loop trajectory tracking control is performed on the AGV to reduce docking errors. in: To track errors, This indicates the current pose and position of the AGV. For reference trajectory state, To control the output speed, As the reference velocity vector, This is the feedback gain matrix; Based on the target material thickness field and real-time material thickness information, the material shortage area is determined and the target landing point set is generated. The release point and target landing point of the winnowing basket are determined in the world coordinate system. Under the given flight time condition, the equivalent initial velocity vector of the material discharge is solved according to the ballistic relationship. Then, the dual cooperative robotic arms are controlled to coordinately adjust the tilt angle and back and forth shaking action of the winnowing basket to throw the mash into the spatial area corresponding to the target landing point set. Meanwhile, based on the relationship between the dispersion width and the scoop tilt angle, the amplitude of the shaking angle and the shaking frequency, a cost function is constructed with the uniformity of the landing point density distribution, the deviation between the actual thickness field and the target thickness field, and the penalty for the steam escape area as indicators. The scoop tilt angle, shaking angle amplitude, and shaking frequency are updated online based on the cost function to achieve light and even spreading and reduce the probability of covering areas where steam escapes.

8. The automatic steamer loading method based on AGV collaborative robots according to claim 7, characterized in that, The process involves determining the under-material area based on the target material thickness field and real-time material thickness information, generating a target landing point set, determining the scoop release point and target landing point in the world coordinate system, and solving the equivalent initial velocity vector of the discharged material according to the ballistic relationship under a given flight time condition. This is followed by controlling the dual-cooperative robotic arms to collaboratively adjust the tilt angle and back-and-forth shaking motion of the scoop, scattering the mash into the spatial area corresponding to the target landing point set. This includes: in: The position of the target landing point in the world coordinate system. This represents the position of the sieve's discharge port in the world coordinate system. The equivalent initial velocity vector for material discharge. For release duration, This is the acceleration due to gravity.

9. The automatic steamer loading method based on AGV collaborative robots according to claim 8, characterized in that, Simultaneously, based on the relationship between the dispersion width and the scoop tilt angle, the shaking angle amplitude, and the shaking frequency, a cost function is constructed with indicators such as the uniformity of the landing point density distribution, the deviation between the actual thickness field and the target thickness field, and the penalty for the steam escape area. This function includes: in; The width of the spread per unit of spray. The angle of inclination of the winnowing basket. The amplitude of the jitter angle. The frequency of the jitter. The calibrator coefficient for the tilt angle of the winnowing basket. This is the calibration coefficient for the jitter angle amplitude. This is the jitter frequency calibration coefficient. For learning rate, Cost function The gradient of the control parameters, The coefficient of variation of density distribution; the smaller the value, the more uniform the distribution. For the density distribution of landing points, The root mean square error of thickness deviation characterizes thinness and flatness. For the actual thickness of the stockpile, This is a penalty for areas where steam escapes; the closer to an area with strong steam coverage, the greater the penalty. The weights are the coefficients of variation of the density distribution. The root mean square error weight for thickness deviation The penalty weight for the steam escape region.

10. The automatic steamer loading method based on AGV collaborative robots according to claim 9, characterized in that, The process of updating the thickness field and steam boundary after scattering, and determining whether to replenish or terminate the cycle based on height and coverage criteria, to complete the cleaning process and evacuation, includes: After each spill, the visual detection device is driven to collect point cloud and / or distance information of the steam pot opening area again to update the actual stockpile thickness field, and to collect thermal imaging temperature field again to update the steam escape area and steam boundary curve. The loading process is judged based on the loading completion criteria, which include at least: the maximum value of the actual material thickness field reaches a preset maximum material height threshold, and the proportion of grid points in the loading grid point set that satisfy the condition that the actual material thickness is not less than the target material thickness minus the allowable error band reaches a preset coverage threshold. in: This represents the maximum value of the actual thickness field. This indicates that for all grid points Take the maximum. The actual thickness of the material at the grid points The value at that location, The maximum allowable stacking height threshold, For the number of grid points, For indicator functions, For the target paving thickness at the grid points The value at that location, For thickness tolerance zone, The coverage threshold is the minimum required level. If the loading completion criterion is not met, update the set of material shortage area and target landing point and the process cannot continue; When the loading completion criterion is met, a clean-up and finishing action is performed to shake off / recycle the residual material in the hopper and drive the AGV to leave the target wine still station.

Images