Discharge robot for rare earth metal electrolysis furnace and end compliance control method and system thereof

Abstract

The application provides a rare earth metal electrolytic furnace discharging robot and an end compliance control method and system thereof, wherein the method comprises the following steps: a rare earth discharging simulation environment for simulating a discharging scene is established, and a rare earth discharging control model for simulating a discharging tool motion track is trained and constructed based on the simulation environment through reinforcement learning; contact force sensation information between a discharging tool and an electrolytic furnace in a rare earth metal electrolytic furnace discharging process is acquired in real time and input into the rare earth discharging control model to obtain a real-time discharging tool motion track; and the real-time discharging tool motion track is input into a discharging robot controller to realize compliance control of the end of the discharging robot. The application realizes a robot discharging operation method suitable for a rare earth metal electrolytic furnace based on reinforcement learning and force sensing, controls the end of the rare earth metal electrolytic furnace through a force sensor for sensing the force of the tool and the electrolytic furnace and a reinforcement learning method, and thus realizes a humanized discharging operation.

Description

Technical Field

[0001] This invention relates to the field of special robot technology, specifically to a rare earth metal electrolytic furnace discharge robot based on reinforcement learning and end-effector force perception, and its end-effector compliance control method and system. Background Technology

[0002] In rare earth metal production, traditional manual unloading methods suffer from high labor intensity and poor safety. With rising labor costs, enterprises urgently need to automate the production process to reduce labor costs and improve safety. The characteristics of rare earth unloading scenarios are high temperature and unstructured environments. First, the electrolytic furnace temperature exceeds 1000 degrees Celsius, placing high demands on the electromechanical system design. Second, during electrolysis, the cathode and bottom crucible on the electrolytic furnace undergo shape changes due to adhering metal substances. Existing unloading robots do not consider these changes in the internal shape of the electrolytic furnace, resulting in poor stability and reliability in unloading operations. Furthermore, the high-temperature electrolyte makes it difficult to perceive the contents of the electrolytic furnace using visual sensors. Therefore, existing unloading robots are unsuitable for rare earth metal electrolytic furnace unloading operations.

[0003] Currently, no descriptions or reports of technologies similar to this invention have been found, and no similar information has been collected domestically or internationally. Summary of the Invention

[0004] To address the aforementioned shortcomings in the prior art, this invention provides a rare earth metal electrolytic furnace discharge robot and its end-effector compliant control method and system. This method and system, based on reinforcement learning and end-effector force sensing, achieves intelligent control of rare earth discharge operations and is applicable to robot discharge operations in rare earth metal electrolytic furnaces.

[0005] According to one aspect of the present invention, a method for end-effector compliance control of a rare earth metal electrolysis furnace discharge robot is provided, comprising:

[0006] A rare earth discharge simulation environment is established to simulate discharge scenarios, and based on the simulation environment, a rare earth discharge control model is constructed to simulate the motion trajectory of the discharge tool through reinforcement learning training.

[0007] The contact force information between the discharge tool and the electrolytic furnace during the discharge process of rare earth metal electrolytic furnace is acquired in real time and input into the rare earth discharge control model to obtain the real-time movement trajectory of the discharge tool.

[0008] The real-time discharge tool's motion trajectory is input to the discharge robot controller to achieve compliant control of the discharge robot's end effector.

[0009] According to another aspect of the present invention, a compliant end-effector control system for a rare earth metal electrolysis furnace discharge robot is provided, comprising: a force sensing module, a compliant controller, and a reinforcement learning training module; wherein:

[0010] The force sensing module is installed at the end of the discharge robot and is used to acquire contact force information between the discharge tool and the electrolytic furnace in real time during the discharge process of rare earth metal electrolytic furnace.

[0011] The reinforcement learning training module is used to establish a rare earth discharge simulation environment for simulating discharge scenarios, and based on the simulation environment, to construct a rare earth discharge control model for simulating the motion trajectory of the discharge tool through reinforcement learning training.

[0012] The compliant controller is used to carry the rare earth discharge control model and to obtain the contact force information acquired by the force sensing module in real time as the input of the rare earth discharge control model. It obtains the real-time movement trajectory of the discharge tool and outputs it to the discharge robot controller to realize the compliant control of the discharge robot end.

[0013] According to a third aspect of the present invention, a rare earth metal electrolytic furnace discharge robot is provided, comprising: a discharge robot body, wherein the discharge robot body achieves compliant control of the discharge robot end effector through the method or system described above.

[0014] By adopting the above technical solution, the present invention has at least one of the following beneficial effects compared with the prior art:

[0015] The rare earth metal electrolytic furnace unloading robot and its end-effector compliant control method and system provided by this invention utilize the operation mode in which the trajectory of the tool can be adjusted by the contact force between the tool and the electrolytic furnace during the manual unloading process. By sensing the interaction force between the tool and the electrolytic furnace through a force sensor, and by using reinforcement learning method to control the end of the work, human-like unloading operation is achieved.

[0016] This invention provides a rare earth metal electrolysis furnace unloading robot and its end-effector compliant control method and system. The robot's end effector is equipped with a high-precision six-dimensional force moment sensor to perceive force information interacting with the electrolysis furnace in real time, including the magnitude and direction of the contact force. A reinforcement learning-based compliant controller combines this force feedback data with the robot's joint position and velocity information. After processing by a deep neural network, it generates the unloading tool's motion trajectory in real time, thereby obtaining the end effector's adjustment strategy. This enables the robot to flexibly adjust its movements according to changes in contact force, much like a skilled worker, accurately responding to the high-temperature, unstructured, and dynamically changing unloading environment, greatly improving the level of intelligence in the operation and accurately completing the unloading task. Integrating force sensors and reinforcement learning algorithms endows the robot with end-effector force perception and compliant control capabilities, enabling it to adjust its movements in real time according to the contact force with the electrolysis furnace, accurately responding to the high-temperature, unstructured, and dynamically changing rare earth unloading environment. Its operational accuracy and stability far exceed those of traditional robots, significantly improving the level of operational intelligence.

[0017] The rare earth metal electrolysis furnace discharge robot and its end-effector compliant control method and system provided by the present invention, wherein the compliant controller outputs motion control strategy in real time based on force sensor feedback data and joint state information, and plans the discharge motion trajectory according to the control strategy.

[0018] The rare earth metal electrolysis furnace unloading robot and its end-effector compliant control method and system provided by this invention can quickly detect and respond to collision risks. By calculating the motion parameters of each joint based on the input trajectory and controlling the robotic arm joints to complete the movement, collisions and overloads are effectively prevented, ensuring the continuity and stability of the unloading operation. This avoids production interruptions and equipment damage caused by collisions or overloads, improving production efficiency and equipment lifespan. The compliant controller, based on reinforcement learning, relies on force sensor feedback data and robot joint state information to output end-effector adjustment strategies in real time, effectively preventing collisions and jamming, ensuring continuous and stable unloading, significantly improving reliability, and enhancing operational stability.

[0019] The rare earth metal electrolysis furnace discharge robot and its end-effector compliant control method and system provided by this invention significantly reduce manual intervention through automated discharge. With the continuous rise in labor costs, this can substantially reduce enterprises' labor costs in the long run. By reducing manual intervention through automated discharge, and with rising labor costs, enterprises can reduce labor costs in the long term, thereby lowering production costs.

[0020] The rare earth metal electrolysis furnace discharge robot and its end-effector compliant control method and system provided by this invention allow the robot to replace manual labor in high-temperature and high-risk discharge operations, completely eliminating the safety hazards of manual operation. By replacing manual labor in high-temperature and high-risk discharge operations, the safety hazards of manual operation are eliminated, and operational safety is improved.

[0021] This invention provides a rare earth metal electrolysis furnace discharge robot and its end-effector compliant control method and system. Leveraging its high-precision force sensing and compliant control capabilities, the robot can accurately complete the discharge operation, ensuring consistency in the cooling and forming quality of the rare earth metal. This reduces defects caused by human error or equipment malfunction, improves the overall product quality and performance, and helps enhance the company's market competitiveness and product added value. The improved discharge accuracy and stability ensure consistency in the cooling and forming quality of the rare earth metal, reduce defects caused by human error or equipment malfunction, improve the overall product quality and performance, and enhance the discharge quality.

[0022] The rare earth metal electrolysis furnace discharge robot and its end-effector compliant control method and system provided by this invention, based on force feedback and reinforcement learning, enable the robot to autonomously adapt to the discharge requirements of different electrolysis furnace models and processes, reducing equipment replacement and debugging time. It boasts wide adaptability and rapid debugging advantages, improving production flexibility and efficiency. Simultaneously, the collaborative control system can optimize the production process in real time based on the electrolysis furnace operating parameters and the robot's discharge progress, further enhancing production efficiency and enabling enterprises to respond more quickly to market changes and customer needs. Through force feedback and reinforcement learning, the robot autonomously adapts to the discharge requirements of different electrolysis furnace models and processes, reducing equipment replacement and debugging time and improving production efficiency. Furthermore, the collaborative control system optimizes the production process, further enhancing efficiency.

[0023] The rare earth metal electrolytic furnace unloading robot and its end-effector compliant control method and system provided by this invention combine reinforcement learning with end-effector force sensing technology, offering a novel intelligent solution for rare earth metal electrolytic furnace unloading operations. This technological integration not only improves the stability and quality of robot unloading operations but also significantly reduces operational risks and costs. It represents the cutting-edge development direction of special robot technology in the rare earth metal production field and provides valuable reference for the automation upgrade of similar production scenarios.

[0024] Based on the above advantages, the rare earth metal electrolytic furnace discharge robot and its end-effector compliant control method and system provided by the present invention not only improve the stability and quality of the robot's rare earth discharge operation, but also significantly reduce the operation risk and cost, providing a more advanced and reliable technical solution for rare earth electrolytic furnace discharge operations. Attached Figure Description

[0025] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0026] Figure 1 This is a flowchart illustrating the end-effector compliance control method for a rare earth metal electrolytic furnace discharge robot in a preferred embodiment of the present invention.

[0027] Figure 2 This is a flowchart of the rare earth discharge simulation environment training and reinforcement learning training process for the end-effector compliant control method of the rare earth metal electrolytic furnace discharge robot in a preferred embodiment of the present invention.

[0028] Figure 3 This is a schematic diagram of the components of the end-effector compliance control system for a rare earth metal electrolytic furnace discharge robot in a preferred embodiment of the present invention.

[0029] Figure 4 This is a schematic diagram of the installation structure of the force sensing module of the unloading robot in a specific application example of the present invention.

[0030] Figure 5 This is a schematic diagram of the rare earth discharge simulation environment in a specific application example of the present invention. Detailed Implementation

[0031] The embodiments of the present invention are described in detail below: These embodiments are implemented based on the technical solution of the present invention, and provide detailed implementation methods and specific operation processes. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention.

[0032] In rare earth metal production, the traditional manual unloading method has problems such as high labor intensity and poor safety. Due to the high temperature and unstructured nature of rare earth unloading scenarios, existing unloading robots are difficult to apply to rare earth metal electrolytic furnace unloading operations, and there is no existing technology for automation of rare earth metal electrolytic furnace unloading operations.

[0033] To address the aforementioned issues, one embodiment of the present invention provides a compliant end-effector control method for a rare earth metal electrolysis furnace discharge robot. This method is based on reinforcement learning and force perception to achieve a robot discharge operation method suitable for rare earth metal electrolysis furnaces. By sensing the interaction force between the tool and the electrolysis furnace through a force sensor and controlling the working end of the rare earth metal electrolysis furnace through reinforcement learning, a human-like discharge operation is achieved.

[0034] Specifically, such as Figure 1 As shown in this embodiment, the end-effector compliance control method for the rare earth metal electrolysis furnace discharge robot includes:

[0035] S1. Establish a rare earth discharge simulation environment for simulating discharge scenarios, and based on the simulation environment, construct a rare earth discharge control model for simulating the motion trajectory of the discharge tool through reinforcement learning training.

[0036] S2, real-time acquisition of contact force information between the discharge tool and the electrolytic furnace during the discharge process of rare earth metal electrolytic furnace, and input into the rare earth discharge control model to obtain the real-time movement trajectory of the discharge tool;

[0037] S3 inputs the real-time motion trajectory of the discharge tool to the discharge robot controller to achieve compliant control of the discharge robot's end effector.

[0038] In some preferred embodiments, S1 above establishes a rare earth discharge simulation environment for simulating discharge scenarios, such as... Figure 2 As shown, it may further include:

[0039] A virtual simulation environment corresponding to the actual material discharge scenario is established, and the following initialization is performed in the virtual simulation environment to obtain a rare earth material discharge simulation environment for simulating the material discharge scenario:

[0040] S111, the position of the discharge robot arm is set directly above the electrolytic furnace as the initial position;

[0041] S112, by placing obstacles on the inner wall of the electrolytic furnace to simulate the change in the internal shape ratio of the electrolytic furnace during the electrolysis process, the initial environment is randomly generated;

[0042] S113, initialize various parameters of the unloading robot. In some preferred embodiments, S111 above, establishing a virtual simulation environment corresponding to the actual unloading scenario, may further include:

[0043] Build a virtual simulation environment that corresponds to the actual material output scenario, and construct the following key objects within the virtual simulation environment:

[0044] Construct an electrolytic furnace and set parameters such as its size, shape, position, mass, and centroid to ensure that the virtual simulation environment can realistically reflect the physical relationships in the real discharge environment.

[0045] Construct a 3D model of the material discharge robot, and adjust and configure the 3D model according to the specifications of the actual material discharge robot;

[0046] Set up the sensors and ensure that their position, orientation, and parameters match those of the actual material-discharging robot in order to accurately simulate the sensor data output in the simulation.

[0047] Define the action space and the observation space. The action space is the joint motion space of the unloading robot, and the observation space is the end position and posture of the unloading tool and the force feedback force from the force sensor. Based on the dimensions and range of the action space and the observation space, set the control interface and the status monitoring interface of the unloading robot to ensure that the unloading robot can receive action commands and execute corresponding actions, and at the same time, it can acquire the status information in the observation space in real time, providing the necessary input and output interfaces for reinforcement learning training.

[0048] The obtained rare earth discharge simulation environment, such as Figure 5 As shown.

[0049] In some preferred embodiments, S1 above, based on a simulation environment, constructs a rare earth material discharge control model for simulating the end-effector trajectory of a tool through reinforcement learning training, such as... Figure 2 As shown, it may further include:

[0050] A pre-trained reinforcement learning network model (neural network model) is established, and the reinforcement learning network model is trained as follows:

[0051] S114: Based on the current state information of the end of the unloading robot, the next action position is randomly generated through a pre-trained reinforcement learning network model to form a planned path.

[0052] S115: Execute the action position generated by S114 using the simulator, and generate the reward and the next state information to complete one path planning simulation process;

[0053] S116 stores the action position, execution probability, and reward during the simulation process;

[0054] S117, input the data stored in S116 into the reinforcement learning network model, update the parameters of the reinforcement learning network model, and optimize the motion control strategy;

[0055] S118, determine whether the end of the discharge robot has reached the bottom of the electrolytic furnace or whether the maximum number of training rounds has been reached; if yes, end the current round of training and retain the training parameters of the current reinforcement learning network model; if no, return to restart the training process until the training ends and a rare earth discharge control model is constructed.

[0056] In some preferred embodiments, S2 above, which involves acquiring the contact force information between the discharge tool and the electrolytic furnace during the rare earth metal electrolytic furnace discharge process in real time and inputting it into the rare earth discharge control model to obtain the real-time movement trajectory of the discharge tool, may further include:

[0057] The system acquires real-time force sensor feedback data between the discharge tool and the electrolytic furnace during the actual rare earth metal electrolytic furnace discharge process. This data, combined with the joint position and speed information of the discharge robot, serves as the input to the rare earth discharge control model. The model then outputs a corresponding motion control strategy (i.e., the direction and speed of tool movement) and calculates the joint control parameters based on the motion control strategy to obtain the real-time motion trajectory of the discharge tool.

[0058] In some preferred embodiments, S3 above, which inputs the real-time motion trajectory of the unloading tool to the unloading robot controller to achieve compliant control of the unloading robot's end effector, may further include:

[0059] The real-time motion trajectory of the unloading tool is input to the unloading robot controller. The unloading robot calculates the motion parameters of each joint based on the input trajectory and controls the robotic arm joints to complete the movement. It simulates manual unloading by correcting the action in real time based on the contact force, thus achieving compliant control of the unloading robot's end effector.

[0060] The compliant end-effector control method for rare earth metal electrolysis furnace discharge robots provided in the above embodiments of the present invention designs a force sensing system adapted to high-temperature scenarios for electrolysis furnace discharge operations and processes. A rare earth discharge simulation environment is constructed to simulate the discharge scenario for reinforcement learning training of the rare earth discharge control model. During operation, the trained rare earth discharge control model is integrated into a compliant controller. The compliant controller acquires the force information detected by the force sensing system in real time, inputs it into the rare earth discharge control model, calculates the trajectory of the tool's end effector movement through the model, and outputs it to the robot controller, thus achieving compliant end-effector control. This method aims to address the problems existing in rare earth metal electrolysis furnace discharge by leveraging reinforcement learning and end-effector force sensing technology to achieve intelligent, stable, and reliable robot discharge operations, significantly improving operational safety and efficiency while reducing labor costs.

[0061] Based on the same inventive concept, one embodiment of the present invention also provides a compliant end-effector control system for a rare earth metal electrolytic furnace discharge robot.

[0062] Specifically, such as Figure 3 As shown, the end-effector compliance control system for the rare earth metal electrolysis furnace discharge robot provided in this embodiment may include: a force sensing module, a compliance controller, and a reinforcement learning training module; wherein:

[0063] The force sensing module, located at the end of the discharge robot, is used to acquire contact force information between the discharge tool and the electrolytic furnace in real time during the discharge process of rare earth metal electrolytic furnace.

[0064] The reinforcement learning training module is used to establish a rare earth discharge simulation environment for simulating discharge scenarios, and based on the simulation environment, a rare earth discharge control model for simulating the motion trajectory of the discharge tool is constructed through reinforcement learning training.

[0065] The compliant controller is used to carry the rare earth discharge control model and to obtain the contact force information acquired by the force sensing module in real time as the input of the rare earth discharge control model. It obtains the real-time motion trajectory of the discharge tool and outputs it to the discharge robot controller to realize the compliant control of the discharge robot end effector.

[0066] The following describes in detail, with reference to preferred embodiments, the specific implementation methods of each functional module constituting the system provided in the above embodiments of the present invention.

[0067] In some preferred embodiments, the force sensing module described above may further include any one or more of the following:

[0068] High-precision six-dimensional torque sensors are used and distributed on the end effector of the unloading robot;

[0069] Data acquisition is performed using an EtherCAT bus, and the acquired analog signals are converted into digital signals by a transmitter; the data synchronization period is <2ms, preferably 1ms;

[0070] Heat-insulating material is used for installation between the material and the discharge tool.

[0071] In some preferred embodiments, the reinforcement learning training module may further include: a simulation environment construction unit, an environment training unit, and a reinforcement learning training unit. Wherein:

[0072] The simulation environment construction unit establishes a virtual simulation environment corresponding to the actual material output scenario.

[0073] The environment training unit is used to perform the following initialization in the virtual simulation environment:

[0074] The initial position of the discharge robot arm is set directly above the electrolytic furnace.

[0075] By placing obstacles on the inner wall of the electrolytic furnace, the changes in the internal shape ratio of the electrolytic furnace during the electrolysis process are simulated, and the initial environment is randomly generated.

[0076] Initialize all parameters of the unloading robot.

[0077] The reinforcement learning training unit is used to perform the following reinforcement learning training process in a virtual simulation environment to build a rare earth discharge control model for simulating the end-effector trajectory:

[0078] Establish a pre-trained reinforcement learning neural network model;

[0079] Based on the current state information of the end effector of the unloading robot, the next action position is randomly generated through a pre-trained neural network model to form a planned path;

[0080] The simulator is used to execute the generated action positions, and to generate rewards and the next state information, thus completing a path planning simulation process.

[0081] Store the action location, execution probability, and reward during the simulation process;

[0082] The stored data is input into the neural network model to update the neural network model parameters and optimize the control strategy;

[0083] The training ends when the end of the discharge robot reaches the bottom of the electrolytic furnace or after the maximum number of training rounds. The training parameters of the current neural network model are retained to construct the rare earth discharge control model.

[0084] In some preferred embodiments, the simulation environment construction unit, which establishes a virtual simulation environment corresponding to the actual material discharge scenario, may further include:

[0085] Build a virtual simulation environment that corresponds to the actual material output scenario, and construct the following key objects within the virtual simulation environment:

[0086] Construct an electrolytic furnace and set parameters such as its size, shape, position, mass, and centroid to ensure that the virtual simulation environment can realistically reflect the physical relationships in the real discharge environment.

[0087] Construct a 3D model of the material discharge robot, and adjust and configure the 3D model according to the specifications of the actual material discharge robot;

[0088] Set up the sensors and ensure that their position, orientation, and parameters match those of the actual material-discharging robot in order to accurately simulate the sensor data output in the simulation.

[0089] Define the action space and the observation space. The action space is the joint motion space of the unloading robot, and the observation space is the end position and posture of the unloading tool and the force feedback force from the force sensor. Based on the dimensions and range of the action space and the observation space, set the control interface and the status monitoring interface of the unloading robot to ensure that the unloading robot can receive action commands and execute corresponding actions, and at the same time, it can acquire the status information in the observation space in real time, providing the necessary input and output interfaces for reinforcement learning training.

[0090] In some preferred embodiments, the compliant controller may further include:

[0091] The control unit is used to carry the rare earth discharge control model and to acquire the contact force information obtained by the force sensing module in real time as the input of the rare earth discharge control model. It obtains the real-time movement trajectory of the discharge tool and outputs it to the discharge robot controller to realize the compliant control of the discharge robot end effector.

[0092] It should be noted that the steps in the method provided by the present invention can be implemented using the corresponding components in the system. Those skilled in the art can refer to the technical solution of the system to implement the steps of the method, and can also refer to the technical solution of the method to implement the composition of the system. That is, the embodiments in the system and the embodiments in the method can be understood as preferred examples of each other, which will not be elaborated here.

[0093] Based on the same inventive concept, one embodiment of the present invention also provides a rare earth metal electrolysis furnace discharge robot.

[0094] Specifically, the rare earth metal electrolytic furnace discharge robot provided in this embodiment includes: a discharge robot body, which achieves compliant control of the discharge robot end effector through the method or system provided in the above embodiment of the present invention.

[0095] In this embodiment:

[0096] Force sensing module: High-precision six-dimensional torque sensors are distributed and embedded on the robot's end effector to acquire force information of the interaction between the robot and the electrolytic furnace in real time, covering the magnitude and direction of the contact force, providing data support for subsequent compliant control.

[0097] Compliant Controller: Based on reinforcement learning algorithm, it takes force sensor feedback data and robot joint position and speed information as input, processes them through a rare earth material discharge control model based on deep neural network, outputs the motion trajectory of the material discharge tool, and outputs the value to the robot control module.

[0098] The unloading robot controller adjusts the control strategy according to the motion trajectory of the unloading tool, including motion direction correction and force fine-tuning. It simulates the flexible operation of manual unloading by correcting the action in real time based on the contact force, and enhances the robot's adaptability to complex environments.

[0099] The reinforcement learning training module builds a virtual simulation platform that simulates the discharge scenario of a rare earth metal electrolytic furnace. It incorporates key elements such as high temperature, unstructured environment, and dynamic changes in the internal shape of the electrolytic furnace, allowing the robot to learn through trial and error in this environment. The robot optimizes its control strategy based on the feedback of the reward function until it achieves high-precision and high-stability discharge results in the simulated scenario. The mature strategy is then transferred to actual production to ensure the level of intelligence in the robot's discharge operation.

[0100] Discharge Operation Module: This module activates compliant control after the discharge tool enters the electrolysis furnace space. The robot moves along the trajectory of the discharge tool, which gradually approaches the bottom of the electrolysis furnace. During the approach, the end effector force sensing module continuously monitors the contact force between the discharge tool and the electrolysis furnace. The rare earth discharge control model corrects the tool trajectory based on the end effector contact force, ensuring that the tool tip maintains no contact or a low contact force with the inner wall of the electrolysis furnace. When the tool tip reaches the lowest position set by the trajectory or when the bottom contact force is too large to penetrate further, the robot stops for a period of time and then moves upward. During the movement, the system detects the contact force and adjusts the trajectory according to the rare earth discharge control model.

[0101] Force Sensing and Feedback Adjustment Module: Throughout the entire material discharge process, the force sensor at the robot's end effector monitors the interaction forces between the discharge tool and the electrolytic furnace and crucible in real time. This force information is transmitted to the compliant controller at high frequency, and the controller quickly adjusts the robot's end effector according to a preset force feedback control algorithm.

[0102] High-temperature protection design: Heat insulation and heat dissipation measures are taken for the electronic components and connecting lines of the force sensor and compliant controller. High-temperature resistant materials are used to make the protective shell, and the circuit layout is optimized to ensure that the key components operate stably under the high-temperature radiation of the electrolysis furnace, avoid performance degradation or failure caused by overheating, and maintain the continuity of material discharge operation.

[0103] The technical solution provided by the above embodiments of the present invention will be further described in detail below with reference to a specific application example.

[0104] like Figure 4 The diagram shown is a schematic of the installation structure of the force sensing module of the unloading robot. Figure 4 In the middle, 101-102 are force sensor modules, 103-104 are mechanical grippers designed with heat-insulating materials, and 105 is a discharge tool designed with high-temperature and corrosion-resistant metal materials.

[0105] Based on the methods and systems provided in the above embodiments of the present invention, in this specific application example, the following is adopted: Figure 3 The architecture shown features a high real-time compliant controller that receives and processes force sensor data via an EtherCAT bus to generate motion data; analog force sensor signals are acquired by a transmitter and converted into digital signals, which are then sent to the compliant controller via the EtherCAT bus; and three-dimensional force sensor measurement data is directly acquired by the transmitter.

[0106] Reinforcement learning training is conducted as follows: The training scenario is initialized, with the robotic arm position initialized to its initial position, directly above the electrolytic furnace. An initial environment is randomly generated, simulating changes in the furnace's internal shape during electrolysis by placing obstacles on the furnace's inner wall. The agent is initialized, i.e., its initial robot parameters. Training ends once all simulated scenarios are completed. The next action position is randomly generated based on the robot's current position, speed, and acceleration constraints. The simulator executes the generated action position, generating a reward and the next observation state. The action positions, execution probabilities, and observed rewards from the simulation process are stored. The observed data is input into the neural network to update the execution strategy. The current training round ends when the robot reaches the bottom of the electrolytic furnace or after the maximum number of rounds. One training round is defined as the tool descending from above the furnace to the bottom of the furnace.

[0107] Furthermore:

[0108] Force sensing system design: To address the issue of high temperatures in the electrolytic furnace, the torque sensor is installed away from the tool end (i.e., the end entering the electrolytic furnace). To solve the problem of excessively large measured torque, a distributed measurement method is adopted. That is, two three-dimensional force sensors are placed separately to achieve the measurement of large torques. To improve the real-time performance of the measurement, EtherCAT bus is used for data acquisition, with a data synchronization period of 1ms being preferred. Temperature-resistant force sensors are selected, capable of withstanding temperatures up to 80℃, to reduce the impact of high temperatures on the sensor's measurement accuracy. Thermal insulation material is used for installation between the sensor and the discharge tool.

[0109] Simulation Model Construction: A virtual simulation platform similar to the actual production environment is built using simulation software (such as a simulation software combining Gymnasium and Mujoco), including key objects such as electrolytic furnaces, robots, and tools. Parameters such as the size, shape, position, mass, and center of mass of the electrolytic furnace are precisely set to ensure that the simulation environment accurately reflects the physical relationships in the real-world scenario.

[0110] Configure the robot model: Import the robot's 3D model into the simulation software and adjust and configure it according to the actual robot's specifications. Install and set the sensors defined above, ensuring their position, orientation, and parameters are consistent with the actual robot to accurately simulate the sensor data output in the simulation.

[0111] Define the action space and observation space: Define the action space as the robot's joint motion space in the simulation software, and define the observation space as the tool end-effector's position and orientation, and the force sensor feedback force. Based on the dimensions and ranges of the defined action and observation spaces, set up the robot's control interface and state monitoring interface. Ensure that the robot can receive action commands and execute corresponding actions, while simultaneously acquiring real-time state information in the observation space, providing the necessary input / output interfaces for the reinforcement learning algorithm.

[0112] Definition of reinforcement learning rewards: Reinforcement learning rewards include contact force rewards, approach rewards, and control rewards. The smaller the contact force between the tool and the external environment detected by the sensor, the greater the contact force reward; the closer the tool tip is to the bottom of the electrolytic furnace, the greater the approach reward; the smaller the change in joint motion, the greater the control reward.

[0113] Reinforcement learning training: A Deep Q-Network (DQN) model is used for training. At each step, the robot selects an action (such as the linear velocity and angular velocity of the end effector) through the DQN network based on its current state (e.g., joint position, end effector force information). After executing the action in the simulation environment, new states and reward signals are collected (calculated based on task completion and action rationality). State transition information is stored in an experience replay buffer, and then small batches of data are randomly sampled from this buffer to train the DQN network, updating network parameters to optimize the strategy. Training ends after reaching a preset number of training steps or conditions. The training process is as follows: Figure 2 As shown.

[0114] Algorithm Deployment: In the robot control system, code is written to load and save the model parameters, and the model is integrated into the robot's control program. This ensures the model can receive the robot's state information and output corresponding control actions. The deployed system is tested in a simulated or real-world environment to verify whether the model's decisions meet expectations and whether the robot can stably and accurately complete the material unloading task. Simultaneously, the system's response speed, stability, and reliability are checked.

[0115] Material unloading process: After entering the electrolysis furnace space, the tool activates compliant control. The robot moves along the planned path, gradually approaching the bottom of the electrolysis furnace. During the approach, the end effector force sensing system continuously monitors the contact force between the robot and the electrolysis furnace. The algorithm model corrects the robot trajectory based on the end effector contact force, ensuring that the tool tip maintains no contact or a low contact force with the inner wall of the electrolysis furnace. When the tool tip reaches the lowest position set by the trajectory or when the bottom contact force is too large to penetrate further, the robot stops for a period of time and then moves upward. During the movement, the system detects the contact force and adjusts the trajectory according to the algorithm model.

[0116] Force sensing and feedback adjustment: Throughout the material discharge process, the force sensor at the robot's end effector monitors the interaction forces between the robot and the electrolytic furnace and crucible in real time. This force information is transmitted to the compliant controller at high frequency, and the controller quickly calculates the adjustment strategy for the robot's end effector based on a preset force feedback control algorithm.

[0117] The rare earth metal electrolytic furnace discharge robot and its end-effector compliant control method and system provided in the above embodiments of the present invention have the following significant technical effects:

[0118] In terms of operational intelligence: The robot's end effector is equipped with a high-precision six-dimensional force sensor to perceive force information in real time regarding its interaction with the electrolytic furnace, including the magnitude and direction of the contact force. A compliant controller based on reinforcement learning combines this force feedback data with the robot's joint position and velocity information. After processing by a deep neural network, it generates an adjustment strategy for the end effector in real time. This allows the robot to adjust its movements flexibly according to changes in contact force, much like a skilled worker, precisely responding to the high-temperature, unstructured, and dynamically changing discharge environment, greatly improving the level of operational intelligence and accurately completing the discharge task.

[0119] In terms of operational stability: The compliant controller outputs the end effector adjustment strategy in real time based on the force sensor feedback data and joint status information. It can quickly detect and respond to collision risks. The unloading robot calculates the motion parameters of each joint according to the input trajectory and controls the robotic arm joints to complete the movement, effectively preventing collisions and overloads, ensuring the continuity and stability of the unloading operation, avoiding production interruptions and equipment damage caused by collisions or overloads, and improving production efficiency and equipment lifespan.

[0120] Reduced labor costs: Automated material handling solutions significantly reduce manual intervention. With the continuous rise in labor costs, in the long run, this can significantly reduce enterprises' labor cost expenditures and bring considerable economic benefits to enterprises.

[0121] In terms of safe operation: Robots replace human labor in high-temperature and high-risk material unloading operations, completely eliminating the safety hazards of manual operation.

[0122] Regarding output quality: Thanks to its high-precision force sensing and compliant control capabilities, the robot can accurately complete the output operation, ensuring consistency in the cooling and forming quality of rare earth metals. This reduces defective products caused by human error or equipment malfunction, improves the overall product quality and performance, and helps enhance the company's market competitiveness and product added value.

[0123] In terms of adaptability and flexibility: Based on force feedback and reinforcement learning, the robot can autonomously adapt to the discharge requirements of different electrolytic furnace models and processes, reducing equipment replacement and debugging time and improving production flexibility and efficiency. Meanwhile, the collaborative control system can optimize the production process in real time based on the electrolytic furnace operating parameters and the robot's discharge progress, further improving production efficiency and enabling enterprises to respond more quickly to market changes and customer needs.

[0124] In terms of technological advancement: By combining reinforcement learning with end-effector force sensing technology, a novel intelligent solution is provided for the unloading operation of rare earth metal electrolytic furnaces. This technological integration not only improves the stability and quality of robotic unloading operations but also significantly reduces operational risks and costs. It represents the cutting-edge development direction of special robot technology in the rare earth metal production field and provides valuable reference for the automation upgrade of similar production scenarios.

[0125] Any matters not covered in the above embodiments of the present invention are well-known in the art.

[0126] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various modifications or variations within the scope of the claims, which do not affect the essence of the present invention.

Claims

1. A method for end-effector compliance control of a rare earth metal electrolytic furnace discharge robot, characterized in that, include: A rare earth discharge simulation environment is established to simulate discharge scenarios, and based on the simulation environment, a rare earth discharge control model is constructed to simulate the motion trajectory of the discharge tool through reinforcement learning training. The contact force information between the discharge tool and the electrolytic furnace during the discharge process of rare earth metal electrolytic furnace is acquired in real time and input into the rare earth discharge control model to obtain the real-time movement trajectory of the discharge tool. The real-time material discharge tool's motion trajectory is input to the material discharge robot controller to achieve compliant control of the material discharge robot's end effector; The establishment of a rare earth discharge simulation environment for simulating discharge scenarios includes: A virtual simulation environment corresponding to the actual material discharge scenario is established, and the following initialization is performed in the virtual simulation environment to obtain a rare earth material discharge simulation environment for simulating the material discharge scenario: The initial position of the discharge robot arm is set directly above the electrolytic furnace. By placing obstacles on the inner wall of the electrolytic furnace, the changes in the internal shape ratio of the electrolytic furnace during the electrolysis process are simulated, and the initial environment is randomly generated. Initialize all parameters of the unloading robot.

2. The end-effector compliance control method for the rare earth metal electrolytic furnace discharge robot according to claim 1, characterized in that, The establishment of a virtual simulation environment corresponding to the actual material output scenario includes: Build a virtual simulation environment that corresponds to the actual material output scenario, and construct the following key objects in the virtual simulation environment: Construct an electrolytic furnace and set its parameters to ensure that the virtual simulation environment can accurately reflect the physical relationships in the real discharge environment. Construct a 3D model of the material discharge robot, and adjust and configure the 3D model according to the specifications of the actual material discharge robot; Set up sensors and ensure that the position, orientation, and parameters of the sensors are consistent with the actual material discharge robot so as to accurately simulate the data output of the sensors in the simulation; Define an action space and an observation space, wherein the action space is the joint motion space of the unloading robot, and the observation space is the end position and posture of the unloading tool and the force feedback force from the force sensor; according to the dimensions and range of the action space and the observation space, set up a control interface and a status monitoring interface for the unloading robot to receive action commands and execute corresponding actions, and at the same time acquire the status information in the observation space, providing the necessary input and output interfaces for reinforcement learning training.

3. The end-effector compliance control method for the rare earth metal electrolytic furnace discharge robot according to claim 1, characterized in that, The rare earth discharge control model, based on the simulation environment and trained using reinforcement learning, is used to simulate the end-effector trajectory of the tool. This includes: A pre-trained reinforcement learning network model is established, and the reinforcement learning network model is trained as follows: Using the simulation environment, real-time status information of the end effector of the unloading robot is obtained; Based on the current state information of the end effector of the unloading robot, the next action position is randomly generated through the reinforcement learning network model to form a planned path; The simulator generates action positions, rewards, and next state information to complete a path planning simulation process. Store the action position, execution probability, and reward during the simulation process; The stored data is input into the reinforcement learning network model to update the parameters of the reinforcement learning network model and optimize the motion control strategy; Determine whether the end of the discharge robot has reached the bottom of the electrolytic furnace or whether the maximum number of training rounds has been reached. If yes, end the current round of training and retain the training parameters of the current reinforcement learning network model. If no, return to start the training process again until the training ends and a rare earth discharge control model is constructed.

4. The end-effector compliance control method for the rare earth metal electrolytic furnace discharge robot according to claim 1, characterized in that, The real-time acquisition of contact force information between the discharge tool and the electrolytic furnace during the rare earth metal electrolytic furnace discharge process, and inputting it into the rare earth discharge control model to obtain the real-time movement trajectory of the discharge tool, includes: The system acquires real-time force sensor feedback data between the discharge tool and the electrolytic furnace during the actual rare earth metal electrolytic furnace discharge process. This data, combined with the joint position and speed information of the discharge robot, serves as the input to the rare earth discharge control model. The model then outputs a corresponding motion control strategy, and the real-time motion trajectory of the discharge tool is obtained based on this strategy.

5. The end-effector compliance control method for the rare earth metal electrolytic furnace discharge robot according to claim 1, characterized in that, The step of inputting the real-time discharging tool's motion trajectory to the discharging robot controller to achieve compliant control of the discharging robot's end effector includes: The motion trajectory of the real-time unloading tool is input to the unloading robot controller. The unloading robot controller calculates the motion parameters of each joint according to the input trajectory and controls the joints of the robotic arm to complete the movement. It simulates manual unloading and corrects the action in real time according to the contact force, so as to achieve compliant control of the unloading robot end.

6. A compliant end-effector control system for a rare earth metal electrolytic furnace discharge robot used to implement the method of claim 1, characterized in that, include: Force sensing module, compliance controller, reinforcement learning training module; among which: The force sensing module is installed at the end of the discharge robot and is used to acquire contact force information between the discharge tool and the electrolytic furnace in real time during the discharge process of rare earth metal electrolytic furnace. The reinforcement learning training module is used to establish a rare earth discharge simulation environment for simulating discharge scenarios, and based on the simulation environment, to construct a rare earth discharge control model for simulating the motion trajectory of the discharge tool through reinforcement learning training. The compliant controller is used to carry the rare earth discharge control model and to obtain the contact force information acquired by the force sensing module in real time as the input of the rare earth discharge control model. It obtains the real-time movement trajectory of the discharge tool and outputs it to the discharge robot controller to realize the compliant control of the discharge robot end.

7. The end-effector compliance control system for the rare earth metal electrolytic furnace discharge robot according to claim 6, characterized in that, The force sensing module may further include any one or more of the following: A six-dimensional torque sensor is used and distributed on the end effector of the unloading robot; Data acquisition is performed using an EtherCAT bus, and the acquired analog signals are converted into digital signals via a transmitter; the data synchronization period is <2ms. Heat-insulating material is used for installation between the material and the discharge tool.

8. A rare earth metal electrolytic furnace discharge robot, comprising: The material discharge robot body is characterized in that the material discharge robot body achieves compliant control of the end effector of the material discharge robot by means of any one of claims 1-5 or any one of claims 6-7.

Images