Industrial robot material sorting system with intelligent dispatching function

Abstract

The invention relates to the field of industrial automation, in particular to an intelligent dispatching industrial robot material sorting system. The system integrates various sensors such as a visual sensor and a laser radar sensor, and high-precision material identification is realized by applying multi-source information fusion and deep learning; a sorting path is dynamically planned through reinforcement learning, operation of multiple robots is coordinated according to a distributed collaborative optimization algorithm, and the motion stability is improved through self-adaptive control; and by matching with an intelligent monitoring management platform, data acquisition, analysis and early warning diagnosis are realized. Through practical application, the system effectively improves the material recognition accuracy and sorting efficiency, enhances the multi-robot cooperative ability and motion stability, and meets the efficient and accurate sorting requirements of modern industry.

Description

Technical Field

[0001] The present invention belongs to the field of industrial automation, and in particular relates to an intelligently scheduled industrial robot material sorting system. Background Art

[0002] In industrial production, material sorting is an important link, which is widely used in logistics warehousing, electronic manufacturing, food processing and other industries. With the continuous expansion of industrial production scale and the increasing requirements for production efficiency, the traditional manual material sorting method can no longer meet the needs. Industrial robots have gradually become an important tool for material sorting with their advantages of high speed, high precision and repeatability. However, there are still some urgent problems to be solved in the material sorting of industrial robots, which restricts their further development and application in this field.

[0003] Low material recognition accuracy: There are many types of materials in industrial production, with different shapes, colors, and materials, and the differences between some materials are very subtle. Traditional industrial robot material sorting systems mainly rely on a single visual recognition technology, which is easily affected by factors such as lighting conditions, material surface texture, and occlusion, resulting in low material recognition accuracy. In the actual sorting process, misidentification often occurs, which not only reduces the sorting efficiency, but may also cause product quality problems.

[0004] Unreasonable sorting path planning: In complex material sorting scenarios, there are often multiple sorting targets and obstacles, requiring industrial robots to plan an efficient and safe sorting path. Existing sorting path planning methods are mostly based on simple geometric models and fixed rules, which cannot fully consider the robot's motion characteristics, material distribution, and real-time changes in the working environment. This makes it easy for robots to collide and take detours during the sorting process, resulting in low sorting efficiency and increased energy consumption.

[0005] Insufficient collaborative sorting capabilities of multiple robots: In order to improve the efficiency of large-scale material sorting, multiple industrial robots are usually required to work together. However, the existing multi-robot collaborative sorting system lacks an effective task allocation and coordination mechanism, and problems such as work conflicts and repeated operations are prone to occur between robots. At the same time, due to the lack of real-time information interaction and collaborative control, each robot cannot adjust its work strategy in time according to changes in the overall task, making it difficult to achieve efficient collaborative operations, reducing the overall sorting efficiency. Summary of the invention

[0006] The present invention provides an intelligent scheduling industrial robot material sorting system, aiming to solve problems such as efficiency, accuracy, and adaptability faced by industrial robots during the material sorting process. By introducing advanced sensing technologies, intelligent algorithms, and optimized scheduling strategies, the overall performance of industrial robot material sorting is improved to meet the requirements of modern industrial production for efficient and accurate material sorting, including:

[0007] High-precision material recognition module based on multi-source information fusion: Integrate various types of sensors such as vision, lidar, infrared, and tactile sensors on the industrial robot to collect multi-source information of materials in real time. Let the data collected by the vision sensor be V, the data collected by the lidar be L, the data collected by the infrared sensor be I, and the data collected by the tactile sensor be T. Use the weighted fusion formula M = ωVV + ωLL + ωII + ωTT, where ωV, ωL, ωI, and ωT are the weights of the corresponding sensor data, determined by training through machine learning methods. Integrate and process different sensor data to highlight key information, and then combine with the deep learning module to achieve high-precision recognition of materials;

[0008] Dynamic sorting path planning module based on reinforcement learning: Construct the material sorting process as a reinforcement learning environment, and define in detail the state space including information such as the position of materials, the position of obstacles, the position and posture of the robot, the action space covering the robot's movement, rotation, grasping, and placing actions, and the reward function designed based on the robot's sorting efficiency, energy consumption, and collision situation. Through continuous interaction and learning between the robot and the environment, obtain the optimal sorting path planning strategy;

[0009] Multi-robot task allocation and coordination module based on distributed collaborative optimization: Construct multiple industrial robots into a distributed collaborative system. Comprehensively consider factors such as the working ability of each robot, the current task status, and the priority of materials, and construct a distributed objective function and constraint conditions including robot working ability limitations, task sequence constraints, resource constraints, etc. Use the distributed optimization module to collaboratively optimize the task allocation, working paths, and operation sequences of multiple robots;

[0010] Robot motion stability improvement system based on adaptive control: Install various sensors such as accelerometers, gyroscopes, and force sensors at key parts such as the joints and end effectors of the industrial robot to monitor the motion state and force conditions of the robot in real time. Use the adaptive control module to automatically adjust the motion control parameters such as the joint torque, speed, and acceleration of the robot according to the data collected by the sensors;

[0011] Intelligent Monitoring and Management Platform for Material Sorting Process: This platform conducts data interaction with the industrial robot material sorting system, real-time collects and stores various types of data during the sorting process, such as material identification results, sorting paths, robot motion states, task completion status, etc. It uses big data analysis and visualization technologies to achieve intelligent monitoring and management of the sorting process, and has functions of intelligent early warning and fault diagnosis.

[0012] Furthermore, the sensors used in the high-precision material identification module based on multi-source information fusion include visual sensors, lidar, infrared sensors, and tactile sensors.

[0013] Furthermore, the dynamic sorting path planning module based on reinforcement learning adopts a deep Q-network (DQN) or proximal policy optimization module (PPO) reinforcement learning module.

[0014] Furthermore, the multi-robot task allocation and coordination module based on distributed collaborative optimization adopts a distributed particle swarm optimization module (DPSO) or distributed genetic algorithm module (DGA) distributed optimization module.

[0015] Furthermore, the robot motion stability improvement system based on adaptive control installs multiple sensors such as acceleration sensors, gyroscopes, and force sensors on industrial robots.

[0016] Furthermore, in the high-precision material identification module based on multi-source information fusion, a convolutional neural network (CNN) is used for feature extraction.

[0017] Furthermore, the dynamic sorting path planning module based on reinforcement learning allows the robot to continuously try different actions in the environment and adjusts its behavior according to the reward feedback, thus learning an efficient sorting path planning strategy.

[0018] Furthermore, in the multi-robot task allocation and coordination module based on distributed collaborative optimization, each robot exchanges information through a real-time communication network and flexibly adjusts its working strategy according to the global task requirements and its own status.

[0019] Furthermore, the robot motion stability improvement system based on adaptive control can dynamically adjust the control parameters according to the real-time motion and force conditions of the robot to ensure the stable operation of the robot.

[0020] Furthermore, the intelligent monitoring and management platform for the material sorting process can provide valuable information such as sorting efficiency analysis and material identification accuracy statistics for management personnel through the analysis of the stored data.

[0021] Beneficial Effects

[0022] Improve the accuracy of material identification: The high-precision material identification algorithm based on multi-source information fusion can comprehensively utilize the information of multiple sensors to achieve high-precision identification of various materials, improve the accuracy and stability of material identification, reduce misidentification, and improve sorting efficiency and product quality.

[0023] Optimize the sorting path planning: The dynamic sorting path planning algorithm based on reinforcement learning enables industrial robots to dynamically plan the optimal sorting path according to the real-time working environment and task requirements, avoiding collisions and detours, and improving sorting efficiency and energy utilization rate.

[0024] Enhance the collaborative sorting ability of multiple robots: The multi-robot task allocation and coordination algorithm based on distributed collaborative optimization can achieve efficient collaborative work among multiple robots, avoid work conflicts and duplicate operations, improve the overall sorting efficiency, and meet the needs of large-scale material sorting.

[0025] Enhance the motion stability of the robot: The robot motion stability improvement system based on adaptive control can real-time monitor the motion state and force conditions of the robot, automatically adjust the motion control parameters, improve the stability and reliability of the robot motion, and reduce material dropping and sorting errors caused by unstable robot motion.

[0026] Facilitate the monitoring and management of the sorting process: The intelligent monitoring and management platform for the material sorting process provides real-time sorting process monitoring and decision support for managers, has intelligent early warning and fault diagnosis functions, ensures the smooth progress of the sorting process, and improves the production management level. Description of the Drawings

[0027] Figure 1 Flow chart of the system operation principle. Detailed Implementation Manner

[0028] Example 1

[0029] I. System Setup

[0030] In the material sorting workshop of an electronic component manufacturing factory, a set of industrial robot material sorting system based on multi-source information fusion and intelligent scheduling is deployed.

[0031] (I) Perception Layer

[0032] Six industrial robots are selected for material sorting operations. Each robot is equipped with multiple sensors:

[0033] 1. Vision sensor: Install a high-precision industrial camera, model MV-CA050-10GC, with a frame rate of 100fps and a resolution of 2448×2048 pixels, used to obtain the appearance, shape, and color information of the materials.

[0034] 2. LiDAR: The SICK TIM571 LiDAR is adopted, with a scanning range of 360°, a ranging accuracy of ±10 mm, and it can accurately measure the distances between the material, the robot, and the obstacles.

[0035] 3. Infrared sensor: The OPB700 infrared sensor is selected, with a detection distance of 0 - 100 mm, and it can quickly sense the presence and approximate position of the material.

[0036] 4. Tactile sensor: The ATI Nano17 six - axis force / torque sensor is installed at the end - effector of the robot, which can accurately measure the force and torque when grasping the material, and prevent the material from being damaged due to improper grasping force.

[0037] 5. Accelerometer: The ADXL345 accelerometer is installed at the joints of the robot, with a measurement range of ±16 g, and it is used to monitor the acceleration changes during the robot's movement.

[0038] 6. Gyroscope: The MPU6050 gyroscope is adopted, which can obtain the robot's attitude information in real - time to ensure stable movement.

[0039] 7. Force sensor: In addition to the six - axis force / torque sensor at the end - effector, single - direction force sensors are added at key stress parts, such as the FUTEK LSB200 force sensor, to monitor the stress conditions of each part of the robot. These sensors collect data in real - time at a frequency of 500 Hz and transmit the data to the data - processing layer through the industrial Ethernet.

[0040] (2) Data - processing layer

[0041] 1. Multi - source information fusion module: A multi - source information fusion algorithm based on deep learning is used, specifically a fusion network based on the attention mechanism. This module performs fusion processing on the data collected by multi - source sensors to eliminate data conflicts and redundancies. For example, when the vision sensor detects the shape of the material, the LiDAR obtains the distance of the material, and the infrared sensor senses the position of the material, the fusion network automatically learns the correlation relationships between these information sources, highlights the key information, and improves the accuracy of information fusion.

[0042] 2. Feature extraction module: A convolutional neural network (CNN) is used to extract the material features and the robot's motion state information from the fused data. For example, through operations such as convolution and pooling on the vision image data, features such as the edges and textures of the material are extracted; according to the data from the LiDAR, accelerometer, etc., features such as the robot's motion trajectory and speed are extracted.

[0043] 3. Reinforcement Learning Model Training Module: Use the Proximal Policy Optimization (PPO) algorithm to model and train the material sorting process. The state space includes information such as the material position, obstacle position, robot position, and posture; the action space includes actions such as the robot's movement, rotation, grasping, and placing; the reward function is designed based on factors such as the robot's sorting efficiency, energy consumption, and whether there is a collision. After a large number of trainings, the robot gradually learns the optimal sorting path planning strategy.

[0044] 4. Adaptive Control Parameter Adjustment Module: Use the Model Predictive Control (MPC) algorithm to calculate the adjustment amount of the robot's motion control parameters in real time according to the data collected by the sensors. For example, when the acceleration sensor detects abnormal acceleration of the robot's motion, or the force sensor feedbacks insufficient grasping force, this module automatically adjusts the motion control parameters such as the robot's joint torque, speed, and acceleration.

[0045] (III) Decision-making Layer

[0046] 1. High-precision Material Identification Module: Based on the results of the multi-source information fusion module and the feature extraction module, use the high-precision material identification algorithm based on multi-source information fusion to achieve high-precision identification of materials. For example, by comparing the fused material features with the pre-trained material feature library, different models of electronic components can be accurately identified.

[0047] 2. Dynamic Sorting Path Planning Module: Based on the results of the reinforcement learning model training module, combined with the real-time material distribution and environmental information, dynamically plan the optimal sorting path. For example, when new obstacles are added in the workshop or the material position changes, this module can quickly adjust the robot's sorting path to avoid collisions and detours.

[0048] 3. Multi-robot Task Allocation and Coordination Module: According to the global task requirements and the states of each robot, use the multi-robot task allocation and coordination algorithm based on distributed collaborative optimization for task allocation and coordination. For example, according to the priority of the materials, the working ability of the robots, and the current task status, reasonably allocate the sorting tasks of each robot, and coordinate their working paths and operation sequences to avoid work conflicts and duplicate operations.

[0049] 4. Motion Stability Control Module: Control the motion stability of the robot according to the results of the adaptive control parameter adjustment module. For example, when the adaptive control parameter adjustment module calculates that the joint torque needs to be adjusted to stabilize the robot's motion, this module sends the control instruction to the robot in the execution layer to ensure its stable and reliable motion.

[0050] 5. Intelligent Monitoring and Management Module: Responsible for collecting, storing, analyzing, and presenting data during the sorting process. It collects data such as material identification results, sorting paths, robot motion states, and task completion status in real time and stores them in the database. Using big data analysis technology, it analyzes and mines the data, such as calculating sorting efficiency and statistically analyzing the material identification accuracy rate, and presents it to the management personnel through a visualization interface to achieve intelligent monitoring and management of the sorting process.

[0051] (IV) Execution Layer

[0052] Industrial robots execute tasks such as material identification, sorting path planning, material grasping, and placement according to the control instructions generated by the decision-making layer. For example, after the robot identifies the material through a vision sensor, it moves to the material position along the path planned by the dynamic sorting path planning module, grasps the material using the end effector, and places it at the designated position. At the same time, feedback information during the execution process, such as grasping force and motion state, is transmitted back to the data processing layer and the decision-making layer for system adjustment and optimization.

[0053] (V) Application Layer

[0054] The intelligent monitoring and management platform for the material sorting process provides a friendly interaction interface for management personnel. Through this platform, management personnel can monitor the sorting process in real time, view information such as the working status of robots, material identification accuracy rate, and sorting efficiency. For example, on the platform interface, the sorting efficiency changes at different times of the day and the identification accuracy rates of different types of materials are presented in the form of charts. The platform also has intelligent warning and fault diagnosis functions. When a robot fails or the material identification accuracy rate is lower than the set threshold, it automatically issues a warning signal and conducts fault diagnosis to prompt management personnel to take corresponding measures.

[0055] II. Algorithm Implementation

[0056] (I) Implementation of High-Precision Material Identification Algorithm Based on Multi-Source Information Fusion

[0057] 1. Sensor Deployment and Data Collection: According to the characteristics of electronic components and sorting requirements, the installation angles and positions of vision sensors are reasonably adjusted to ensure that the materials can be photographed comprehensively and clearly. The lidar, infrared sensors, and tactile sensors are also optimized for deployment. Each sensor collects data in real time according to the set frequency and transmits it to the multi-source information fusion module in the data processing layer through the network.

[0058] 2. Implementation of Multi-source Information Fusion Algorithm: The multi-source information fusion module adopts a fusion network based on the attention mechanism to fuse different types of sensor data. For example, when identifying a tiny electronic chip, the detailed appearance information of the chip provided by the vision sensor and the precise position information of the chip measured by the lidar are fused through the fusion network to highlight the key features and improve the recognition accuracy of the chip.

[0059] 3. Material Feature Extraction and Recognition: The feature extraction module uses a convolutional neural network (CNN) to analyze and process the fused data. For example, through multi-layer convolution and pooling operations, the unique features of the electronic chip are extracted. The high-precision material recognition module uses these features and combines them with a pre-trained recognition model to accurately identify the model and specifications of the chip.

[0060] (II) Implementation of the Dynamic Sorting Path Planning Algorithm Based on Reinforcement Learning

[0061] 1. Definition of State Space, Action Space, and Reward Function: The state space accurately includes information such as the position of the material on the shelf, the position of obstacles in the workshop, the position and posture of the robot in the working area, etc. The action space covers the movement of the robot in the XYZ axis directions, rotation around the axis, grasping and placing actions of the end effector, etc. The reward function is designed based on factors such as the time for the robot to complete a sorting task, the power consumption, and whether it collides with obstacles. For example, if the robot can accurately complete the sorting task in the shortest time without collision, it will receive a higher reward; conversely, if there is a collision or the sorting time is too long, it will receive a lower reward.

[0062] 2. Training of the Reinforcement Learning Model: The reinforcement learning model training module uses the Proximal Policy Optimization algorithm (PPO) to model and train the material sorting process. During the training process, the robot continuously interacts with the environment, selects actions according to the current state, observes the reward signal and the new state feedback by the environment after executing the actions, and updates the policy network through the optimization algorithm to maximize the cumulative reward. After thousands of trainings, the robot gradually learns the optimal sorting path planning strategy under different material distributions and environmental conditions.

[0063] 3. Execution of the Dynamic Sorting Path Planning: During the actual sorting process, the dynamic sorting path planning module dynamically plans the optimal sorting path according to the real-time material distribution, obstacle positions, and the motion state of the robot, in combination with the trained reinforcement learning model. For example, when a handling device is temporarily added as a new obstacle in the workshop, this module can quickly re-plan the sorting path of the robot according to the reinforcement learning model to ensure the smooth progress of the sorting task.

[0064] (III) Implementation of the Multi-robot Task Allocation and Coordination Algorithm Based on Distributed Cooperative Optimization

[0065] 1. Task analysis and modeling: According to the production orders of the day, conduct a detailed analysis and modeling of the material sorting tasks. Determine the sorting priorities of different types of electronic components, the completion time required by the orders, etc. At the same time, evaluate and model the working capabilities (such as maximum load, movement speed), working ranges, and current task status of each industrial robot.

[0066] 2. Implementation of distributed collaborative optimization algorithm: Use the distributed particle swarm optimization algorithm (DPSO) to construct distributed objective functions and constraint conditions. The objective function is set to maximize the sorting efficiency, that is, to complete all material sorting tasks in the shortest possible time. The constraint conditions include the working capacity limitations of the robots (such as the load of each robot cannot exceed its maximum load), task sequence constraints (such as certain components must be sorted before assembly), resource constraints (such as the space limitation of the shared material placement area), etc. Through the optimization algorithm, collaboratively optimize the task allocation, working paths, and operation sequences of multiple robots to obtain the optimal task allocation plan and collaborative working strategy.

[0067] 3. Execution of multi-robot collaborative work: The multi-robot task allocation and coordination module sends the optimized task allocation plan and collaborative working strategy to each industrial robot. During the actual work process, the robots communicate with each other through a real-time communication network. For example, when a robot completes its current task, it promptly sends a task completion signal to other robots so that other robots can dynamically adjust their working strategies according to the global task requirements and their own status. When abnormal situations such as task changes (such as new emergency orders) and equipment failures occur, the multi-robot task allocation and coordination module can promptly re-allocate and coordinate tasks to ensure the efficiency and stability of multi-robot collaborative work.

[0068] (4) Implementation of a robot motion stability improvement system based on adaptive control

[0069] 1. Sensor deployment and data collection: Accurately install various sensors such as acceleration sensors, gyroscopes, and force sensors at key parts such as the joints and end effectors of industrial robots. These sensors monitor the motion state and force conditions of the robots in real time at high frequencies and transmit the data to the adaptive control parameter adjustment module in the data processing layer. For example, the acceleration sensor monitors the acceleration changes of the robot during motion in real time, and the force sensor accurately measures the force when the robot grasps the material.

[0070] 2. Adaptive control algorithm implementation: The adaptive control parameter adjustment module uses the model predictive control (MPC) algorithm to calculate the adjustment amount of the robot's motion control parameters in real time based on the data collected by the sensors. For example, when the gyroscope detects a deviation in the robot's posture or the force sensor feedback shows unstable grasping force, this module calculates the adjustment amounts of parameters such as joint torque, speed, and acceleration that need to be adjusted through the MPC algorithm. The motion stability control module automatically adjusts the robot's motion control parameters according to the adjustment amounts to adapt to different workloads and working environments.

[0071] 3. Execution of improving the robot's motion stability: The industrial robot at the execution layer adjusts its own motion state according to the control instructions sent by the motion stability control module. For example, after the motion stability control module issues an instruction to adjust the joint torque, the motor of the corresponding joint of the robot adjusts the output torque, thereby improving the stability and reliability of the motion and reducing material dropping and sorting errors caused by unstable motion.

[0072] (V) Implementation of the intelligent monitoring and management platform for the material sorting process

[0073] 1. Data collection and storage: The intelligent monitoring and management module collects various data during the sorting process in real time through the data interface with the industrial robot material sorting system. For example, the material identification results are collected once per minute, the sorting path information is collected once every 5 seconds, the robot motion state data is collected once per second, and the task completion status is collected once per completed order. After formatting the collected data, it is stored in the SQL Server database to establish a detailed digital file of the sorting process.

[0074] 2. Data analysis and visualization: Use big data analysis technology to analyze and mine the data stored in the database.

[0075] For example, by analyzing the sorting data within a week, calculate the average sorting efficiency at different times of each day and count the recognition accuracy of different types of materials. Through visualization technology, display the analysis results in the form of bar charts, line charts, pie charts, etc. on the interface of the intelligent monitoring and management platform. For example, display the sorting quantities of different robots on the same day in a bar chart and show the change trend of the sorting efficiency within a week in a line chart, providing intuitive and clear monitoring and decision-making support for managers.

[0076] 3. Implementation of intelligent warning and fault diagnosis functions: The intelligent monitoring and management platform has intelligent warning and fault diagnosis functions.

[0077] By setting thresholds and analysis models, the data during the sorting process is monitored in real time. For example, set the threshold for the material recognition accuracy rate to 98%. When the recognition accuracy rate is lower than this threshold for 5 consecutive times, the platform automatically issues a warning signal. At the same time, by analyzing the motion state data of the robot, the fault diagnosis model is used to determine whether the robot has faults, such as overheating of the motor, jamming of the joints, etc., and prompts the management staff to take corresponding maintenance measures to ensure the smooth progress of the sorting process.

Claims

1. An intelligent scheduling industrial robot material sorting system, characterized in that, Including: High-precision material recognition module based on multi-source information fusion: Integrate various types of sensors such as vision, lidar, infrared, and tactile sensors on the industrial robot to collect multi-source information of the material in real time. Let the data collected by the vision sensor be V, the data collected by the lidar be L, the data collected by the infrared sensor be I, and the data collected by the tactile sensor be T. Use the weighted fusion formula M = ωVV + ωLL + ωII + ωTT, where ωV, ωL, ωI, and ωT are the weights of the corresponding sensor data, which are determined by training through machine learning methods. Integrate and process the data of different sensors to highlight key information, and then combine with the deep learning module to achieve high-precision recognition of the material; Dynamic sorting path planning module based on reinforcement learning: Construct the material sorting process as a reinforcement learning environment, and define in detail the state space including information such as the position of the material, the position of obstacles, the position and posture of the robot, the action space covering the actions of the robot moving, rotating, grasping, and placing, and the reward function designed based on the sorting efficiency, energy consumption, and collision situation of the robot. Through the continuous interaction and learning between the robot and the environment, obtain the optimal sorting path planning strategy; Multi-robot task allocation and coordination module based on distributed collaborative optimization: Construct multiple industrial robots into a distributed collaborative system. Comprehensively consider factors such as the working ability of each robot, the current task status, and the priority of the material, and construct a distributed objective function and constraint conditions including robot working ability limitations, task sequence constraints, resource constraints, etc. Use the distributed optimization module to perform collaborative optimization on the task allocation, working path, and operation sequence of multiple robots; Robot motion stability improvement system based on adaptive control: Install various sensors such as accelerometers, gyroscopes, and force sensors at key parts such as the joints and end effectors of the industrial robot to monitor the motion state and force conditions of the robot in real time. Use the adaptive control module to automatically adjust the motion control parameters such as the joint torque, speed, and acceleration of the robot according to the data collected by the sensors; Intelligent monitoring and management platform for the material sorting process: This platform conducts data interaction with the industrial robot material sorting system, collects and stores various types of data during the sorting process in real time, such as material recognition results, sorting paths, robot motion states, task completion situations, etc. Use big data analysis and visualization technologies to achieve intelligent monitoring and management of the sorting process, and have functions of intelligent early warning and fault diagnosis.

2. The system according to claim 1, characterized in that, The sensors used in the high-precision material recognition module based on multi-source information fusion include vision sensors, lidar, infrared sensors, and tactile sensors.

3. The system according to claim 1, wherein The dynamic sorting path planning module based on reinforcement learning adopts a deep Q-network (DQN) or proximal policy optimization module (PPO) reinforcement learning module.

4. The system according to claim 1, wherein The multi-robot task allocation and coordination module based on distributed collaborative optimization adopts a distributed particle swarm optimization module (DPSO) or distributed genetic module (DGA) distributed optimization module.

5. The system according to claim 1, wherein The robot motion stability improvement system based on adaptive control installs various sensors such as accelerometers, gyroscopes, and force sensors on the industrial robot.

6. The system according to claim 1, wherein In the high-precision material identification module based on multi-source information fusion, a convolutional neural network (CNN) is used for feature extraction.

7. The system according to claim 1, wherein The dynamic sorting path planning module based on reinforcement learning enables the robot to continuously try different actions in the environment and adjust its behavior according to the reward feedback, thus learning an efficient sorting path planning strategy.

8. The system according to claim 1, wherein The multi-robot task allocation and coordination module based on distributed cooperative optimization allows each robot to interact information through a real-time communication network and flexibly adjust its working strategy according to the global task requirements and its own status.

9. The system according to claim 1, characterized in that, The robot motion stability improvement system based on adaptive control can dynamically adjust the control parameters according to the real-time motion and force conditions of the robot to ensure the stable operation of the robot.

10. The system according to claim 1, wherein The intelligent monitoring and management platform for the material sorting process can provide valuable information such as sorting efficiency analysis and material identification accuracy statistics for managers through the analysis of stored data.

Images