Intelligent factory self-organizing multi-robot collaborative scheduling production method and system

Abstract

This invention discloses a method for self-organizing multi-robot collaborative scheduling production in a smart factory, comprising: acquiring historical production data of the factory and preprocessing it; inputting the preprocessed historical production data into a trained quantum machine learning model, outputting prediction results and recording them on a blockchain, wherein the prediction results include task execution time, robot movement path, and path length; acquiring real-time environmental information of the factory and inputting it into a trained deep reinforcement learning model, outputting task allocation and robot path planning, and recording the real-time environmental information and the corresponding task allocation and robot path planning on a blockchain; retrieving the real-time environmental information and the prediction results of the quantum machine learning model from the blockchain, and locally updating the deep reinforcement learning model; and sharing the locally updated model parameters with all robots for global model updates based on a distributed learning framework.

Description

Technical Field

[0001] This invention belongs to the field of intelligent scheduling production, specifically a method and system for intelligent factory self-organizing multi-robot collaborative scheduling production. Background Technology

[0002] In smart factories, robots need intelligent decision-making and learning capabilities. Using technologies such as reinforcement learning and deep learning, robots can make intelligent scheduling decisions based on the constantly changing production environment. They also learn optimal strategies to adapt to complex production environments. The Internet of Things (IoT) connects various devices, sensors, and robots in the factory, enabling them to share data via the internet and providing real-time information for robot scheduling. Meanwhile, artificial intelligence (AI) and machine learning technologies provide smart factories with intelligent decision-making and learning capabilities. By analyzing large amounts of historical data, machine learning algorithms can predict patterns and trends in the production environment, and by rapidly processing real-time data generated by sensors, the system can make timely decisions, thereby optimizing robot scheduling and task allocation.

[0003] Currently, Siemens' Industrial Internet of Things (IIoT) comprehensively monitors and manages production lines by connecting sensors, devices, and control systems to an IoT platform through various communication protocols, such as MQTT (Message Queuing Telemetry Transport). Real-time data is collected from various devices and sensors, including robots, sensors, and production equipment on the production line. Edge computing technology is deployed at the device level or on the production line, enabling some data processing to be performed on-site while a large amount of data is transmitted to the cloud. The cloud computing platform then performs in-depth analysis of the data, applying machine learning and data mining techniques to extract valuable information. Through data analysis and machine learning, production plans are dynamically optimized, and production tasks are adjusted based on real-time demand and resource status. This enables intelligent task scheduling, ensuring collaborative work among devices and improving overall production efficiency.

[0004] Existing methods for collaborative scheduling of multi-robot production in smart factories have the following drawbacks: 1. Limitations and singularity Some traditional production scheduling systems may rely too much on classical computing models and struggle to handle complex global optimization problems, especially when facing large-scale, highly dynamic factory environments.

[0005] 2. Data security and transparency Some systems have security vulnerabilities in data interaction and sharing, lacking effective trust mechanisms and data transparency, which may lead to information leakage or untrusted data transmission.

[0006] 3. Real-time performance and adaptability Some systems may struggle to monitor and adapt to changes in the production environment in real time, causing task scheduling and robot path planning to fail to respond to dynamic changes in a timely manner, thus reducing the system's flexibility and real-time performance.

[0007] 4. Learning efficiency Some traditional learning algorithms may be inefficient when processing large-scale data and have difficulty adapting to the differences in data quality and importance among different robots. Summary of the Invention

[0008] To address the shortcomings of traditional scheduling systems in handling complex global optimizations, as well as their security and transparency deficiencies in data interaction and sharing, and to meet the challenges of real-time adaptability and collaborative operation of multiple robots in the dynamic environment of smart factories, this invention provides a self-organizing multi-robot collaborative scheduling production method and system for smart factories. By employing quantum machine learning, establishing a blockchain trust mechanism, and modeling an adaptive scheduling algorithm based on deep reinforcement learning, this invention achieves more efficient, flexible, secure, and cost-effective self-organizing multi-robot collaborative scheduling in smart factories.

[0009] According to one aspect of the present invention, a method for self-organizing multi-robot collaborative scheduling production in a smart factory is provided, comprising: Acquire and preprocess historical production data from the factory; The preprocessed historical production data is input into the trained quantum machine learning model, and the prediction results are output and recorded on the blockchain. The prediction results include the task execution time, robot movement path and path length. The system acquires real-time environmental information from the factory and inputs it into a trained deep reinforcement learning model. It outputs task allocation and robot path planning and records the real-time environmental information and the corresponding task allocation and robot path planning on the blockchain. The real-time environmental information and prediction results of the quantum machine learning model are retrieved from the blockchain, and the deep reinforcement learning model is updated locally. Based on a distributed learning framework, locally updated model parameters are shared with all robots for global model updates.

[0010] As a further technical solution, the training of the quantum machine learning model includes: Quantum support vector machine was selected as the machine learning algorithm for quantum computing to build a quantum machine learning model; The preprocessed task execution time, robot movement path, production task complexity, and environmental factors are input into the quantum machine learning model, and the model is trained based on the quantum computing framework. The parameters of a quantum machine learning model are optimized using a quantum approximation optimization algorithm. The optimized model parameters are used to predict and output the task execution time, robot motion path, and path length.

[0011] As a further technical solution, a quantum approximation optimization algorithm is used to optimize the parameters of the quantum machine learning model, including: The weights, learning rate, and number of iterations of the quantum machine learning model are optimized.

[0012] As a further technical solution, based on the optimized model parameters, the method for predicting and outputting task execution time, robot motion path, and path length also includes: A coding mapping strategy is developed to map specific properties of quantum states to classical coding forms.

[0013] As a further technical solution, the method also includes: Choose a blockchain platform for the Ethereum smart factory scenario, design the data structure recorded in the blockchain, and write smart contracts to define collaboration rules and execute tasks. Record the interactions and task scheduling information between robots on the blockchain; Consensus algorithms and network mechanisms are used to keep the blockchain updated and synchronized in real time, ensuring that all participants receive the latest interaction and task scheduling information.

[0014] As a further technical solution, the training of the deep reinforcement learning model includes: A deep reinforcement learning model is constructed. The input of the deep reinforcement learning model is the robot's current position, task execution status, and production equipment status. The output is task allocation and robot path planning. Build a training dataset to train the model and output the trained deep reinforcement learning model.

[0015] As a further technical solution, the training of the deep reinforcement learning model also includes: An experience replay mechanism is adopted, which periodically samples randomly from historical experience data and uses the randomly sampled data to train the model.

[0016] As a further technical solution, the method also includes: Build a distributed learning framework and use a federated learning algorithm for global model updates; A weighted local training strategy is used to implement a distributed training strategy; The update strategy of federated learning algorithm is implemented by using gradient descent or model fusion.

[0017] According to one aspect of the present invention, a smart factory self-organizing multi-robot collaborative scheduling production system is provided, comprising: The data acquisition module is used to collect data from robots and production equipment in the factory in real time and record it on the blockchain; The quantum computing module is used to acquire historical production data from the factory and preprocess it. The preprocessed historical production data is then input into the trained quantum machine learning model, which outputs prediction results and records them on the blockchain. The prediction results include task execution time, robot movement path, and path length. The adaptive scheduling module is used to acquire real-time environmental information of the factory and input it into the trained deep reinforcement learning model, output task allocation and robot path planning, and record the real-time environmental information and corresponding task allocation and robot path planning on the blockchain; retrieve the real-time environmental information and the prediction results of the quantum machine learning model from the blockchain to update the deep reinforcement learning model locally; and based on the distributed learning framework, share the locally updated model parameters with all robots for global model update.

[0018] According to one aspect of the present invention, a non-transitory computer read storage medium is provided, the non-transitory computer read storage medium storing computer instructions that cause the computer to execute the intelligent factory self-organizing multi-robot collaborative scheduling production method.

[0019] Compared with the prior art, the beneficial effects of the present invention are: (1) This invention adopts cutting-edge technologies such as quantum machine learning, blockchain trust mechanism and deep reinforcement learning, and makes full use of the advantages of each technology to form a comprehensive and efficient production scheduling system.

[0020] (2) This invention uses a quantum machine learning model to process global optimization problems, giving full play to the advantages of quantum computing and improving the efficiency of processing complex optimization problems.

[0021] (3) This invention establishes a trust mechanism through blockchain technology, records the interaction and task scheduling information between robots, ensures the security and transparency of data, and provides a reliable information sharing platform for multi-party participation.

[0022] (4) This invention utilizes a deep reinforcement learning model to adjust task allocation and robot path planning in real time, enabling the system to adapt to the dynamic changes in the actual production environment and improving the system's adaptability.

[0023] (5) The present invention adopts a federated learning approach, which enables the robot to train the model locally and share the model parameters through a distributed learning framework, thereby improving learning efficiency and the optimization level of the global model.

[0024] (6) This invention utilizes quantum machine learning models and deep reinforcement learning models to dynamically optimize the production environment, monitor system performance in real time, and upload system feedback information to the blockchain through blockchain smart contracts to ensure information security.

[0025] In summary, this invention can effectively improve key performance indicators such as production efficiency, reduce task completion time, and optimize robot path length, making smart factories more intelligent, efficient, and safe. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the intelligent factory self-organizing multi-robot collaborative scheduling production method according to an embodiment of the present invention.

[0027] Figure 2 This is a schematic diagram illustrating the principle of the intelligent factory self-organizing multi-robot collaborative scheduling production method according to an embodiment of the present invention.

[0028] Figure 3 This is a framework diagram of the intelligent factory self-organizing multi-robot collaborative scheduling production system according to an embodiment of the present invention. Detailed Implementation

[0029] To address the challenges of traditional scheduling systems in handling complex global optimizations, their deficiencies in data interaction and sharing regarding security and transparency, and the real-time and adaptability issues arising from the dynamic changes in smart factory environments and the collaborative cooperation of multiple robots, this invention proposes a self-organizing multi-robot collaborative scheduling production method for smart factories. This method mainly includes the following parts: 1. Quantum Machine Learning Model Training: Based on historical data, quantum machine learning algorithms (such as QSVM or QNN) are used in conjunction with quantum computers for model training, and parameters are optimized using quantum optimization algorithms.

[0030] 2. Blockchain establishes a trust mechanism: By selecting a suitable blockchain platform, designing data structures and smart contracts, and recording the interactions and task scheduling information between robots, data security and transparency can be ensured.

[0031] 3. Adaptive scheduling algorithm modeling based on deep reinforcement learning: Deploy sensors on each robot and production equipment, and use deep reinforcement learning models to adjust task allocation and robot path planning in real time to adapt to the dynamic changes in the actual production environment.

[0032] 4. Distributed learning framework: It adopts a federated learning approach, which enables robots to train models locally and receive and update global model parameters through a central server. At the same time, it takes into account the data quality and importance of different robots to implement a weighted local training strategy.

[0033] 5. System Optimization and Feedback: The production environment is dynamically optimized using quantum machine learning models and deep reinforcement learning models. Quantum machine learning models dynamically optimize the production environment by processing complex data. Deep reinforcement learning models can adjust task allocation and robot path planning in real time during runtime to adapt to changes in the actual production environment. Through continuous learning and updating, it improves the system's adaptability.

[0034] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] This invention utilizes cutting-edge technologies such as quantum machine learning, establishing blockchain trust mechanisms, and adaptive scheduling algorithm modeling based on deep reinforcement learning to construct a self-organizing multi-robot collaborative scheduling production method for smart factories. Please refer to [link to relevant documentation]. Figure 1 The process includes: acquiring and preprocessing historical production data from the factory; inputting the preprocessed historical production data into a trained quantum machine learning model, outputting prediction results and recording them on a blockchain, wherein the prediction results include task execution time, robot movement path, and path length; acquiring real-time environmental information from the factory and inputting it into a trained deep reinforcement learning model, outputting task allocation and robot path planning, and recording the real-time environmental information and corresponding task allocation and robot path planning on a blockchain; retrieving the real-time environmental information and the prediction results from the quantum machine learning model from the blockchain, and locally updating the deep reinforcement learning model; and sharing the locally updated model parameters with all robots for global model updates based on a distributed learning framework.

[0036] As one implementation method, please refer to Figure 2 The specific technical process is as follows: 1. Quantum machine learning model training; 2. Blockchain trust mechanism establishment; 3. Modeling of adaptive scheduling algorithm based on deep reinforcement learning; 4. Distributed learning framework; 5. System optimization and feedback.

[0037] In the training of quantum machine learning models, historical data is first collected and preprocessed; then the quantum machine learning model is trained.

[0038] 1.1 Collect past production data from the factory, including task execution time, robot movement paths, task complexity, environmental factors (temperature and humidity, internal layout of the plant), etc., and preprocess the data. Preprocessing includes data cleaning, noise reduction, and normalization to ensure data consistency and accuracy. Historical data collection and preprocessing are the foundation of the entire process, providing high-quality data input for subsequent quantum machine learning model training. Normalization includes: , c represents the normalized data. and These represent the maximum and minimum values ​​of the original data, respectively.

[0039] 1.2 In the training of the quantum machine learning model, a machine learning algorithm for quantum computing is selected: Quantum Support Vector Machine (QSVM). The machine learning model is trained using a quantum computer, employing the quantum machine learning framework Qiskit to simplify the training process and predict tasks such as execution time, robot movement paths, and path lengths. The output of the trained quantum machine learning model (predicted task execution time, robot movement paths, and path lengths) will serve as input to the blockchain smart contract and scheduling algorithm to ensure the accuracy of task scheduling and path planning.

[0040] Specifically, the input data for the quantum machine learning model is: 1) Task execution time: The execution time of each task, including start and end times. This data helps the model understand the time distribution and execution efficiency of tasks.

[0041] 2) Robot movement path: The robot's movement path, including the start point, end point, and path length. By analyzing this path data, the robot's movement strategy can be optimized.

[0042] 3) Complexity of production tasks: Information on the complexity of different production tasks, including the type of task, the resources required, and the complexity of the execution steps. More complex tasks may require more time and resources.

[0043] 4) Environmental factors: The main environmental factors affecting production include temperature and humidity, equipment status, and factory layout. These factors can affect the production process and robot scheduling.

[0044] The output data is: 1) Predict task execution time: The model predicts the execution time of each production task, helping to optimize task scheduling and resource allocation.

[0045] 2) Robot movement path: The model predicts the best movement path for the robot during the task execution process, ensuring the shortest path and the highest efficiency.

[0046] 3) Path length: The total length of the robot's predicted movement path, used to evaluate and optimize the overall efficiency of the robot's motion.

[0047] The model training steps are as follows: Quantum Support Vector Machine (QSVM) was chosen as the primary algorithm. QSVM can effectively handle complex high-dimensional data and leverages the advantages of quantum computing to improve computational efficiency.

[0048] QSVM models are built and trained using quantum computing frameworks such as Qiskit. Qiskit provides tools for building and running quantum algorithms. Preprocessed input data (task execution time, robot movement path, task complexity, environmental factors) is fed into the QSVM model for training. The model is trained and evaluated using Qiskit, and its parameters are continuously adjusted to improve prediction accuracy.

[0049] 3) The parameters of the QSVM machine learning model are optimized using the Quantum Approximation Optimization Algorithm (QAOA). QAOA is a hybrid optimization algorithm combining classical and quantum computing, suitable for solving complex combinatorial optimization problems. QAOA optimizes the model's parameter settings through quantum computing, including weights (e.g., rotations and entanglements of each qubit are considered weights in the network), learning rate (a hyperparameter controlling the step size of parameter adjustments in each iteration), and number of iterations (the number of iterations the optimization algorithm runs, determining the depth of the search for the optimal solution), finding a parameter setting that maximizes the model's prediction accuracy. Ultimately, this enables the quantum machine learning model to optimally predict outputs such as task execution time, robot motion path, and path length.

[0050] In this embodiment of the invention, a dimensionality reduction operation is performed as a preprocessing step before quantum computing to reduce the burden on the quantum computer.

[0051] Preferably, after obtaining the results from the quantum computer, the method further includes: interpreting the results obtained from the quantum computer and translating them into a form that can be understood by a classical computer. Simultaneously, an encoding mapping strategy is formulated to map specific properties of the quantum state to classical encoding, such as mapping the phase of the quantum state to a part of classical encoding.

[0052] , and These correspond to the sine and cosine terms in classical encoding, respectively. It is a complex factor representing the possible global phase, with 0 and 1 representing the ground state of the qubit, respectively. In this example, the embodiment of the invention uses binary encoding to represent the two possible states of the classical bit.

[0053] In establishing a trust mechanism in a blockchain, this embodiment of the invention includes: 2.1 When selecting a blockchain platform for the Ethereum smart factory scenario, factors such as platform performance, privacy, scalability, and smart contract support should be considered. The blockchain establishes a trust mechanism by recording the prediction results and real-time scheduling information of the quantum machine learning model on the blockchain, ensuring data transparency and security. Simultaneously, the data on the blockchain provides reliable historical data and real-time information for deep reinforcement learning algorithms.

[0054] 2.2 Design the data structure recorded in the blockchain, including interactions between robots, task scheduling information, and related metadata. Simultaneously, write smart contracts to define collaboration rules and execute tasks.

[0055] 2.3 Record the interactions and task scheduling information between robots on the blockchain, including key information such as task start time, end time, and robot execution status. Simultaneously record the interactions and task scheduling information between robots to ensure data security and transparency.

[0056] 2.4 Consensus algorithms and network mechanisms are used to maintain the real-time updates and synchronization of the blockchain, ensuring that all participants can obtain the latest interaction and scheduling information.

[0057] The adaptive scheduling algorithm decision based on deep reinforcement learning in this embodiment of the invention includes: 3.1 Data Acquisition and Real-time Processing: Sensors are deployed on each robot and production equipment to collect data in real time, including the robot's current position, task execution status, and production equipment status. The collected data is processed in real time to provide the deep reinforcement learning model with the latest environmental information, ensuring that decisions are based on the actual state of the current production line.

[0058] 3.2 An experience replay mechanism is used to periodically sample randomly from historical experience to improve training efficiency and stability, thereby enabling the model to be continuously updated to adapt to changes in the production environment. This random sampling from historical experience can be considered as random sampling from historical operational data and other recorded data of the production management system.

[0059] 3.3 Each robot perceives changes in the production environment in real time, including task status, robot position, and production equipment status. Based on the current state, a trained deep reinforcement learning model is used to select the optimal action, i.e., adjusting task allocation and robot path planning.

[0060] 3.4 Regularly train the model offline, using the latest data to improve its performance. Train the real-time task allocation and path planning models using parameters provided by blockchain and quantum machine learning models. Deep reinforcement learning algorithms continuously update the model using real-time data and historical experience collected in step 3.1, improving the accuracy and adaptability of scheduling decisions.

[0061] The distributed learning framework of this invention includes: 4.1 The distributed learning framework ensures that multiple robots can train models locally and share model parameters through federated learning to form a globally optimized model. The result of the distributed learning framework is a more efficient model for the adaptive scheduling algorithm, improving the overall system performance. Using the federated learning distributed learning framework, robots can train models locally and then share the updated model parameters with other robots. The central server receives the local model parameters from all robots and uses the federated learning algorithm to update the global model. , These are global model parameters. The global model parameters are obtained by summing the local model parameters of all robots and dividing by the number of robots N, which represents the number of robots.

[0062] 4.2 Simultaneously, a distributed training strategy is implemented. When the robot is trained locally, considering that different robots may have different data quality and importance, a weighted local training strategy is adopted (these weights can be based on the size of the data, the representativeness of the data, and the reliability of the robot): , This represents the quality and importance of robot data. This represents the robot's local data. This represents the weight of different data from the robot.

[0063] 4.3 Implement update strategies in federated learning algorithms through gradient descent, model fusion, and other methods.

[0064] System optimization and feedback in this embodiment of the invention include: 5.1 Task allocation and path planning: Based on real-time environmental status data (such as the robot's current position, task execution status, and production equipment status), deep reinforcement learning algorithms dynamically adjust task allocation and robot paths. This ensures that tasks are allocated and executed in the most efficient way and optimizes the robot's motion path.

[0065] 5.2 Adaptive Adjustment: Deep reinforcement learning algorithms adjust their strategies and parameters (i.e., update the parameters of the reinforcement learning model) based on feedback from real-time data to adapt to changes in the production environment.

[0066] 5.3 The system's feedback information and the dynamic optimization results of the quantum machine learning model are recorded on the blockchain via smart contracts to ensure information security. Through the combined application of deep reinforcement learning and quantum machine learning models, the production environment is dynamically optimized. The system's feedback information and optimization results are recorded on the blockchain via smart contracts to ensure information security and provide optimized data for the next cycle's scheduling.

[0067] Based on the aforementioned steps, this invention achieves a more efficient, flexible, safe, and cost-effective intelligent factory self-organizing multi-robot collaborative scheduling production.

[0068] Based on the same inventive concept as the foregoing embodiments, this invention also provides a smart factory self-organizing multi-robot collaborative scheduling production system, such as... Figure 3 As shown, the system includes: a data acquisition module, installed on robots and production equipment, for real-time acquisition of data from robots and production equipment within the factory and recording it on a blockchain; a quantum computing module, implemented using a quantum computer, for acquiring historical production data from the factory and preprocessing it, inputting the preprocessed historical production data into a trained quantum machine learning model, outputting prediction results and recording them on the blockchain, the prediction results including task execution time, robot movement path, and path length; an adaptive scheduling module, which can use a computer as an adaptive scheduling device, for acquiring real-time environmental information from the factory and inputting it into a trained deep reinforcement learning model, outputting task allocation and robot path planning, and recording the real-time environmental information and corresponding task allocation and robot path planning on the blockchain; retrieving the real-time environmental information and the prediction results of the quantum machine learning model from the blockchain to locally update the deep reinforcement learning model; and sharing the locally updated model parameters with all robots based on a distributed learning framework for global model updates.

[0069] The core of this invention is the use of cutting-edge technologies such as quantum machine learning, edge computing and blockchain, adaptive scheduling algorithms based on deep reinforcement learning, and distributed learning frameworks to construct a self-organizing multi-robot collaborative scheduling production system for intelligent factories. Through the comprehensive application of these advanced technologies, this invention constructs an intelligent factory scheduling system that enables multiple robots to collaboratively and efficiently cope with complex production environments, achieving self-organizing scheduling and dynamic optimization.

[0070] Based on the same inventive concept as the foregoing embodiments, this embodiment of the invention also provides a non-transitory computer read-only storage medium, which stores computer instructions that cause the computer to execute the energy consumption prediction method for a low-carbon construction production line. Acquire and preprocess historical production data from the factory; The preprocessed historical production data is input into the trained quantum machine learning model, and the prediction results are output and recorded on the blockchain. The prediction results include the task execution time, robot movement path and path length. The system acquires real-time environmental information from the factory and inputs it into a trained deep reinforcement learning model. It outputs task allocation and robot path planning and records the real-time environmental information and the corresponding task allocation and robot path planning on the blockchain. The real-time environmental information and prediction results of the quantum machine learning model are retrieved from the blockchain, and the deep reinforcement learning model is updated locally. Based on a distributed learning framework, locally updated model parameters are shared with all robots for global model updates.

[0071] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, located in one place, or distributed across multiple network units. The purpose of this embodiment is achieved by selecting some or all of the modules according to actual needs. Those skilled in the art will understand and implement this without any inventive effort.

[0072] It should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0073] In summary, the key points and areas to be protected in this invention are as follows: 1. Training a quantum machine learning model This study utilizes a quantum machine learning framework to protect the use of quantum computers for model training and to apply quantum optimization algorithms for parameter optimization. Global optimization is achieved, including aspects such as production scheduling and resource allocation.

[0074] 2. Blockchain Trust Mechanism The blockchain trust mechanism ensures data security and transparency within the blockchain, especially in robot interaction and task scheduling.

[0075] 3. Adaptive Scheduling Algorithm By utilizing an adaptive scheduling algorithm based on deep reinforcement learning, the robot is adapted in real time and dynamically to changes in the actual production environment. This enables it to continuously learn and improve its adaptability over time, demonstrating its real-time adaptability.

[0076] 4. Distributed learning framework By employing a federated learning and distributed training strategy, weighted local training and weighted local training are used to instruct different robots on how to contribute to the global model based on their data quality and importance.

[0077] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent factory self-organizing multi-robot collaborative scheduling production, characterized in that, include: Acquire and preprocess historical production data from the factory; The preprocessed historical production data is input into the trained quantum machine learning model, and the prediction results are output and recorded on the blockchain. The prediction results include the task execution time, robot movement path and path length. The system acquires real-time environmental information from the factory and inputs it into a trained deep reinforcement learning model. It outputs task allocation and robot path planning and records the real-time environmental information and the corresponding task allocation and robot path planning on the blockchain. The real-time environmental information and prediction results of the quantum machine learning model are retrieved from the blockchain, and the deep reinforcement learning model is updated locally. Based on a distributed learning framework, locally updated model parameters are shared with all robots for global model updates.

2. The intelligent factory self-organizing multi-robot collaborative scheduling production method according to claim 1, characterized in that, The training of the quantum machine learning model includes: Quantum support vector machine was selected as the machine learning algorithm for quantum computing to build a quantum machine learning model; The preprocessed task execution time, robot movement path, production task complexity, and environmental factors are input into the quantum machine learning model, and the model is trained based on the quantum computing framework. The parameters of a quantum machine learning model are optimized using a quantum approximation optimization algorithm. The optimized model parameters are used to predict and output the task execution time, robot motion path, and path length.

3. The intelligent factory self-organizing multi-robot collaborative scheduling production method according to claim 2, characterized in that, The parameters of a quantum machine learning model are optimized using a quantum approximation optimization algorithm, including: The weights, learning rate, and number of iterations of the quantum machine learning model are optimized.

4. The intelligent factory self-organizing multi-robot collaborative scheduling production method according to claim 2, characterized in that, The optimized model parameters are used to predict and output the task execution time, robot motion path, and path length, and also include: A coding mapping strategy is developed to map specific properties of quantum states to classical coding forms.

5. The intelligent factory self-organizing multi-robot collaborative scheduling production method according to claim 1, characterized in that, The method further includes: Choose a blockchain platform for the Ethereum smart factory scenario, design the data structure recorded in the blockchain, and write smart contracts to define collaboration rules and execute tasks. Record the interactions and task scheduling information between robots on the blockchain; Consensus algorithms and network mechanisms are used to keep the blockchain updated and synchronized in real time, ensuring that all participants receive the latest interaction and task scheduling information.

6. The intelligent factory self-organizing multi-robot collaborative scheduling production method according to claim 1, characterized in that, The training of the deep reinforcement learning model includes: A deep reinforcement learning model is constructed. The input of the deep reinforcement learning model is the robot's current position, task execution status, and production equipment status. The output is task allocation and robot path planning. Build a training dataset to train the model and output the trained deep reinforcement learning model.

7. The intelligent factory self-organizing multi-robot collaborative scheduling production method according to claim 6, characterized in that, The training of the deep reinforcement learning model also includes: An experience replay mechanism is adopted, which periodically samples randomly from historical experience data and uses the randomly sampled data to train the model.

8. The intelligent factory self-organizing multi-robot collaborative scheduling production method according to claim 1, characterized in that, The method further includes: Build a distributed learning framework and use a federated learning algorithm for global model updates; A weighted local training strategy is used to implement a distributed training strategy; The update strategy of federated learning algorithm is implemented by using gradient descent or model fusion.

9. A smart factory self-organizing multi-robot collaborative scheduling production system, characterized in that, include: The data acquisition module is used to collect data from robots and production equipment in the factory in real time and record it on the blockchain; The quantum computing module is used to acquire historical production data from the factory and preprocess it. The preprocessed historical production data is then input into the trained quantum machine learning model, which outputs prediction results and records them on the blockchain. The prediction results include task execution time, robot movement path, and path length. The adaptive scheduling module is used to acquire real-time environmental information of the factory and input it into the trained deep reinforcement learning model, output task allocation and robot path planning, and record the real-time environmental information and corresponding task allocation and robot path planning on the blockchain; retrieve the real-time environmental information and the prediction results of the quantum machine learning model from the blockchain to update the deep reinforcement learning model locally; and based on the distributed learning framework, share the locally updated model parameters with all robots for global model update.

10. A non-transitory computer read storage medium, characterized in that, The non-transitory computer reads and stores computer instructions on the storage medium, which cause the computer to execute the intelligent factory self-organizing multi-robot collaborative scheduling production method according to any one of claims 1 to 8.

Images