Robot assembly optimization method and system based on federal learning

Abstract

The invention discloses a robot assembly optimization method and system based on federated learning, and the method comprises the steps: obtaining the image data and point cloud data of a robot, and carrying out the data feature fusion recognition processing, and obtaining the comprehensive feature data of the robot; obtaining force sense information and visual information of the robot, and performing motion error optimization processing based on deep reinforcement learning in combination with the comprehensive feature data of the robot to obtain the comprehensive feature data of the robot after error compensation; performing model training on the robot through a federated learning optimization algorithm, and constructing a robot with a global assembly control model; and based on the meta-learning strategy, performing communication cooperation on the robot with the global assembly control model to execute an assembly task, and obtaining a robot assembly optimization result. According to the embodiment of the invention, the environment perception capability of the robot can be improved, and the response speed and adaptability of the robot are effectively improved in a multi-task switching scene. The method can be widely applied to the technical field of robot assembly.

Description

Technical Field

[0001] The present application relates to the field of robot assembly technology, and in particular to a robot assembly optimization method and system based on federated learning. Background Art

[0002] With the rapid development of robotics technology, industrial robots are increasingly used in the field of precision assembly. Especially in the assembly process of tiny parts, robots need to have high precision and flexibility to cope with complex and changing environments and assembly requirements. However, at present, most robot assembly systems rely on a single sensor to perceive the assembly environment. Due to the limitations of different sensors in obtaining information, the use of a single sensor is easily interfered by factors such as ambient light, reflection, and occlusion, resulting in insufficient environmental perception accuracy. Even if some solutions introduce sensor fusion technology, due to the complexity and imperfection of the data fusion algorithm, the accuracy and robustness of the fused information are still difficult to fully guarantee. In the precision assembly process, in the existing assembly system, error compensation is usually based on simple feedback control rather than complex real-time learning models. The feedback mechanism lags and the error compensation effect is limited. Especially in dynamic environments or complex tasks, it is difficult for the system to adjust and optimize quickly and adaptively. The current deep learning models used in robot precision assembly are mostly trained in offline environments. The assembly tasks and environmental conditions faced by each robot may be different, and a single model is difficult to adapt to the needs of all scenarios. Moreover, once the model is deployed, it is difficult to update, which makes it impossible for robots to learn and adapt to new situations in real time in actual applications. In addition, there is a lack of effective knowledge sharing and collaboration mechanisms between robots. Each robot can only rely on its own data and experience to operate, resulting in low learning efficiency and poor task execution. Most related federated learning algorithms rely on synchronous update strategies, and there are still delay problems in the processing of large-scale, multimodal perception data. The real-time performance is insufficient, and they are easily affected by Byzantine attacks, data pollution and other problems. In addition, there is a lack of sufficient security protection mechanisms, which can easily lead to data leakage or system crashes.

[0003] In summary, the technical problems existing in the relevant technologies need to be improved. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to propose a robot assembly optimization method and system based on federated learning, which can enhance the robot's ability to perceive the environment and effectively improve the robot's response speed and adaptability in multi-task switching scenarios.

[0005] To achieve the above objectives, one aspect of an embodiment of the present application proposes a robot assembly optimization method based on federated learning, the method comprising:

[0006] Acquire the robot's image data and point cloud data and perform data feature fusion and recognition processing to obtain the robot's comprehensive feature data;

[0007] Acquire force information and visual information of the robot, and perform motion error optimization processing based on deep reinforcement learning in combination with the comprehensive feature data of the robot to obtain comprehensive feature data of the robot after error compensation;

[0008] Based on the comprehensive feature data of the robot after error compensation, a model of the robot is trained by a federated learning optimization algorithm to construct a robot with a global assembly control model;

[0009] Based on the meta-learning strategy, the robots with the global assembly control model communicate and collaborate to perform assembly tasks, and obtain robot assembly optimization results.

[0010] In some embodiments, the acquiring of the robot's image data and point cloud data and performing data feature fusion recognition processing to obtain the robot's comprehensive feature data includes:

[0011] Calibrate the camera and the laser radar to obtain a calibrated camera and a calibrated laser radar;

[0012] Acquire image data of the robot through the calibrated camera and acquire point cloud data of the robot through the calibrated laser radar;

[0013] Performing data preprocessing and data fusion processing on the image data of the robot and the point cloud data of the robot to obtain fused robot data;

[0014] Introduce image branch module, point cloud branch module and feature fusion layer to build a deep learning fusion model;

[0015] Based on the deep learning fusion model, feature recognition and fusion processing is performed on the fused robot data to obtain comprehensive feature data of the robot.

[0016] In some embodiments, the acquiring of the robot's force information and the robot's visual information, and combining the robot's comprehensive feature data with the robot's comprehensive feature data to perform motion error optimization processing based on deep reinforcement learning to obtain the robot's comprehensive feature data after error compensation, includes:

[0017] Acquiring force information of the robot through a force sensor and acquiring visual information of the robot through a visual sensor;

[0018] Defining a state space of deep reinforcement learning according to the comprehensive feature data of the robot, defining an action space of deep reinforcement learning according to the force information of the robot and the visual information of the robot, setting an experience replay pool, and performing interactive learning to obtain preliminary optimized comprehensive feature data of the robot;

[0019] Controlling the robot to perform simulation according to the preliminary optimized comprehensive characteristic data of the robot to obtain a robot motion deviation vector;

[0020] The robot motion deviation vector is error compensated by an error compensation algorithm, and is optimized in combination with a deep reinforcement learning algorithm to obtain comprehensive feature data of the robot after error compensation.

[0021] In some embodiments, the robot model is trained based on the error-compensated robot comprehensive feature data by using a federated learning optimization algorithm to construct a robot with a global assembly control model, including:

[0022] Controlling the robot to perform an assembly task according to the robot comprehensive feature data after error compensation, and constructing a local data set according to the assembly execution result;

[0023] Performing data preprocessing on the local data set to obtain a preprocessed local data set;

[0024] A robot assembly control model is constructed based on a multi-sensor fusion model and a motion control model, and trained using a small batch gradient descent algorithm to obtain a trained robot assembly control model.

[0025] The pre-processed local data set is compressed and uploaded to the trained robot assembly control model through an AES symmetric encryption algorithm;

[0026] Based on the trained robot assembly control model, weight parameters are assigned to the preprocessed local data set and aggregation processing is performed through an asynchronous federated averaging algorithm to obtain a global assembly control model;

[0027] The global assembly control model is sent to a robot control terminal to obtain the robot having the global assembly control model.

[0028] In some embodiments, the robot with the global assembly control model is communicated and coordinated to perform the assembly task based on the meta-learning strategy to obtain the robot assembly optimization result, including:

[0029] Based on the industrial communication protocol, the robots having the global assembly control model are networked for communication;

[0030] A personalized network model is introduced to perform global network fine-tuning processing on the robot having the global assembly control model to obtain a fine-tuned robot;

[0031] A meta-learning strategy is introduced to perform adaptive training on the fine-tuned robot to obtain an adaptively trained robot;

[0032] Based on a continuous learning mechanism, dynamically adjusting the adaptively trained robot to obtain a robot with multi-task compatibility;

[0033] The robot assembly optimization result is obtained by executing the assembly task based on the robot with multi-task compatibility.

[0034] In some embodiments, the personalized network model includes a plurality of shared layers and a plurality of dedicated layers, and output ends of the plurality of shared layers are connected to input ends of the plurality of dedicated layers, wherein:

[0035] The shared layers are used to extract common features of the robot;

[0036] The dedicated layers are used to extract the personalized demand features of the robot.

[0037] In some embodiments, the introducing of a meta-learning strategy to adaptively train the fine-tuned robot to obtain an adaptively trained robot includes:

[0038] Initializing the fine-tuned robot by using a meta-learning algorithm to obtain an initialized robot;

[0039] The end control and force feedback parameters of the initialized robot are adjusted to obtain the adaptively trained robot.

[0040] In some embodiments, the method of dynamically adjusting the adaptively trained robot based on a continuous learning mechanism to obtain a robot with multi-task compatibility includes:

[0041] Based on a continuous learning mechanism, dynamically adjusting the learning rate of the adaptively trained robot to obtain an adjusted robot;

[0042] The adjusted robot is subjected to knowledge transfer learning between models through a knowledge distillation algorithm to obtain the robot with multi-task compatibility.

[0043] In some embodiments, the method further includes adding noise during the process of communicating and cooperating with the robots having the global assembly control model to perform assembly tasks.

[0044] To achieve the above objectives, another aspect of the embodiments of the present application provides a robot assembly optimization system based on federated learning, the system comprising:

[0045] The first module is used to obtain the robot's image data and point cloud data and perform data feature fusion recognition processing to obtain the robot's comprehensive feature data;

[0046] The second module is used to obtain the force information and visual information of the robot, and to perform motion error optimization processing based on deep reinforcement learning in combination with the comprehensive feature data of the robot to obtain the comprehensive feature data of the robot after error compensation;

[0047] The third module is used to perform model training on the robot through a federated learning optimization algorithm based on the comprehensive feature data of the robot after the error compensation, so as to construct a robot with a global assembly control model;

[0048] The fourth module is used to communicate and cooperate with the robots with the global assembly control model to perform assembly tasks based on a meta-learning strategy to obtain robot assembly optimization results.

[0049] The embodiments of the present application include at least the following beneficial effects: The present application provides a robot assembly optimization method and system based on federated learning. The solution acquires the robot's image data and point cloud data and performs data feature fusion recognition processing. Through multi-sensor fusion and deep learning control algorithms, the robot has high-precision environmental perception capabilities in complex assembly tasks, and then acquires the robot's force information and the robot's visual information, and combines the robot's comprehensive feature data to perform motion error optimization processing based on deep reinforcement learning. Through real-time feedback from force sensors and visual sensors, assembly errors are detected and automatically compensated. The error detection module combines force and visual information to calculate the deviation vector in real time. The compensation algorithm quickly corrects the motion path through the controller to ensure high-precision operation of the robot during the assembly process. The error compensation module predicts historical error patterns through deep learning and makes adjustments in advance to improve the speed and accuracy of compensation. The robot model is further trained through a federated learning optimization algorithm based on the comprehensive feature data of the robot after error compensation. In the multi-robot collaboration framework, distributed model training and sharing of each robot are achieved through federated learning. Finally, based on the meta-learning strategy, robots with a global assembly control model communicate and collaborate to perform assembly tasks. The trained robot model has the ability to adapt quickly, so that it can quickly adjust to new task requirements through a small number of training steps when switching tasks or changing environments. This effectively improves the response speed and adaptability of the system in multi-task switching scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1is a flow chart of a robot assembly optimization method based on federated learning provided in an embodiment of the present application;

[0051] Figure 2 It is a structural diagram of a robot assembly optimization system based on federated learning provided in an embodiment of the present application. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are only examples of systems and methods consistent with some aspects of the embodiments of the present application as detailed in the attached claims.

[0053] It is understood that the terms "first", "second", etc. used in this application can be used to describe various concepts in this article, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another concept. For example, without departing from the scope of the embodiment of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein can be interpreted as "at the time of" or "when" or "in response to determination".

[0054] The terms "at least one", "multiple", "each", "any", etc. used in this application, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0056] There are some deficiencies in the related technology, such as:

[0057] 1) Reliance on a single sensor and insufficient perception capabilities.

[0058] 2) Lack of efficient error compensation mechanism;

[0059] 3) Limitations of deep learning models and insufficient localized adaptability.

[0060] 4) Multi-robot collaboration is inefficient and lacks an effective knowledge sharing mechanism.

[0061] 5) Real-time and asynchrony issues in federated learning applications.

[0062] 6) Security and reliability issues.

[0063] In view of this, a robot assembly optimization method based on federated learning is provided in the embodiment of the present application. By distributing training and aggregating models without sharing data among nodes (robots), data privacy can be protected while enhancing the generalization ability and learning effect of the model. In the robot precision assembly scenario, federated learning can optimize the model using the perception and assembly data of different robots in their respective working environments, thereby improving the accuracy, real-time and adaptability of the overall assembly.

[0064] Reference Figure 1 , Figure 1 A flowchart of a robot assembly optimization method based on federated learning provided by an embodiment of the present invention, referring to Figure 1 , the method comprises the following steps:

[0065] S100, acquiring the image data and point cloud data of the robot and performing data feature fusion recognition processing to obtain comprehensive feature data of the robot;

[0066] It should be noted that, in some embodiments, step S100 may include steps S110 to S150.

[0067] S110, performing calibration processing on the camera and the laser radar to obtain a calibrated camera and a calibrated laser radar;

[0068] In step S110 of some embodiments, illustratively, the sensor is calibrated, the internal and external parameters of the camera are calibrated using a calibration plate or a specific marker, the intrinsic parameter matrix and distortion parameters of the camera are obtained, the lidar is calibrated, the conversion relationship between its coordinate system and the robot coordinate system is obtained, and the precise coordinate transformation matrix between the camera, lidar and the robot end effector is determined through a hand-eye calibration method.

[0069] S120, acquiring image data of the robot through the calibrated camera and acquiring point cloud data of the robot through the calibrated laser radar;

[0070] In step S120 of some embodiments, illustratively, the camera and the lidar start data acquisition at the same time, the camera captures image frames, the lidar scans to obtain point cloud data, and the camera image and lidar data are synchronized through timestamps to ensure the time consistency of multi-sensor data.

[0071] S130, performing data preprocessing and data fusion processing on the image data and the point cloud data of the robot to obtain fused robot data;

[0072] In step S130 of some embodiments, the image data is preprocessed, illustratively, and operations such as image denoising, graying, and enhancement are performed to improve image quality. ROI (region of interest) extraction is used to focus on the assembly area to reduce the amount of calculation, and then the point cloud data is preprocessed, downsampled, filtered, and noise points and abnormal points are removed. According to the assembly task, the key point cloud area is extracted, and finally spatial fusion is performed. Using known sensor calibration parameters, the point cloud data of the lidar is projected into the image coordinate system of the camera, or the image data is mapped to the point cloud coordinate system to achieve spatial alignment.

[0073] S140, introduce an image branch module, a point cloud branch module and a feature fusion layer to build a deep learning fusion model;

[0074] S150. Perform feature recognition and fusion processing on the fused robot data based on the deep learning fusion model to obtain comprehensive feature data of the robot.

[0075] In steps S140-S150 of some embodiments, illustratively, a fusion convolutional neural network (Fusion-CNN) is used, which is composed of an image processing branch and a point cloud processing branch. The image branch uses classic network structures such as ResNet and VGG to extract image features. The point cloud branch uses networks such as PointNet and PointCNN to extract point cloud features. The feature fusion layer fuses the image features and the point cloud features to generate a comprehensive feature representation, and then constructs a training set containing multimodal data, with labels such as the position, posture, category and other information of the assembly object. Use supervised learning methods to minimize the loss function (such as cross entropy and mean square error) between the predicted results and the true labels. Use data enhancement technology to improve the robustness of the model, generate high-precision position information and posture information of the assembly object, and detection results of environmental obstacles, and provide data support for subsequent motion planning.

[0076] S200, acquiring force information and visual information of the robot, and performing motion error optimization processing based on deep reinforcement learning in combination with comprehensive feature data of the robot to obtain comprehensive feature data of the robot after error compensation;

[0077] It should be noted that, in some embodiments, step S200 may include steps S210 to S240.

[0078] S210, acquiring force information of the robot through a force sensor and acquiring visual information of the robot through a visual sensor;

[0079] In step S210 of some embodiments, illustratively, a force sensor is installed on the robot end effector, which can monitor the force and torque applied during the assembly process in real time. When an abnormal force change is detected, the system determines that there may be an assembly error or collision. Through force feedback, the system can capture tiny mechanical changes that are difficult to detect with the naked eye, which is crucial for error detection in precision assembly. For example, when the position of the part clamped by the actuator is slightly offset during the assembly process, the force sensor will sense that the force on the actuator is no longer balanced. The system will determine the direction and size of the error based on this information, so as to compensate for it and capture the tiny displacement changes in the assembly process in real time through a high-resolution camera. The visual sensor can detect whether the assembly object is in the correct predetermined position and whether the posture is accurate. When the assembly object is offset or the posture is incorrect, the system will immediately issue an alarm and record the offset for error analysis and correction. For example, during precision assembly, the image captured by the camera can provide real-time feedback on whether the current assembly parts are aligned. If a small rotational deviation is found, the system can issue a correction instruction through visual feedback.

[0080] S220, defining a state space of deep reinforcement learning according to the comprehensive feature data of the robot, defining an action space of deep reinforcement learning according to the force information and visual information of the robot, setting an experience replay pool, and performing interactive learning to obtain preliminary optimized comprehensive feature data of the robot;

[0081] In step S220 of some embodiments, illustratively, a state space is set, which defines the key information that the robot needs to perceive at each moment. In precision assembly tasks, the state space includes the following elements: Robot joint angles and speeds: This information describes the current posture and motion state of the robot arm. Position and posture of the assembly object: The precise position and posture of the assembly object are obtained through sensors (such as cameras, lidar). Environmental obstacle information: To prevent collisions during the assembly process, the system will identify and record the position information of surrounding obstacles.

[0082] An action space is set, which defines the specific actions that the robot can perform in each state. In the present invention, the action space includes: Joint angular velocity instructions: adjust the angular velocity of each joint of the robot arm to control the motion trajectory of the robot. End effector displacement instructions: directly control the displacement of the robot end effector in space to ensure that the actuator reaches the target position accurately. The design of the action space needs to balance flexibility and control accuracy to ensure that the robot can accurately perform assembly operations in complex environments.

[0083] Set the reward function. The reward function is the core part of reinforcement learning. It determines how the robot adjusts its actions to achieve the best task execution effect. In precision assembly, the reward function is designed based on the following factors: Task completion: Each time the robot successfully completes an assembly task, the system will give a higher reward. Error size: When the robot approaches the target position, the smaller the error, the higher the reward, which encourages the robot to improve the accuracy in assembly. A negative reward is set. When the task takes too long or consumes too much energy, the robot will be punished, thereby guiding it to choose more efficient actions in future operations. The design of the reward function should take into account both assembly accuracy and efficiency, ensuring that the robot completes the task as quickly and efficiently as possible while meeting the accuracy requirements. In order to reduce the exploration risk in the real environment, the initial model is pre-trained in a simulation environment. Through simulation, the robot can accumulate experience in a virtual scene and learn how to optimize the motion strategy. Specific training algorithms include: Deep Q-Network (DQN): Select the optimal action through a discrete action space. Proximal Policy Optimization (PPO): Applicable to continuous action spaces, which can adjust the robot's actions more finely. After the training in the simulation phase is completed, fine-tuning is performed based on data from actual assembly tasks to improve the model's adaptability and generalization capabilities, making it more robust in real-world environments.

[0084] S230, controlling the robot to perform simulation according to the preliminary optimized comprehensive feature data of the robot, and obtaining a robot motion deviation vector;

[0085] In step S230 of some embodiments, illustratively, after error detection, the system first needs to analyze the error. By combining the data of force feedback and visual feedback, the system can calculate the deviation vector between the current position and the target position. The deviation vector not only describes the offset in position, but also includes the deviation in posture (rotation). Through precise error analysis, the system can understand the gap between the current state of the robot and the ideal state, thereby providing a basis for compensation. If the visual feedback detects that the assembly object has a 1mm offset in the XY plane, the force sensor detects that an additional 0.1N torque force occurs during the assembly process. The system will generate a comprehensive deviation vector by combining these two error information to guide subsequent compensation actions.

[0086] S240. Perform error compensation on the robot motion deviation vector through an error compensation algorithm, and optimize it in combination with a deep reinforcement learning algorithm to obtain comprehensive feature data of the robot after error compensation.

[0087] In step S240 of some embodiments, illustratively, the compensation algorithm calculates the motion instructions that need to be adjusted based on the error analysis results. In this system, the compensation algorithm uses a classic PID (proportional-integral-differential) controller or model predictive control (MPC) to correct the robot motion in real time. PID controller: The PID controller is a commonly used error correction algorithm that calculates the output correction amount based on the current error (proportional term), the accumulated error (integral term) and the error change rate (differential term). The advantage of the PID controller is that it is simple and can correct the slight deviation of the system in real time, and is suitable for most static or slowly changing assembly scenarios. Model predictive control (MPC): Compared with the PID controller, MPC is more intelligent. MPC not only considers the current error, but also predicts the future error development trend by establishing a mathematical model of the robot's kinematics and dynamics, and calculates the optimal control strategy in advance. MPC is suitable for complex assembly tasks, especially in a rapidly changing dynamic environment, and can effectively reduce errors and optimize the assembly process. In addition, the system adds an error compensation module to the deep learning model and uses historical error data for learning. By analyzing past assembly data, the system can predict possible error patterns and make corresponding adjustments in advance. For example, if certain assembly tasks are usually prone to errors in specific steps, the system can make appropriate adjustments in advance to prevent deviations from occurring, and then execute adjustment strategies. After the error compensation algorithm calculates the corrected motion instructions, the system sends these instructions to the robot controller to adjust the motion trajectory and posture of the robot arm in real time. In this process, the system ensures that the corrected action is accurate and smooth, and does not cause additional interference to the current assembly task. During the adjustment process, the system will also continue to monitor force and visual feedback to ensure the execution effect of the compensation action. For example, when the robot actuator is compensating for the offset action, the force sensor will monitor the force applied by the actuator to ensure that the adjusted force distribution is still balanced. At the same time, the camera will capture the position changes of the assembly object to verify whether the adjusted accuracy meets the expected goal.

[0088] In summary, the robot's motion control and real-time error correction during the precision assembly process ensure that the robot can complete the assembly task with high precision. During the assembly process, the robot needs to adjust its motion trajectory and posture in real time according to the data fed back by the sensor to adapt to various environmental changes and small displacements of the assembly object. To this end, an embodiment of the present invention proposes a motion control model based on deep reinforcement learning, and combines force perception with visual feedback to achieve error detection and compensation. In order to enable the robot to perform high-precision operations in a complex assembly environment, a deep reinforcement learning (DRL) strategy is adopted to achieve motion control. The deep reinforcement learning model can continuously optimize its action strategy through interaction with the environment, so that the robot can dynamically adapt to environmental changes when performing tasks.

[0089] S300, based on the comprehensive feature data of the robot after error compensation, the robot model is trained through the federated learning optimization algorithm to build a robot with a global assembly control model;

[0090] It should be noted that, in some embodiments, step S300 may include steps S310 to S360.

[0091] S310, controlling the robot to perform an assembly task according to the robot comprehensive feature data after error compensation, and constructing a local data set according to the assembly execution result;

[0092] In step S310 of some embodiments, illustratively, each robot collects sensor data (such as visual data, lidar data, force data, etc.) and assembly results (including the position, posture, force feedback, etc. of the assembly object) in real time when performing an assembly task. These data constitute a local data set, reflecting the perception and control information of each robot in a specific task. For example, when a robot completes an electronic component assembly task, the collected visual data can record the alignment of the components, and the force data can record the changes in the clamping force. Combined with the success and failure results of the assembly task, these data provide valuable feedback for model training.

[0093] S320, performing data preprocessing on the local data set to obtain a preprocessed local data set;

[0094] In step S320 of some embodiments, the collected data is exemplarily preprocessed (such as denoising, normalization, feature extraction, etc.) to construct a local data set suitable for deep learning model training. The label of the data set is usually an evaluation of the assembly result, such as success or failure, specific assembly error, etc.

[0095] S330, constructing a robot assembly control model based on a multi-sensor fusion model and a motion control model, and training the model through a small batch gradient descent algorithm to obtain a trained robot assembly control model;

[0096] In step S330 of some embodiments, illustratively, based on the local data set, the robot trains or fine-tunes the multi-sensor fusion model and the motion control model. The initial model is usually issued by the global server. After fine-tuning with local data, it can better adapt to the unique environment of each robot in a specific task. In order to efficiently update the model parameters, a mini-batch gradient descent (Mini-Batch Gradient Descent) or Adam optimizer is used. These optimization algorithms can converge quickly and adapt to the training of small-scale data sets. For example, using the Adam optimizer, the learning rate can be dynamically adjusted to adapt to data fluctuations in complex assembly scenarios, so that the model can learn key perception and control features faster. Local training is usually performed after each assembly task is completed, or a certain time interval is set according to the complexity of the task and the amount of data to ensure that the model can adapt to the new working environment and assembly tasks in a timely manner.

[0097] S340, compressing the preprocessed local data set and uploading it to the trained robot assembly control model through the AES symmetric encryption algorithm;

[0098] In step S340 of some embodiments, illustratively, in order to prevent the model parameters from being intercepted or tampered with during the transmission process, the system uses the AES symmetric encryption algorithm to encrypt the model parameters. AES is an efficient encryption algorithm that can maintain high transmission efficiency while ensuring data security. In this way, even if the communication network is attacked, the privacy and integrity of the model parameters can be ensured. The data is transmitted through a secure transmission protocol (such as TLS / SSL), further ensuring the security and reliability of the data transmission process. In order to reduce the occupation of communication bandwidth and speed up the transmission speed, the model parameters are compressed using model compression technology. Common compression methods include weight quantization and model pruning. Weight quantization: convert floating point weights into low-precision integer representations to reduce the storage and transmission costs of the model. Model pruning: by removing redundant parameters in the model, the size of the model is reduced while maintaining the performance of the model. During the model upload process, the system will dynamically adjust according to the network conditions and bandwidth to ensure that the compressed model parameters can be transmitted to the central server in the shortest possible time.

[0099] S350, based on the trained robot assembly control model, assigning weight parameters to the preprocessed local data set and performing aggregation processing through an asynchronous federated averaging algorithm to obtain a global assembly control model;

[0100] In step S350 of some embodiments, illustratively, after the central server receives the model updates uploaded by multiple robots, it is necessary to aggregate these parameters to generate a global model. Since robots perform different tasks in different environments, there may be differences in the uploaded model updates, so it is necessary to optimize the global model through a reasonable aggregation strategy. In order to avoid the delay caused by the synchronous waiting of the robot during the transmission process, the central server adopts an asynchronous federated averaging algorithm (FedAsync). This method allows the server to aggregate immediately after receiving the model update of any robot, without having to wait for the parameters of all robots to arrive. This can effectively improve the real-time performance of the system and ensure that the global model can be updated in time. During the aggregation process, the central server assigns different weights to the model parameters of each robot. These weights can be dynamically adjusted according to factors such as the amount of local data of each robot, model performance, and task importance. For example, the amount of data is weighted, and the robot with more local data may contribute more to the model, so its update weight is higher. Performance weighting, the model performance is evaluated according to indicators such as the assembly success rate and error size of the robot. The better the performance, the greater the impact of the model update on the global model.

[0101] To further prevent model updates of some robots from being abnormal (such as due to network errors, equipment failures, or malicious attacks), the server will perform anomaly detection on the received model parameters. By analyzing the model's gradient or parameter update amount, it is detected whether any model deviates from the normal range of global updates. For detected abnormal models, the server will use methods to defend against Byzantine attacks (such as the Krum algorithm) to filter them. The Krum algorithm selects a part of the models with the smallest update difference for aggregation, preventing malicious or erroneous data from affecting the global model, thereby improving the robustness of the system.

[0102] S360: Send the global assembly control model to the robot control terminal to obtain a robot having the global assembly control model.

[0103] In step S360 of some embodiments, illustratively, after completing model aggregation, the central server needs to send the global model to each robot and update their local models so that each robot can share the latest learning results. Similar to model uploading, when sending the global model, the server will also encrypt the model parameters to ensure the security of data during transmission. AES encryption and TLS / SSL protocols ensure that the model parameters will not be tampered with or intercepted during transmission.

[0104] Then update the local model. After receiving the global model, the robot needs to fuse it with the local model to maintain the balance between the system's personalization and global sharing. After receiving the global model, the robot fuses the global model parameters with the local personalized parameters. The first few layers of the model are usually used to extract common features (such as visual features, object recognition, etc.). These layers are updated using the global model. The following layers are dedicated layers for local personalization, which are used to adapt to specific task scenarios or assembly conditions. These layers retain the robot's own local parameters. Personalized adjustments are made. During the fusion process, the robot can make appropriate fine-tuning of the global model according to its actual task requirements to ensure that the model can share global knowledge and adapt to local specific work scenarios; Continuous learning and online updating: After fusing the global model, the local model still maintains the ability to learn continuously. The robot can continue to collect data in the assembly task and perform incremental training on the model to ensure that the model can continuously adapt to new environments and task changes.

[0105] In summary, the robot's multi-sensor fusion model and motion control model are continuously optimized through the federated learning framework. The core advantage of federated learning is that it allows each robot to share model updates through distributed training without transmitting local raw data. This not only protects data privacy, but also improves the generalization ability of the model, making it better adaptable to diverse assembly tasks and environmental changes.

[0106] S400, based on the meta-learning strategy, the robot with the global assembly control model communicates and collaborates to perform the assembly task, and obtains the robot assembly optimization result;

[0107] It should be noted that, in some embodiments, step S400 may include steps S410 to S450.

[0108] S410, based on the industrial communication protocol, network communication is performed between robots having the global assembly control model;

[0109] In step S410 of some embodiments, illustratively, first, the network architecture adopts a distributed network architecture, supports direct communication between the robot and the central server, and between robots, based on the TCP / IP protocol stack, combined with industrial communication protocols (such as OPC UA, EtherCAT), to achieve high-reliability communication, and the security mechanism uses the SSL / TLS protocol to establish a secure communication channel. Perform identity authentication and authority management to prevent unauthorized access.

[0110] Through the central dispatch system, assembly tasks are intelligently allocated according to the status, load and capabilities of each robot. Robots share environmental information, assembly progress, fault conditions, etc. to coordinate work. In the scenario of multi-robot collaboration, collision avoidance algorithms are used to prevent physical conflicts between robots.

[0111] S420, introducing a personalized network model, performing global network fine-tuning processing on the robot having the global assembly control model, and obtaining a fine-tuned robot;

[0112] In step S420 of some embodiments, illustratively, personalized models and meta-learning are key modules for improving the adaptability and rapid response capabilities of robots in multiple tasks and multiple environments. Under the federated learning framework, although the global model can provide shared basic knowledge for all robots, personalized model fine-tuning is crucial because each robot faces different task scenarios and environmental conditions. By combining meta-learning strategies, robots can quickly adapt to new tasks after receiving the global model, while maintaining continuous learning capabilities to cope with dynamically changing working environments.

[0113] In order to balance global knowledge sharing and local task adaptability, personalized models usually adopt a hierarchical structure. In this design, the first few layers of the model are shared layers, which focus on extracting common features, such as visual feature extraction, basic motion patterns, etc., which are common to different tasks. The shared layers are updated by the global model of federated learning to ensure that the robot learns common perception and control knowledge from other robots. The last few layers of the model are dedicated layers, which focus on handling the personalized needs of robots in specific tasks or environments. For example, for different assembly tasks, some robots may need to handle different object types, postures, or constraints. These dedicated layers are trained and fine-tuned based on the local data of each robot to ensure that the model can adapt well to the uniqueness of local tasks.

[0114] When updating the model, the system will give priority to updating the shared layer to ensure that all robots can use the globally optimized common features. For the dedicated layer, the robot retains or fine-tunes the parameters according to the specific circumstances of the local task. This approach can enhance its personalized processing capabilities while maintaining the overall consistency of the model. For example, for a robot that performs electronic component assembly tasks, its shared layer is mainly responsible for basic visual feature extraction and posture recognition, while the dedicated layer is optimized according to the specific electronic component type and assembly position requirements faced by the robot to ensure accurate execution of the task.

[0115] S430, introducing a meta-learning strategy, performing adaptive training on the fine-tuned robot, and obtaining an adaptively trained robot;

[0116] Specifically: S431, the fine-tuned robot is initialized through a meta-learning algorithm to obtain an initialized robot; S432, the end control and force feedback parameters of the initialized robot are adjusted to obtain an adaptively trained robot.

[0117] In step S430 of some embodiments, illustratively, meta-learning is a learning method for training models to quickly adapt to new tasks. Unlike traditional training methods, meta-learning enables the model to quickly update and adapt when faced with new tasks by learning "how to learn". In the present invention, the Model-Agnostic Meta-Learning (MAML) algorithm is used to initialize the model. The core idea of ​​MAML is to train a model so that its initial parameters are close to the "optimal adaptation point", that is, when the model faces a new task, only a small amount of gradient updates are required to quickly adapt to the new task. For example, after the model is trained on multiple different assembly tasks, its parameters will be in a "globally optimal" position. When a robot faces a new assembly task, the model trained based on MAML only needs a small amount of local data training (such as a few iterations) to adapt to the requirements of the task.

[0118] The advantage of the meta-learning strategy is that after receiving the global model, when the model faces new tasks or environmental changes, only a small amount of training data and a small number of training steps are needed to enable the model to quickly adjust to the new task. This feature is particularly suitable for dynamically changing industrial scenarios. For example, a robot originally performs a screw assembly task and then needs to switch to an electronic component welding task. Under the MAML strategy, the model can quickly adjust its end control strategy and force feedback adjustment parameters to adapt to different task requirements without the need to train a new model from scratch. This significantly improves the robot's task switching efficiency and adaptability.

[0119] S440, based on a continuous learning mechanism, dynamically adjusting the robot after adaptive training to obtain a robot with multi-task compatibility;

[0120] Specifically, S441, based on the continuous learning mechanism, the learning rate of the robot after adaptive training is dynamically adjusted to obtain the adjusted robot; S442, the knowledge distillation algorithm is used to perform knowledge transfer learning between models on the adjusted robot to obtain a robot with multi-task compatibility.

[0121] In step S440 of some embodiments, illustratively, in order to ensure that the robot always has the latest knowledge and optimal model performance in a dynamic environment, the system has designed a continuous learning mechanism, so that the robot can continuously learn and update its model during the task execution to avoid the "model forgetting" phenomenon.

[0122] During the execution of the task, the robot continuously collects new sensor data, assembly results, and operation feedback, and uses this new data to fine-tune the model online. Through this continuous learning mechanism, the model can continue to evolve to adapt to changes in the environment and tasks. In order to prevent the model from overfitting or forgetting during the update process, online updates usually adopt an incremental learning strategy to ensure that the model retains existing knowledge when introducing new data. For example, during the welding process, if the robot encounters new assembly materials or environmental changes, the temperature, force and other information fed back by the sensor will be updated to the model in real time. The model adjusts the welding force or time based on these new data to ensure that the welding quality is always in the optimal state.

[0123] During the online update process, the system will dynamically adjust the learning rate to ensure the stability and robustness of the model. For new significant changes (such as changes in material properties or assembly tool switching), the system will increase the learning rate to speed up adaptation; for minor environmental fluctuations, the system will reduce the learning rate to avoid drastic changes in the model.

[0124] In the process of continuous updating, in order to prevent the model from forgetting the knowledge of old tasks when encountering new tasks, the system introduces knowledge distillation technology. The core idea of ​​knowledge distillation is to utilize the knowledge transfer between the new and old models, retain the important knowledge in the old model through soft labels, and avoid model degradation.

[0125] Specifically, when the model is trained on new task data, the system uses the output of the old model as an auxiliary target, so that the new model can adapt to new tasks and maintain the performance of old tasks at the same time. This method can effectively solve the "catastrophic forgetting" problem of the model and ensure that the robot always has comprehensive task execution capabilities when facing diversified tasks. For example, the robot initially focused on precision parts assembly. When receiving a completely new type of assembly task, knowledge distillation ensures that the learning of the new task does not affect the robot's mastery of the precision parts assembly task, allowing it to switch smoothly between new and old tasks.

[0126] Through knowledge distillation, the model can retain important features in multiple tasks, which is particularly important for robots that need to perform multiple assembly tasks at the same time. The system enables the robot to quickly switch and retain key knowledge of other tasks while performing one task, thereby improving task compatibility and response speed.

[0127] S450, performing assembly tasks based on a robot with multi-task compatibility to obtain robot assembly optimization results.

[0128] In addition, it should be noted that during the model update process, noise is added to protect the privacy of local data, and strict permission management is further set. Only authorized entities can access the model and data, and then the system status is monitored in real time, and faults are promptly reported. Hot backup and automatic switching of the system are supported to ensure continuous operation. Finally, a fault-tolerant mechanism is set. During the federated learning process, if a robot is offline or fails, the system can still perform model aggregation and updates normally.

[0129] In summary, the embodiments of the present invention have the following improvements over the prior art:

[0130] 1) Use multi-sensor fusion technologies such as high-resolution cameras, lidars, and force sensors to achieve high-precision real-time perception of the assembly environment. Cameras are used to obtain visual information of assembly objects, lidars are used for three-dimensional environment modeling, and force sensors are used to monitor the force feedback of the end effector to ensure that the robot has comprehensive perception capabilities for changes in position, angle, and force during the assembly process. The time synchronization, spatial alignment, and data fusion algorithms of multi-sensor data are used to perform multi-modal information processing through deep learning models to improve the accuracy and robustness of environmental perception.

[0131] 2) In the framework of multi-robot collaboration, distributed model training and sharing of each robot are realized through federated learning. Each robot trains the multi-sensor fusion model and motion control model locally, encrypts the updated model parameters and uploads them to the central server, and performs weighted aggregation through the asynchronous federated average algorithm to generate a global model and send it to each robot to realize knowledge sharing. The asynchronous federated average (FedAsync) strategy is adopted to weighted aggregate the model parameters uploaded by each robot, combined with data encryption and compression transmission technology to ensure data security and model privacy. During the aggregation process, the Byzantine defense algorithm (such as the Krum algorithm) is applied to filter abnormal data to ensure the stability and security of the global model.

[0132] 3) The personalized model adopts a personalized structural design of shared layers and dedicated layers. The first few layers are shared layers that extract common features, and the latter few layers are dedicated layers that optimize the personalized needs of robots in specific assembly tasks and environments, making the model both universal and task adaptable. The shared layer updates use global model optimization of federated learning, and the dedicated layer is fine-tuned through local data to enhance the robot's responsiveness and adaptability to specific tasks. The personalized model structure realizes the task customization of individual robots without affecting the shared knowledge of federated learning.

[0133] 4) During the execution of the assembly task, the assembly error is detected and automatically compensated through the real-time feedback of the force sensor and the visual sensor. The force sensor monitors the force changes during the assembly process, and the visual sensor monitors the position and posture of the assembly object. According to the error calculation results, the system automatically adjusts the motion trajectory of the robot arm through the PID controller or the model predictive control (MPC) algorithm. The error detection module combines force and visual information to calculate the deviation vector in real time. The compensation algorithm quickly corrects the motion path through the controller to ensure high-precision operation of the robot during the assembly process. The error compensation module predicts historical error patterns through deep learning, makes adjustments in advance, and improves the speed and accuracy of compensation.

[0134] 5) Based on the Model-Agnostic Meta-Learning (MAML) meta-learning algorithm, the robot model is trained to have the ability to adapt quickly, so that when the task switches or the environment changes, it can quickly adjust to the new task requirements through a small number of training steps. The meta-learning algorithm plays a role in the global model initialization process, so that the model parameters are close to the "optimal adaptation point", ensuring that the model can be quickly updated and adapted when facing new tasks. The meta-learning strategy effectively improves the system's response speed and adaptability in multi-task switching scenarios.

[0135] 6) During the execution of the assembly task, the robot continuously collects new data for incremental training through online learning to prevent the "model forgetting" phenomenon. At the same time, the knowledge distillation technology is used to realize the knowledge transfer between the new and old models to prevent the model from forgetting the old task knowledge when learning the new task. The robot system applies knowledge distillation when updating the model, and uses the soft labels of the old model as the auxiliary targets of the new model to achieve compatibility between the new and old tasks, ensuring that the model always maintains stable performance in multi-task execution.

[0136] 7) In multi-robot collaboration scenarios, the central server conducts intelligent analysis and allocation of task status. The system schedules and allocates tasks based on the robot's respective task load, assembly success rate, and real-time feedback to ensure task balance and overall assembly efficiency. The task allocation system dynamically adjusts the task allocation strategy based on the robot's real-time status information, historical data, and environmental feedback. Through real-time information sharing, the robot can actively avoid collisions and dynamically adjust its movements during the collaboration process to ensure efficient and safe collaborative work.

[0137] 8) The robot system uses the AES encryption algorithm during data transmission to ensure the security of model parameters during upload and download, and combines the SSL / TLS protocol to establish an encrypted communication channel to avoid data leakage and attacks, encrypt the transmission of model parameters, and prevent unauthorized access and data tampering. Through a secure communication protocol, the privacy of federated learning model updates is protected, and the data security and system integrity of the multi-robot system are guaranteed.

[0138] See also Figure 2 The embodiment of the present application further provides a robot assembly optimization system based on federated learning, which can implement the above-mentioned robot assembly optimization method based on federated learning. The system includes:

[0139] The first module 201 is used to obtain the image data and point cloud data of the robot and perform data feature fusion recognition processing to obtain comprehensive feature data of the robot;

[0140] The second module 202 is used to obtain the force information and the visual information of the robot, and perform motion error optimization processing based on deep reinforcement learning in combination with the comprehensive feature data of the robot to obtain the comprehensive feature data of the robot after error compensation;

[0141] The third module 203 is used to perform model training on the robot through a federated learning optimization algorithm based on the comprehensive feature data of the robot after error compensation, and to construct a robot with a global assembly control model;

[0142] The fourth module 204 is used to communicate and cooperate with the robots with the global assembly control model to perform assembly tasks based on the meta-learning strategy, so as to obtain the robot assembly optimization result.

[0143] It can be understood that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0144] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A robot assembly optimization method based on federated learning, characterized in that: The method comprises the following steps: Acquire the robot's image data and point cloud data and perform data feature fusion and recognition processing to obtain the robot's comprehensive feature data; Acquire force information and visual information of the robot, and perform motion error optimization processing based on deep reinforcement learning in combination with the comprehensive feature data of the robot to obtain comprehensive feature data of the robot after error compensation; Based on the comprehensive feature data of the robot after error compensation, a model of the robot is trained by a federated learning optimization algorithm to construct a robot with a global assembly control model; Based on the meta-learning strategy, the robots with the global assembly control model communicate and collaborate to perform assembly tasks, and obtain robot assembly optimization results.

2. The method according to claim 1, characterized in that The method of acquiring the image data and point cloud data of the robot and performing data feature fusion recognition processing to obtain comprehensive feature data of the robot includes: Calibrate the camera and the laser radar to obtain a calibrated camera and a calibrated laser radar; Acquire image data of the robot through the calibrated camera and acquire point cloud data of the robot through the calibrated laser radar; Performing data preprocessing and data fusion processing on the image data of the robot and the point cloud data of the robot to obtain fused robot data; Introduce image branch module, point cloud branch module and feature fusion layer to build a deep learning fusion model; Based on the deep learning fusion model, feature recognition and fusion processing is performed on the fused robot data to obtain comprehensive feature data of the robot.

3. The method according to claim 1, characterized in that The acquiring of the force information and the visual information of the robot, and combining the comprehensive feature data of the robot to perform motion error optimization processing based on deep reinforcement learning to obtain the comprehensive feature data of the robot after error compensation, include: Acquiring force information of the robot through a force sensor and acquiring visual information of the robot through a visual sensor; Defining a state space of deep reinforcement learning according to the comprehensive feature data of the robot, defining an action space of deep reinforcement learning according to the force information of the robot and the visual information of the robot, setting an experience replay pool, and performing interactive learning to obtain preliminary optimized comprehensive feature data of the robot; Controlling the robot to perform simulation according to the preliminary optimized comprehensive characteristic data of the robot to obtain a robot motion deviation vector; The robot motion deviation vector is error compensated by an error compensation algorithm, and is optimized in combination with a deep reinforcement learning algorithm to obtain comprehensive feature data of the robot after error compensation.

4. The method according to claim 1, characterized in that: Based on the comprehensive feature data of the robot after error compensation, the robot is trained by a federated learning optimization algorithm to construct a robot with a global assembly control model, including: Controlling the robot to perform an assembly task according to the robot comprehensive feature data after error compensation, and constructing a local data set according to the assembly execution result; Performing data preprocessing on the local data set to obtain a preprocessed local data set; A robot assembly control model is constructed based on a multi-sensor fusion model and a motion control model, and trained using a small batch gradient descent algorithm to obtain a trained robot assembly control model. The pre-processed local data set is compressed and uploaded to the trained robot assembly control model through an AES symmetric encryption algorithm; Based on the trained robot assembly control model, weight parameters are assigned to the preprocessed local data set and aggregation processing is performed through an asynchronous federated averaging algorithm to obtain a global assembly control model; The global assembly control model is sent to a robot control terminal to obtain the robot having the global assembly control model.

5. The method according to claim 1, characterized in that: The meta-learning strategy is used to communicate and collaborate with the robot with the global assembly control model to perform the assembly task, thereby obtaining the robot assembly optimization result, including: Based on the industrial communication protocol, the robots having the global assembly control model are networked for communication; A personalized network model is introduced to perform global network fine-tuning processing on the robot having the global assembly control model to obtain a fine-tuned robot; A meta-learning strategy is introduced to perform adaptive training on the fine-tuned robot to obtain an adaptively trained robot; Based on a continuous learning mechanism, dynamically adjusting the adaptively trained robot to obtain a robot with multi-task compatibility; The robot assembly optimization result is obtained by executing the assembly task based on the robot with multi-task compatibility.

6. The method according to claim 5, characterized in that The personalized network model includes a plurality of shared layers and a plurality of dedicated layers, and the output ends of the plurality of shared layers are connected to the input ends of the plurality of dedicated layers, wherein: The shared layers are used to extract common features of the robot; The dedicated layers are used to extract the personalized demand features of the robot.

7. The method according to claim 5, characterized in that The introducing of the meta-learning strategy and performing adaptive training on the fine-tuned robot to obtain an adaptively trained robot includes: Initializing the fine-tuned robot by using a meta-learning algorithm to obtain an initialized robot; The end control and force feedback parameters of the initialized robot are adjusted to obtain the adaptively trained robot.

8. The method according to claim 5, characterized in that The method of dynamically adjusting the adaptively trained robot based on the continuous learning mechanism to obtain a robot with multi-task compatibility includes: Based on a continuous learning mechanism, dynamically adjusting the learning rate of the adaptively trained robot to obtain an adjusted robot; The adjusted robot is subjected to knowledge transfer learning between models through a knowledge distillation algorithm to obtain the robot with multi-task compatibility.

9. The method according to claim 1, characterized in that: The method also includes adding noise to the robot having the global assembly control model during the process of communicating and cooperating to perform the assembly task.

10. A robot assembly optimization system based on federated learning, characterized in that: The system comprises: The first module is used to obtain the robot's image data and point cloud data and perform data feature fusion recognition processing to obtain the robot's comprehensive feature data; The second module is used to obtain the force information and visual information of the robot, and to perform motion error optimization processing based on deep reinforcement learning in combination with the comprehensive feature data of the robot to obtain the comprehensive feature data of the robot after error compensation; The third module is used to perform model training on the robot through a federated learning optimization algorithm based on the comprehensive feature data of the robot after the error compensation, so as to construct a robot with a global assembly control model; The fourth module is used to communicate and cooperate with the robots with the global assembly control model to perform assembly tasks based on a meta-learning strategy to obtain robot assembly optimization results.

Images