Robot control method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, can be applied to service system platforms of medical health, financial science and technology and the like, and discloses a robot control method and device, computer equipment and a storage medium. Cleaning, normalizing and enhancing the received data and the action data, and integrating to obtain standby data; training a pre-constructed deep learning model by using the standby data to generate a diffusion strategy model; based on the diffusion strategy model, constructing a single-step diffusion strategy model, and migrating knowledge of the diffusion strategy model to the single-step diffusion strategy model through a knowledge distillation technology; processing a task instruction input by the user, and controlling the target robot to execute a target task according to a processing result; the response speed and adaptability of the robot during task execution can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a robot control method, device, computer equipment, and computer-readable storage medium. Background Art

[0002] With the rapid development of artificial intelligence and robotics, robots are increasingly being used in a variety of fields, including industry, services, healthcare, and financial technology. To achieve efficient and precise task execution, robots often need to be able to make autonomous decisions and generate actions. Existing robot control methods primarily rely on complex multi-step policy models that generate high-quality action sequences through step-by-step denoising and optimization. However, these multi-step policy models suffer from the following issues:

[0003] 1. Long response time: Multi-step strategy models usually require multiple iterations to generate the final action sequence, resulting in long response time and difficulty meeting the requirements of real-time tasks.

[0004] 2. High consumption of computing resources: Multi-step strategy models usually require a lot of computing resources to complete complex denoising and optimization processes, which limits their application in resource-constrained environments.

[0005] 3. Poor adaptability: Existing multi-step strategy models often need to be retrained or parameter adjusted when faced with changes in the environment or task requirements, and have poor adaptability.

[0006] In healthcare, robots are used for surgical assistance, rehabilitation, and drug delivery. In fintech, they are used for intelligent customer service, transaction monitoring, and risk warning. Similarly, robots in both healthcare and fintech face challenges when performing tasks, such as long response times, limited real-time performance, and poor adaptability.

[0007] Based on this, how to provide a robot control method, device, computer equipment and computer-readable storage medium that can effectively improve the response speed and adaptability of the robot when performing tasks is an urgent problem to be solved by technical personnel in this field. Summary of the Invention

[0008] In view of the above-mentioned deficiencies in the prior art, the object of the present invention is to provide a robot control method, apparatus, computer equipment and computer-readable storage medium, aiming to solve the problem of how to effectively improve the response speed and adaptability of the robot when performing tasks.

[0009] In order to achieve the above object, the present invention adopts the following technical solutions:

[0010] In a first aspect, the present invention provides a robot control method, comprising:

[0011] Collecting received data and action data when the robot performs a task, cleaning, normalizing, and enhancing the received data and the action data, and integrating them to obtain stand-by data;

[0012] Using the stand-by data, a pre-built deep learning model is trained to generate a diffusion strategy model;

[0013] Based on the diffusion strategy model, a single-step diffusion strategy model is constructed, and the knowledge of the diffusion strategy model is transferred to the single-step diffusion strategy model through knowledge distillation technology;

[0014] The single-step diffusion strategy model is used to process the task instructions input by the user, and the target robot is controlled to perform the target task according to the processing results.

[0015] In a second aspect, the present invention provides a robot control device, comprising:

[0016] A collection module is used to collect reception data and action data when the robot performs tasks, clean, normalize and enhance the reception data and the action data, and integrate them to obtain stand-by data;

[0017] A training module, configured to train a pre-built deep learning model using the stand-by data to generate a diffusion strategy model;

[0018] A migration module, configured to construct a single-step diffusion strategy model based on the diffusion strategy model, and migrate the knowledge of the diffusion strategy model to the single-step diffusion strategy model through knowledge distillation technology;

[0019] The control module is used to use the single-step diffusion strategy model to process the task instructions input by the user, and control the target robot to perform the target task according to the processing results.

[0020] In a third aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the robot control method as described above when executing the computer program.

[0021] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program implements the robot control method as described above when executed by a processor.

[0022] Compared with the prior art, the present invention provides a robot control method, device, computer equipment and computer-readable storage medium, wherein the received data and action data when the robot performs a task are collected, the received data and the action data are cleaned, normalized and enhanced, and integrated to obtain stand-by data; the stand-by data is used to train a pre-built deep learning model to generate a diffusion strategy model; based on the diffusion strategy model, a single-step diffusion strategy model is constructed, and the knowledge of the diffusion strategy model is transferred to the single-step diffusion strategy model through knowledge distillation technology; the single-step diffusion strategy model is used to process the task instructions input by the user, and the target robot is controlled to perform the target task according to the processing results; thereby, the present invention can effectively improve the response speed and adaptability of the robot when performing tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0024] Figure 1 A schematic diagram of an application environment of a robot control method provided by one embodiment of the present invention.

[0025] Figure 2 A schematic flow chart of a robot control method provided by one embodiment of the present invention.

[0026] Figure 3 A schematic diagram of a program module of a robot control device provided by one embodiment of the present invention.

[0027] Figure 4 A schematic diagram of the structure of a computer device provided in one embodiment of the present invention.

[0028] Figure 5 Another structural diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0030] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0031] It will also be understood that the term "and / or" used in the present description and appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0032] As used in the present specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0033] In addition, in the description of the present specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0034] References to "one embodiment" or "some embodiments" in the present specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present invention. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in yet other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0035] It should be understood that the order of execution of the steps in the following embodiments does not necessarily mean the order in which they are executed. The order in which each process is executed should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0036] In order to illustrate the technical solution of the present invention, specific embodiments are provided below.

[0037] An embodiment of the present invention provides a robot control method that can be applied in Figure 1In the application environment shown, the client and server communicate via a network. The client includes, but is not limited to, PDAs, desktop computers, laptops, ultra-mobile personal computers (UMPCs), netbooks, cloud computing devices, personal digital assistants (PDAs), and other computer devices. The server can be a standalone server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0038] See also Figure 2 An embodiment of the present invention provides a robot control method, wherein the method comprises the following steps:

[0039] S100, collecting received data and action data when the robot performs a task, cleaning, normalizing, and enhancing the received data and the action data, and integrating them to obtain stand-by data;

[0040] S200: Using the stand-by data, training a pre-built deep learning model to generate a diffusion strategy model;

[0041] S300: Based on the diffusion strategy model, construct a single-step diffusion strategy model, and migrate the knowledge of the diffusion strategy model to the single-step diffusion strategy model through knowledge distillation technology;

[0042] S400: Utilize the single-step diffusion strategy model to process the task instruction input by the user, and control the target robot to execute the target task according to the processing result.

[0043] In specific implementation, the robot control method of this embodiment effectively improves the response speed and adaptability of the robot when performing tasks through a series of innovative steps and technical means. The specific analysis is as follows:

[0044] 1. Application of the single-step diffusion strategy model

[0045] Traditional multi-step policy models require multiple iterations to generate the final action sequence, resulting in long response times. However, the one-step diffusion policy (OneDP) model constructed in this method directly maps Gaussian noise to the target action, eliminating the complex multi-step denoising and optimization process. This single-step generation mechanism significantly reduces action generation time, significantly improving the robot's response speed. For example, in real-time interactive tasks, the robot can generate and execute actions more quickly, meeting the demand for rapid response.

[0046] 2. Optimization of knowledge distillation technology

[0047] Through knowledge distillation, the knowledge of the complex diffusion strategy model is transferred to the single-step diffusion strategy model. This process not only retains the high performance of the multi-step model but also optimizes the parameters of the single-step model, enabling it to generate high-quality action sequences while maintaining high efficiency. Knowledge distillation ensures that the single-step model, despite its simplified structure, still inherits the key features and performance of the multi-step model, thus achieving rapid response without sacrificing quality.

[0048] 3. Data preprocessing and enhancement

[0049] During the data preparation phase, the collected reception and motion data are cleaned, normalized, and enhanced. These preprocessing steps not only improve data quality but also increase data diversity through data augmentation (such as random transformations and noise injection). This diverse data allows the trained model to better adapt to different task scenarios and environmental changes, thereby improving the robot's adaptability. For example, when faced with new lighting conditions or changes in object position, the augmented data can help the model adjust and adapt more quickly.

[0050] 4. Real-time feedback and dynamic adjustment

[0051] During the execution of a task, a real-time feedback mechanism can be introduced when controlling the robot using the action sequence generated by the single-step diffusion strategy model. The robot collects feedback information during the execution process in real time through sensors (such as vision, torque, and position sensors) and dynamically adjusts its action strategy based on this information. This real-time feedback and adjustment capability enables the robot to quickly respond to environmental changes in complex and changing environments, ensuring the smooth progress of the task. For example, in medical scenarios such as surgical assistance or rehabilitation treatment, the robot can adjust the accuracy and strength of its movements based on real-time feedback to adapt to the patient's specific condition.

[0052] 5. Multimodal input support

[0053] This method also supports users inputting task instructions through multiple methods (such as voice, text, and gestures). This multimodal input mechanism not only increases the flexibility of user interaction but also enables the robot to better understand the user's intent. In different task scenarios, users can choose the most convenient input method based on their specific situation, and the robot can quickly interpret and execute the task, further improving the system's adaptability.

[0054] In this embodiment, by constructing a single-step diffusion strategy model, applying knowledge distillation techniques, optimizing data preprocessing, and introducing a real-time feedback mechanism, the method improves the robot's response speed while significantly enhancing its adaptability in complex environments. This comprehensive optimization not only increases task execution efficiency but also ensures that the robot can quickly adjust and optimize in response to changing environments and task requirements, thus enabling efficient and accurate task execution in a variety of fields, including industry, healthcare, and finance.

[0055] It is understandable that the robot control method provided in the embodiment of the present invention can be applied to robot control scenarios related to the medical and health field. The following is a specific example:

[0056] In the field of medical health, the robot control method provided by the embodiment of the present invention can be applied to surgical assistance robots. For example, in minimally invasive surgery, the surgical robot needs to operate surgical instruments quickly and accurately according to the doctor's instructions. Through this method:

[0057] Data collection and preprocessing: Visual images, instrument position data, and doctor's operating instructions during the operation are collected, and then cleaned, normalized, and enhanced to adapt to different surgical scenarios and patient signs.

[0058] Model training and migration: Use preprocessed data to train a deep learning model to generate a diffusion strategy model. Then, use knowledge distillation technology to migrate its knowledge to a single-step diffusion strategy model, ensuring that the robot can quickly generate accurate action sequences.

[0059] Real-time control and feedback: During surgery, the surgeon inputs task instructions through voice or gestures. The single-step diffusion strategy model generates action sequences in real time, controlling the robot to perform the surgical operation. Simultaneously, the robot uses vision and torque sensors to provide real-time feedback on the position and force of surgical instruments, dynamically adjusting its action strategy to ensure surgical safety and precision.

[0060] This application not only improves the response speed of the surgical robot, but also enhances its adaptability and stability in complex surgical environments, significantly improving the success rate and safety of the operation.

[0061] It is understandable that the robot control method provided in the embodiment of the present invention can also be applied to robot control scenarios related to the field of financial technology. The following is a specific example:

[0062] In the field of financial technology, the robot control method provided by the embodiment of the present invention can be applied to intelligent customer service robots. For example, in the customer service center of a bank or financial institution, an intelligent customer service robot needs to quickly respond to customer inquiries and provide accurate solutions. Through this method:

[0063] Data collection and preprocessing: Collect text data, voice data of customer inquiries, as well as related financial transaction records and human-computer interaction information, and perform cleaning, normalization, and enhancement processing to adapt to different customer issues and business scenarios.

[0064] Model training and migration: Use preprocessed data to train a deep learning model to generate a diffusion strategy model. Then, use knowledge distillation technology to migrate its knowledge to a single-step diffusion strategy model to ensure that the robot can quickly generate accurate response strategies.

[0065] Real-time interaction and feedback: During the customer service process, customers input their questions via voice or text. The single-step diffusion strategy model generates responses in real time, controlling the robot's interaction with the customer. Simultaneously, the robot analyzes customer feedback in real time and dynamically adjusts its response strategy to ensure that customer questions are answered quickly and accurately.

[0066] This application not only improves the response speed of intelligent customer service robots, but also enhances their adaptability and accuracy when facing complex customer problems, significantly improving customer satisfaction and financial service efficiency.

[0067] Furthermore, in one embodiment, the robot control method, wherein the collecting of received data and action data when the robot performs a task, cleaning, normalizing and enhancing the received data and the action data, and integrating them to obtain stand-by data, specifically comprises the steps of:

[0068] Using a multimodal sensor network, collecting reception data and action data of the robot when performing a task, and performing cleaning, normalization, and enhancement processing on the reception data and the action data;

[0069] The processed received data and the action data are integrated, and the integrated data are marked to obtain stand-by data.

[0070] During specific implementation, the specific implementation process of the steps in this embodiment is roughly as follows:

[0071] 1. Multimodal Data Collection

[0072] A multimodal sensor network is used to collect both reception and motion data as the robot performs its tasks. The reception data includes visual images, torque sensor data, and position sensor data; the motion data includes information such as the angles, velocities, and accelerations of the robot's joints. This data is collected in real time via the sensor network, providing rich information for subsequent processing.

[0073] 2. Data Preprocessing

[0074] The collected reception and motion data are cleaned to remove noise and outliers, ensuring data accuracy and completeness. Subsequently, the data is normalized to scale it to a uniform range for easier model training. Furthermore, the visual data is enhanced through random cropping, rotation, and color adjustment, while the motion data is smoothed to increase data diversity and robustness.

[0075] 3. Data Integration

[0076] Integrate preprocessed received data and motion data to form a structured dataset. Synchronize data across modalities using timestamps to ensure temporal consistency. Fuse multimodal data such as vision, torque, and position to form a complete data record, providing high-quality data for subsequent model training.

[0077] 4. Data Annotation

[0078] The integrated data is annotated with information about the task type (e.g., grasping, handling, welding), environmental conditions (e.g., light intensity, background noise), and action objectives (e.g., target location, action accuracy). Detailed annotation information helps the model better understand the data context, improving its training effectiveness and generalization capabilities.

[0079] 5. Generate data to be used

[0080] The labeled data is stored in a format suitable for deep learning frameworks (such as TensorFlow's TFRecord or PyTorch's Dataset). The stored data is verified to ensure its integrity and consistency. This processed and labeled data ultimately becomes the ready-to-use data for subsequent model training.

[0081] Through the above process, this embodiment realizes the complete flow from data collection to data labeling, and provides high-quality input data for the training of the robot control model.

[0082] Furthermore, in one embodiment, the robot control method, wherein the using of the standby data to train a pre-built deep learning model to generate a diffusion strategy model, specifically comprises the steps of:

[0083] Based on the deep learning architecture, build a deep learning model according to the model generation requirements;

[0084] According to the data partitioning strategy, the to-be-used data is divided into a training set, a validation set and a test set;

[0085] Using the training set, the deep learning model is trained using a multi-stage training strategy, wherein the multi-stage training strategy includes a pre-training stage and a fine-tuning stage;

[0086] Evaluate the trained deep learning model using the validation set, and optimize the deep learning model based on the evaluation results;

[0087] The optimized deep learning model is tested on the test set, and when the deep learning model meets the preset performance requirements, a diffusion strategy model is generated.

[0088] During specific implementation, the specific implementation process of the steps in this embodiment is roughly as follows:

[0089] 1. Deep Learning Model Construction

[0090] Based on the model generation requirements, build a deep learning model based on a deep learning architecture. Choose an architecture suitable for the robotics task, such as a diffusion model or Transformer, and customize the configuration based on the task requirements. Initialize the model parameters and use pretrained weights (if available) to accelerate convergence.

[0091] 2. Data Partitioning

[0092] Based on the data partitioning strategy, the data to be used is divided into training, validation, and test sets. Typically, the ratio is 70% training, 15% validation, and 15% test. This ensures that the data distribution is uniform and representative, so that the model can perform well on different datasets.

[0093] 3. Multi-stage training

[0094] Using the training set, a multi-stage training strategy is used to train the deep learning model. The multi-stage training strategy includes a pre-training stage and a fine-tuning stage:

[0095] Pre-training phase: Use a large amount of unsupervised data to pre-train the model and learn general feature representations.

[0096] Fine-tuning phase: Based on pre-training, the model is fine-tuned using labeled task-related data to optimize the model's performance on specific tasks.

[0097] 4. Model Evaluation and Optimization

[0098] Evaluate the trained deep learning model using the validation set and monitor the model's performance metrics (such as accuracy, recall, F1 score, etc.). Based on the evaluation results, adjust the model's hyperparameters (such as learning rate, regularization parameters, etc.) and structure (such as the number of layers, number of hidden units, etc.) to optimize model performance.

[0099] 5. Model testing and generation of diffusion strategy models

[0100] The optimized deep learning model is tested on the test set to evaluate its performance on unseen data. When the model meets the preset performance requirements (such as accuracy above 90% and response time below 100ms), it is used as the diffusion strategy model for subsequent single-step diffusion strategy model construction and task execution.

[0101] Through the above process, from model construction to data partitioning, multi-stage training, model evaluation and optimization, this embodiment finally generates a diffusion strategy model that meets performance requirements, providing a foundation for efficient task execution of robots.

[0102] Furthermore, in one embodiment, the robot control method, wherein the step of constructing a single-step diffusion strategy model based on the diffusion strategy model and migrating the knowledge of the diffusion strategy model to the single-step diffusion strategy model through knowledge distillation technology, specifically comprises the following steps:

[0103] Constructing a single-step diffusion strategy model based on the architecture of the diffusion strategy model, and initializing parameters of the single-step diffusion strategy model using the weights of the diffusion strategy model;

[0104] defining a loss function for measuring the difference in action distribution between the single-step diffusion strategy model and the diffusion strategy model;

[0105] Calculating a loss value between the single-step diffusion strategy model and the diffusion strategy model using the loss function;

[0106] By minimizing the loss value, the parameters of the single-step diffusion strategy model are optimized, so that the knowledge of the diffusion strategy model is transferred to the single-step diffusion strategy model.

[0107] During specific implementation, the specific implementation process of the steps in this embodiment is roughly as follows:

[0108] 1. Construction of single-step diffusion strategy model

[0109] Based on the architecture of the diffusion policy model, a single-step diffusion policy model is constructed. This single-step diffusion policy model consists of two main components: a one-step action generator and a generator scoring network. The one-step action generator is responsible for directly mapping from Gaussian noise to target actions, while the generator scoring network is used to evaluate the quality of the generated actions. To ensure the initial performance of the single-step diffusion policy model, it is initialized using the weights of the diffusion policy model.

[0110] 2. Definition of Action Distribution

[0111] Define the action distributions for the single-step diffusion strategy model and the diffusion strategy model. The action distribution refers to the probability distribution of the model's output actions, reflecting the likelihood and diversity of the actions generated by the model. By comparing the action distributions of the two models, we can quantify their differences and provide a basis for knowledge distillation.

[0112] 3. Calculation of Kullback-Leibler Divergence

[0113] Compute the Kullback-Leibler (KL) divergence between the action distribution of the single-step diffusion policy model and the diffusion policy model. KL divergence is a tool for measuring the difference between two probability distributions. By minimizing KL divergence, we can ensure that the action distribution generated by the single-step diffusion policy model is as close as possible to the action distribution of the diffusion policy model.

[0114] 4. Parameter Optimization

[0115] Based on the calculated KL divergence, the parameters of the single-step diffusion policy model are optimized. Through the backpropagation algorithm, the gradient of the KL divergence is transferred to the parameters of the single-step diffusion policy model, and the parameters are adjusted to minimize the difference in action distribution between the two models. This process can be seen as transferring knowledge from the diffusion policy model to the single-step diffusion policy model.

[0116] 5. Verification of knowledge transfer

[0117] During the optimization process, the performance of the single-step diffusion strategy model is regularly evaluated to ensure that the quality and response speed of the actions it generates meet the expected goals. The effectiveness of knowledge transfer is verified by comparing the performance of the single-step diffusion strategy model with the diffusion strategy model on the validation set. The knowledge transfer process is complete when the performance of the single-step diffusion strategy model reaches or approaches that of the diffusion strategy model.

[0118] Through the above process, this embodiment, from the construction and initialization of the single-step diffusion strategy model to the calculation of KL divergence and parameter optimization, ultimately realizes the migration of knowledge from the diffusion strategy model to the single-step diffusion strategy model, providing an optimized model foundation for the efficient task execution of the robot.

[0119] Furthermore, in one embodiment, the robot control method, wherein the single-step diffusion strategy model is used to process the task instruction input by the user, and the target robot is controlled to perform the target task according to the processing result, specifically comprises the steps of:

[0120] receiving a task instruction input by a user, and formatting the task instruction to convert the task instruction into a vector form required for input of the single-step diffusion strategy model;

[0121] Inputting the formatted task instructions into the single-step diffusion strategy model to generate an action sequence to be executed by the target robot;

[0122] Based on the action sequence, the target robot is controlled to perform a target task.

[0123] Furthermore, the robot control method, wherein the controlling the target robot to perform the target task based on the action sequence, specifically comprises the steps of:

[0124] Smoothing the motion sequence, and checking whether the smoothed motion sequence conforms to the motion parameters of the target robot;

[0125] If it meets the requirements, the target robot is controlled to perform the target task according to the smoothed action sequence.

[0126] Furthermore, the robot control method, wherein, after smoothing the action sequence and checking whether the smoothed action sequence conforms to the motion parameters of the target robot, further specifically comprises the steps of:

[0127] If not, adjusting the smoothed action sequence according to the motion parameters;

[0128] The target robot is controlled to perform the target task according to the adjusted action sequence.

[0129] During specific implementation, the specific implementation process of the steps in this embodiment is roughly as follows:

[0130] 1. Receive and format task instructions

[0131] Receive task instructions from the user, which can be in the form of voice, text, or gestures. Format the task instructions and convert them into the vector form required by the single-step diffusion strategy model. Formatting includes natural language processing (NLP) technology to parse the instructions and convert them into numerical vectors that the model can understand.

[0132] 2. Generate action sequences

[0133] The formatted task instructions are fed into the single-step diffusion strategy model. Based on the input task instructions, the model generates a sequence of actions for the target robot to perform. The action sequence includes parameters such as the robot's joint angles, the position and velocity of the end effector, and other parameters when performing the task.

[0134] 3. Smoothing and motion parameter checking

[0135] The generated motion sequence is smoothed to reduce sudden changes and discontinuities in the motion and ensure smoothness. Smoothing can be done using filters or other mathematical methods. The smoothed motion sequence is then checked to ensure it complies with the target robot's motion parameters, such as joint velocity limits, acceleration limits, and torque limits.

[0136] 4. Action sequence adjustment (if necessary)

[0137] If the smoothed motion sequence does not conform to the target robot's motion parameters, the motion sequence is adjusted based on the robot's motion parameters. Adjustments may include replanning the motion path, adjusting the motion speed, or modifying the torque output to ensure that the motion sequence is within the robot's motion capabilities.

[0138] 5. Control the robot to perform tasks

[0139] According to the adjusted action sequence, the target robot is controlled to perform the target task. The robot receives the action sequence through its control system and performs the corresponding actions according to the instructions in the action sequence to complete the task specified by the user (target task).

[0140] Through the above process, this embodiment ensures the generation, smoothing and adjustment of action sequences from receiving user instructions to controlling the robot to perform tasks, so as to adapt to the movement capabilities of the target robot, thereby achieving efficient and accurate task execution.

[0141] As can be seen from the above method embodiments, the robot control method provided by the present invention includes: collecting received data and action data when the robot performs a task, cleaning, normalizing, and enhancing the received data and the action data, and integrating them to obtain stand-by data; using the stand-by data to train a pre-built deep learning model to generate a diffusion strategy model; based on the diffusion strategy model, constructing a single-step diffusion strategy model, and migrating the knowledge of the diffusion strategy model to the single-step diffusion strategy model through knowledge distillation technology; using the single-step diffusion strategy model to process the task instructions input by the user, and controlling the target robot to perform the target task based on the processing results. In this way, the method of the present invention can effectively improve the response speed and adaptability of the robot when performing tasks.

[0142] It should be understood that although the present application provides method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-creative work, and these operation steps are not necessarily performed in the order of the embodiments or flowcharts. The order of steps listed in the embodiments or flowcharts is only one way of executing the steps among many steps and does not represent the only execution order. It should be noted that there is not necessarily a certain order between the above steps. Those of ordinary skill in the art can understand from the description of the embodiments of the present invention that in different embodiments, the above steps may have different execution orders, that is, they may be executed in parallel, or they may be executed in an interchangeable manner, etc. Moreover, at least a portion of the steps in the embodiments or flowcharts may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but may be executed at different times. The execution order of these sub-steps or stages is not necessarily to be performed in sequence, but may be executed in turn, alternately or synchronously with other steps or at least a portion of the sub-steps or stages of other steps.

[0143] Based on the above method embodiment, please refer to Figure 3 Another embodiment of the present invention further provides a robot control device, wherein the device comprises:

[0144] The collection module 11 is used to collect the received data and action data when the robot performs a task, clean, normalize and enhance the received data and the action data, and integrate them to obtain the data to be used;

[0145] A training module 12 is configured to train a pre-built deep learning model using the stand-by data to generate a diffusion strategy model;

[0146] A migration module 13 is configured to construct a single-step diffusion strategy model based on the diffusion strategy model, and migrate the knowledge of the diffusion strategy model to the single-step diffusion strategy model through a knowledge distillation technique;

[0147] The control module 14 is used to use the single-step diffusion strategy model to process the task instructions input by the user, and control the target robot to perform the target task according to the processing results.

[0148] Furthermore, in one embodiment, the robot control device, wherein the collecting of received data and action data when the robot performs a task, cleaning, normalizing and enhancing the received data and the action data, and integrating them to obtain the ready-to-use data, specifically includes:

[0149] Using a multimodal sensor network, collecting reception data and action data of the robot when performing a task, and performing cleaning, normalization, and enhancement processing on the reception data and the action data;

[0150] The processed received data and the action data are integrated, and the integrated data are marked to obtain stand-by data.

[0151] Furthermore, in one embodiment, the robot control device, wherein the using of the standby data to train a pre-built deep learning model to generate a diffusion strategy model, specifically includes:

[0152] Based on the deep learning architecture, build a deep learning model according to the model generation requirements;

[0153] According to the data partitioning strategy, the to-be-used data is divided into a training set, a validation set and a test set;

[0154] Using the training set, the deep learning model is trained using a multi-stage training strategy, wherein the multi-stage training strategy includes a pre-training stage and a fine-tuning stage;

[0155] Evaluate the trained deep learning model using the validation set, and optimize the deep learning model based on the evaluation results;

[0156] The optimized deep learning model is tested on the test set, and when the deep learning model meets the preset performance requirements, a diffusion strategy model is generated.

[0157] Furthermore, in one embodiment, the robot control device, wherein the step of constructing a single-step diffusion strategy model based on the diffusion strategy model and migrating the knowledge of the diffusion strategy model to the single-step diffusion strategy model through knowledge distillation technology, specifically includes:

[0158] Constructing a single-step diffusion strategy model based on the architecture of the diffusion strategy model, and initializing parameters of the single-step diffusion strategy model using the weights of the diffusion strategy model;

[0159] defining a loss function for measuring the difference in action distribution between the single-step diffusion strategy model and the diffusion strategy model;

[0160] Calculating a loss value between the single-step diffusion strategy model and the diffusion strategy model using the loss function;

[0161] By minimizing the loss value, the parameters of the single-step diffusion strategy model are optimized, so that the knowledge of the diffusion strategy model is transferred to the single-step diffusion strategy model.

[0162] Furthermore, in one embodiment, the robot control device, wherein the processing of the task instruction input by the user using the single-step diffusion strategy model and controlling the target robot to perform the target task according to the processing result, specifically includes:

[0163] receiving a task instruction input by a user, and formatting the task instruction to convert the task instruction into a vector form required for input of the single-step diffusion strategy model;

[0164] Inputting the formatted task instructions into the single-step diffusion strategy model to generate an action sequence to be executed by the target robot;

[0165] Based on the action sequence, the target robot is controlled to perform a target task.

[0166] Furthermore, the robot control device, wherein the control of the target robot to perform the target task based on the action sequence specifically includes:

[0167] Smoothing the motion sequence, and checking whether the smoothed motion sequence conforms to the motion parameters of the target robot;

[0168] If it meets the requirements, the target robot is controlled to perform the target task according to the smoothed action sequence.

[0169] Furthermore, the robot control device, wherein, after smoothing the action sequence and checking whether the smoothed action sequence conforms to the motion parameters of the target robot, further specifically includes:

[0170] If not, adjusting the smoothed action sequence according to the motion parameters;

[0171] The target robot is controlled to perform the target task according to the adjusted action sequence.

[0172] It should be noted that, in the embodiment of the device of the present invention, the information interaction, execution process and other contents between the above modules are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the aforementioned method embodiment part and will not be repeated here.

[0173] Based on the above method embodiment, another embodiment of the present invention further provides a computer device, which can be a server, and its internal structure diagram can be as follows: Figure 4As shown. The computer device includes a processor, a memory, a network interface, and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the functions or steps on the server side of the robot control method in any of the above-mentioned method embodiments are implemented.

[0174] Based on the above method embodiment, another embodiment of the present invention further provides a computer device, which can be a client, and its internal structure diagram can be as follows: Figure 5 As shown. The computer device includes a processor, memory, network interface, display screen, and input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the functions or steps of the client side of the robot control method in any of the above-mentioned method embodiments are implemented.

[0175] Those skilled in the art will understand that Figure 4 and Figure 5 The structural diagram shown in the figure is only a schematic diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied. The specific computer device may include more components than shown in the figure, or combine certain components, or have a different component arrangement.

[0176] The processor referred to herein may be a CPU, other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor, or any conventional processor, etc.

[0177] The memory includes a readable storage medium, an internal memory, etc., wherein the internal memory can be the memory of a computer device, and the internal memory provides an environment for the operation of the operating system and computer-readable instructions in the readable storage medium. The readable storage medium can be a hard disk of the computer device, and in other embodiments, it can also be an external storage device of the computer device, for example, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device. Furthermore, the memory can also include both an internal storage unit of the computer device and an external storage device. The memory is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of a computer program. The memory can also be used to temporarily store data that has been output or is about to be output.

[0178] Based on the above method embodiments, another embodiment of the present invention further provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the robot control method described in any of the above method embodiments. The computer-readable storage medium may be non-volatile or volatile.

[0179] It should be noted that the above-mentioned functions or steps that can be implemented by computer-readable storage media or computer devices, and the technical effects brought about by the functions / steps, can be found in the relevant descriptions in the aforementioned method embodiments. To avoid repetition, they will not be described one by one here.

[0180] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM). The disclosed memory components or memories of the operating environments described herein are intended to comprise one or more of these and / or any other suitable types of memory.

[0181] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, in the embodiment of the device of the present invention, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual application, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other and are not used to limit the scope of protection of the present invention. The specific working process of the units and modules in the above-mentioned device can refer to the corresponding process in the above-mentioned method embodiment, which will not be repeated here. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0182] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0183] In the embodiments provided by the present invention, it should be understood that the disclosed apparatus / computer equipment and methods can be implemented in other ways. For example, the apparatus / computer equipment embodiments described above are merely illustrative. For example, the division of modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0184] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0185] It should be noted that if software tools or components other than those of the Company appear in the embodiments of this application, they are merely for illustration and do not represent actual use. The above embodiments are intended only to illustrate the technical solutions of the present invention, not to limit them. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some of the technical features therein with equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A robot control method, characterized in that: include: Collecting received data and action data when the robot performs a task, cleaning, normalizing, and enhancing the received data and the action data, and integrating them to obtain stand-by data; Using the stand-by data, a pre-built deep learning model is trained to generate a diffusion strategy model; Based on the diffusion strategy model, a single-step diffusion strategy model is constructed, and the knowledge of the diffusion strategy model is transferred to the single-step diffusion strategy model through knowledge distillation technology; The single-step diffusion strategy model is used to process the task instructions input by the user, and the target robot is controlled to perform the target task according to the processing results.

2. The robot control method according to claim 1, characterized in that: The collecting of received data and action data when the robot performs a task, cleaning, normalizing and enhancing the received data and the action data, and integrating them to obtain stand-by data, includes: Using a multimodal sensor network, collecting reception data and action data of the robot when performing a task, and performing cleaning, normalization, and enhancement processing on the reception data and the action data; The processed received data and the action data are integrated, and the integrated data are marked to obtain stand-by data.

3. The robot control method according to claim 1, characterized in that: The method of using the stand-by data to train a pre-built deep learning model to generate a diffusion strategy model includes: Based on the deep learning architecture, build a deep learning model according to the model generation requirements; According to the data partitioning strategy, the to-be-used data is divided into a training set, a validation set and a test set; Using the training set, the deep learning model is trained using a multi-stage training strategy, wherein the multi-stage training strategy includes a pre-training stage and a fine-tuning stage; Evaluate the trained deep learning model using the validation set, and optimize the deep learning model based on the evaluation results; The optimized deep learning model is tested on the test set, and when the deep learning model meets the preset performance requirements, a diffusion strategy model is generated.

4. The robot control method according to claim 1, wherein: The step of constructing a single-step diffusion strategy model based on the diffusion strategy model and migrating the knowledge of the diffusion strategy model to the single-step diffusion strategy model through knowledge distillation technology includes: Constructing a single-step diffusion strategy model based on the architecture of the diffusion strategy model, and initializing parameters of the single-step diffusion strategy model using the weights of the diffusion strategy model; defining a loss function for measuring the difference in action distribution between the single-step diffusion strategy model and the diffusion strategy model; Calculating a loss value between the single-step diffusion strategy model and the diffusion strategy model using the loss function; By minimizing the loss value, the parameters of the single-step diffusion strategy model are optimized, so that the knowledge of the diffusion strategy model is transferred to the single-step diffusion strategy model.

5. The robot control method according to claim 1, characterized in that: The method of using the single-step diffusion strategy model to process the task instruction input by the user and controlling the target robot to perform the target task according to the processing result includes: receiving a task instruction input by a user, and formatting the task instruction to convert the task instruction into a vector form required for input of the single-step diffusion strategy model; Inputting the formatted task instructions into the single-step diffusion strategy model to generate an action sequence to be executed by the target robot; Based on the action sequence, the target robot is controlled to perform a target task.

6. The robot control method according to claim 5, characterized in that: The step of controlling the target robot to perform a target task based on the action sequence includes: Smoothing the motion sequence, and checking whether the smoothed motion sequence conforms to the motion parameters of the target robot; If it meets the requirements, the target robot is controlled to perform the target task according to the smoothed action sequence.

7. The robot control method according to claim 6, characterized in that: After smoothing the action sequence and checking whether the smoothed action sequence conforms to the motion parameters of the target robot, the method further includes: If not, adjusting the smoothed action sequence according to the motion parameters; The target robot is controlled to perform the target task according to the adjusted action sequence.

8. A robot control device, characterized in that: include: A collection module is used to collect reception data and action data when the robot performs tasks, clean, normalize and enhance the reception data and the action data, and integrate them to obtain stand-by data; A training module, configured to train a pre-built deep learning model using the stand-by data to generate a diffusion strategy model; A migration module, configured to construct a single-step diffusion strategy model based on the diffusion strategy model, and migrate the knowledge of the diffusion strategy model to the single-step diffusion strategy model through knowledge distillation technology; The control module is used to use the single-step diffusion strategy model to process the task instructions input by the user, and control the target robot to perform the target task according to the processing results.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the robot control method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the robot control method according to any one of claims 1 to 7 is implemented.

Images