Fine grabbing-oriented dexterous hand adaptive admittance control system and method

Abstract

The invention discloses a dexterous hand self-adaptive admittance control system and method for fine grabbing, and relates to the technical field of robots, and the method comprises the following steps: presetting initial parameters of a dexterous hand, calibrating the state of an intelligent material by using a self-adaptive calibration algorithm, and when the dexterous hand executes an action, adjusting the initial parameters of the dexterous hand; the energy collecting device converts kinetic energy of the dexterous hand into electric energy to be stored in the battery; the method comprises the following steps: comprehensively scanning an operation environment by using a sensor to obtain entity data and environment data of a target object, and preprocessing the entity data and the environment data; and analyzing the preprocessed entity data through an advanced algorithm, identifying the specific characteristics of the target object, and calculating an optimal approaching path by using a mixed path planning method according to the specific characteristics of the target object and the environmental data. According to the method, the initial parameters of the dexterous hand are preset, and the state of the intelligent material integrated in the fingers is calibrated by applying the self-adaptive calibration algorithm, so that the optimal performance of the dexterous hand in different operation environments is ensured.

Description

Technical Field

[0001] The present invention relates to the field of robotics technology, and in particular to an adaptive admittance control system and method for a dexterous hand oriented to fine grasping. Background Art

[0002] With the rapid development of robotics, dexterous hands are increasingly being used for precision grasping tasks. Traditional dexterous hand systems typically rely on pre-programmed path planning and fixed control strategies to achieve grasping operations on objects of specific shapes and sizes. However, these systems are often limited to static environments and known objects, and lack the ability to adapt to complex and changing environments.

[0003] In recent years, researchers have been committed to developing more intelligent and adaptive control systems, integrating multiple sensors and advanced algorithms to enhance the dexterity and precision of dexterous hands. For example, admittance control methods based on vision and force feedback have improved the success rate and stability of grasping to a certain extent. However, existing technologies still have shortcomings in energy management, dynamic environment perception, and real-time adjustment strategies, limiting their effectiveness in practical applications. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an adaptive admittance control method for dexterous hands for fine grasping to solve the problems of low energy utilization efficiency, incomplete environmental perception and insufficient real-time adjustment strategy of dexterous hands in fine grasping tasks.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides an adaptive admittance control method for a dexterous hand for fine grasping, comprising: presetting initial parameters of the dexterous hand, calibrating the state of the smart material using an adaptive calibration algorithm, and converting the kinetic energy of the dexterous hand into electrical energy and storing it in a battery when the dexterous hand performs an action; Use sensors to comprehensively scan the operating environment, obtain the physical data and environmental data of the target object, and pre-process the physical data and environmental data; Advanced algorithms analyze pre-processed physical data, identify the specific features of the target object, and use hybrid path planning methods to calculate the optimal approach path and formulate a grasping strategy based on the specific features of the target object and environmental data; The dexterous hand moves along the calculated approach path, adjusts the pressure applied by the fingers based on force feedback from sensors, and uses machine learning algorithms to adjust the grasping strategy based on environmental data. The grasping task is performed according to the grasping strategy. After completing the grasping task, the grasping force is relaxed, the object is put down, and the dexterous hand is reset to the initial parameters. The dexterous hand generates a log based on historical data and transmits it to the storage device.

[0007] As a preferred solution of the adaptive admittance control method for a dexterous hand for fine grasping described in the present invention, the method includes: presetting the initial parameters of the dexterous hand, using an adaptive calibration algorithm to calibrate the state of the smart material, and when the dexterous hand performs an action, the energy harvesting device converts the kinetic energy of the dexterous hand into electrical energy and stores it in a battery, including the following steps: An automated script presets the initial position of the dexterous hand, the expected maximum grasping force, and the minimum safe distance, activates an adaptive calibration algorithm, and performs status detection on the smart material integrated in the dexterous finger to obtain the physical properties of the smart material. Based on the physical properties of smart materials, the adaptive calibration algorithm adjusts the smart material parameters based on built-in historical data. When it detects that the material properties deviate from the ideal value, it automatically adjusts the excitation signal to restore the material to its optimal working state. After the material returns to its optimal working condition, the energy harvesting device is activated. When the dexterous hand generates kinetic energy, the energy device converts the kinetic energy into electrical energy. The converted electrical energy is stored in a high-efficiency battery, and the power level and charging rate are monitored in real time through the energy management panel.

[0008] As a preferred solution of the adaptive admittance control method for dexterous hands for fine grasping described in the present invention, wherein: using sensors to comprehensively scan the operating environment to obtain the entity data and environmental data of the target object, and pre-processing the entity data and environmental data, the following steps are included: Activate all sensors on the dexterous hand, initialize each sensor, and use visual sensors to scan the target object and its surroundings from multiple angles to generate a series of high-resolution images; Use distance sensors to measure the distance between the target object and other objects and construct point cloud data; A filtering algorithm is used to filter noise from high-resolution images, computer vision technology is used to extract environmental data from the filtered high-resolution images, and data normalization is used to uniformly format all data.

[0009] As a preferred solution of the adaptive admittance control method for dexterous hands for fine grasping described in the present invention, wherein: using advanced algorithms to analyze pre-processed entity data, identifying specific features of the target object, using a hybrid path planning method to calculate the optimal approach path based on the specific features of the target object and environmental data, and formulating a grasping strategy, the following steps are included: Use a deep learning model to extract the 3D features of the physical data. Based on the 3D features of the target object, a pre-trained machine learning model is used to classify and identify the target object. Combined with the point cloud data provided by the range sensor, a 3D spatial model of the surrounding environment is constructed. According to the characteristics of the target object and the needs of the operation task, the basic requirements for grasping are determined, and the intelligent material morphology of the dexterous hand is adjusted according to the shape and surface characteristics of the target object; According to the characteristics of the target object and the three-dimensional model of the surrounding environment, a hybrid path planning method is selected and combined with multi-objective optimization technology to calculate an optimal approach path, and a grasping strategy is formulated based on the characteristics of the target object and the three-dimensional model of the surrounding environment.

[0010] As a preferred solution of the adaptive admittance control method for a dexterous hand for fine grasping described in the present invention, the dexterous hand is moved according to a calculated approach path, and the pressure applied by the finger is adjusted based on the force feedback information provided by the sensor, including the following steps: Based on the optimal approach path, the control algorithm is used to drive the dexterous hand to move to the target location. During the movement, the visual sensor is used to continuously monitor the surrounding environment. When the dexterous fingers touch the target object, real-time contact force data is obtained through force sensors installed on the fingertips. An external computer analyzes the force feedback data to determine whether the current pressure applied to the object is suitable for stable grasping without causing damage. Based on the judgment result, the pressure applied by each finger of the dexterous hand is adjusted in real time. Use temperature sensors to detect the working environment temperature, use humidity sensors to collect ambient humidity, combine data provided by visual sensors and distance sensors, and use data visualization and user interaction to build a real-time updated model of the operating environment.

[0011] As a preferred solution of the adaptive admittance control method for dexterous hands for fine grasping described in the present invention, wherein: according to environmental data, a machine learning algorithm is used to output a grasping strategy, including the following steps: According to the environmental data extracted by the sensor, a feature extraction algorithm is used to extract features from the environmental data to obtain environmental features; Select a machine learning algorithm based on the task requirements, collect and label a large number of training samples, obtain a training data set, use the training data set to train the selected machine learning model, input the environmental features into the trained model, and based on the input data, the machine learning model outputs the best grasping strategy for the current situation, expressed as, ; in, For the best crawling strategy, is the number of environmental features, is the number of training samples, Indicates the A function of environmental characteristics, is the environmental characteristic variable, is the upper limit of the integral, is the lower limit of the integral, is a Gaussian filter, is the mean, is the standard deviation, Indicates the The objective function of the training samples is is the training sample variable, is the average value of the objective function of the training samples.

[0012] As a preferred solution of the adaptive admittance control method for dexterous hands for fine grasping described in the present invention, wherein: a grasping task is performed according to a grasping strategy, after completing the grasping task, the grasping force is relaxed, the object is put down, and the dexterous hands are reset to the initial parameters. The dexterous hands generate logs according to all steps and transmit them to a storage device, including the following steps: The calculated approach path and grasping strategy are loaded, and the dexterous hand is controlled to move to the target object along the approach path. Based on the finger posture, contact points, and applied pressure values ​​set in the grasping strategy, the position and pressure of each finger are adjusted to achieve stable grasping of the target object. After completing the grasping task, the total pressure applied to the object is reduced to maintain sufficient grip to prevent the object from falling. The visual sensor is used to identify the exact location of the placement area to ensure that the object can be accurately placed on it. The pressure of each finger is reduced in a predetermined order, and the object is moved towards the placement point. When on the designated surface, the speed is further reduced to ensure a smooth placement. After the grasping task is completed, all joints of the dexterous hand are reset to their initial parameters, and a comprehensive self-diagnosis is performed on the dexterous hand to check whether the functions of each sensor and actuator are normal; During the entire crawling task cycle, key events and parameter changes are continuously recorded, logs are generated based on historical data, and the generated log files are transmitted to a remote storage device through a network interface for storage.

[0013] In a second aspect, the present invention provides an adaptive admittance control system for a dexterous hand for fine grasping, comprising an initialization and calibration module, a data processing module, a path planning module, a grasping strategy module, and a log recording module; The initialization and calibration module presets the initial parameters of the dexterous hand and uses an adaptive calibration algorithm to calibrate the state of the smart material. When the dexterous hand performs an action, the energy harvesting device converts the kinetic energy of the dexterous hand into electrical energy and stores it in the battery; The data processing module uses sensors to perform a comprehensive scan of the operating environment to obtain physical data and environmental data of the target object, and pre-processes the physical data and environmental data; The path planning module analyzes the pre-processed entity data through advanced algorithms, identifies the specific features of the target object, calculates the optimal approach path based on the specific features of the target object and environmental data, and formulates a grasping strategy; The grasping strategy module moves the dexterous hand according to the calculated approach path, adjusts the pressure applied by the fingers based on the force feedback information provided by the sensor, and uses a machine learning algorithm to adjust the grasping strategy according to the environmental data; The log recording module performs a grasping task according to the grasping strategy. After completing the grasping task, it relaxes the grasping force, puts down the object, and resets the dexterous hand to the initial parameters. The dexterous hand generates a log based on historical data and transmits it to the storage device.

[0014] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the adaptive admittance control method for dexterous hands for fine grasping as described in the first aspect of the present invention is implemented.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the adaptive admittance control method for dexterous hands for fine grasping as described in the first aspect of the present invention.

[0016] The beneficial effects of the present invention are as follows: by presetting the initial parameters of the dexterous hand and applying an adaptive calibration algorithm to calibrate the state of the smart materials integrated in the fingers, the optimal performance of the dexterous hand in different operating environments is ensured, a variety of sensors are used to perform an all-round scan of the operating environment to obtain detailed information about the target object and its surrounding environment, and the collected data is preprocessed to improve the quality and availability of the data, and advanced algorithms are used to conduct in-depth analysis of the preprocessed data to identify the specific characteristics of the target object, and a hybrid path planning method is used in combination with environmental data to calculate the optimal approach path, formulate a detailed grasping strategy, and ensure the efficiency and safety of the grasping path. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 This is a flow chart of the adaptive admittance control method for dexterous hands for fine grasping in Example 1.

[0019] Figure 2 This is a module diagram of the adaptive admittance control method for dexterous hands for fine grasping in Example 1. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0021] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0023] Example 1, with reference to Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides an adaptive admittance control method for a dexterous hand for fine grasping, comprising the following steps: S1. Preset the initial parameters of the dexterous hand and use the adaptive calibration algorithm to calibrate the state of the smart material. When the dexterous hand performs an action, the energy harvesting device converts the kinetic energy of the dexterous hand into electrical energy and stores it in the battery. The process includes the following steps: Start the dexterous hand and run the automated script. Based on the task requirements, set the initial angles and positions of each joint of the dexterous hand to ensure that the fingers can accurately approach the target object. Set the expected maximum grasping force based on the material properties to avoid damage caused by excessive force. Activate the adaptive calibration algorithm to detect the state of the smart material integrated in the dexterous finger and measure the current physical properties of the smart material, including hardness and elastic modulus. The adaptive calibration algorithm refers to the built-in historical data set, which contains the ideal working state of the smart material under different conditions and the corresponding excitation signals. Based on the state detection results, if the physical properties of the smart material deviate from the ideal value, the adaptive calibration algorithm automatically adjusts the excitation signal to restore the material to the optimal working state. The algorithm uses closed-loop control logic to continuously monitor the material response and adjust the excitation signal in real time based on the feedback until the optimal state is achieved. The calibrated smart material is then retested to ensure that its performance meets the expected requirements. When the dexterous hand begins to perform an action, the built-in energy harvesting device is activated, converting the kinetic energy generated by the dexterous hand into electrical energy. During this process, the energy harvesting device captures the mechanical motion and converts it into storable electrical energy, which is then stored in a high-performance battery. Simultaneously, the energy management panel monitors the power level and charging rate in real time to ensure a stable power supply.

[0024] S2. Use sensors to fully scan the operating environment to obtain the entity data and environmental data of the target object, and pre-process the entity data and environmental data, including the following steps: Activate all sensors on the dexterous hand, including but not limited to visual sensors, distance sensors, and force sensors, and initialize each sensor to ensure they are in optimal working condition, including but not limited to adjusting the resolution, frame rate, and sensitivity to suit the current task requirements; Use vision sensors to scan the target object and its surroundings from multiple angles, generating a series of high-resolution images. Panoramic photography technology can be used to change the viewing angle through the movement of the robotic arm to ensure comprehensive coverage. Simultaneously, distance sensors are used to measure the distance between the target object and other objects, constructing point cloud data for a 3D spatial model. This point cloud data provides detailed information about the object's shape and position. Timestamps are used to ensure that data from all sensors accurately corresponds to the same moment, guaranteeing data consistency and accuracy. First, a filtering algorithm is used to filter the raw data to remove unnecessary signals caused by the sensor itself and external interference. Gaussian blur and bilateral filtering techniques are used to denoise the high-resolution image to reduce random noise in the image. Statistical outlier removal and voxel grid downsampling methods are applied to the point cloud data to remove isolated points and abnormal values, thereby improving data quality. Use computer vision technology to extract useful features from images and point cloud data. For distance sensor data, focus on analyzing the relative position relationship between objects. Standardize numerical data to make them have similar scales, convert different types of data into the same scale and unit for subsequent processing, uniformly format the preprocessed ambient data, and record the key parameters and results of the entire data collection and preprocessing process.

[0025] S3. Analyze the pre-processed entity data through advanced algorithms, identify the specific features of the target object, calculate the optimal approach path based on the specific features of the target object and environmental data, and formulate a grasping strategy, including the following steps: The data augmentation technology is used on the preprocessed surrounding environment data to increase the number and diversity of training samples to obtain the surrounding environment dataset. The surrounding environment dataset is divided into a training set, a validation set, and a test set, with a ratio of 70% training set, 15% validation set, and 15% test set; Select the deep learning model architecture based on the task requirements, select the loss function based on the task type, select the optimization algorithm based on the task type to update the model parameters, and determine the learning rate, batch size, and number of iterations of the deep learning model through grid search; Start training the deep learning model by extracting a batch of data from the training set and inputting it into the model. Calculate the loss and backpropagate the error to update the model parameters. Repeat this process until the predetermined number of iterations is reached. Use the validation set to evaluate the model's performance. Based on the validation results, further optimize the model, including but not limited to adjusting the model structure, increasing the amount of data, improving the data augmentation strategy, and adjusting hyperparameters. Use an independent test set to evaluate the final performance of the model. The data in the test set should not be seen by the model during the training process to provide an unbiased performance evaluation. Continuously collect new data based on feedback from actual applications and retrain the model regularly to maintain its effectiveness in dynamic environments. Ultimately, a trained deep learning model is obtained. Use a trained deep learning model to extract key features of target objects, including but not limited to shape, color, and texture, from high-resolution images provided by visual sensors. The model can automatically learn and extract features that help distinguish different objects. Based on the extracted features, a pre-trained machine learning model is used to classify and identify the target objects, determine the object type, and provide a preliminary estimate of its physical properties. Combined with the point cloud data provided by the range sensor, a 3D model of the surrounding environment is constructed to help identify the location and size of obstacles. Point cloud processing tools are used to process the point cloud data to generate an accurate 3D spatial model. Determine the basic requirements for grasping based on the specific characteristics of the target object and the needs of the manipulation task. For example, for fragile objects, set a lower maximum grasping force. For heavy objects, consider increasing the contact area to distribute the pressure. When the dexterous hand integrates smart materials, the finger shape can be adjusted according to the shape and surface characteristics of the target object. Based on the complexity of the task and the characteristics of the environment, a hybrid path planning method is selected, combined with a genetic algorithm, to calculate a path that can avoid obstacles and reach the target location efficiently, thus obtaining the optimal approach path. In the path planning process, in addition to considering the shortest path, other factors must be taken into account, including but not limited to minimizing energy consumption and maximizing motion smoothness. Based on the three-dimensional model of the target object characteristics and the surrounding environment, a specific grasping strategy is formulated, including but not limited to finger posture adjustment and applied pressure control, real-time monitoring of environmental changes, and dynamic adjustment of the path planning scheme accordingly to ensure the effectiveness and safety of the path even in a dynamic environment.

[0026] S4, moving the dexterous hand according to the calculated approach path, and adjusting the pressure applied by the finger based on the force feedback information provided by the sensor, including the following steps: The optimal approach path is loaded and a control algorithm is used to drive the dexterous hand to move to the target location along the predetermined path, ensuring a stable speed and direction during movement. During the movement, the visual sensor is used to continuously monitor the surrounding environment to promptly detect new obstacles or environmental changes that may affect the execution of the task. When the fingers of the dexterous hand touch the target object, real-time contact force data is obtained through force sensors installed on the fingertips. These sensors can provide accurate force feedback information. An external computer analyzes the force feedback data to determine whether the current pressure applied to the object is suitable for stable grasping without causing damage. For example, a threshold can be set to determine the maximum allowable pressure value to ensure that fragile objects are not damaged by excessive force. Based on the analysis results, the pressure applied by each finger of the dexterous hand is adjusted in real time. Use temperature sensors to detect the working environment temperature to prevent extreme conditions from affecting equipment performance. Use humidity sensors to collect ambient humidity to prevent electrical failures caused by excessive humidity. Combined with data provided by vision sensors and distance sensors, as well as data from temperature and humidity sensors, a multi-dimensional environmental monitoring panel is formed. Use LiDAR to generate 3D point cloud data of the environment, and use the point cloud library for processing to extract feature points and plane information. Synchronous positioning and map construction - technology is used to simultaneously estimate the robot's position in an unknown environment and build an environmental map. It combines visual SLAM and LiDAR data to improve accuracy. By analyzing the differences between consecutive frames, moving objects are identified and separated from the static background, achieving incremental updates, only updating the parts that have changed instead of recalculating the entire environmental model each time.

[0027] S5. Based on the environmental data, the machine learning algorithm is used to adjust the crawling strategy, including the following steps: Select an appropriate machine learning algorithm based on task requirements, collect and annotate training samples from environmental data to form a training dataset, and use the resulting training dataset to train the selected machine learning model; During the training process, the model performance can be evaluated through cross-validation methods, and model parameters can be adjusted as needed to optimize performance. The extracted environmental features are used as input, and the corresponding optimal grasping strategy is used as the output label. The learning rate and regularization coefficient are adjusted to improve the generalization ability and accuracy of the model. The environmental data collected in real time is input into the trained model after feature extraction. Based on the input data, the machine learning model outputs the best grasping strategy for the current situation. The grasping strategy based on the model output is executed and dynamically adjusted according to the actual effect. At the same time, the key data of each operation is recorded for model improvement, which is expressed as, ; in, For the best crawling strategy, is the number of environmental features, which is used to represent the total number of different environmental features. is the number of training samples, which is used to represent the total number of samples for training the model. Indicates the A function of environmental characteristics, used to determine how specific environmental characteristics affect the final grasping strategy. is an environmental characteristic variable, used to represent different environmental characteristics. is the upper limit of the integral, which is used to define the upper limit of the effective range of the environmental characteristics. is the lower limit of the integral, which is used to define the lower limit of the effective range of the environmental characteristics. is a Gaussian filter used to smooth and weight the data. is the mean, which is used to represent the central tendency of environmental characteristic data. is the standard deviation, which is used to indicate the fluctuation range of environmental characteristic data. Indicates the The objective function of each training sample is used to evaluate the ideal capture effect of each training sample under given features. is a training sample variable, used to represent different training samples. It is the average value of the objective function of the training samples, and its use in data of different scales and ranges will not affect the final path results.

[0028] S6. Perform the grasping task according to the grasping strategy. After completing the grasping task, relax the grasping force, put down the object, and reset the dexterous hand to the initial parameters. The dexterous hand generates a log based on all steps and transmits it to the storage device, including the following steps: Using an automation panel to load the optimal approach path calculated using a hybrid path planning method and the grasping strategy adjusted using a machine learning algorithm, these strategies include but are not limited to information about the finger posture, contact points, and applied pressure values. The dexterous hand is controlled to precisely move to the target object along the pre-planned path, and the position and pressure of each finger are gradually adjusted based on the finger posture, contact points, and applied pressure values ​​set in the grasping strategy. Use visual sensors to monitor and confirm in real time that the object is firmly grasped, checking for signs of slippage and instability. When preparing to place the object, first reduce the total pressure applied to the object while maintaining sufficient grip to prevent the object from falling. Use visual sensors to identify the edges and center points of the placement area to ensure that the object can be accurately placed there. Reduce the pressure of each finger in a predetermined sequence, slowly move the object toward the placement point, and further reduce the speed when approaching the designated surface to ensure a smooth placement. After placement is completed, a notification is immediately sent to the operator and the control panel to inform them that the grasping task has been completed. After the grasping task is completed, all joints of the dexterous hand are reset to their initial states, including but not limited to finger positions, joint angles, sensor calibration, energy status, and smart material status. The dexterous hand test program is run to verify the working status of each component and check whether the functions of each sensor and actuator are normal. During the entire grasping task cycle, key events and parameter changes are continuously recorded, including but not limited to sensor readings, control instructions, and timestamps. The data in the log is automatically analyzed, important indicators are extracted, and the performance of the task is evaluated accordingly. The generated log files are transmitted to a remote storage device through a network interface for storage to ensure data security and accessibility.

[0029] This embodiment also provides a dexterous hand adaptive admittance control system for fine grasping, comprising: an initialization and calibration module, a data processing module, a path planning module, a grasping strategy module, and a log recording module; The initialization and calibration module presets the initial parameters of the dexterous hand and uses an adaptive calibration algorithm to calibrate the state of the smart material. When the dexterous hand performs an action, the energy harvesting device converts the kinetic energy of the dexterous hand into electrical energy and stores it in the battery. The data processing module uses sensors to comprehensively scan the operating environment, obtain the physical data and environmental data of the target object, and preprocess the physical data and environmental data. The path planning module analyzes the preprocessed physical data through advanced algorithms, identifies the specific characteristics of the target object, and uses a hybrid path planning method to calculate the optimal approach path and formulate a grasping strategy based on the specific characteristics of the target object and the environmental data. The grasping strategy module moves the dexterous hand according to the calculated approach path, adjusts the pressure applied by the fingers based on the force feedback information provided by the sensor, and uses a machine learning algorithm to adjust the grasping strategy based on the environmental data. The log recording module performs the grasping task according to the grasping strategy. After completing the grasping task, the gripping force is relaxed, the object is put down, and the dexterous hand is reset to the initial parameters. The dexterous hand generates logs based on historical data and transmits them to the storage device.

[0030] This embodiment also provides a computer device suitable for the adaptive admittance control method of dexterous hands for fine grasping, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the adaptive admittance control method of dexterous hands for fine grasping proposed in the above embodiment.

[0031] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0032] This embodiment also provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the program implements the adaptive admittance control method for a dexterous hand for fine grasping proposed in the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0033] In summary, the present invention ensures the optimal performance of the dexterous hand in different operating environments by: presetting the initial parameters of the dexterous hand and applying an adaptive calibration algorithm to calibrate the state of the smart materials integrated in the fingers; uses a variety of sensors to perform an all-round scan of the operating environment to obtain detailed information about the target object and its surrounding environment, and preprocesses the collected data to improve the quality and availability of the data; applies advanced algorithms to conduct in-depth analysis of the preprocessed data to identify the specific characteristics of the target object; and uses a hybrid path planning method in combination with environmental data to calculate the optimal approach path, formulates a detailed grasping strategy, and ensures the efficiency and safety of the grasping path.

[0034] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An adaptive admittance control method for a dexterous hand for fine grasping, characterized by: include, The initial parameters of the dexterous hand are preset, and the state of the smart material is calibrated using an adaptive calibration algorithm. When the dexterous hand performs an action, the energy harvesting device converts the kinetic energy of the dexterous hand into electrical energy and stores it in a battery. Use sensors to comprehensively scan the operating environment, obtain the physical data and environmental data of the target object, and pre-process the physical data and environmental data; Advanced algorithms analyze pre-processed physical data, identify the specific features of the target object, and use hybrid path planning methods to calculate the optimal approach path and formulate a grasping strategy based on the specific features of the target object and environmental data; The dexterous hand moves along the calculated approach path, adjusts the pressure applied by the fingers based on force feedback from sensors, and uses machine learning algorithms to adjust the grasping strategy based on environmental data. The grasping task is performed according to the grasping strategy. After completing the grasping task, the grasping force is relaxed, the object is put down, and the dexterous hand is reset to the initial parameters. The dexterous hand generates a log based on historical data and transmits it to the storage device.

2. The adaptive admittance control method for dexterous hands for fine grasping according to claim 1, characterized in that: The initial parameters of the dexterous hand are preset, and the state of the smart material is calibrated using an adaptive calibration algorithm. When the dexterous hand performs an action, the energy harvesting device converts the kinetic energy of the dexterous hand into electrical energy and stores it in the battery. The following steps are included: An automated script presets the initial position of the dexterous hand, the expected maximum grasping force, and the minimum safe distance, activates an adaptive calibration algorithm, and performs status detection on the smart material integrated in the dexterous finger to obtain the physical properties of the smart material. Based on the physical properties of smart materials, the adaptive calibration algorithm adjusts the smart material parameters based on built-in historical data. When it detects that the material properties deviate from the ideal value, it automatically adjusts the excitation signal to restore the material to its optimal working state. After the material returns to its optimal working condition, the energy harvesting device is activated. When the dexterous hand generates kinetic energy, the energy device converts the kinetic energy into electrical energy. The converted electrical energy is stored in a high-efficiency battery, and the power level and charging rate are monitored in real time through the energy management panel.

3. The adaptive admittance control method for dexterous hands for fine grasping according to claim 2, characterized in that: The sensor is used to fully scan the operating environment to obtain the entity data and environmental data of the target object, and the entity data and environmental data are pre-processed. The following steps are included: Activate all sensors on the dexterous hand, initialize each sensor, and use visual sensors to scan the target object and its surroundings from multiple angles to generate a series of high-resolution images; Use distance sensors to measure the distance between the target object and other objects and construct point cloud data; A filtering algorithm is used to filter noise from high-resolution images, computer vision technology is used to extract environmental data from the filtered high-resolution images, and data normalization is used to uniformly format all data.

4. The method for adaptive admittance control of a dexterous hand for fine grasping according to claim 3, characterized in that: The method uses advanced algorithms to analyze pre-processed entity data, identify the specific features of the target object, calculate the optimal approach path based on the specific features of the target object and environmental data, and formulate a grasping strategy. The following steps are included: Use a deep learning model to extract the 3D features of the physical data. Based on the 3D features of the target object, a pre-trained machine learning model is used to classify and identify the target object. Combined with the point cloud data provided by the range sensor, a 3D spatial model of the surrounding environment is constructed. According to the characteristics of the target object and the needs of the operation task, the basic requirements for grasping are determined, and the intelligent material morphology of the dexterous hand is adjusted according to the shape and surface characteristics of the target object; According to the characteristics of the target object and the three-dimensional model of the surrounding environment, a hybrid path planning method is selected and combined with multi-objective optimization technology to calculate an optimal approach path, and a grasping strategy is formulated based on the characteristics of the target object and the three-dimensional model of the surrounding environment.

5. The adaptive admittance control method for dexterous hands for fine grasping according to claim 4, characterized in that: The dexterous hand is moved according to the calculated approach path, and the pressure applied by the finger is adjusted based on the force feedback information provided by the sensor. The following steps are included: Based on the optimal approach path, the control algorithm is used to drive the dexterous hand to move to the target location. During the movement, the visual sensor is used to continuously monitor the surrounding environment. When the dexterous fingers touch the target object, real-time contact force data is obtained through force sensors installed on the fingertips. An external computer analyzes the force feedback data to determine whether the current pressure applied to the object is suitable for stable grasping without causing damage. Based on the judgment result, the pressure applied by each finger of the dexterous hand is adjusted in real time. Use temperature sensors to detect the working environment temperature, use humidity sensors to collect ambient humidity, combine data provided by visual sensors and distance sensors, and use data visualization and user interaction to build a real-time updated model of the operating environment.

6. The adaptive admittance control method for dexterous hands for fine grasping according to claim 5, characterized in that: The method of adjusting the crawling strategy using a machine learning algorithm based on environmental data includes the following steps: According to the environmental data extracted by the sensor, a feature extraction algorithm is used to extract features from the environmental data to obtain environmental features; Select a machine learning algorithm based on task requirements, collect and label a large number of training samples to obtain a training data set, use the training data set to train the selected machine learning model, input environmental features into the trained model, and based on the input data, the machine learning model outputs the best grasping strategy for the current situation.

7. The method for adaptive admittance control of a dexterous hand for fine grasping according to claim 6, characterized in that: The grasping task is performed according to the grasping strategy. After completing the grasping task, the grasping force is relaxed, the object is put down, and the dexterous hand is reset to the initial parameters. The dexterous hand generates a log based on all steps and transmits it to the storage device. The following steps are included: The calculated approach path and grasping strategy are loaded, and the dexterous hand is controlled to move to the target object along the approach path. Based on the finger posture, contact points, and applied pressure values ​​set in the grasping strategy, the position and pressure of each finger are adjusted to achieve stable grasping of the target object. After completing the grasping task, the total pressure applied to the object is reduced to maintain sufficient grip to prevent the object from falling. The visual sensor is used to identify the exact location of the placement area to ensure that the object can be accurately placed on it. The pressure of each finger is reduced in a predetermined order, and the object is moved towards the placement point. When on the designated surface, the speed is further reduced to ensure a smooth placement. After the grasping task is completed, all joints of the dexterous hand are reset to their initial parameters, and a comprehensive self-diagnosis is performed on the dexterous hand to check whether the functions of each sensor and actuator are normal; During the entire crawling task cycle, key events and parameter changes are continuously recorded, logs are generated based on historical data, and the generated log files are transmitted to a remote storage device through a network interface for storage.

8. An adaptive admittance control system for a dexterous hand for fine grasping, based on the adaptive admittance control method for a dexterous hand for fine grasping according to any one of claims 1 to 7, characterized in that: Including initialization and calibration module, data processing module, path planning module, grasping strategy module and log recording module; The initialization and calibration module presets the initial parameters of the dexterous hand and uses an adaptive calibration algorithm to calibrate the state of the smart material. When the dexterous hand performs an action, the energy harvesting device converts the kinetic energy of the dexterous hand into electrical energy and stores it in the battery; The data processing module uses sensors to perform a comprehensive scan of the operating environment to obtain physical data and environmental data of the target object, and pre-processes the physical data and environmental data; The path planning module analyzes the pre-processed entity data through advanced algorithms, identifies the specific features of the target object, calculates the optimal approach path based on the specific features of the target object and environmental data, and formulates a grasping strategy; The grasping strategy module moves the dexterous hand according to the calculated approach path, adjusts the pressure applied by the fingers based on the force feedback information provided by the sensor, and uses a machine learning algorithm to adjust the grasping strategy according to the environmental data; The log recording module performs a grasping task according to the grasping strategy. After completing the grasping task, it relaxes the grasping force, puts down the object, and resets the dexterous hand to the initial parameters. The dexterous hand generates a log based on historical data and transmits it to the storage device.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the adaptive admittance control method for dexterous hands for fine grasping according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the adaptive admittance control method for dexterous hands for fine grasping according to any one of claims 1 to 7 are implemented.

Images