Parameter configuration method of dexterous hand and related equipment

Abstract

The embodiment of the invention provides a parameter configuration method for a dexterous hand and related equipment, and the method comprises the steps: obtaining a working parameter when the dexterous hand executes a target specified action, and collecting an assistance parameter of a data glove, the assistance parameter being used for instructing an operator to operate the data glove and assist the dexterous hand, executing parameters of the data glove when the target specified action is executed; according to the assistance parameter, determining an action compensation amount when the operator assists the dexterous hand to execute the target specified action; and optimizing the parameters of the dexterous hand according to the action compensation amount and the working parameters. According to the method, the parameter optimization efficiency of the dexterous hand is improved.

Description

Technical Field

[0001] This application relates to the field of robotics, and in particular to a method for configuring parameters of a dexterous hand and related equipment. Background Technology

[0002] Today, by utilizing high-precision tactile sensors, advanced control algorithms, and precise mechanical design, the dexterity and freedom of the hand have been effectively improved. It can not only complete repetitive tasks that traditional industrial robotic arms can perform, but also cope with more complex and ever-changing operating environments.

[0003] However, due to fundamental differences between the human hand and the dexterous hand in terms of kinematic structure, degree of freedom configuration, and joints, continuous optimization of the dexterous hand's parameters is necessary to achieve similar functionality to the human hand and improve the dexterous hand's performance in executing actions. Currently, this is often achieved by issuing control commands to the dexterous hand and collecting the parameters output by the dexterous hand when executing the corresponding actions, and then optimizing the dexterous hand's parameters based on these parameters.

[0004] However, due to problems such as insufficient accuracy, excessive noise, or systematic bias in the parameters collected by the dexterous hand, the optimization efficiency of the collected parameters is also low without changing the dexterous hand itself. Summary of the Invention

[0005] This application proposes a parameter configuration method and related equipment for a dexterous hand to solve the problem of low efficiency in optimizing parameters collected by a dexterous hand.

[0006] In a first aspect, embodiments of this application provide a method for optimizing the parameters of a dexterous hand, including: The system acquires the working parameters of the dexterous hand when performing the target-specified action, and collects the assistance parameters of the data glove. The assistance parameters are used to instruct the operator to operate the data glove to assist the dexterous hand in performing the target-specified action. Based on the assistance parameters, determine the motion compensation amount when the operator assists the dexterous hand in performing the target-specified action; The parameters of the dexterous hand are optimized based on the motion compensation amount and the working parameters.

[0007] In one embodiment, optimizing the parameters of the dexterous hand based on the motion compensation amount and the working parameters includes: Determine the degree of completion of the dexterous hand in performing the specified action on the target, and compare the degree of completion with a completion threshold; When the completion rate is greater than or equal to the completion rate threshold, the parameters of the dexterous hand are optimized based on the comparison results between the working parameters, the action compensation amount and the compensation amount threshold.

[0008] In one embodiment, optimizing the parameters of the dexterous hand based on the comparison result between the working parameters, the motion compensation amount, and the compensation amount threshold includes: When the comparison result indicates that the action compensation amount is greater than or equal to the sub-compensation amount threshold, the current training sample is generated based on the action compensation amount, the target specified action, and the working parameters. The preset model is trained based on multiple target training samples to obtain the target mapping relationship between the operation instructions required for the dexterous hand to perform the action and the parameters output by the dexterous hand when performing the operation instructions. The multiple target training samples include current training samples and historical training samples. The historical training samples are determined based on the parameters collected when the operator operates the data glove to assist the dexterous hand in performing other specified actions. Replace the current mapping relationship in the dexterous hand with the target mapping relationship.

[0009] In one embodiment, optimizing the parameters of the dexterous hand based on the comparison result between the working parameters, the motion compensation amount, and the compensation amount threshold includes: When the comparison result indicates that the motion compensation amount is less than the sub-compensation amount threshold, the expected parameters output by the dexterous hand when it completes the target specified motion are determined according to the working parameters and the motion compensation amount. Based on the expected parameters, the current mapping relationship of the dexterous hand is optimized.

[0010] In one embodiment, optimizing the parameters of the dexterous hand based on the motion compensation amount and the working parameters includes: The parameters to be compensated for the dexterous hand are identified based on the motion compensation amount; The parameters to be compensated are corrected based on the operating parameters.

[0011] In one embodiment, identifying the parameters to be compensated for the dexterous hand based on the motion compensation amount includes: Based on each sub-compensation quantity in the motion compensation quantity, multiple feature values ​​are determined. Each sub-compensation quantity is the compensation quantity corresponding to each moment within the time period during which the operator assists the dexterous hand. The multiple feature values ​​include the mean, standard deviation, coefficient of variation, and maximum value of each sub-compensation quantity. Based on each of the aforementioned feature values, the parameters to be compensated for by the dexterous hand are identified.

[0012] In one embodiment, identifying the parameters to be compensated for the dexterous hand based on the motion compensation amount includes: Based on the motion compensation amount, the activation intensity of each joint component of the dexterous hand is determined; The parameters to be compensated for in the dexterous hand are identified based on the activation intensity of each joint component.

[0013] In one embodiment, identifying the parameters to be compensated for the dexterous hand based on the motion compensation amount includes: Based on the motion compensation amount, determine the contribution of each joint component in the dexterous hand to the execution of the target-specified motion; Based on the contribution, a target joint component is determined among the various joint components, wherein the contribution of the target joint component is greater than a preset threshold. Based on the target joint components, determine the motion pattern to be adjusted for the dexterous hand; Based on the motion pattern, the parameters to be compensated for the dexterous hand are determined.

[0014] In one embodiment, identifying the parameters to be compensated for the dexterous hand based on the motion compensation amount includes: The motion compensation amount is decomposed into sub-compensation amounts for each joint component of the dexterous hand by the operator; The target features are extracted from the sub-compensation amounts of the joint components, and the target features are time-domain features or frequency-domain features; Based on the target features of each of the joint components, the parameters to be compensated for by the dexterous hand are identified.

[0015] In one embodiment, optimizing the parameters of the dexterous hand based on the motion compensation amount and the working parameters includes: Based on the motion compensation amount and the working parameters, determine the trigger factor for the dexterous hand to perform compensation; A correction strategy is determined based on the triggering factor, and the parameters of the dexterous hand are optimized based on the correction strategy.

[0016] In one embodiment, optimizing the parameters of the dexterous hand based on the motion compensation amount and the working parameters includes: Based on the preset compensation function, the action compensation amount, and the working parameters, a first operation instruction is determined for the operator to assist the dexterous hand in performing the target-specified action; Based on the first operation instruction and the second operation instruction output by the dexterous hand when performing the target-specified action, the current mapping relationship of the dexterous hand is corrected. The current mapping relationship is the mapping relationship between the operation instruction output by the dexterous hand and the parameters output by the dexterous hand when performing the operation instruction.

[0017] In one embodiment, optimizing the parameters of the dexterous hand based on the motion compensation amount and the working parameters includes: Based on the motion compensation amount, determine the compensation behavior by which the operator assists the dexterous hand in performing the target-specified motion; The parameters to be corrected for the dexterous hand are determined based on the compensation behavior, and the parameters to be corrected for the dexterous hand are optimized based on the motion compensation amount and the working parameters.

[0018] The acquisition module is used to acquire the working parameters of the dexterous hand when performing the target-specified action, and to collect the assistance parameters of the data glove. The assistance parameters are used to instruct the operator to operate the data glove to assist the dexterous hand in performing the target-specified action. The determining module is used to determine the motion compensation amount when the operator assists the dexterous hand in performing the target-specified action, based on the assistance parameters. An optimization module is used to optimize the parameters of the dexterous hand based on the motion compensation amount and the working parameters.

[0019] Thirdly, embodiments of this application provide an electronic device, including: One or more processors; A memory having stored one or more programs that, when executed by one or more processors, cause the one or more processors to implement, as described in the first aspect, the parameter optimization method for a dexterous hand.

[0020] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements, as described in the first aspect, the parameter optimization method for a dexterous hand.

[0021] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements, for example, the parameter optimization method for a dexterous hand described in the first aspect.

[0022] In this embodiment, the working parameters of the dexterous hand performing a specified action are obtained, and the assistance parameters of the data glove are collected. Based on the assistance parameters, the motion compensation amount when the operator assists the dexterous hand in performing the specified action is determined. Therefore, the parameters of the dexterous hand are optimized based on the motion compensation amount and the working parameters. In this application, when the dexterous hand performs a specified action, the data glove assists the dexterous hand in completing the specified action. The assistance parameters of the data glove accurately determine the motion compensation amount when the operator assists the dexterous hand in performing the specified action. This motion compensation amount then precisely guides the dexterous hand in parameter optimization, improving the efficiency of parameter optimization. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0024] Figure 1 One of the flowcharts illustrating an embodiment of the parameter optimization method for the dexterous hand provided in this application.

[0025] Figure 2 This is a second flowchart illustrating an embodiment of the parameter optimization method for the dexterous hand provided in this application.

[0026] Figure 3 The third flowchart illustrates an embodiment of the parameter optimization method for the dexterous hand provided in this application.

[0027] Figure 4 The fourth flowchart illustrates an embodiment of the parameter optimization method for the dexterous hand provided in this application.

[0028] Figure 5 The fifth flowchart illustrates an embodiment of the parameter optimization method for the dexterous hand provided in this application.

[0029] Figure 6 This is the sixth flowchart illustrating an embodiment of the parameter optimization method for the dexterous hand provided in this application.

[0030] Figure 7 The seventh flowchart illustrates an embodiment of the parameter optimization method for the dexterous hand provided in this application.

[0031] Figure 8 This is the eighth flowchart illustrating an embodiment of the parameter optimization method for the dexterous hand provided in this application.

[0032] Figure 9 The ninth flowchart illustrates an embodiment of the parameter optimization method for the dexterous hand provided in this application.

[0033] Figure 10 The tenth flowchart illustrates an embodiment of the parameter optimization method for the dexterous hand provided in this application.

[0034] Figure 11 A schematic diagram of the functional modules of the parameter optimization device for a dexterous hand provided in an embodiment of this application.

[0035] Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Specific Implementation To enable those skilled in the art to better understand the technical solutions of this application, the technical solutions provided in this application will be described in detail below with reference to the accompanying drawings.

[0037] Exemplary embodiments will be described more fully below with reference to the accompanying drawings; however, the described exemplary embodiments may be embodied in different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will enable those skilled in the art to fully understand the scope of this application.

[0038] As used herein, the term "and / or" includes any and all combinations of one or more related enumerated purposes.

[0039] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. As used herein, the singular forms “a” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that when the terms “comprising” and / or “made of” are used in this specification, the presence of a feature, integral, step, operation, element, and / or component is specified, but the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof is not excluded.

[0040] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0041] Unless otherwise specified, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and this application, and will not be interpreted as having an idealized or overly formal meaning, unless expressly so defined in the embodiments of this application.

[0042] Today, by utilizing high-precision tactile sensors, advanced control algorithms, and precise mechanical design, the dexterity and freedom of the hand have been effectively improved. It can not only complete repetitive tasks that traditional industrial robotic arms can perform, but also cope with more complex and ever-changing operating environments.

[0043] However, due to fundamental differences between the human hand and the dexterous hand in terms of kinematic structure, degree of freedom configuration, and joints, continuous optimization of the dexterous hand's parameters is necessary to achieve similar functionality to the human hand and improve the dexterous hand's performance in executing actions. Currently, this is often achieved by issuing control commands to the dexterous hand and collecting the parameters output by the dexterous hand when executing the corresponding actions, and then optimizing the dexterous hand's parameters based on these parameters.

[0044] A fundamental challenge in optimizing parameters in a dexterous hand is proving that the data collected by the dexterous hand can be remapped into corresponding control commands, and that the hand can perform its designated actions based on these mapped commands. This data-to-control-command mapping, whereby the control commands instruct the dexterous hand to perform a specified action, involves two deeply coupled variables: 1. Quality issues of remapping algorithms Because of the fundamental differences between humans and various types of dexterous hands in terms of kinematic structure, degree of freedom configuration, and joint constraints, it is quite difficult to construct a perfect mapping function.

[0045] 2. Data acquisition accuracy issues with dexterous hands Even if a dexterous hand has an ideal mapping function, factors such as the resolution, sampling frequency, and signal stability of the sensor in the dexterous hand may cause problems such as insufficient accuracy, excessive noise, or systematic bias in the data collected by the dexterous hand, making it unable to effectively complete the specified action.

[0046] When a dexterous hand fails to perform the action specified in the task, the two problems mentioned above become intertwined, forming a typical ill-posed problem, namely, it is impossible to accurately determine whether the failure is due to insufficient data quality or an imperfect mapping function.

[0047] In exemplary techniques, benchmarking, trajectory tracking, or statistical analysis methods are used to verify the effectiveness of the mapping function, and the parameters of the dexterous hand are optimized based on the verification results. However, none of the above methods can effectively resolve the coupling between the two problems mentioned above. More importantly, the exemplary techniques cannot verify a key issue: whether there exists a possible mapping method that can successfully drive the dexterous hand to complete the task using the collected data.

[0048] To address this problem, the inventors of this application conceived of a method that uses a data glove operated by an operator to assist the dexterous hand in completing a designated action. This method can be used for dexterous hand parameter optimization. Its core lies in: collecting assistance parameters through the data glove as the operator assists the dexterous hand in completing the designated action; using these assistance parameters to represent human perception and movement, compensating for the dexterous hand's shortcomings; and optimizing the dexterous hand's parameters through this compensation. Specifically, the operator acts as an adaptive mapping function for the dexterous hand; the operator's supplementary behavior in assisting the dexterous hand in completing the designated action externalizes the dexterous hand's limitations; and the dexterous hand completing the designated action with the assistance of these parameters demonstrates the usability of these parameters within the dexterous hand.

[0049] The following detailed description of the parameter optimization parameters for the dexterous hand provided in this application is provided through various embodiments.

[0050] Please see Figure 1 This is one of the flowcharts illustrating the parameter optimization method for a dexterous hand provided in the embodiments of this application. For example... Figure 1 As shown, the parameter optimization method for this dexterous hand includes: Step S101: Obtain the working parameters of the dexterous hand when performing the target specified action, and collect the assistance parameters of the data glove. The assistance parameters are used to instruct the operator to operate the data glove to assist the dexterous hand in performing the target specified action.

[0051] In this embodiment, when it is necessary to optimize the parameters of the dexterous hand, the operator can wear a data glove to assist the dexterous hand in completing the target-specified action based on the operator's control of the data glove.

[0052] Operators play multiple roles in assisting dexterous hands to complete the target-specific actions: As a sensor: the operator's vision system provides high-bandwidth, low-latency observation of the dexterity hand's state, a capability that far surpasses existing automatic state estimation algorithms for the dexterity hand; As a controller: the operator's motion control system can generate appropriate control strategies based on task requirements and the current state of the dexterous hand. The control strategies include complex prediction, planning and execution capabilities. As a learning system, locking the operator's assistance to the dexterous hand allows it to gradually learn and adapt to the characteristics of the dexterous hand, enabling the operator to find more effective control strategies for the dexterous hand.

[0053] The designated target movements can be multiple, specifically comprising three levels. The first level of designated target movements verifies the basic motor abilities of the dexterous hand, including, but not limited to: single-finger flexion and extension, finger-to-finger opposition, and wrist rotation. The second level of designated target movements verifies the coordinated motor abilities of the dexterous hand, including, but not limited to: multi-finger coordinated grasping, fine manipulation, and intramanual manipulation. The third level of designated target movements verifies the task completion abilities of the dexterous hand, including, but not limited to: object grasping and placement, tool use, and ambidextrous coordination.

[0054] When a dexterous hand performs a designated action, it collects its own operational parameters. Meanwhile, an operator assists the dexterous hand in performing the same action using a data glove. The data glove collects assistance parameters and transmits these parameters back to the dexterous hand. These assistance parameters refer to the parameters of the data glove when the operator manipulates it to assist the dexterous hand in performing the designated action.

[0055] Step S102: Determine the motion compensation amount when the operator assists the dexterous hand in performing the target-specified action based on the assistance parameters.

[0056] Step S103: Optimize the parameters of the dexterous hand based on the motion compensation amount and working parameters.

[0057] After obtaining the assistance parameters, the motion compensation amount is determined based on these parameters when the operator assists the dexterous hand in performing the target-specified action. This motion compensation amount, along with the working parameters, optimizes the dexterous hand's parameters. For example, there is a mapping relationship between the assistance parameters and the motion compensation amount. This mapping relationship, along with the assistance parameters, allows the motion compensation amount to be determined. The dexterous hand can optimize its working parameters using the motion compensation amount, thus achieving parameter optimization. For instance, if the current working parameters indicate that the dexterous hand's finger bend is 50°, and the parameter controlling this bend is parameter A, inputting the working parameters into the mapping function between parameter A and finger bend yields a finger bend of 50°. After operator assistance, the finger bend becomes 70°. The motion compensation amount compensates for the bend by 20°, therefore the mapping function needs correction. The correction is: inputting the working parameters into the mapping function between parameter A and finger bend yields a finger bend of 70°.

[0058] In this embodiment, from an information perspective, the operator's assistance can be understood as injecting additional information into the dexterous hand. The information transmission of the dexterous hand can be represented as follows: I(G;R)=H(R) H(R∣G); Where I(H;R|G) represents the additional information provided by the operator input H to the dexterity state R given the data of the dexterity hand.

[0059] Specifically, H(R) represents the entropy of the dexterous hand's state, which means: the total uncertainty of all possible states of the dexterous hand. For example, if the dexterous hand has 1000 possible states, and these states are uniformly distributed, then H(R) ≈ log2(1000) ≈ 10 bits.

[0060] H(R|G) refers to the conditional entropy given the data of a dexterous hand. H(R|G) means that even if the data of the dexterous hand is known, there are still uncertainties in the state of the dexterous hand. For example, the dexterous hand shows "clenched fist", but the dexterous hand may perform "light grip" or "tight grip". This ambiguity is characterized by H(R|G).

[0061] H(R|G,H) refers to the remaining uncertainties given the data of the dexterous hand and H(R|G). For example, even with the dexterous hand's "clench your fist" + the voice "be gentle," the dexterous hand may still have execution errors.

[0062] I(G;R)=H(R) The physical meaning of H(R|G) is how much uncertainty of the dexterous hand can be eliminated by the given data of the dexterous hand. Examples: - H(R) = 10 bits (uncertainty of the dexterous hand without given data) - H(R|G) = 6 bits (uncertainty after the dexterous hand has given data) - I(G;R) = 4 bits (the dexterous hand provides 4 bits of information).

[0063] The meaning of I(H;R|G) = H(R|G) - H(R|G,H): Based on the existing data of the dexterous hand, how much additional information is provided by the operator's input. Examples: - H(R|G) = 6 bits (only the uncertainty of the dexterous hand's data) - H(R|G,H) = 1 bit (adding the uncertainty after the operator's input) - I(H;R|G) = 5 bits (the operator adds 5 bits of information).

[0064] The following explanation uses scenarios to illustrate this: The grasping task of a dexterous hand: Initial state: The robot does not know what to grab or how to grab it, so H(R) is very large; Input from a dexterous hand: The operator made a grasping gesture → This reduced some uncertainty, but it is still not precise enough; Operator compensation: Operators say "be gentle" or use eye contact to indicate specific locations to further reduce uncertainty; Final state: The dexterous hand accurately executes the specified action, and H(R|G,H) is very small.

[0065] The above scenario illustrates why operator compensation for the dexterous hand is necessary, specifically when the data I(G;R) given to the dexterous hand is insufficient to fully complete the task.

[0066] In this embodiment, the working parameters of the dexterous hand performing the target specified action are obtained, and the assistance parameters of the data glove are collected. Based on the assistance parameters, the motion compensation amount when the operator assists the dexterous hand in performing the target specified action is determined. Therefore, the parameters of the dexterous hand are optimized based on the motion compensation amount and the working parameters. In this embodiment, when the dexterous hand performs the specified action, the data glove assists the dexterous hand in completing the specified action. The assistance parameters of the data glove accurately determine the motion compensation amount when the operator assists the dexterous hand in performing the specified action. This motion compensation amount then precisely guides the dexterous hand in parameter optimization, improving the efficiency of parameter optimization.

[0067] See Figure 2 , Figure 2 This is the second flowchart illustrating the parameter optimization method for a dexterous hand provided in this application embodiment, based on... Figure 1 In the embodiment shown, step S103 includes: Step S201: Determine the degree of completion of the dexterous hand in performing the specified action of the target, and compare the degree of completion with the completion threshold.

[0068] In this embodiment, the completion rate of the dexterous hand in performing the target-specified action is recorded. For example, after the dexterous hand completes the target-specified action, the current state of the dexterous hand is compared with the expected state to obtain the completion rate. For instance, if the dexterous hand needs to bend its fingers to reach 90° after performing the target-specified action, but the actual finger bend is only 85°, then the completion rate = (85 / 90) × 100% = 94.44%.

[0069] After obtaining the execution completion rate, compare it with the completion rate threshold. The completion rate threshold can be any suitable value, for example, a completion rate threshold of 95%.

[0070] Step S202: When the completion rate is greater than or equal to the completion rate threshold, optimize the parameters of the dexterous hand based on the comparison results between the work parameters, the motion compensation amount and the compensation amount threshold.

[0071] When the completion rate is greater than or equal to the completion rate threshold, the parameters of the dexterous hand are optimized based on the comparison between the work parameters, the motion compensation amount and the compensation amount threshold.

[0072] In one example, when the comparison result indicates that the motion compensation amount is greater than or equal to the compensation threshold, it can be determined that the operator's compensation scenario for the dexterous hand is high-compensation, high-success, indicating that a large amount of complex compensation is required for the dexterous hand to complete the task. In this case, it indicates that the parameters of the dexterous hand contain necessary information but have systematic problems, requiring a more complex mapping function. The mapping function of the dexterous hand can be improved through the operator's strategy. To this end, the current training sample is generated by the motion compensation amount, the target specified motion, and the working parameters. Then, multiple target training samples are obtained, and the preset model is trained based on each target training sample to obtain the target mapping relationship between the operation instructions required for the dexterous hand to perform the motion and the parameters output by the dexterous hand when performing the operation instructions. The target training samples include the current training samples and historical training samples. The historical training samples are determined based on the parameters collected by the operator's operation data glove assisting the dexterous hand in performing other specified operations. That is, the historical training samples include historical motion compensation amounts, historical assistance parameters, historical working parameters, and other specified motions. After obtaining the target mapping relationship, the dexterous hand replaces the current mapping relationship with the target mapping relationship. Understandably, by extracting the motion compensation amount obtained from the operator's assistance with the dexterous hand using various target training samples, this compensation is used as a supervisory signal to train the neural network in the pre-set model to mimic the operator's assistance behavior with the dexterous hand. The pre-set model can be a deep learning model.

[0073] In another example, when the comparison result indicates that the motion compensation amount is less than the compensation threshold, it can be determined that the operator's compensation scenario for the dexterous hand is low compensation with high success, indicating that only minor, smooth adjustments are needed for the dexterous hand to complete the task. Therefore, the dexterous hand's parameter quality is good, a simple and effective mapping algorithm exists, and the dexterous hand is close to practical application. To address this, based on the working parameters and the motion compensation amount, the expected parameters output by the dexterous hand when completing the target specified action are determined, and the current mapping relationship of the dexterous hand is optimized based on these expected parameters. For example, if the dexterous hand, with the operator's assistance, has an execution completion rate less than 100%, then the motion compensation amount needs to be appropriately increased to obtain the target compensation amount. Then, using the target compensation amount and the working parameters, the parameters output by the dexterous hand when 100% of the target specified action are simulated as the expected parameters. The mapping relationship can be the relationship between the working parameters and the control command. Therefore, the error between the expected parameters and the working parameters is determined. The mapping relationship before adjustment is: working parameters mapped to control command A; the mapping relationship after adjustment is: working parameters + error mapped to control command A.

[0074] Furthermore, if the completion rate is less than the completion rate threshold, the dexterous hand will not complete the target specified action. In this case, it indicates that even if the operator tries to compensate, the dexterous hand cannot complete the task. Therefore, it is determined that there is a fundamental limitation in the hardware capability of the dexterous hand, and some of the specified actions are beyond the capability range of the dexterous hand. In this case, it is necessary to redesign the dexterous hand sensor layout or improve the accuracy of the sensors.

[0075] In this embodiment, when the dexterous hand's completion rate of the target specified action is greater than or equal to the completion rate threshold, the parameters of the dexterous hand are adaptively optimized by comparing the working parameters, the action compensation amount, and the compensation amount threshold, thereby improving the intelligence level of the dexterous hand.

[0076] See Figure 3 , Figure 3 This is the third flowchart illustrating the parameter optimization method for a dexterous hand provided in this application embodiment, based on... Figure 1 In the embodiment shown, step S103 includes: Step S301: Identify the parameters to be compensated for the dexterous hand based on the motion compensation amount.

[0077] In this embodiment, in order to comprehensively evaluate the performance of the dexterous hand in completing the target-specified action, it is necessary to establish a multi-level indicator system. These indicators not only reflect the final task completion status, but also quantify the degree and characteristics of the operator's compensation to the dexterous hand.

[0078] Specifically, the primary indicator in the indicator system is the execution completion rate, which is determined based on success rate, completion time, and accuracy indicators.

[0079] Success rate: Ps = N success / N total ; Completion Time: T avg = (1 / N) ; Accuracy index: E avg = (1 / N) (1 / N) .

[0080] Where Ps is the task success rate, representing the proportion of successfully completed tasks out of the total number of tasks (dimensionless, value range [0,1]); N success Number of tasks successfully completed (times); N total Total number of tasks executed (times); T avg The average task completion time (seconds); N is the number of statistical samples (times). E represents the completion time (in seconds) of the i-th task; avg The average position accuracy error represents the average deviation (in meters or millimeters) between the actual position and the target position. It is the actual end position (vector, usually 3D spatial coordinates) in the i-th task. It is the target end position (vector, usually 3D spatial coordinates) in the i-th task; the || operator represents absolute value or vector norm operation.

[0081] If the completion time is less than the time threshold and the accuracy index is higher than the accuracy threshold, then the success can be determined, i.e., N. success Increase by 1.

[0082] The secondary indicators are compensation characteristics. Specifically, the compensation indicators include time-domain analysis indicators of the operator's compensation amount:

[0083] Frequency domain analysis metrics:

[0084] In the frequency domain, the low-frequency components reflect systematic deviations in dexterity, while the high-frequency components reflect noise and instability.

[0085] This refers to the operator's time-domain compensation amount, representing the average intensity of the operator's compensation action per unit time; T is the length of the observation time window (seconds). It refers to the operator's compensation vector for the dexterous hand at time t, representing the additional control amount applied by the operator to correct the dexterous hand's movements (the dimension is the same as the dexterous hand's degrees of freedom, usually joint angles or end-effector positions). This represents the definite integral over the time interval [0, T]. This represents the frequency domain compensation energy, and the total energy of the compensation signal across the entire frequency spectrum. represents the Nyquist angular frequency, which is equal to half the sampling frequency, and represents the highest analyzable frequency (radians / second); F[] represents the Fourier transform operator, which converts the time-domain signal into a frequency-domain representation; W represents the angular frequency variable (radians / second). This represents the power spectral density of the compensation signal at angular frequency ω; the low-frequency component represents the portion of the frequency close to 0, reflecting systematic deviations and slowly changing errors; the high-frequency component represents the portion of the frequency close to the Nyquist frequency, reflecting noise, oscillations, and system instability. The third-level indicator is cognitive load, which is defined as the degree of naturalness of movement based on the optimal deviation of the operator's motor control:

[0086] Cognitive complexity is represented by information entropy, specifically as follows:

[0087] in, It represents the smoothness index of motion, based on the principle of minimum jerk (the smaller the value, the smoother and more natural the motion); This represents jerk, which is the third derivative of acceleration with respect to time, and represents the rate of change of acceleration (m / s³ or joint angle / s³). This represents the square norm of the jerk vector; "-" makes the index positively correlated with smoothness (the larger the value, the smoother the surface). Representing cognitive complexity, it measures the complexity (in bits) of human compensatory actions based on Shannon information entropy. This represents the i-th type of compensation action (e.g., discrete action categories such as position adjustment, posture correction, force change, etc.). represents the probability of the i-th compensation action occurring; log represents the logarithmic function, usually base 2 (information entropy standard) or the natural logarithm.

[0088] The dexterous hand identifies multiple compensation modes using the aforementioned indicators characterizing the amount of motion compensation, thereby determining the parameters to be compensated based on these modes. The compensation modes include constant bias compensation, linear scaling compensation, nonlinear compensation, and time-varying compensation. For constant bias compensation, it indicates a systematic calibration error, requiring recalibration of the sensor's zero point in the dexterous hand; therefore, the parameters to be compensated are the sensor parameters. For linear scaling compensation, it indicates a gain mismatch in the dexterous hand, requiring adjustment of the scaling factor of the factor function; the parameters to be compensated are the scaling factor of the mapping function. For nonlinear compensation, it indicates a complex mapping problem requiring a more complex mapping relationship; the parameters to be compensated are the mapping relationship. For time-varying compensation, it indicates drift or instability in the dexterous hand system, requiring online modification of the adaptive algorithm; therefore, the parameters to be compensated are the parameters corresponding to the adaptive algorithm.

[0089] Step S302: Correct the parameters to be compensated based on the working parameters.

[0090] Once the parameters to be compensated are determined, the dexterous hand can adjust them based on the operating parameters. For example, a correction value is determined based on the operating parameters, and then the parameters to be compensated are adjusted based on the correction value.

[0091] In this embodiment, the dexterous hand identifies the parameters to be compensated based on the motion compensation amount, and then accurately compensates the parameters to be compensated based on the working parameters.

[0092] See Figure 4 , Figure 4 This is the fourth flowchart illustrating the parameter optimization method for a dexterous hand provided in this application embodiment, based on... Figure 3 In the embodiment shown, step S301 includes: Step S401: Based on each sub-compensation quantity in the motion compensation quantity, determine multiple characteristic values. Each sub-compensation quantity is the compensation quantity corresponding to each moment within the time period during which the operator assists the dexterous hand. The multiple characteristic values ​​include the mean, standard deviation, coefficient of variation, and maximum value of each sub-compensation quantity.

[0093] In this embodiment, the parameters to be compensated can be identified using amplitude feature analysis. For example, the motion compensation amount includes compensation amounts at different times, and each compensation amount at a different time is defined as a sub-compensation amount. The Euclidean norm of each sub-compensation amount is calculated to obtain the time series of the compensation amplitude for each sub-compensation amount. Then, multiple statistics are extracted based on each compensation amplitude time series. These statistics include, but are not limited to, the mean, standard deviation, coefficient of variation, and the maximum value among the sub-compensation amounts. Each statistical measure is defined as an eigenvalue.

[0094] Step S402: Identify the parameters to be compensated for by the dexterous hand based on each feature value.

[0095] After obtaining multiple feature values, the parameters to be compensated for by the dexterous hand can be determined based on each feature value.

[0096] In one example, when the mean is greater than the first threshold and the standard deviation is less than the second threshold, it can be determined that the parameters of the dexterous hand are continuously and significantly off due to mapping error. Therefore, the mapping relationship is used as the parameter to be compensated.

[0097] In another example, when the mean is less than or equal to the first threshold and the standard deviation is greater than or equal to the second threshold, it is determined that the system stability of the dexterous hand is poor. Therefore, it is necessary to address the tremor of the dexterous hand, that is, the parameter to be compensated is the parameter to adjust the tremor.

[0098] In another example, when the mean is less than or equal to the first threshold and the standard is less than the second threshold, only fine-tuning of the dexterous hand is needed, and the default parameter that needs to be adjusted is used as the parameter to be compensated.

[0099] In this embodiment, amplitude analysis is used to accurately identify the parameters to be compensated for by the dexterous hand.

[0100] See Figure 5 , Figure 5 This is the fifth flowchart illustrating the parameter optimization method for a dexterous hand provided in this application embodiment, based on... Figure 3 In the embodiment shown, step S301 includes: Step S501: Determine the activation intensity of each joint component of the dexterous hand based on the motion compensation amount.

[0101] Step S502: Identify the parameters to be compensated for by the dexterous hand based on the activation intensity of each joint component.

[0102] In this embodiment, the parameters to be compensated for in a dexterous hand can be identified through the joint components of the dexterous hand. Specifically, the activation intensity of each joint component of the dexterous hand is determined based on the motion compensation amount, and the parameters to be compensated for in the dexterous hand are identified through the activation intensity of each joint component.

[0103] For example, the root mean square (RMS) values ​​of the joint components of a dexterous hand are calculated, where the RMS value characterizes the activation intensity of key components. Among the various joint components, those with activation intensities greater than a threshold are extracted as active joint components, with the threshold being, for example, 50%. By identifying the active joint components, the parameters to be compensated for in the dexterous hand can be determined.

[0104] In one example, the active joint components are the thumb and index finger, and fine-grasping adjustments are required for the dexterous hand. Therefore, the parameters to be compensated are the dexterous hand's grasping parameters.

[0105] In another example, the active joint is the wrist joint, so dexterity hand posture compensation is required, and the parameters to be compensated are the parameters that control the movement of the dexterity hand.

[0106] In another example, the active joint components are all the fingers, so the grip strength of the dexterous hand needs to be adjusted, that is, the parameter to be compensated is the parameter that controls the grip strength of the dexterous hand.

[0107] In another example, if the active joint component is a single finger, then independent finger control compensation is required, and the parameter to be compensated is the control parameter of that finger.

[0108] In this embodiment, the activation intensity of the joint components of the dexterous hand can be determined through simple calculations. The calculation complexity of the dexterous hand is low, and the parameters to be compensated can be determined quickly.

[0109] See Figure 6 , Figure 6 This is the sixth flowchart illustrating the parameter optimization method for a dexterous hand provided in this application embodiment, based on... Figure 3 In the embodiment shown, step S301 includes: Step S601: Determine the contribution of each joint component in the dexterous hand to the execution of the target specified action based on the motion compensation amount.

[0110] In this embodiment, principal component analysis can be used to identify the parameters to be compensated for in a dexterous hand. The dexterous hand first determines the contribution of each joint component in performing the target action through motion compensation. For example, the activation intensity of each joint component can be determined; the greater the activation intensity, the greater the contribution of that joint component.

[0111] Step S602: Based on the contribution, determine the target joint component among all joint components, where the contribution of the target joint component is greater than a preset threshold.

[0112] After determining the contribution of each joint component, a target joint component is identified among them, whose contribution is greater than a preset threshold. For example, a joint component with a contribution of 80% or higher is designated as the target joint component. Step S603: Determine the motion pattern to be adjusted for the dexterous hand based on the target joint component.

[0113] After identifying the target joint component, obtain the collaboration weights of other joint components to the target joint component. For example, if the target joint component is the thumb, the index finger has the most collaboration with the thumb, and the little finger has the least collaboration with the thumb. Set collaboration weights based on the amount of collaboration of other joint components to the target key component. The greater the amount of collaboration, the greater the collaboration weight. Each collaboration weight can form an n-dimensional principal component vector, where n refers to the number of other joint components.

[0114] After determining the principal component vectors, eigenvalues ​​are calculated for the principal component vectors, and the principal component vectors are sorted from largest to smallest according to the eigenvalues. The target joint components corresponding to the top N principal component vectors are used as the reference joint components to determine the motion mode.

[0115] In one example, when the reference joint components are the distal joints of the thumb and the distal joints of the index finger, the motion model is a fine grasping mode.

[0116] In another example, when the reference joint components are the proximal joints of all fingers, the motion pattern is a force-grabbing pattern.

[0117] In another example, if the reference joint components are wrist flexion / extension, radioulnar deviation, and pronation / supination joints, then the movement pattern is the wrist posture pattern.

[0118] In another example, if each reference joint component contains a large number of proximal joints, the motion mode is a proximal adjustment mode; if each reference joint component contains both distal and proximal joints, the motion mode is a distal adjustment mode.

[0119] In other examples, if the reference joint component is the joint of a finger, the movement mode is a single-finger independent mode.

[0120] In another example, if the reference joint component is the opposing gripping posture of the thumb and other fingers, then the movement pattern is the opposing movement pattern.

[0121] In other examples, the ratio of the number of reference joint components to the total number of joint components can also be calculated. If the ratio is less than a first ratio threshold, the motion mode is a sparse mode, and the first ratio threshold is, for example, 30%. If the ratio is greater than a second ratio threshold, the motion mode is a dispersed mode, and the second ratio threshold is greater than the first ratio threshold, for example, 60%.

[0122] In other examples, the number of joints with positive weights and the number of joints with negative weights are calculated. If all joints are positive weights or all joints are negative weights, the motion mode is unidirectional; if it includes both positive and negative weight joints, the motion mode is bidirectional.

[0123] Step S604: Determine the parameters to be compensated for for the dexterous hand based on the motion pattern.

[0124] After determining the motion pattern, the parameters to be compensated are determined based on the motion pattern. For example, if the movement mode is fine grasping mode, the parameter to be compensated is the grasping parameter of the dexterous hand; if the movement mode is force grasping mode, the parameter to be compensated is the parameter for adjusting grip strength in the dexterous hand; if the movement mode is wrist posture mode, the parameter to be compensated is the parameter for adjusting wrist strength in the dexterous hand; if the movement mode is proximal adjustment mode, the parameter to be compensated is the parameter for adjusting finger direction; if the movement mode is distal adjustment mode, the parameter to be compensated is the parameter for adjusting finger posture; if the movement mode is single-finger independent mode, the parameter to be compensated is the parameter for controlling the movement of that finger; if the movement mode is opposing movement mode, the parameter to be compensated is the parameter for controlling the relative movement of the fingers; if the movement mode is sparse mode, the parameter to be compensated is the fine adjustment of the dexterous hand, that is, fine-tuning the parameters of the dexterous hand; if the movement mode is dispersed mode, the parameter to be compensated is the overall coordination compensation parameter; if the operation mode is unidirectional mode, the parameter to be compensated is the parameter for the dexterous hand to perform unidirectional movement; if the operation mode is bidirectional mode, the parameter to be compensated is the adversarial coordination parameter or posture parameter of the dexterous hand.

[0125] In this embodiment, the movement pattern to be adjusted for the dexterous hand is determined by the contribution of the joint components of the dexterous hand, thereby accurately determining the compensation parameters of the dexterous hand based on the movement pattern.

[0126] See Figure 7 , Figure 7 This is the seventh flowchart illustrating the parameter optimization method for a dexterous hand provided in this application embodiment, based on... Figure 3 In the embodiment shown, step S301 includes: Step S701: Decompose the motion compensation amount into sub-compensation amounts for each joint component of the dexterous hand by the operator.

[0127] In this embodiment, the motion compensation amount is decomposed into sub-compensation amounts for each joint component of the dexterous hand by the operator.

[0128] Step S702: Extract target features from the sub-compensation quantities of the joint components. The target features are time-domain features or frequency-domain features.

[0129] The dexterous hand extracts target features from the sub-compensation quantities of joint components, such as time-domain features or frequency-domain features.

[0130] Step S703: Identify the parameters to be compensated for by the dexterous hand based on the target features of each joint component.

[0131] After determining the target features of the joint components, the parameters to be compensated for by the dexterous hand are identified based on each target feature.

[0132] In one example, the target feature is a frequency domain feature. Power spectral density analysis is performed on the frequency domain feature to obtain the power spectrum estimation parameters of the joint components. The dominant frequency components and their power are identified through the power spectrum estimation parameters, and then the periodic intensity is calculated, which is the ratio between the peak power and the average power. If the active frequency is between 4 and 12 Hz and the periodic intensity is greater than 5, it can be determined that the dexterity hand has physiological tremor, and the parameters to be compensated are the parameters for adjusting physiological tremor. If the dominant frequency is between 0.1 and 1.1 Hz, it is low-frequency drift, caused by slow mismatch of mapping parameters and cumulative error in attitude estimation. This can be addressed by introducing an adaptive correction mechanism, i.e., the parameters to be compensated are the parameters corresponding to the adaptive correction mechanism. If the dominant frequency is greater than 12 Hz, it is high-frequency oscillation, caused by instability in the control system and sensor noise. This can be addressed by adding a low-pass filter and checking the control parameters, i.e., the parameters to be compensated are the control parameters.

[0133] In another example, when the target feature is a time-domain feature, a rate of change analysis is performed on the time-domain feature, and the rate of change is quantified by first-order difference (to obtain velocity) and second-order difference (to obtain acceleration). The average velocity is obtained through the first-order difference of each joint component, and the average acceleration is obtained through each second-order difference. The jerkiness is also obtained, which is the rate of change of acceleration and reflects the smoothness of motion. The average velocity reflects the speed of compensation adjustment, while the average acceleration reflects the urgency of the adjustment. The parameters to be compensated are determined using the average velocity, average acceleration, and jerkiness. For example, abrupt change points can be identified using these three parameters, and the cause of the abrupt change is determined based on the associated task events of the abrupt change points. The parameter causing the abrupt change is the parameter to be compensated. An abrupt change point refers to the time point at which the sub-compensation quantity of a certain joint component jumps.

[0134] In this embodiment, the operator's assistance to the dexterous hand has significant characteristics in the time domain or frequency domain. Therefore, the parameters to be compensated can be accurately determined through analysis in the time domain and frequency domain.

[0135] In one embodiment, the parameters to be compensated can be determined through unsupervised clustering. The principle is that without pre-setting the pattern type, the algorithm automatically discovers the cluster centers of typical compensation patterns from the data, which is suitable for exploratory analysis and building a pattern library.

[0136] First, the time series of motion compensation quantities is standardized, such as to zero mean and unit variance. A sliding window is used to extract time segments, with each segment representing a sample. A cluster number k is set, and the standardized compensation quantity samples are clustered to obtain k cluster centers, each representing a typical compensation pattern. For each cluster center, principal component analysis is applied to obtain multiple motion patterns, i.e., the aforementioned principal component analysis determines the motion patterns. The number of samples at each cluster center is counted to determine the degree of separation and distribution characteristics of the clusters, thereby determining the parameters to be compensated based on the degree of separation and distribution characteristics.

[0137] In this embodiment, there is no need to predefine pattern types, it is highly adaptable, automatically builds a pattern library for specific tasks or operators, and can identify rare or abnormal compensation behaviors.

[0138] In one embodiment, the compensation pattern can be determined through task context semantic interpretation. The principle is that the same compensation pattern may represent different meanings at different task stages. By fusing task context information (task stage, contact state, object attributes, etc.), the identified pattern is semantically interpreted, and improvement suggestions are provided.

[0139] Specifically, before the dexterous hand performs the target-specified action, contextual information about the dexterous hand is first collected. Simultaneously with the acquisition of compensation data, the following is recorded: Task phase tags: Approach, Prepare to grab, Grab, Lift, Transport, Place, Release; Contact status: Whether it is in contact with an object (detected by force sensor or vision). Object properties: type (cylinder, sphere, cube, etc.), size, weight (if known); Timestamp: Used for alignment.

[0140] Establish a rule base to describe the meaning of patterns in different contexts. The rule base is as follows:

[0141] For each moment or time window: extract the task context of the current moment; call the aforementioned layers (second or fourth layer) to identify the current compensation mode type; query the rule base to match [mode type + task stage + contact state].

[0142] Output: Semantic description, i.e., what behavior the pattern represents; possible causes, i.e. why the compensation occurs; improvement suggestions, how to optimize the mapping or control.

[0143] Pattern evolution tracking: Tracking changes in the same pattern at different task stages; identifying pattern switching caused by task transition points (such as approach → grasp); analyzing the smoothness (abrupt or gradual) of pattern transitions.

[0144] Based on the above description, the various stages of this embodiment will be briefly explained as follows: Phase 1: Rapid screening: Real-time monitoring of the statistical characteristics of compensation amounts to quickly identify anomalies (such as sudden large-scale compensation); determine whether further analysis is needed.

[0145] Phase 2: Core Identification: Perform PCA analysis on the collected data within a time window (e.g., 30 seconds to 2 minutes) to identify the main collaborative compensation patterns and generate pattern descriptions and visualizations.

[0146] Phase 3: Specific Analysis: If high variance or periodicity is found, perform spectral analysis to detect dynamic problems such as tremors and drift.

[0147] Phase 4: Pattern Library Establishment: After collecting a large amount of data, perform cluster analysis; establish a pattern library for specific tasks or operators for subsequent rapid pattern matching.

[0148] Phase 5: Semantic Interpretation: Interpret each identified pattern in the context of the task; generate a readable analysis report; and propose specific improvement suggestions.

[0149] In this embodiment, the technical identification results are transformed into actionable improvement suggestions, and it can distinguish between "normal human skills" and "forced compensation caused by system defects," supporting personalized mapping strategy optimization.

[0150] See Figure 8 , Figure 8 This is the eighth flowchart illustrating the parameter optimization method for a dexterous hand provided in this application embodiment, based on... Figure 1 or Figure 2 In the embodiment shown, step S103 includes: Step S801: Based on the preset compensation function, motion compensation amount, and working parameters, determine the first operation instruction for the operator to assist the dexterous hand in performing the target-specified motion.

[0151] Step S802: Based on the first operation instruction and the second operation instruction output by the dexterous hand when performing the target-specified action, the current mapping relationship of the dexterous hand is corrected. The current mapping relationship is the mapping relationship between the operation instruction output by the dexterous hand and the parameters output by the dexterous hand when performing the operation instruction.

[0152] In this embodiment, the operator, data glove, and dexterous hand constitute a closed-loop system, which can be mathematically represented by the following formula: r(t)=fmapping(g(t))+Ψ(e(t),e˙(t),∫e(τ)dτ); Where g(t) is the assist parameter characterizing the data glove set; fmapping is the basic mapping function; Ψ is the operator's motion compensation amount, which depends on the error e(t) and its derivative and integral; and r(t) is the final control command of the dexterous hand.

[0153] As algorithms and accuracy improve, mapping quality continuously increases, reducing the need for operator compensation by the dexterous hand, ultimately achieving automated control. r(t)=fmapping(g(t))↑+Ψ(e(t),e˙(t),∫e(τ)dτ)↓.

[0154] Understandably, based on the preset compensation function (fmapping(g(t))), the motion compensation amount Ψ, and the e(t) determined by the working parameters, the first operation instruction for the operator to assist the dexterous hand in performing the target-specified action is determined. The first operation instruction is: r(t)=fmapping(g(t))+Ψ(e(t),e˙(t),∫e(τ)dτ), which means the first operation instruction is the operation instruction before optimization.

[0155] Based on the first operation instruction and the second operation instruction output by the dexterous hand when performing the target-specified action, the current mapping relationship of the dexterous hand is corrected. The current mapping relationship is the mapping relationship between the operation instructions output by the dexterous hand and the parameters output by the dexterous hand when performing the operation instructions.

[0156] The second operation instruction is: r(t)=fmapping(g(t))↑+Ψ(e(t),e˙(t),∫e(τ)dτ)↓.

[0157] In this embodiment, the operation instructions are accurately determined by the compensation function, the motion compensation amount, and the working parameters, so as to correct the current mapping relationship of the dexterous hand.

[0158] See Figure 9 , Figure 9 This is the ninth flowchart illustrating the parameter optimization method for a dexterous hand provided in this application embodiment, based on... Figure 1 or Figure 2 In the embodiment shown, step S103 includes: Step S901: Based on the motion compensation amount, determine the compensation behavior of the operator assisting the dexterous hand in performing the target-specified motion.

[0159] Step S902: Determine the parameters to be corrected for the dexterous hand based on the compensation behavior, and optimize the parameters to be corrected for the dexterous hand based on the motion compensation amount and working parameters.

[0160] In this embodiment, the dexterous hand can determine the compensation behavior of the operator assisting the dexterous hand in performing the target specified action through the motion compensation amount. The dexterous hand's parameters to be corrected are determined through the compensation behavior, and then the parameters to be corrected of the dexterous hand are optimized based on the motion compensation amount and the working parameters.

[0161] For example, the dexterous hand identifies multiple compensation behaviors using the aforementioned indicators characterizing the amount of motion compensation, thereby determining the parameters to be corrected by the dexterous hand through these compensation behaviors. These compensation behaviors include constant bias compensation, linear scaling compensation, nonlinear compensation, and time-varying compensation. For constant bias compensation, it indicates a systematic calibration error, requiring recalibration of the sensor's zero point in the dexterous hand; therefore, the parameters to be corrected are the sensor parameters. For linear scaling compensation, it indicates a gain mismatch in the dexterous hand, requiring adjustment of the scaling factor of the factor function; the parameters to be corrected are the scaling factor of the mapping function. For nonlinear compensation, it indicates a complex mapping problem requiring a more complex mapping relationship; the parameters to be corrected are the mapping relationship. For time-varying compensation, it indicates that the dexterous hand's system exhibits drift or instability, requiring online modification of the adaptive algorithm; therefore, the parameters to be corrected are the parameters corresponding to the adaptive algorithm.

[0162] In this embodiment, by determining the compensation behavior, the parameters to be corrected are accurately determined, thereby optimizing the parameters to be corrected of the dexterous hand accurately based on the working parameters and the amount of motion compensation.

[0163] See Figure 10 , Figure 10 This is the tenth flowchart illustrating the parameter optimization method for a dexterous hand provided in this application embodiment, based on... Figure 1 or Figure 2 In the embodiment shown, step S103 includes: Step S1001: Determine the trigger factor for the dexterous hand to perform compensation based on the motion compensation amount and working parameters.

[0164] In this embodiment, the dexterous hand can determine the triggering factors for compensation by using the motion compensation amount and working parameters, that is, identify which factors directly trigger the compensation behavior and eliminate the interference of mixed factors.

[0165] Step S1002: Determine the correction strategy based on the triggering factor, and optimize the parameters of the dexterous hand based on the correction strategy.

[0166] In this embodiment, the dexterous hand can determine the triggering factors for compensation by using motion compensation amounts and working parameters. This means identifying which factors directly trigger the compensation behavior and eliminating interference from confounding factors. A correction strategy is then determined based on these triggering factors, allowing for the optimization of the dexterous hand's parameters.

[0167] Example 1: "Thumb-index finger fine grip compensation" is mainly triggered by "contact force error". This means that after the operator contacts the object, he or she quickly adjusts the posture of the thumb and index finger to optimize the grip based on the perceived force feedback error.

[0168] Example 2: "Wrist posture compensation" is triggered by multiple factors. The primary triggering factor is wrist mapping error, and the secondary triggering factor is operating speed. This indicates that wrist compensation is mainly used to correct systematic biases in the mapping algorithm, and that the accumulation of mapping errors during high-speed movement leads to an increased need for compensation.

[0169] Example 3: "Operating speed" indirectly affects "whole-hand grip strength compensation" through "accumulated mapping error". The direct path is: Operating speed ↑ → Mapping error ↑ → Grip strength compensation ↑. This indicates that rapid operation causes a tracking delay in the mapping algorithm, leading to inaccurate grip strength, forcing the operator to compensate. Recommendation: Add predictive feedforward control during high-speed operations.

[0170] Example 4: The triggering mechanism of "thumb compensation" depends on the task phase: In the "approach phase", it is dominated by "visual tracking error" and the operator pre-adjusts the thumb posture based on visual information. In the "grasping phase", it is dominated by "contact force error" and the operator finely adjusts the grip force based on tactile feedback. This reflects the strategy of human multimodal perception fusion: visual-dominated feedforward control + tactile-dominated feedback control.

[0171] In this embodiment, the triggering factor for the dexterous hand to perform compensation is determined by the motion compensation amount and working parameters. Then, the correction strategy is accurately determined based on the triggering factor, and the parameters of the dexterous hand are optimized based on the correction strategy.

[0172] Corresponding to the above-mentioned method for optimizing the parameters of a dexterous hand, this application also provides a device for optimizing the parameters of a dexterous hand. Figure 11 This is a schematic diagram of a parameter optimization device for a dexterous hand provided in an embodiment of this application. The parameter optimization device 1100 for a dexterous hand provided in this embodiment of the application includes: The acquisition module 1110 is used to acquire the working parameters of the dexterous hand when performing the target specified action, and to collect the assistance parameters of the data glove. The assistance parameters are used to instruct the operator to operate the data glove to assist the dexterous hand in performing the target specified action. The determination module 1120 is used to determine the motion compensation amount when the operator assists the dexterous hand in performing the target specified action, based on the assistance parameters; The optimization module 1130 is used to optimize the parameters of the dexterous hand based on the motion compensation amount and working parameters.

[0173] In some implementations, the parameter optimization device 1100 of the dexterous hand is also used for: Determine the degree of completion of the dexterous hand in performing the specified action on the target, and compare the degree of completion with the completion threshold; When the completion rate is greater than or equal to the completion rate threshold, the parameters of the dexterous hand are optimized based on the comparison results between the work parameters, the motion compensation amount and the compensation amount threshold.

[0174] In some implementations, the parameter optimization device 1100 of the dexterous hand is also used for: When the comparison result indicates that the action compensation amount is greater than or equal to the sub-compensation amount threshold, the current training sample is generated based on the action compensation amount, the target specified action, and the working parameters. The preset model is trained based on multiple target training samples to obtain the target mapping relationship between the operation instructions required for the dexterous hand to perform the action and the parameters output by the dexterous hand when performing the operation instructions. The multiple target training samples include the current training samples and historical training samples. The historical training samples are determined based on the parameters collected when the operator's operation data glove assists the dexterous hand in performing other specified actions. Replace the current mapping relationship in the dexterous hand with the target mapping relationship.

[0175] In some implementations, the parameter optimization device 1100 of the dexterous hand is also used for: When the comparison result indicates that the motion compensation amount is less than the sub-compensation amount threshold, the expected parameters output by the dexterous hand when it completes the target specified motion are determined based on the working parameters and the motion compensation amount. Based on the expected parameters, the current mapping relationship of the dexterous hand is optimized.

[0176] In some implementations, the parameter optimization device 1100 of the dexterous hand is also used for: Identify the parameters to be compensated for in a dexterous hand based on the amount of motion compensation; Based on the operating parameters, the parameters to be compensated are corrected.

[0177] In some implementations, the parameter optimization device 1100 of the dexterous hand is also used for: Based on each sub-compensation quantity in the motion compensation quantity, multiple characteristic values ​​are determined. Each sub-compensation quantity is the compensation quantity corresponding to each moment within the time period during which the operator assists the dexterous hand. The multiple characteristic values ​​include the mean, standard deviation, coefficient of variation, and maximum value of each sub-compensation quantity. Based on each feature value, identify the parameters to be compensated for in a dexterous hand.

[0178] In some implementations, the parameter optimization device 1100 of the dexterous hand is also used for: Based on the amount of motion compensation, determine the activation intensity of each joint component of the dexterous hand; Based on the activation intensity of each joint component, the parameters to be compensated for in the dexterous hand are identified.

[0179] In some implementations, the parameter optimization device 1100 of the dexterous hand is also used for: Based on the motion compensation amount, determine the contribution of each joint component in the dexterous hand to the execution of the target-specified motion; Based on the contribution, a target joint component is identified among all joint components, where the contribution of the target joint component is greater than a preset threshold. Based on the target joint components, determine the movement pattern to be adjusted for the dexterous hand; Based on the movement pattern, determine the parameters to be compensated for the dexterous hand.

[0180] In some implementations, the parameter optimization device 1100 of the dexterous hand is also used for: The motion compensation amount is broken down into sub-compensation amounts for each joint component of the dexterous hand by the operator. Extract target features from the sub-compensation quantities of joint components. The target features are either time-domain features or frequency-domain features. Based on the target features of each joint component, identify the parameters to be compensated for in the dexterous hand.

[0181] In some implementations, the parameter optimization device 1100 of the dexterous hand is also used for: Based on the motion compensation amount and working parameters, determine the triggering factor for the dexterous hand to perform compensation; The correction strategy is determined based on the triggering factor, and the parameters of the dexterous hand are optimized based on the correction strategy.

[0182] In some implementations, the parameter optimization device 1100 of the dexterous hand is also used for: Based on the preset compensation function, motion compensation amount, and working parameters, determine the first operation instruction for the operator to assist the dexterous hand in performing the target-specified motion; Based on the first operation instruction and the second operation instruction output by the dexterous hand when performing the target-specified action, the current mapping relationship of the dexterous hand is corrected. The current mapping relationship is the mapping relationship between the operation instructions output by the dexterous hand and the parameters output by the dexterous hand when performing the operation instructions.

[0183] In some implementations, the parameter optimization device 1100 of the dexterous hand is also used for: Based on the motion compensation amount, determine the compensation behavior of the operator in assisting the dexterous hand to perform the target-specified motion; The parameters to be corrected for the dexterous hand are determined based on the compensation behavior, and then optimized based on the amount of motion compensation and the working parameters.

[0184] The dexterity hand parameter optimization device and the dexterity hand parameter optimization method provided in the above embodiments of this application belong to the same application concept and can execute the dexterity hand parameter optimization method provided in any of the above embodiments of this application. They possess the corresponding functional modules and beneficial effects for executing the dexterity hand parameter optimization method. Technical details not described in detail in this embodiment can be found in the specific processing content of the dexterity hand parameter optimization method provided in the above embodiments of this application, and will not be repeated here. The functions implemented by each module in the dexterity hand parameter optimization device can be implemented by the same or different processors, and this application embodiment does not limit this.

[0185] It should be understood that the modules in the above-described dexterous hand parameter optimization device can be implemented in the form of processor calling firmware. For example, the system includes a processor connected to a memory containing instructions. The processor calls the instructions stored in the memory to implement any of the above methods or to implement the functions of each module of the device. The processor can be a general-purpose processor, such as a CPU or microprocessor, and the memory can be internal to the device or external to the system. Alternatively, the modules in the system can be implemented in the form of hardware circuits. By designing the hardware circuits, some or all of the module functions can be implemented. The hardware circuits can be understood as one or more processors. For example, in one implementation, the hardware circuit is an ASIC, and the functions of some or all of the above modules are implemented by designing the logical relationships of the components within the circuit. In another implementation, the hardware circuit can be implemented using a PLD, such as an FPGA, which can include a large number of logic gates. The connection relationships between the logic gates are configured through a configuration file, thereby implementing the functions of some or all of the above modules. All modules of the above-described dexterous hand parameter optimization device can be implemented entirely through processor calling firmware, entirely through hardware circuits, or partially through processor calling firmware with the remaining parts implemented through hardware circuits.

[0186] In this application embodiment, a processor is a circuit with signal processing capabilities. In one implementation, the processor can be a circuit with instruction reading and execution capabilities, such as a CPU, microprocessor, GPU, or DSP. In another implementation, the processor can implement certain functions through the logical relationships of hardware circuits. These logical relationships are fixed or reconfigurable. For example, the processor may be a hardware circuit implemented as an ASIC or PLD, such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document and configuring the hardware circuit can be understood as the processor loading instructions to implement the functions of some or all of the above modules. Furthermore, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as an NPU, TPU, or DPU.

[0187] As can be seen, each module in the above-mentioned dexterous hand parameter optimization device can be one or more processors (or processing circuits) configured to implement the above methods, such as: CPU, GPU, NPU, TPU, DPU, microprocessor, DSP, ASIC, FPGA, or a combination of at least two of these processor forms.

[0188] Furthermore, the modules in the above-mentioned dexterous hand parameter optimization device can be integrated in whole or in part, or they can be implemented independently. In one implementation, these modules are integrated together and implemented in the form of a System-on-Chip (SoC). The SoC may include at least one processor for implementing any of the above methods or implementing the functions of the modules of the device. The at least one processor can be of different types, such as CPU and FPGA, CPU and artificial intelligence processor, CPU and GPU, etc.

[0189] Please see Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 12 As shown, the electronic device 1200 includes: One or more processors 1210; The memory 1220 stores one or more programs that, when executed by one or more processors 1210, cause the one or more processors 1210 to implement the parameter optimization method for the dexterous hand described in any of the above embodiments.

[0190] Memory 1220, as a non-transitory network system, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory 1220 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 1220 may optionally include remotely located memories 1220 relative to processor 1210, which can be connected to processor 1210 via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0191] The memory 1220 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1220 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1220 and is called and executed by the processor 1210.

[0192] The processor 1210 can be implemented using a general-purpose central processing unit (CPU), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0193] In some embodiments, the electronic device further includes: Input / output interfaces are used to implement information input and output; The communication interface is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). A bus is a channel for inputting information between various components of a device, such as processor 1210, memory 1220, input / output interfaces, and communication interfaces. The processor 1210, memory 1220, input / output interface, and communication interface can communicate with each other within the device via a bus.

[0194] One embodiment of this application also provides a computer-readable storage medium storing computer-executable instructions for performing a parameter optimization method for implementing any of the embodiments described above.

[0195] An embodiment of this application also provides a computer program product, including a computer program or computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer program or computer instructions from the computer-readable storage medium and executes the computer program or computer instructions, causing the computer device to perform a parameter optimization method for a dexterous hand as described in any of the above embodiments.

[0196] The system architecture and application scenarios described in this application are for the purpose of more clearly illustrating the technical solutions of this application, and do not constitute a limitation on the technical solutions provided in this application. Those skilled in the art will know that as system architectures evolve and new application scenarios emerge, the technical solutions provided in this application are also applicable to similar technical problems.

[0197] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0198] It will be understood by those skilled in the art that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0199] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0200] The above description, with reference to the accompanying drawings, illustrates some embodiments of this application, but does not limit the scope of the invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and spirit of this invention should be considered within the scope of this application.

Claims

1. A method for optimizing the parameters of a dexterous hand, characterized in that, include: The system acquires the working parameters of the dexterous hand when performing the target-specified action, and collects the assistance parameters of the data glove. The assistance parameters are used to instruct the operator to operate the data glove to assist the dexterous hand in performing the target-specified action. Based on the assistance parameters, determine the motion compensation amount when the operator assists the dexterous hand in performing the target-specified action; The parameters of the dexterous hand are optimized based on the motion compensation amount and the working parameters.

2. The parameter optimization method for a dexterous hand according to claim 1, characterized in that, The optimization of the parameters of the dexterous hand based on the motion compensation amount and the working parameters includes: Determine the degree of completion of the dexterous hand in performing the specified action on the target, and compare the degree of completion with a completion threshold; When the completion rate is greater than or equal to the completion rate threshold, the parameters of the dexterous hand are optimized based on the comparison results between the working parameters, the action compensation amount and the compensation amount threshold.

3. The parameter optimization method for a dexterous hand according to claim 2, characterized in that, The optimization of the parameters of the dexterous hand based on the comparison results between the working parameters, the motion compensation amount, and the compensation amount threshold includes: When the comparison result indicates that the action compensation amount is greater than or equal to the sub-compensation amount threshold, the current training sample is generated based on the action compensation amount, the target specified action, and the working parameters. The preset model is trained based on multiple target training samples to obtain the target mapping relationship between the operation instructions required for the dexterous hand to perform the action and the parameters output by the dexterous hand when performing the operation instructions. The multiple target training samples include current training samples and historical training samples. The historical training samples are determined based on the parameters collected when the operator operates the data glove to assist the dexterous hand in performing other specified actions. Replace the current mapping relationship in the dexterous hand with the target mapping relationship.

4. The parameter optimization method for a dexterous hand according to claim 2, characterized in that, The optimization of the parameters of the dexterous hand based on the comparison results between the working parameters, the motion compensation amount, and the compensation amount threshold includes: When the comparison result indicates that the motion compensation amount is less than the sub-compensation amount threshold, the expected parameters output by the dexterous hand when it completes the target specified motion are determined according to the working parameters and the motion compensation amount. Based on the expected parameters, the current mapping relationship of the dexterous hand is optimized.

5. The parameter optimization method for a dexterous hand according to claim 1, characterized in that, The optimization of the parameters of the dexterous hand based on the motion compensation amount and the working parameters includes: The parameters to be compensated for the dexterous hand are identified based on the motion compensation amount; The parameters to be compensated are corrected based on the operating parameters.

6. The parameter optimization method for a dexterous hand according to claim 5, characterized in that, The step of identifying the parameters to be compensated for the dexterous hand based on the motion compensation amount includes: Based on each sub-compensation quantity in the motion compensation quantity, multiple feature values ​​are determined. Each sub-compensation quantity is the compensation quantity corresponding to each moment within the time period during which the operator assists the dexterous hand. The multiple feature values ​​include the mean, standard deviation, coefficient of variation, and maximum value of each sub-compensation quantity. Based on each of the aforementioned feature values, the parameters to be compensated for by the dexterous hand are identified.

7. The parameter optimization method for a dexterous hand according to claim 5, characterized in that, The step of identifying the parameters to be compensated for the dexterous hand based on the motion compensation amount includes: Based on the motion compensation amount, the activation intensity of each joint component of the dexterous hand is determined; Based on the activation intensity of each joint component, the parameters to be compensated for the dexterous hand are identified.

8. The parameter optimization method for a dexterous hand according to claim 5, characterized in that, The step of identifying the parameters to be compensated for the dexterous hand based on the motion compensation amount includes: Based on the motion compensation amount, determine the contribution of each joint component in the dexterous hand to the execution of the target-specified motion; Based on the contribution, a target joint component is determined among the various joint components, wherein the contribution of the target joint component is greater than a preset threshold. Based on the target joint component, determine the motion pattern to be adjusted for the dexterous hand; Based on the motion pattern, the parameters to be compensated for the dexterous hand are determined.

9. The parameter optimization method for a dexterous hand according to claim 5, characterized in that, The step of identifying the parameters to be compensated for the dexterous hand based on the motion compensation amount includes: The motion compensation amount is decomposed into sub-compensation amounts for each joint component of the dexterous hand by the operator; The target features are extracted from the sub-compensation amounts of the joint components, and the target features are time-domain features or frequency-domain features; Based on the target features of each of the joint components, the parameters to be compensated for by the dexterous hand are identified.

10. An electronic device, characterized in that, include: One or more processors; A memory that stores one or more programs, which, when executed by one or more processors, cause the one or more processors to perform the following: The parameter configuration method for a dexterous hand according to any one of claims 1-9.

Images