An underactuated gripper end force coarse perception method based on driver current information

Abstract

This invention belongs to the field of mechanical gripper technology and discloses a method for coarsely sensing the end effector force of an underactuated gripper based on actuator current information. This method includes prediction of the underactuated gripper's attitude and coarse estimation of the end effector gripping force. In the underactuated gripper attitude prediction, an underactuated gripper is constructed and modeled, and the current information is analyzed and processed. In the end effector gripping force information, the current information is mapped to the end effector gripping force, and the influence of the mechanical structure on the end effector force error is considered. This invention fills the gap in sensorless force sensing for underactuated gripping mechanical structures. This method can be applied to end effectors in various environments, especially in scenarios where there are certain requirements for gripping force. While ensuring gripping force control with a certain accuracy, it eliminates the need for force sensors, making the gripper structure simpler and reducing its cost.

Description

Technical Field

[0001] This invention belongs to the field of mechanical gripper technology, and particularly relates to a method for coarsely sensing the end force of an underactuated gripper based on driver current information. Background Technology

[0002] Currently, end-effector force sensing in industrial robots and automation systems is crucial for safe, accurate, and efficient operation. However, in some cases, the high cost and complex installation of end-effector force sensors limit their widespread application in certain fields. Summary of the Invention

[0003] The purpose of this invention is to provide a method for coarsely sensing the end force of an underactuated gripper based on driver current information, so as to solve the above-mentioned technical problems.

[0004] To address the aforementioned technical problems, the specific technical solution of the present invention, a method for coarsely sensing the end force of an underactuated gripper based on driver current information, is as follows:

[0005] A method for coarsely sensing the end force of an underactuated gripper based on driver current information includes the following steps:

[0006] S1: Kinematic / Dynamic Modeling of Underactuated Gripper: A theoretical model is constructed for the structure of the underactuated gripper. The position and torque information of the actuator of the underactuated gripper are used as inputs, and the position and coupled motion state of the end effector of the underactuated gripper are used as outputs.

[0007] S2: Analysis of current information of underactuated gripper actuator driver: The position information output by the driver is directly obtained from the encoder of the driver, and the output torque of the driver is obtained by using the current constant of the driver. The current information of the driver is processed by a filter.

[0008] S3: Underactuated gripper attitude prediction: The attitude of the finger is predicted by combining the changes in current information with the kinematic and dynamic models of the underactuated gripper.

[0009] S4: Optimization of end force accuracy of underactuated gripper: Introducing machine learning to compensate for uncertain errors, thereby improving prediction accuracy;

[0010] S5: Mapping underactuated gripper current information with end effector force: Substitute the actual input torque of the underactuated gripper actuator driver into the predicted gripper posture to obtain end effector gripping force information.

[0011] Furthermore, S1 uses an underactuated gripper to construct a kinematic and dynamic model, and the process is as follows:

[0012]

[0013] Where t represents the input torque vector from the actuator and the torsion spring mounted on the joint; ω a This represents the angular velocity vector of the torsion springs installed in each joint of the underactuated gripper; F is the contact force, the positive direction of which points towards the object; v represents the projected velocity vector at the contact point; τ and These refer to the vectors of torque and angular velocity of the joint, respectively; Δθ represents the initial compression angle of the torsion spring; and K represents the stiffness coefficient installed on the torsion spring.

[0014] Furthermore, after the driver in S2 is set with its operating speed and maximum operating current, when the current of the driver after overcoming the load is less than the set maximum operating current, the actual operating speed of the driver will remain unchanged at the set value, while the actual operating current of the driver will increase with the increase of the load; when the current of the driver after overcoming the load is greater than the set maximum operating current, the actual operating speed of the driver will gradually decrease until the driver enters a stall state, and the actual current of the motor when stalled will fluctuate around the set maximum operating current value.

[0015] Furthermore, S3 includes: classifying the gripping situation and analyzing its erect structure, decoupling the motion of the three joints of the underactuated gripper, and classifying the gripping mode switching situation into three categories according to the order in which the object contacts the three joints: the distal finger joint contacts the object first, the middle finger joint contacts the object first, and the proximal finger joint contacts the object first.

[0016] When the distal interphalangeal joint first contacts the object: the underactuated gripper will always maintain the parallel gripping mode. At this time, the distal spring remains stationary, θ3 remains unchanged, and no load is generated when the motor moves. However, the proximal spring will move, and θ1 and θ2 will change, and the two will remain complementary. During this stage, an accurate mapping relationship is established between the actuator position information and θ1 and θ2. At this time, the proximal spring will gradually relax from a taut state when the gripper closes, so the generated load gradually decreases.

[0017] When the middle finger joint first contacts the object: the underactuated gripper maintains a parallel gripping mode before contacting the object, the distal spring remains stationary, and the proximal spring moves; after the middle finger joint contacts the object, the underactuated gripper switches to an enveloping gripping mode, the distal spring begins to move and θ3 changes, and the distal spring will gradually tighten, and the generated load will gradually increase, while at this time the proximal spring will stop moving, θ1 and θ2 remain unchanged, and no longer generate load. During this stage, an accurate mapping relationship is established between the actuator position information and θ3.

[0018] When the proximal interphalangeal joint first contacts the object: the underactuated gripper maintains a parallel gripping mode before contacting the object, the distal spring remains stationary, and the proximal spring moves; after the proximal interphalangeal joint contacts the object, the underactuated gripper switches to an enveloping gripping mode, the distal spring begins to move and couples with the movement of the proximal spring, θ2 and θ3 change and are mutually coupled, while θ1 remains unchanged. This mode corresponds to the contact situation in case 1. In actual design, by increasing the spring constant of the proximal spring, only the distal spring moves after the underactuated gripper switches to the enveloping gripping mode, optimizing it so that only θ3 changes while θ1 and θ2 remain unchanged. i Establish an accurate mapping relationship between θ3 and θ3.

[0019] Furthermore, S3 classifies the grasping process, thereby establishing an accurate mapping relationship between the driver position information and θ1, θ2, and θ3 in both parallel grasping mode and wrapping grasping mode. However, when the two modes switch, θ1 changes from a variable to a fixed value. Therefore, obtaining the critical value of θ1 during mode switching becomes the key to closely connecting the transformation of the mapping relationship between the driver position information and θ1, θ2, and θ3 between the two modes.

[0020] By combining the characteristic of the driver current changing with the load with the movement of the far-end and near-end springs during gripping mode switching, the critical point of gripping mode switching is predicted by the change in current, and then the values ​​of θ1, θ2, and θ3 are predicted during mode switching: In parallel gripping mode, as the overall load gradually decreases, the actual change in motor current is also small. At this time, there is an accurate mapping relationship between the driver position information and θ1 and θ2. When switching from parallel gripping to enveloping gripping mode, the overall load gradually increases and the current surges. Therefore, when the current surges, it is considered that the underactuated gripping actuator is in contact with the object. At this time, there is an accurate mapping relationship between the driver position information and θ3.

[0021] The LSTM forget gate is used to combine information from before and after the current time step to comprehensively determine whether the interval with large current fluctuations has been forgotten. In the LSTM network model, current information is collected when θ1 forms different angles during grasping, forming a dataset. The collected information includes the actual current value, the first derivative of the current, and the actual angle of θ1. After data acquisition, the filtered current value and the first derivative of the current are input into the network. During model training, time points are marked according to the actual angle of θ1; when the actual angle of θ1 is less than the grasped object, that time point is marked as...

[0022] "0" indicates that the actual angle θ1 is greater than the object being grasped, and this time point is marked as "1". This means the dataset is subjected to binary classification, and the model uses the time point's label as output. If the output for a time point is "0", then...

[0023] This indicates that the underactuated gripper is in parallel gripping mode. If the output at a time point is "1", it means that the underactuated gripper is in enveloping gripping mode. In the actual use of this LSTM model, the current information will be input in real time and the time point will be marked. When the output changes from "0" to "1", it means that the gripping mode has been switched. The corresponding value of θ1 at this time is the predicted value when the gripping mode is switched. Furthermore, S4 uses a mathematical model to fit and compensate for the deviation between θ1 predicted by LSTM and the actual θ1. It collects the critical value of θ1 when the predicted mode switches at different rotation speeds and when grasping objects that cause θ1 to form different angles. It also collects the θ1 data predicted by the LSTM network when the underactuated grasping actuator grasps different objects at different rotation speeds. The data is then fitted sequentially on the xz plane and yz plane, using quadratic and quartic polynomials respectively to achieve the fitting without overfitting. The resulting fitted plane and partial projections in the x and y directions are used. In practical applications, the value of θ1 that needs to be compensated is obtained by comparing the actuator speed with the predicted θ1.

[0024] Furthermore, in step S5, the torque information obtained by the driver of the underactuated gripper based on the processed current information and current constant, as well as the predicted gripper posture, are substituted into formulas (1) and (2) to obtain the end gripping force information.

[0025] The coarse sensing method for the end force of an underactuated gripper based on driver current information, as proposed in this invention, has the following advantages:

[0026] This invention infers the clamping force at the end of the gripper by monitoring the actuator current. This method leverages the correlation between actuator current and end-effector force to achieve coarse sensing of the end-effector force without the need for an additional end-effector force sensor. The advantages of this technique lie in its simplicity and cost-effectiveness; by utilizing existing actuator current monitoring devices, end-effector force sensing can be achieved without additional hardware costs. In summary, this invention provides a simple and effective method for achieving coarse sensing of end-effector force in underactuated grippers, thus offering a new solution for the safety and performance of industrial robots and automation systems. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the underactuated gripping actuator of the present invention;

[0028] Figure 2 This is a schematic diagram of the parallel grasping mode structure of the underactuated grasping actuator of the present invention;

[0029] Figure 3 This is a schematic diagram of the underactuated gripping actuator's enveloping gripping mode structure according to the present invention;

[0030] Figure 4 This is a schematic diagram of the process for the coarse sensing method of the underactuated gripper end force based on driver current information according to the present invention.

[0031] Figure 5 This is a schematic diagram of the LSTM network structure of the present invention;

[0032] Figure 6 This is a schematic diagram illustrating the data performance results of θ1 predicted by the LSTM network according to the present invention;

[0033] Figure 7 This is a schematic diagram of the fitting plane of the present invention and its partial projection in the x and y directions. Detailed Implementation

[0034] To better understand the purpose, structure, and function of this invention, the following description, in conjunction with the accompanying drawings, provides a more detailed account of a method for coarsely sensing the end force of an underactuated gripper based on driver current information.

[0035] The coarse force sensing method at the end of the underactuated gripper based on driver current information in this invention is constructed based on the underactuated gripper actuator in the paper "Construction of a Multiple-DOF Underactuated Gripper with Force-Sensing via Deep Learning", such as... Figure 1-3 As shown, each rotary joint of the linkage of the underactuated gripper is connected by pins. To further reduce gripping resistance, roller bearings are installed at the pin mating points. At the same time, to facilitate the relative position configuration of the fingers, torsion springs and mechanical stops are used to ensure their initial relative position. It is worth noting that the left and right torsion springs are configured completely independently. When gripping asymmetrical objects, the gripping effect can be improved by adjusting the stiffness of the left and right torsion springs separately.

[0036] Based on the distance of each phalanx from the palm, the nearest phalanx is called the proximal phalanx 1, the middle phalanx is called the middle phalanx 2, and the distal phalanx is called the distal phalanx 3. All three phalanxes are driven by the outer connecting rod 4. A torsion spring is installed at the intersection of each phalanx. The torsion spring installed at the intersection of the distal and middle phalanxes is called the distal spring 5, and the torsion spring installed at the intersection of the proximal and middle phalanxes is called the proximal spring 6. The installed springs enable the underactuated gripper to switch between parallel gripping mode and enveloping gripping mode.

[0037] When the distal knuckle first contacts the object, the three joints of the gripper will be constrained by two springs, maintaining a parallel gripping pattern. Figure 2As shown); when the middle or proximal interphalangeal joint first contacts the object, the knuckles of the gripper will overcome the constraint of the spring, and the gripper will switch to an envelope gripping mode. Figure 3 As shown in the figure, since the gripper is driven by only one drive link, it is impossible to establish a mapping relationship between each finger joint in the envelope gripping mode and the drive link. Therefore, it is impossible to obtain the posture of each finger joint in the envelope gripping mode. This invention will use the current information of the underactuated gripping actuator driver to roughly estimate the posture of each finger joint in the envelope gripping mode.

[0038] like Figure 4 As shown, the present invention provides a method for coarsely sensing the end force of an underactuated gripper based on driver current information, comprising the following steps:

[0039] S1: Kinematic / Dynamic Modeling of the Underactuated Gripper. A theoretical model is constructed based on the actual structure of the underactuated gripper. The position and torque information of the actuator are used as inputs, and the attitude and coupled motion state of the end effector are used as outputs. The kinematic and dynamic models are constructed using the underactuated gripper described above, and the process is as follows:

[0040]

[0041] Where t represents the input torque vector from the actuator and the torsion spring mounted on the joint; ω a This represents the angular velocity vector of the torsion springs installed in each joint of the underactuated gripper; F is the contact force, the positive direction of which points towards the object; v represents the projected velocity vector at the contact point; τ and These refer to the vectors of torque and angular velocity of the joint, respectively; Δθ represents the initial compression angle of the torsion spring; and K represents the stiffness coefficient installed on the torsion spring.

[0042] S2: Underactuated gripper actuator driver current information analysis. This method requires the ability to read the driver's position, speed, and current information from the driver's encoder, and to set the driver's maximum operating current, operating speed, and position. It can also obtain the driver's output torque using the driver's current constant (in N*m). However, since the actual output current of the driver is a fluctuating curve, a filter is used to process the driver's current information. After setting the operating speed and maximum operating current, when the current of the driver overcoming the load is less than the set maximum operating current, the actual operating speed of the driver will remain unchanged at the set value, while the actual operating current will increase with the increase of the load. When the current of the driver overcoming the load is greater than the set maximum operating current, the actual operating speed of the driver will gradually decrease until the driver enters a stall state (driver speed is 0 rpm), and the actual current of the motor during stall will fluctuate around the set maximum operating current value.

[0043] S3: Attitude prediction of underactuated gripper. Since the degrees of freedom of an underactuated gripper are greater than the number of actuators, the multiple joints of the actuator's fingers cannot move independently. It is impossible to obtain the movement of each finger joint through the input of a single actuator. This method uses changes in current information combined with the kinematic and dynamic models of the underactuated gripper to predict the finger's attitude.

[0044] The torsion springs installed at the intersections of each joint allow the underactuated gripper to switch between parallel gripping and enveloping gripping modes. However, due to the coupling relationship between θ1, θ2, and θ3, it is impossible to establish an accurate mapping relationship between the actuator position information and θ1, θ2, and θ3 during gripper movement. Therefore, by classifying the gripping situations and analyzing the supporting structure, the movement of the three joints of the underactuated gripper is decoupled. Based on the order in which the object contacts the three joints, the gripping mode switching situations are divided into three categories:

[0045] 1. The distal joint contacts the object first: The underactuated gripper will always maintain the parallel gripping mode. At this time, the distal spring remains stationary (θ3 remains unchanged) and does not generate load when the motor moves, while the proximal spring moves (θ1 and θ2 change and remain complementary). During this stage, an accurate mapping relationship can be established between the actuator position information and θ1 and θ2. At this time, the proximal spring will gradually relax from a taut state when the gripper closes, so the generated load gradually decreases.

[0046] 2. The middle finger joint contacts the object first: Before contacting the object, the underactuated gripper maintains a parallel gripping mode, with the distal spring remaining stationary and the proximal spring moving; after the middle finger joint contacts the object, the underactuated gripper switches to an enveloping gripping mode, the distal spring begins to move (θ3 changes), and the distal spring will gradually tighten, gradually increasing the load, while the proximal spring will stop moving (θ1 and θ2 remain unchanged), no longer generating a load. This stage can establish an accurate mapping relationship between the actuator position information and θ3.

[0047] 3. Proximal knuckle contacts the object first: Before contacting the object, the underactuated gripper maintains a parallel gripping mode, with the distal spring remaining stationary and the proximal spring moving. Once the proximal knuckle contacts the object, the underactuated gripper switches to an enveloping gripping mode, the distal spring begins to move, and its movement is coupled with that of the proximal spring (θ2 and θ3 change and are coupled, while θ1 remains unchanged). This mode corresponds to the contact situation in case 1. Extensive gripping experiments revealed that the case of the proximal knuckle contacting the object first is relatively rare. In practical design, by increasing the spring constant of the proximal spring, only the distal spring moves after the underactuated gripper switches to the enveloping gripping mode (optimized to only change θ3, while θ1 and θ2 remain unchanged), θ can be reduced. i Establish an accurate mapping relationship between θ3 and θ3.

[0048] By classifying the crawling, an accurate mapping relationship between the driver position information and θ1, θ2, and θ3 was established in both parallel crawling mode and wrapping crawling mode. However, when switching between the two modes, θ1 changed from a variable to a fixed value. Therefore, obtaining the critical value of θ1 during mode switching became the key to closely connecting the transformation of the mapping relationship between the driver position information and θ1, θ2, and θ3 between the two modes.

[0049] By combining the characteristic of the driver current changing with the load with the movement of the far-end and near-end springs during gripping mode switching, the critical point of gripping mode switching can be predicted by the change in current, and thus the values ​​of θ1, θ2, and θ3 during mode switching can be predicted: In parallel gripping mode, as the overall load gradually decreases, the actual change in motor current is also small. At this time, there is an accurate mapping relationship between the driver position information and θ1 and θ2. When switching from parallel gripping to enveloping gripping mode, the overall load gradually increases and the current surges. Therefore, when the current surges, it is considered that the underactuated gripping actuator is in contact with the object. At this time, there is an accurate mapping relationship between the driver position information and θ3.

[0050] However, due to random fluctuations in current, when the current fluctuation is large, there may be situations where the actual current approaches or even exceeds the mode switching current threshold even if the underactuated grasping actuator does not switch grasping modes. This could affect the prediction accuracy of θ1 during mode switching, or even cause misjudgment. The forget gate of LSTM can combine information before and after the current time step to comprehensively determine whether the interval with large current fluctuations has been forgotten. In the LSTM network model of this experiment, current information was collected when grasping caused θ1 to form different angles, forming a dataset. The collected information included the actual current value, the first derivative of the current, and the actual angle of θ1. After the data collection was completed, the filtered current value and the first derivative of the current were input into the network. During model training, time points were marked according to the actual angle of θ1. When the actual angle of θ1 was less than the grasped object, the time point was marked as "0", and when the actual angle of θ1 was greater than the grasped object, the time point was marked as "1", that is, the dataset was classified into two categories. The model outputs time-point markers. A time-point output of "0" indicates the underactuated gripper is in parallel gripping mode, while a time-point output of "1" indicates it's in enveloping gripping mode. In practical use of this LSTM model, real-time current information is input, and time-point markers are output. When the output changes from "0" to "1", it indicates a gripping mode switch, and the corresponding value of θ1 is the predicted value at the time of the gripping mode switch. The LSTM network structure is as follows: Figure 5 As shown.

[0051] In summary, the critical value of θ1 during mode switching can be predicted by observing changes in current.

[0052] S4: Optimization of the end effector force accuracy of the underactuated gripper. The θ1 directly predicted by the LSTM network model still deviates from the actual critical value of θ1 during mode switching due to factors such as current delay caused by mean filtering and non-geometric errors in the mechanical structure (e.g., backlash, pin fit). Therefore, machine learning is introduced to compensate for these uncertainties, thereby improving prediction accuracy. A mathematical model is used to fit and compensate for the deviation between the LSTM-predicted θ1 and the actual θ1. The critical value of θ1 during mode switching is predicted when gripping objects at different rotational speeds and angles that cause θ1 to form different shapes. Data on θ1 predicted by the LSTM network is collected for a certain number of underactuated gripper actuators gripping different objects at different rotational speeds. The results are as follows. Figure 6As shown in the diagram. The x-axis represents the motor speed, the y-axis represents the size of the object being grasped, and the z-axis represents the deviation between the predicted θ1 and the actual θ1 obtained through LSTM. The data is fitted sequentially in the xz and yz planes. Using quadratic and quartic polynomials respectively achieves good fitting results without overfitting. The resulting fitted planes and their partial projections in the x and y directions are shown in the diagram. Figure 7 As shown. In practical applications, the value of θ1 that needs to be compensated can be obtained by comparing the drive speed with the predicted θ1.

[0053] S5: Mapping the current information of the underactuated gripper to the end force. Substitute the torque information obtained by the driver of the underactuated gripper based on the processed current information and current constant (obtained in S2), and the predicted gripper posture (obtained in S4 and S5) into formulas (1) and (2) to obtain the end gripping force information.

[0054] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A method for coarsely sensing the end force of an underactuated gripper based on driver current information, characterized in that, Includes the following steps: S1: Kinematic / Dynamic Modeling of Underactuated Gripper: A theoretical model is constructed for the structure of the underactuated gripper. The position and torque information of the actuator of the underactuated gripper are used as inputs, and the position and coupled motion state of the end effector of the underactuated gripper are used as outputs. S2: Analysis of current information of underactuated gripper actuator driver: The position information output by the driver is directly obtained from the encoder of the driver, and the output torque of the driver is obtained by using the current constant of the driver. The current information of the driver is processed by a filter. S3: Underactuated gripper attitude prediction: The attitude of the finger is predicted by combining the changes in current information with the kinematic and dynamic models of the underactuated gripper. S4: Optimization of end force accuracy of underactuated gripper: Introducing machine learning to compensate for uncertain errors, thereby improving prediction accuracy; S5: Mapping underactuated gripper current information with end effector force: Substitute the actual input torque of the underactuated gripper actuator driver into the predicted gripper posture to obtain end effector gripping force information.

2. The method for coarsely sensing the end force of an underactuated gripper based on driver current information according to claim 1, characterized in that, S1 uses an underactuated gripper to build a kinematic and dynamic model, and the process is as follows: Where t represents the input torque vector from the actuator and the torsion spring mounted on the joint; ω a This represents the angular velocity vector of the torsion springs installed in each joint of the underactuated gripper; F is the contact force, the positive direction of which points towards the object; v represents the projected velocity vector at the contact point; τ and These refer to the vectors of torque and angular velocity of the joint, respectively; Δθ represents the initial compression angle of the torsion spring; and K represents the stiffness coefficient installed on the torsion spring.

3. The method for coarsely sensing the end force of an underactuated gripper based on driver current information according to claim 1, characterized in that, After the driver of S2 is set with its operating speed and maximum operating current, when the current of the driver after overcoming the load is less than the set maximum operating current, the actual operating speed of the driver will remain unchanged at the set value, while the actual operating current of the driver will increase with the increase of the load. When the current value of the driver overcoming the load is greater than the set maximum operating current, the actual operating speed of the driver will gradually decrease until the driver enters a stall state, and the actual current of the motor when stalled will fluctuate around the set maximum operating current value.

4. The method for coarsely sensing the end force of an underactuated gripper based on driver current information according to claim 1, characterized in that, The S3 includes: classifying the gripping situation and analyzing its erect structure, decoupling the motion of the three joints of the underactuated gripper, and classifying the gripping mode switching situation into three categories according to the order in which the object contacts the three joints: the distal finger joint contacts the object first, the middle finger joint contacts the object first, and the proximal finger joint contacts the object first. When the distal interphalangeal joint first contacts the object: the underactuated gripper will always maintain the parallel gripping mode. At this time, the distal spring remains stationary, θ3 remains unchanged, and no load is generated when the motor moves. However, the proximal spring will move, and θ1 and θ2 will change, and the two will remain complementary. During this stage, an accurate mapping relationship is established between the actuator position information and θ1 and θ2. At this time, the proximal spring will gradually relax from a taut state when the gripper closes, so the generated load gradually decreases. When the middle finger joint first contacts the object: the underactuated gripper maintains a parallel gripping mode before contacting the object, the distal spring remains stationary, and the proximal spring moves; after the middle finger joint contacts the object, the underactuated gripper switches to an enveloping gripping mode, the distal spring begins to move and θ3 changes, and the distal spring will gradually tighten, and the generated load will gradually increase, while at this time the proximal spring will stop moving, θ1 and θ2 remain unchanged, and no longer generate load. During this stage, an accurate mapping relationship is established between the actuator position information and θ3. When the proximal interphalangeal joint first contacts the object: the underactuated gripper maintains a parallel gripping mode before contacting the object, the distal spring remains stationary, and the proximal spring moves; after the proximal interphalangeal joint contacts the object, the underactuated gripper switches to an enveloping gripping mode, the distal spring begins to move and couples with the movement of the proximal spring, θ2 and θ3 change and are mutually coupled, while θ1 remains unchanged. This mode corresponds to the contact situation in case 1. In actual design, by increasing the spring constant of the proximal spring, only the distal spring moves after the underactuated gripper switches to the enveloping gripping mode, optimizing it so that only θ3 changes while θ1 and θ2 remain unchanged. i Establish an accurate mapping relationship between θ3 and θ3.

5. The method for coarsely sensing the end force of an underactuated gripper based on driver current information according to claim 1, characterized in that, The above-mentioned S3 classifies the grasping process, thereby establishing an accurate mapping relationship between the driver position information and θ1, θ2, and θ3 in both parallel grasping mode and wrapping grasping mode. However, when the two modes switch, θ1 changes from a variable to a fixed value. Therefore, obtaining the critical value of θ1 during mode switching becomes the key to closely connecting the change in the mapping relationship between the driver position information and θ1, θ2, and θ3 between the two modes. By combining the characteristic of the driver current changing with the load with the movement of the far-end and near-end springs during gripping mode switching, the critical point of gripping mode switching is predicted by the change in current, and then the values ​​of θ1, θ2, and θ3 are predicted during mode switching: In parallel gripping mode, as the overall load gradually decreases, the actual change in motor current is also small. At this time, there is an accurate mapping relationship between the driver position information and θ1 and θ2. When switching from parallel gripping to enveloping gripping mode, the overall load gradually increases and the current surges. Therefore, when the current surges, it is considered that the underactuated gripping actuator is in contact with the object. At this time, there is an accurate mapping relationship between the driver position information and θ3. The LSTM forget gate is used to combine information from before and after the current time step to comprehensively determine whether the interval with large current fluctuations has been forgotten. In the LSTM network model, current information is collected when θ1 forms different angles during grasping, forming a dataset. The collected information includes the actual current value, the first derivative of the current, and the actual angle of θ1. After data acquisition, the filtered current value and the first derivative of the current are input into the network. During model training, time points are marked according to the actual angle of θ1; when the actual angle of θ1 is less than the grasped object, that time point is marked as... When the actual angle of θ1 is greater than the object being grasped, the time point is marked as "1". This means that the dataset is classified into two categories. The model uses the time point marking as the output. If the output of the time point is "0", it means that the underactuated grasping actuator is in parallel grasping mode. If the output of the time point is "1", it means that the underactuated grasping actuator is in enveloping grasping mode. In the actual use of this LSTM model, the current information is input in real time and the time point marking is output. When the output changes from "0" to "1", it means that the grasping mode has switched. The value of θ1 at this time is the predicted value when the grasping mode is switched.

6. The method for coarsely sensing the end force of an underactuated gripper based on driver current information according to claim 1, characterized in that, S4 uses a mathematical model to fit and compensate for the deviation between θ1 predicted by LSTM and the actual θ1. It collects the critical value of θ1 when the predicted mode switches at different rotation speeds and when grasping objects that form different angles with θ1. It also collects the θ1 data predicted by the LSTM network when the underactuated grasping actuator grasps different objects at different rotation speeds. The data is then fitted to the xz plane and yz plane in turn, using quadratic and quartic polynomials respectively to achieve the fitting without overfitting. The resulting fitted plane and partial projections in the x and y directions are used. In practical applications, the value of θ1 that needs to be compensated is obtained by comparing the actuator speed with the predicted θ1.

7. The method for coarsely sensing the end force of an underactuated gripper based on driver current information according to claim 2, characterized in that, S5 substitutes the torque information obtained by the driver of the underactuated gripper based on the processed current information and current constant, as well as the predicted gripper posture, into formulas (1) and (2) to obtain the end gripping force information.

Images