A biomimetic robot dexterous hand tactile calibration method, device and equipment

Abstract

The application provides a biomimetic robot dexterous hand tactile calibration method, device and equipment, through adopting closed cavity precise limiting clamping and full-coated flexible air bag self-adaptive curved surface uniform pressure fitting, multi-stage gradient air pressure steady-state pressure control and full-channel synchronous data acquisition, single-channel independent quadratic polynomial fitting and closed-loop error checking, the problems that the existing rigid calibration method cannot adapt to the special-shaped curved surface sensor of the dexterous hand, there are many blind areas in calibration, the pressure distribution of the sensor is uneven, the flexible sensing structure is easy to be damaged, the calibration error is large, the pressure control precision is low, the multi-channel calibration consistency is poor, and the global calibration efficiency is extremely low are solved, the standardized precise calibration of the biomimetic robot dexterous hand five fingers is realized, the calibration benchmark of all sensing channels is effectively unified, the multi-channel data dispersion is reduced, and the calibration precision and mass production calibration efficiency of the flexible curved surface tactile sensor are improved.

Description

Technical Field

[0001] This application relates to the field of bionic robot tactile sensor calibration technology, and in particular to a calibration method, apparatus and equipment for the tactile sense of a bionic robot's dexterous hand. Background Technology

[0002] For flexible curved surface array sensors for dexterous hands, it is essential to ensure uniform force application, full-area pressure application, and precise and controllable pressure during the calibration process to obtain reliable calibration datasets. Traditional manual calibration processes, such as alignment, pressure adjustment, state determination, and data recording, are prone to random errors. Differences in human operation can lead to large fluctuations and poor consistency in sensor channel and batch data, weakening sensor stability. Currently, the industry's calibration methods for dexterous hand tactile sensors still mainly rely on traditional rigid contact calibration, including mechanical rigid indenter point pressing calibration, weight-based pressure calibration, and single-point pneumatic pressure calibration.

[0003] The aforementioned traditional calibration methods have many unavoidable technical defects: poor surface adaptability, a large number of calibration blind spots, and the sensor's curved sides and concave transition areas are in a suspended and pressureless state for a long time, making it impossible to achieve synchronous calibration of the entire channel; they are prone to damaging flexible sensors, and the rigid pressure head has a small contact area. Long-term repeated calibration will cause irreversible plastic deformation of the sensor, resulting in problems such as channel drift, sensitivity attenuation, and channel failure, which will significantly reduce the sensor yield and service life.

[0004] The pressure control accuracy is low and the calibration error is large. Traditional weights and mechanical indenters rely on mechanical displacement to apply pressure. Affected by material deformation and mechanical return clearance, the pressure fluctuates greatly and cannot output a stable and controllable standard pressure. This makes it difficult to meet the data acquisition requirements of multi-level static calibration of flexible sensors. The fitting calibration curve has large errors and poor linearity. The calibration efficiency is extremely low and the channel consistency is poor. Existing technologies generally adopt single-finger and single-point successive calibration methods. For a single finger, it is necessary to repeatedly align and apply pressure multiple times. The five-index calibration takes a very long time. At the same time, the loading posture and contact position cannot be kept consistent each time, resulting in poor consistency between channels. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide a calibration method, device, and equipment for the tactile sense of a bionic robot's dexterous hand. By employing a sealed cavity for precise positioning and clamping, and a fully enclosed flexible airbag for adaptive curved surface pressure equalization and bonding, multi-level gradient air pressure steady-state control, and synchronous data acquisition across all channels, this method solves the problems of existing rigid calibration methods, such as inability to adapt to irregular curved surface sensors of dexterous hands, large blind zones in calibration, uneven pressure distribution on sensors, easy damage to flexible sensing structures, large calibration errors, low pressure control accuracy, poor consistency of multi-channel calibration, and extremely low global calibration efficiency. This method achieves standardized and accurate calibration of the five fingers of the bionic robot's dexterous hand simultaneously, with no blind zones across the entire curved surface, and without stress or damage. It effectively unifies the calibration benchmark of all sensing channels, reduces the dispersion of multi-channel data, and improves the calibration accuracy and mass production calibration efficiency of flexible curved surface tactile sensors.

[0006] This application provides a calibration method for the tactile sensation of a bionic robot's dexterous hand. The calibration method is applied to a dexterous hand tactile sensation calibration device, which includes a sealed calibration cavity, a fully enclosed curved flexible airbag module, a pressure stabilization and control unit, and a multi-channel data acquisition unit. The fully enclosed curved flexible airbag module is fixed inside the sealed calibration cavity, and the fully enclosed curved flexible airbag module is provided with multiple independent finger-enclosing cavities. The calibration method includes: The dexterous hand of the target bionic robot to be calibrated is fixed on the positioning structure of the sealed calibration cavity, and each finger of the dexterous hand to be calibrated is controlled to extend into each finger cavity of the fully covered curved flexible airbag module, so that the flexible curved tactile sensor on the surface of each finger is in contact with the inner wall of the airbag of the finger cavity. The multi-channel data acquisition unit is zero-point calibrated, and the air pressure stabilization control unit is controlled to pressurize each finger covering cavity step by step according to a preset calibration pressure gradient sequence. In the pressure steady state stage corresponding to each calibration pressure value in the calibration pressure gradient sequence, the multi-channel data acquisition unit is used to acquire the analog-to-digital conversion value corresponding to each sensing channel of each flexible curved surface tactile sensor. For each of the aforementioned sensing channels, a mapping curve is fitted between the analog-to-digital conversion value corresponding to each of the aforementioned calibrated pressure values ​​and the corresponding calibrated pressure values ​​to obtain the calibration coefficient result corresponding to the sensing channel. Based on the calibration coefficient results corresponding to each of the sensor channels, a tactile calibration file corresponding to the dexterous hand to be calibrated is generated, and the pressure relief and reset of each finger cavity is controlled.

[0007] Furthermore, the zero-point calibration of the multi-channel data acquisition unit includes: The fully enclosed curved flexible airbag module is completely depressurized, and the flexible curved tactile sensor is kept in a static, pressure-free state. In response to the fact that the fully enclosed curved flexible airbag module has been completely depressurized and the flexible curved tactile sensor is in a static and pressureless state, the original analog-to-digital conversion initial value corresponding to each sensing channel configured for each of the flexible curved tactile sensors is acquired by the multi-channel data acquisition unit. The original analog-to-digital conversion initial value is determined as the zero-point reference value for each of the sensing channels to complete the zero-point calibration of the multi-channel data acquisition unit.

[0008] Furthermore, in the steady-state stage corresponding to each calibrated pressure value in the calibrated pressure gradient sequence, the analog-to-digital conversion value corresponding to each sensing channel configured for each flexible curved surface tactile sensor is acquired using the multi-channel data acquisition unit, including: In the steady-state stage of pressure corresponding to each calibrated pressure value in the calibrated pressure gradient sequence, the original analog-to-digital conversion value corresponding to each sensing channel configured for each of the flexible curved surface tactile sensors is acquired by the multi-channel data acquisition unit. The difference between the original analog-to-digital conversion value and the zero-point reference value of the sensing channel is used to obtain the analog-to-digital conversion value corresponding to the sensing channel.

[0009] Furthermore, the calibration coefficient results include quadratic term coefficients, linear term coefficients, constant term coefficients, equivalent sensitivity coefficients, and zero-point offset coefficients; for each of the sensing channels, the process of performing a mapping curve fitting between the analog-to-digital conversion value corresponding to each calibration pressure value and the corresponding calibration pressure value to obtain the calibration coefficient results corresponding to that sensing channel includes: For each of the aforementioned sensing channels, based on the analog-to-digital conversion value and the corresponding calibration pressure value of the sensing channel at each calibration pressure value, the least squares method is used to perform matrix operations on the preset quadratic polynomial formula to obtain the quadratic term coefficients, the linear term coefficients, and the constant term coefficients corresponding to the sensing channel. Based on the quadratic term coefficient, the linear term coefficient, the constant term coefficient, the analog-to-digital conversion value and the corresponding calibration pressure value of the sensing channel under each calibration pressure value, and the quadratic polynomial formula, the calibration curve corresponding to the sensing channel is fitted, and the calibration curve is verified for error to obtain the curve verification result. In response to the curve verification result indicating that the calibration curve corresponding to the sensing channel has passed the error verification, the instantaneous slope of the curve corresponding to the calibration curve is determined, and the instantaneous slope of the curve corresponding to the analog-to-digital conversion value below the midpoint of the range of the calibration curve is determined as the equivalent sensitivity coefficient corresponding to the sensing channel. Based on the zero-point reference value of the sensing channel, the zero-point offset coefficient corresponding to the sensing channel is determined using the calibration curve corresponding to the sensing channel.

[0010] Furthermore, the step of verifying the calibration curve to obtain the curve verification result includes: The air pressure stabilization control unit pressurizes the finger covering cavity according to a preset intermediate verification pressure value, and during the pressure steady-state stage corresponding to the intermediate verification pressure value, the multi-channel data acquisition unit acquires the analog-to-digital conversion sampling value corresponding to the sensing channel. Based on the analog-to-digital conversion sampled value, the test pressure value is determined using the calibration curve corresponding to the sensing channel, and based on the test pressure value, the intermediate verification pressure value, and the preset full-scale pressure value, the error result corresponding to the calibration curve corresponding to the sensing channel is determined. Determine whether the error result is less than a preset error threshold; If the error result is less than the preset error threshold, then the curve verification result corresponding to the calibration curve of the sensing channel is determined to be passed the error verification. If the error result is greater than or equal to the preset error threshold, then the curve verification result is determined to be a failure of error verification.

[0011] This application embodiment also provides a calibration device for the tactile sense of a bionic robot's dexterous hand, the calibration device comprising: The clamping and positioning control module is used to control the dexterous hand of the target bionic robot to be calibrated to be fixed on the positioning structure of the sealed calibration cavity, and to control each finger of the dexterous hand to be calibrated to be inserted into each finger cavity of the fully covered curved flexible airbag module, so that the flexible curved tactile sensor on the surface of each finger is in contact with the inner wall of the airbag of the finger cavity. The calibration data acquisition module is used to perform zero-point calibration on the multi-channel data acquisition unit and control the air pressure stabilization control unit to pressurize each finger covering cavity step by step according to the preset calibration pressure gradient sequence. In the pressure steady state stage corresponding to each calibration pressure value in the calibration pressure gradient sequence, the multi-channel data acquisition unit is used to acquire the analog-to-digital conversion value corresponding to each sensing channel of each flexible curved surface tactile sensor. The mapping curve fitting module is used to perform mapping curve fitting on the analog-to-digital conversion value and the corresponding calibration pressure value for each of the sensing channels under each calibration pressure value, so as to obtain the calibration coefficient result corresponding to the sensing channel. The calibration file generation module is used to generate a tactile calibration file corresponding to the dexterous hand to be calibrated based on the calibration coefficient results corresponding to each of the sensing channels, and to control the depressurization and reset of each finger covering cavity.

[0012] This application embodiment also provides a dexterous hand tactile calibration device, which includes a sealed calibration cavity, a fully enclosed curved flexible airbag module, a pressure stabilization control unit, a multi-channel data acquisition unit, and a bionic robot dexterous hand tactile calibration device applying the above-described bionic robot dexterous hand tactile calibration method; the fully enclosed curved flexible airbag module is fixed inside the sealed calibration cavity; the fully enclosed curved flexible airbag module is provided with multiple independent finger enclosed cavities; the multi-channel data acquisition unit includes multiple acquisition channels, each acquisition channel corresponding to each sensing channel configured for each flexible curved tactile sensor.

[0013] Furthermore, the finger-covering cavity is made of highly elastic and highly ductile thermoplastic polyurethane elastomer or silicone flexible material; in the inflated state of the finger-covering cavity, the finger-covering cavity adaptively conforms to the irregular outer curved surface of each finger of the dexterous hand to be calibrated, so that the finger-covering cavity completely wraps the curved surface of each finger, so that the flexible curved surface tactile sensor on the surface of each finger is uniformly pressed and there are no stress concentration points.

[0014] Furthermore, the sealed calibration chamber is a rigid cylindrical structure, and the top of the sealed calibration chamber is provided with a positioning structure for positioning the dexterous hand to be calibrated. The sealed calibration chamber is also provided with a sealing ring and a limiting seat. The air pressure stabilization and control unit includes a micro air pump, a precision pressure regulating valve, a pressure sensor, a pressure stabilizing air tank, and an electromagnetic control valve group.

[0015] This application also provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, the steps of the above-described bionic robot dexterity hand tactile calibration method are performed.

[0016] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the above-described bionic robot dexterity hand tactile calibration method.

[0017] This application provides a method, apparatus, and device for calibrating the tactile sensation of a bionic robot's dexterous hand. The calibration method is applied to a dexterous hand tactile calibration device, which includes a sealed calibration cavity, a fully enclosed curved flexible airbag module, a pressure stabilization control unit, and a multi-channel data acquisition unit. The fully enclosed curved flexible airbag module is fixed inside the sealed calibration cavity, and the fully enclosed curved flexible airbag module has multiple independent finger enclosed cavities. The calibration method includes: controlling the dexterous hand of the target bionic robot to be calibrated to be fixed on the positioning structure of the sealed calibration cavity, and controlling each finger of the dexterous hand to be calibrated to be inserted into each finger enclosed cavity of the fully enclosed curved flexible airbag module, so that the flexible curved tactile sensor on the surface of each finger corresponds to the inner wall of the airbag of the finger enclosed cavity. The system is fitted together; the multi-channel data acquisition unit is zero-point calibrated, and the air pressure stabilization control unit is controlled to pressurize each finger cavity step by step according to a preset calibration pressure gradient sequence. During the steady-state stage corresponding to each calibration pressure value in the calibration pressure gradient sequence, the multi-channel data acquisition unit acquires the analog-to-digital conversion value corresponding to each sensing channel of each flexible curved surface tactile sensor. For each sensing channel, the analog-to-digital conversion value corresponding to each calibrated pressure value is mapped to the corresponding calibration pressure value to obtain the calibration coefficient result corresponding to the sensing channel. Based on the calibration coefficient result corresponding to each sensing channel, a tactile calibration file corresponding to the dexterous hand to be calibrated is generated, and each finger cavity is controlled to depressurize and reset.

[0018] Compared to existing technologies that primarily rely on traditional rigid contact calibration, including mechanical rigid indenter point-press calibration, weight-based pressure calibration, and single-point pneumatic pressure calibration, this new method employs a sealed cavity for precise positioning and clamping, combined with a fully enclosed flexible airbag for adaptive curved surface pressure equalization. It utilizes multi-level gradient air pressure steady-state control and simultaneous data acquisition across all channels, along with independent quadratic polynomial fitting and closed-loop error verification for each channel. This addresses issues such as the inability of existing rigid calibration methods to adapt to irregularly shaped curved surface sensors in dexterous hands, numerous blind spots in calibration, uneven pressure distribution on sensors, susceptibility to damage to flexible sensing structures, large calibration errors, low pressure control accuracy, poor consistency in multi-channel calibration, and extremely low overall calibration efficiency. This achieves standardized and precise calibration of the bionic robot's dexterous hand with synchronized five fingers, no blind spots across the entire curved surface, and no stress or damage. It effectively unifies the calibration benchmarks for all sensing channels, reduces the dispersion of multi-channel data, and improves the calibration accuracy and mass production calibration efficiency of flexible curved surface tactile sensors.

[0019] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the structure of a dexterous hand tactile calibration device provided in an embodiment of this application; Figure 2 A flowchart illustrating a method for calibrating the tactile sense of a bionic robot's dexterous hand, as provided in an embodiment of this application. Figure 3 A schematic diagram of the structure of a calibration device for the tactile sense of a bionic robot dexterous hand provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. Based on the embodiments of this application, every other embodiment obtained by those skilled in the art without inventive effort falls within the scope of protection of this application.

[0023] To achieve intelligent operation functions such as human-like precise grasping, object slip detection, material texture recognition, and contact force feedback, biomimetic robot dexterous hands are generally equipped with flexible thin-film tactile sensors and flexible piezoresistive tactile sensors. These sensors are fabricated using flexible polymer substrates and flexible conductive materials, possessing advantages such as softness, bendability, conformability to irregular curved surfaces, lightweight design, and high integration. They can be tightly fitted and fixed to the irregular curved surfaces of each finger of the dexterous hand. Currently, mainstream high-fidelity tactile dexterous hands collect contact pressure, contact deformation, and force distribution information at various points on the curved surfaces of the fingers through a globally distributed sensor array, thereby completing precise force control and intelligent sensing operations.

[0024] Compared to traditional rigid pressure sensors, flexible curved surface tactile sensors are made of soft materials, have thin films, and are prone to elastic deformation. However, they also have inherent characteristics such as nonlinear output, large individual channel differences, significant temperature drift, and poor consistency. Therefore, they must rely on professional calibration equipment to complete pressure calibration in order to ensure the accuracy of tactile perception. At the same time, since the fingers of dexterous hands have biomimetic irregular curved surface structures, the sensor bends and covers the surface of the fingers with the substrate, resulting in complex curved surface structures such as arc surfaces, side surfaces, and interfinal transition surfaces. This places extremely high demands on the calibration loading method, loading uniformity, and contact fit. Therefore, for this type of flexible curved surface array sensor, it is necessary to ensure uniform force, full-area pressure on the curved surface, and precise and controllable pressure during the calibration process in order to obtain a reliable calibration dataset.

[0025] Research has revealed that traditional manual calibration processes, such as alignment, pressure adjustment, state determination, and data recording, are prone to random errors. Differences in human operation can cause significant fluctuations and poor consistency in sensor channel and batch data, weakening sensor stability. Currently, the industry primarily uses traditional rigid contact calibration methods for dexterous hand tactile sensors, including mechanical rigid indenter point-press calibration, weight-based pressure calibration, and single-point pneumatic pressure calibration. These traditional calibration methods have many unavoidable technical drawbacks: Poor surface adaptability results in numerous calibration blind spots. Traditional indenters are mostly planar or rigid point-like structures, which can only apply pressure to local planar areas of the finger and cannot conform to the curved surface of the finger. The sides and concave transition areas of the sensor surface are in a suspended, pressureless state for a long time, making it impossible to achieve synchronous calibration of the entire channel.

[0026] Rigid pressure heads can easily damage flexible sensors. The small contact area of ​​rigid pressure heads leads to localized stress concentration during pressure application, which can cause wrinkles, material delamination, and conductive layer misalignment in flexible sensing films. Repeated calibration over a long period can cause irreversible plastic deformation of the sensor, resulting in problems such as channel drift, sensitivity decay, and channel failure, significantly reducing sensor yield and lifespan.

[0027] The pressure control accuracy is low and the calibration error is large. Traditional weights and mechanical indenters rely on mechanical displacement to apply pressure. Affected by material deformation and mechanical return clearance, the pressure fluctuates greatly and cannot output a stable and controllable standard pressure. This makes it difficult to meet the data acquisition requirements of multi-level static calibration of flexible sensors, and the fitting calibration curve has large error and poor linearity.

[0028] The calibration efficiency is extremely low and the channel consistency is poor. Existing technologies generally adopt a single-finger, single-point successive calibration method. For a single finger with multiple sensing channels, repeated alignment and multiple pressure applications are required. The calibration of five fingers with multiple sensing channels is extremely time-consuming. At the same time, the loading posture and contact position cannot be kept consistent each time, which introduces alignment errors and results in inconsistent calibration benchmarks for each channel. The poor consistency between channels significantly increases the difficulty of subsequent haptic fusion algorithms. Traditional calibration methods take a long time to calibrate all channels at once.

[0029] Based on this, this application provides a calibration method for the tactile sense of a bionic robot's dexterous hand. By employing a sealed cavity for precise positioning and clamping, and a fully enclosed flexible airbag for adaptive curved surface pressure equalization and bonding, multi-level gradient air pressure steady-state control, and synchronous data acquisition across all channels, this method addresses the problems of existing rigid calibration methods, such as inability to adapt to irregular curved surface sensors of dexterous hands, numerous blind spots in calibration, uneven pressure distribution on sensors, easy damage to flexible sensing structures, large calibration errors, low pressure control accuracy, poor consistency of multi-channel calibration, and extremely low global calibration efficiency. This method achieves standardized and accurate calibration of the five fingers of the bionic robot's dexterous hand simultaneously, with no blind spots across the entire curved surface, and without stress or damage. It effectively unifies the calibration benchmarks of all sensing channels, reduces the dispersion of multi-channel data, and improves the calibration accuracy and mass production calibration efficiency of flexible curved surface tactile sensors.

[0030] Please see Figure 1 , Figure 1 This is a schematic diagram of the structure of a dexterous hand tactile calibration device provided in an embodiment of this application. Figure 1 As shown, the dexterous hand tactile calibration device 10 includes a sealed calibration chamber 110, a fully enclosed curved flexible airbag module 120, a pneumatic pressure stabilization control unit 130, a multi-channel data acquisition unit 140, and a bionic robot dexterous hand tactile calibration device 300 that applies the bionic robot dexterous hand tactile calibration method provided in the embodiments of this application. Each component or unit in the device is connected and interacts through air paths, circuits, and data links to jointly achieve full-range, non-destructive, and high-precision automated calibration of the bionic dexterous hand flexible curved tactile sensor.

[0031] Here, the five independent output air paths of the air pressure stabilization control unit 130 pass through the side wall sealing interface of the sealed calibration cavity 110 and are respectively connected to multiple independent finger-covering cavities of the fully covered curved flexible airbag module 120; the calibration device 300 of the bionic robot dexterous hand sends control commands to the air pressure stabilization control unit 130 to regulate the start and stop of the air pump, the valve opening degree and the pressure threshold. The internal pressure sensor acquires the air pressure data of the air path in real time and feeds it back to the control circuit board to form a closed-loop pressure stabilization regulation, automatically corrects the pressure deviation, and ensures that the pressure inside the airbag is stable within the target accuracy range.

[0032] Furthermore, the flexible curved tactile sensor of the dexterous hand to be calibrated is led out of the sealed calibration cavity through a shielded ribbon cable and connected to the corresponding acquisition channel of the multi-channel data acquisition unit 140; the multi-channel data acquisition unit 140 communicates with the calibration device 300 through a data bus and uploads the acquired analog-to-digital conversion values ​​of all channels to the calibration device 300 in real time for subsequent zero-point calibration, curve fitting and error verification.

[0033] The sealed calibration cavity 110 contains the fully enclosed curved flexible airbag module 120. The sealed calibration cavity 110 is a rigid cylindrical structure. The top of the sealed calibration cavity 110 is provided with a positioning structure for positioning the dexterous hand to be calibrated. The sealed calibration cavity 110 also contains a sealing ring and a limiting seat (not shown in the figure) for fixing the palm part of the dexterous hand to be calibrated.

[0034] Here, the sealed calibration chamber 110 constructs a closed and stable calibration work space, enabling precise positioning and fixation of the dexterous hand to be calibrated, ensuring that the five fingers of the dexterous hand hang naturally and maintain a naturally curved biomimetic shape, avoiding initial deformation errors caused by external force torsion of the flexible sensor; at the same time, it ensures the airtightness of the internal air circuit environment, for example, the overall airtightness of the chamber under calibration conditions is ≤0.02kPa / min, providing a sealed basis for the pressure stabilization and pressurization of the airbag.

[0035] The fully enclosed curved flexible airbag module 120 is provided with multiple independent finger-enclosing cavities. The finger-enclosing cavities are made of highly elastic and highly ductile thermoplastic polyurethane elastomer (TPU) or silicone flexible material. When the finger-enclosing cavities are inflated, they adaptively conform to the irregular outer curved surface of each finger of the dexterous hand to be calibrated, so that the flexible curved tactile sensor on the surface of each finger is uniformly pressed and there are no stress concentration points.

[0036] Here, the fully enclosed curved flexible airbag module 120 is an integrated flexible airbag structure. The inside of the airbag is divided into multiple independent finger enclosed cavities through a hot-pressing separation process. Each finger enclosed cavity corresponds to the five fingers of the dexterous hand to be calibrated. Each finger enclosed cavity is equipped with an independent air vent. The cavities are completely isolated from each other and there is no cross-flow of air.

[0037] In this embodiment, the fully enclosed curved flexible airbag module 120 serves as a flexible pressurization execution medium. When inflated, it can adaptively conform to the irregular outer curved surface of the fingers of a dexterous hand (including the fingertip arc surface, the finger side curved surface, and the interdigital transition recess area), achieving 360° full-curved surface enclosed pressurization. It relies on the uniform internal air pressure to apply surface pressure to the sensor surface, with no rigid contact or local stress concentration throughout the process, ensuring that all sensing areas on the finger surface are uniformly pressurized, and completely eliminating the pressure blind zone present in traditional rigid calibration.

[0038] The pneumatic pressure stabilizing control unit 130 includes a miniature air pump, a precision pressure regulating valve, a pressure sensor, a pressure stabilizing air tank, and an electromagnetic control valve group (not shown in the figure). It also includes a dedicated control circuit board. The air circuits are connected in sequence in the order of "miniature air pump, pressure stabilizing air tank, precision pressure regulating valve, and electromagnetic control valve group", and finally output five independent and controllable air circuits.

[0039] Here, the fully enclosed curved flexible airbag module 120 is provided with precise and stable multi-level air pressure power. For example, the controllable pressure range covers 0 to 1000 kPa, and the pressure stabilization accuracy can reach ±0.1 kPa. It can receive instructions from the calibration device 300 and automatically complete the entire process of air pressure regulation, such as gradient pressure increase, steady-state pressure holding, and slow pressure release. It can maintain a constant pressure for a long time, ensure the deformation stability of the flexible sensor, and meet the pressure accuracy requirements of static calibration.

[0040] The multi-channel data acquisition unit 140 includes multiple acquisition channels, each of which corresponds to a sensing channel configured for each flexible curved surface tactile sensor.

[0041] For example, a single finger corresponds to 1 to 64 acquisition channels, and the five fingers together support a maximum of 320 channels for simultaneous acquisition; the acquisition board of the multi-channel data acquisition unit 140 is electrically connected to each flexible curved surface tactile sensor of the dexterous hand through a shielded ribbon cable, and has the characteristics of low noise and high sampling rate, and the sampling rate can be adjusted in the range of 1 to 500Hz.

[0042] Here, during the steady-state stage of the pressure corresponding to each calibration pressure value, the original analog-to-digital (AD) electrical signals output by each sensing channel of all flexible curved surface tactile sensors are simultaneously collected, and uploaded to the calibration device 300 after completing the analog-to-digital conversion; this ensures that the time reference and pressure reference of all channel data acquisition are completely consistent, eliminating the reference error caused by time-division acquisition.

[0043] The calibration device 300 serves as the control center and data processing core of the dexterous hand tactile calibration device 10. It sends air pressure regulation commands to the air pressure stabilization control unit 130 and receives sensor data uploaded by the multi-channel data acquisition unit 140. It performs single-channel independent polynomial fitting, error closed-loop verification, and calibration parameter summary calculation, and finally generates a standard tactile calibration file, realizing automated control and data processing of the entire calibration process.

[0044] In this embodiment, when the dexterous hand tactile calibration device 10 performs calibration on the dexterous hand to be calibrated, the dexterous hand to be calibrated is fixed to the positioning structure at the top of the sealed calibration cavity 110, and the five fingers are respectively inserted into the corresponding finger-covering cavities of the airbag module 120; the calibration device 300 coordinates the entire process: first, it controls the air pressure stabilization control unit 130 to perform zero-point calibration, and then it pressurizes and stabilizes the pressure step by step according to the preset calibration pressure gradient sequence. At each steady state, the multi-channel data acquisition unit 140 is synchronously triggered to collect the analog-to-digital conversion value corresponding to the sensor signal; after the collected data is transmitted back to the calibration device 300, single-channel independent fitting, error verification and calibration file generation are completed; after the calibration is completed, the air path is slowly depressurized and reset, and finally the fully automated calibration process is completed.

[0045] An example of the overall assembly and debugging of a dexterous hand tactile calibration device 10: 1. Fabricate and prepare a sealed calibration chamber 110. The calibration chamber is made of aviation aluminum alloy and is cylindrical in shape. The height of the chamber is 100~280mm and the inner diameter of the cylinder is 60~100mm. A palm mounting and positioning port with a diameter of 30~80mm is opened at the top of the chamber. A nitrile rubber sealing ring and a nylon limiting seat are installed to ensure that the palm of the dexterous hand is fixed without shaking, and the air tightness of the seal is ≤0.02kPa / min.

[0046] 2. Preparation of the fully enclosed curved flexible airbag module 120: A highly transparent TPU flexible film with a thickness of about 0.2~0.4mm is selected and heat-sealed to form an integrated airbag structure; the airbag is divided into five independent finger-shaped enclosed cavities by heat-pressing partitioning process, with an inner diameter of 10~20mm for each cavity and an independent ventilation channel for each cavity; the maximum inflation ratio of the airbag is 1~2 times, and the maximum pressure resistance can reach about 3MPa.

[0047] 3. Construct a pneumatic pressure stabilization and control unit 130: equipped with a miniature DC air pump, a 100~200mL pressure-stabilizing gas tank, a high-precision electro-proportional valve, and a five-way independent solenoid control valve group; the pressure sensor has a measurement accuracy of ±0.1kPa, a pressure sampling frequency of 1~50Hz, and a controllable pressure range of 0~1000kPa. Connect the air pump, pressure-stabilizing gas tank, pressure regulating valve, and solenoid valve sequentially through PU gas tubing to ensure that the air pressure of each gas path is independently controllable.

[0048] 4. Equipped with a multi-channel data acquisition unit 140: It adopts a 1~320 channel synchronous AD acquisition board, with 1~64 acquisition channels assigned to a single finger, and the sampling rate can be set from 1~50Hz; the acquisition board is electrically connected to the dexterous hand tactile sensor through a shielded ribbon cable to reduce signal noise interference during the high voltage calibration process.

[0049] 5. Perform assembly calibration and steady-state pressure test on the dexterous hand tactile calibration device 10: Fix the five-finger flexible dexterous hand to the limiting seat at the top of the cavity, let the five fingers hang naturally and place them into the independent encapsulation cavity of the airbag respectively; control the air path to pressurize step by step, set the pressure gradient to 0kPa, 20kPa, 40kPa, 60kPa, 80kPa, 100kPa, and maintain each pressure level for 5~10s to ensure that the deformation of the flexible curved surface sensor reaches a steady state; during the pressure stabilization stage, record the real-time AD raw data of all channels simultaneously to form a pressure-AD corresponding dataset.

[0050] 6. Data Fitting and Calibration File Generation: The host computer collects and stores multi-level steady-state datasets, and uses polynomial fitting to fit the data of each sensing channel to solve for the offset coefficient, sensitivity coefficient, and nonlinear correction coefficient of each channel; it summarizes all channel calibration parameters and generates a calibration file in the corresponding format, which can be directly imported into the dexterous hand tactile analysis control system to complete parameter writing.

[0051] Please see Figure 2 , Figure 2 This is a flowchart illustrating a method for calibrating the tactile sense of a biomimetic robot's dexterous hand, as provided in an embodiment of this application. Figure 2 As shown in the embodiments of this application, the calibration method for the tactile sense of a bionic robot's dexterous hand is typically applied to, for example... Figure 1 In the dexterity hand tactile calibration device 10 shown, the calibration method includes: S101. The dexterous hand of the target bionic robot to be calibrated is fixed on the positioning structure of the sealed calibration cavity, and each finger of the dexterous hand to be calibrated is controlled to extend into each of the finger-covering cavities in the fully covered curved flexible airbag module, so that the flexible curved tactile sensor on the surface of each finger is in contact with the inner wall of the airbag of the finger-covering cavity.

[0052] Here, the dexterous hand to be calibrated is a bionic robotic five-fingered dexterous hand equipped with a flexible curved surface array tactile sensor, which is the calibration object of this application embodiment. It has irregularly shaped curved finger surfaces, and each finger is equipped with multiple independent sensing channels. The flexible curved surface tactile sensor is a flexible piezoresistive or thin-film sensing array that covers the surface of the dexterous hand's fingers. It is made of soft and bendable material and is used to sense the contact pressure and deformation signal of the finger curved surface. It has inherent characteristics such as nonlinear output and individual channel differences.

[0053] In this step, before officially starting the calibration operation, the equipment is first tested under no-load conditions to confirm that the air pressure system and the data acquisition system are communicating normally and without any faults. Then, the dexterous hand of the target bionic robot to be calibrated is placed steadily on the positioning structure of the limiting card seat at the top of the sealed calibration cavity. The limiting slot of the card seat completes the horizontal and vertical bidirectional positioning. The sealing ring is slowly tightened to achieve complete fixation of the palm part and sealing of the cavity port, preventing problems such as palm shaking and cavity leakage during the calibration process.

[0054] Furthermore, after fixation, the overall posture of the dexterous hand is precisely adjusted to ensure that the five fingers remain in a naturally relaxed, biomimetic hanging state, without deliberate bending or external pulling. This allows each finger to smoothly and steadily extend into the five independent finger-covering cavities within the fully enclosed curved flexible airbag module. Finally, fine-tuning of the posture is performed, adjusting the fit between each finger and its corresponding airbag one by one, thoroughly smoothing the flexible walls of the airbag. This ensures that the flexible curved tactile sensor covering the entire surface of each finger can completely and tightly fit against the inner wall of the airbag within the finger's covering cavity, avoiding abnormal states such as sensor twisting, airbag wrinkles, and gaps where the finger is partially suspended. Ultimately, this completes the high-precision, stress-free, and accurate positioning operation at the calibration station.

[0055] This achieves stress-free fixation of the dexterous hand, maintaining the biomimetic shape of the naturally curved fingers and avoiding initial deformation errors caused by external force twisting of the sensor. It also ensures that the sensor is uniformly pressurized across its entire area during subsequent airbag inflatation, completely eliminating calibration blind spots. The dexterous hand's palm is fixed without wobbling, the cavity's airtightness meets standards, and the curved surfaces of the fingers completely conform to and cover the airbag, providing the foundation for uniform calibration across the entire area and preventing calibration failures caused by localized suspension.

[0056] S102. Zero-point calibration is performed on the multi-channel data acquisition unit, and the air pressure stabilization control unit is controlled to pressurize each finger covering cavity step by step according to the preset calibration pressure gradient sequence. In the pressure steady state stage corresponding to each calibration pressure value in the calibration pressure gradient sequence, the multi-channel data acquisition unit is used to acquire the analog-to-digital conversion value corresponding to each sensing channel configured by each flexible curved surface tactile sensor.

[0057] In this embodiment, zero-point calibration is the initial calibration process. By collecting the original electrical signal of the sensor under no-pressure conditions, calibration errors caused by residual pressure of the equipment, clamping stress, and initial offset of the sensor are eliminated, and the calibration benchmark of all sensing channels is unified.

[0058] Here, the calibration pressure gradient sequence is a preset multi-level stepped standard pressure array, which serves as the pressure input reference calibrated in this application embodiment. For example, the calibration pressure gradient sequence includes gradient pressure value nodes of 0 kPa, 10 kPa, 20 kPa, 40 kPa, 60 kPa, 80 kPa, and 100 kPa. The pressure steady-state stage is the stage after the air pressure stabilization control unit pressurizes to the calibration pressure value and then continuously stabilizes the pressure for a preset stabilization time, so that the airbag pressure and the deformation of the flexible sensor are completely stable and without fluctuations, which is the effective stage for data acquisition.

[0059] Among them, the analog-to-digital conversion value (AD value) is the digital signal obtained by analog-to-digital conversion of the analog electrical signal output by the flexible tactile sensor after deformation under force, which is the core raw data for calibration and fitting.

[0060] In the embodiments of this application, the sensing channel is an independent sensing unit of the flexible curved surface tactile sensor, and each channel independently outputs a pressure sensing signal; for example, a single finger is configured with 1 to 64 sensing channels, and the five fingers together have a maximum of 320 independent channels.

[0061] In this step, after completing the dexterous hand clamping and positioning and the equipment self-test reset, the first step is to start the full-domain zero-point calibration program of the multi-channel data acquisition unit to eliminate various system errors such as finger contact stress, residual trace air pressure inside the air pipeline, and initial parameter offset of the flexible sensor itself generated during the assembly process, and to unify the initial reference of all sensing channels.

[0062] Furthermore, after the zero-point calibration process is completed and the verification is qualified, the preset calibration pressure gradient sequence execution command is issued, and the air pressure stabilization control unit starts the gradient pressurization operation. According to the preset parameters, the five independent finger covering cavities are synchronously and uniformly pressurized step by step. The air paths of the five fingers are synchronously regulated and the pressure rise and fall are completely synchronized. Here, each pressurization process adopts a uniform and slow rise mode to avoid the impact of sudden air pressure rise.

[0063] Furthermore, when the real-time pressure reported by the air pressure monitoring sensor reaches the target calibrated pressure value, the system automatically locks the air pressure output parameters and continuously stabilizes the pressure for a preset stabilization time (e.g., 5 to 10 seconds), giving the airbag deformation and sensor elastic deformation sufficient buffer stabilization time. Once the internal pressure of the airbag is completely balanced, the deformation state of the flexible sensor is completely stable, and the pressure value is stable, the data acquisition command is immediately triggered.

[0064] Furthermore, by utilizing the synchronous sampling characteristics of the multi-channel data acquisition unit, all flexible curved surface tactile sensors and the real-time analog-to-digital conversion values ​​corresponding to each independent sensing channel are acquired simultaneously. At the same time, the acquired global data is bound to the standard calibration pressure value of the current stage, classified and stored to form standardized calibration sample data. The pressurization, stabilization, acquisition, and storage process is executed in a step-by-step loop until the data acquisition operation of all gradient pressures is completed.

[0065] This approach unifies the zero-pressure reference across all sensor channels, eliminating initial systematic errors. Through multi-stage precise steady-state pressurization, a complete sample dataset corresponding to pressure and analog-to-digital conversion (AD) signals is constructed, providing sufficient, accurate, and fluctuation-free raw data for subsequent curve fitting. It avoids systematic errors caused by manual clamping and initial equipment conditions; each pressure level exhibits high stabilization accuracy with minimal pressure fluctuations, stable sensor deformation, synchronous data acquisition across all channels, and a completely unified pressure reference across all channels with minimal dispersion between channels.

[0066] In one possible implementation of this application, the step of zero-point calibration of the multi-channel data acquisition unit in step S102 may include: S1021. Control the fully enclosed curved flexible airbag module to completely depressurize, and control the flexible curved tactile sensor to be in a static, pressureless state.

[0067] In this step, at the initial stage of the single calibration process, the equipment reset and depressurization self-test logic is executed first. A complete depressurization reset command is issued to the air pressure stabilization control unit. Simultaneously, all five electromagnetic control valves corresponding to the five finger-covering cavities are opened to keep the air passage unobstructed. The residual compressed gas left in the airbag and air passage is completely and quickly discharged, so that the fully covered curved flexible airbag module is completely retracted and reset, and the overall pressure returns to the normal pressure standard state.

[0068] Furthermore, after the pressure is released, the system is continuously kept still for a preset pressure stabilization period to monitor its stability. The system collects air pressure values ​​in real time to confirm that there is no residual pressure rise and no passive deformation of the airbag. This ensures that the flexible curved tactile sensor inside the airbag is completely free from any external pressure, restraint, or pre-tightening, and is in a stress-free, naturally relaxed, and absolutely pressure-free static reference state. This provides a pure, interference-free, and absolutely zero-pressure reference environment for subsequent zero-point raw data acquisition.

[0069] S1022. In response to the fact that the fully enclosed curved flexible airbag module has been completely depressurized and the flexible curved tactile sensor is in a static and pressureless state, the original analog-to-digital conversion initial value corresponding to each sensing channel configured for each flexible curved tactile sensor is acquired by the multi-channel data acquisition unit.

[0070] Here, the original analog-to-digital conversion initial value is the original analog-to-digital conversion signal collected by the flexible curved surface tactile sensor under pressureless and static conditions. It has not undergone any correction and is the original data source of the zero-point reference value.

[0071] In this step, after the system completes static pressure monitoring and confirms that the airbag is completely depressurized and the flexible curved tactile sensor is in a stable and pressure-free state, the multi-channel data acquisition unit immediately triggers the full-domain synchronous acquisition command to start the high-speed synchronous sampling operation.

[0072] Furthermore, the hardware channel addresses of all flexible curved surface tactile sensors of the five fingers are retrieved simultaneously, the acquisition function of all sensing channels is activated at once, the raw electrical signals output in real time by each sensing channel are collected synchronously, and the onboard analog-to-digital conversion module completes the real-time conversion to accurately obtain the original analog-to-digital conversion initial values ​​corresponding to each channel.

[0073] Furthermore, the entire domain's raw data is cached in real time, categorized and stored, and channel-bound, fully recording and retaining all uncorrected raw sampling data to ensure that the initial zero-voltage characteristic data of each channel can be fully traced.

[0074] S1023. The original analog-to-digital conversion initial value is determined as the zero-point reference value for each of the sensing channels to complete the zero-point calibration of the multi-channel data acquisition unit.

[0075] Here, the zero-point reference value is the zero-pressure reference parameter of each channel determined by the original analog-to-digital conversion initial value, which is used to offset the initial offset of the sensor and the clamping stress error.

[0076] In this step, after the initial value of the original analog-to-digital conversion is collected and stored, the channel number is automatically matched, and the exclusive initial value of the original analog-to-digital conversion collected by each sensing channel is bound and set as the exclusive zero-point reference value of that independent sensing channel, so as to realize the independent zero-point assignment of a single channel.

[0077] Furthermore, after the assignment is completed, a globally unified real-time zero-point correction rule is generated and embedded into the underlying logic of the acquisition program. This ensures that all raw analog-to-digital conversion values ​​acquired during subsequent calibration processes must be subtracted in real-time from the corresponding channel's specific zero-point reference value to achieve dynamic zero-point compensation and completely eliminate initial errors. The zero-point parameters are then solidified, and the correction rule takes effect.

[0078] In one possible embodiment of this application, in specific implementation, step S102, during the steady-state stage corresponding to each calibrated pressure value in the calibrated pressure gradient sequence, may include: The step of acquiring the analog-to-digital conversion value corresponding to each sensing channel of each flexible curved surface tactile sensor using the multi-channel data acquisition unit during this stage may include: S1024. In the steady-state stage of the pressure corresponding to each calibrated pressure value in the calibrated pressure gradient sequence, for each sensing channel configured by each flexible curved surface tactile sensor, the original analog-to-digital conversion value corresponding to the sensing channel is acquired by the multi-channel data acquisition unit.

[0079] In this step, after each target calibration pressure value in the calibration pressure gradient sequence has been pressurized and entered the pressure steady state stage, the air pressure fluctuation status is continuously monitored. Once the pressure fluctuation is confirmed to be stable within ±0.1 kPa and the sensor deformation is completely stable and unchanged, the full-domain synchronous data acquisition process is immediately started.

[0080] Furthermore, through the high-speed synchronous sampling function of the multi-channel data acquisition unit, synchronous signal acquisition is carried out without delay, omission, and time difference for each bionic finger, each flexible curved surface tactile sensor deployed on the finger surface, and each independent sensing channel corresponding to the sensor. The original analog-to-digital conversion value output by each channel under the current standard steady-state pressure is captured in real time, and the real-time response data of the entire channel is completely recorded, ensuring that the time reference and pressure reference of all channel data are completely consistent.

[0081] S1025. Subtract the original analog-to-digital conversion value from the zero-point reference value of the sensing channel to obtain the analog-to-digital conversion value corresponding to the sensing channel.

[0082] In this step, after the raw analog-to-digital conversion value is collected across the entire domain, the dedicated zero-point reference value for each sensing channel stored in the database is retrieved in real time, and differential calibration calculation is performed for each channel. The raw analog-to-digital conversion value of the channel collected under the current steady-state pressure is compared with the pre-calibrated and fixed zero-point reference value for precise differential calculation. Various systematic fixed errors such as sensor initial zero-point offset, residual clamping stress, residual pressure in the equipment air path, and circuit baseline drift are eliminated in real time. Finally, the accurate and effective analog-to-digital conversion value after eliminating all interference is calculated.

[0083] Here, the values ​​after zero-point difference correction serve as the only valid sample data for subsequent polynomial curve fitting operations, ensuring the purity and accuracy of the fitted data.

[0084] S103. For each of the sensing channels, perform a mapping curve fitting between the analog-to-digital conversion value and the corresponding calibration pressure value for each of the calibrated pressure values ​​to obtain the calibration coefficient result for that sensing channel.

[0085] Here, the mapping curve fitting is the process of establishing a correspondence between the standard pressure value and the sensor's analog-to-digital conversion value through mathematical algorithms. This is used to correct the sensor's nonlinear error and construct an accurate pressure-signal mapping model.

[0086] In this embodiment of the application, the calibration coefficient results are a complete set of parameters obtained after single-channel fitting, including quadratic term coefficients, linear term coefficients, constant term coefficients, equivalent sensitivity coefficients, and zero-point offset coefficients.

[0087] In this step, after all the global gradient sample data acquisition and storage are completed, the single-channel independent fitting operation logic is started to perform data analysis and modeling operations on each independent sensing channel one by one; the dedicated dataset of a single sensing channel is retrieved, and the accurate analog-to-digital conversion value corresponding to the channel under all calibration pressure gradient nodes after zero-point calibration is accurately extracted, as well as the standard calibration pressure value that matches it. Scattered abnormal noise data that appeared during the acquisition process are removed, and the effective sample data is retained to form a sample dataset for single-channel dedicated fitting.

[0088] Furthermore, based on this sample dataset, the least squares algorithm is used to perform quadratic polynomial matrix iterative operations. By continuously minimizing the sum of squared sample errors, a unique pressure-to-analog conversion value mapping calibration curve that perfectly matches the output characteristics of the channel is accurately fitted and generated. Simultaneously, the complete set of calibration coefficients for the channel is calculated and obtained. The nonlinear error correction and unique characteristic modeling of all sensing channels are completed channel by channel, completely avoiding the problems of poor channel adaptability and low fitting accuracy caused by the traditional unified fitting method.

[0089] This approach, by specifically adapting to the individual nonlinear differences of each sensor channel, abandons the crude method of batch calibration using a uniform formula, accurately corrects single-channel output deviations, and significantly improves single-channel calibration accuracy and multi-channel consistency. Single-channel fitting nonlinearity errors are smaller, precisely matching the inherent output characteristics of each sensor unit, completely solving the problems of large channel differences and high data dispersion in traditional calibration.

[0090] In one possible implementation of this application, step S103 may include: S1031. For each of the sensing channels, based on the analog-to-digital conversion value and the corresponding calibration pressure value of the sensing channel under each calibration pressure value, the least squares method is used to perform matrix operations on the preset quadratic polynomial formula to obtain the quadratic term coefficient, the linear term coefficient and the constant term coefficient corresponding to the sensing channel.

[0091] Here, the least squares method is a core algorithm for curve fitting. It minimizes the sum of squared sample errors through matrix operations, solves for the optimal calibration coefficients, and ensures the accuracy and applicability of the fitted curve.

[0092] In this step, for each independent sensing channel, the effective analog-to-digital conversion value of that channel after zero-point calibration under all gradient calibration pressures, as well as the standard calibration pressure value that is precisely matched with it, are automatically summarized and organized. Abnormal acquisition points and fluctuation noise are eliminated to build a sample dataset with accurate and reliable single-channel dedicated fitting.

[0093] Furthermore, the sample dataset is substituted into the preset quadratic polynomial fitting formula, and high-precision matrix iterative calculation is carried out using the least squares method core algorithm. With the minimum sum of squared sample errors as the optimal solution objective, the parameters are continuously iterated and optimized. Finally, the quadratic term coefficient, linear term coefficient, and constant term coefficient that uniquely fit the characteristics of the channel are accurately solved, thus completing the solution of the basic parameters for single-channel mathematical modeling.

[0094] In this embodiment of the application, the expression of the preset quadratic polynomial formula is as follows.

[0095] .

[0096] in, To calibrate the pressure value, This is the analog-to-digital conversion value. , , These are the coefficients of the quadratic term, the linear term, and the constant term, respectively.

[0097] S1032. Based on the quadratic term coefficient, the linear term coefficient, the constant term coefficient, the analog-to-digital conversion value and the corresponding calibration pressure value of the sensing channel under each calibration pressure value, and the quadratic polynomial formula, fit the calibration curve corresponding to the sensing channel, and verify the error of the calibration curve to obtain the curve verification result.

[0098] Here, the calibration curve is a single-channel pressure-signal mapping curve obtained by fitting multi-level calibration pressure values ​​and analog-to-digital conversion values, which uniquely corresponds to the output characteristics of a single sensing channel.

[0099] In this step, the quadratic, linear, and constant coefficients of the single channel obtained by iterative solution are completely substituted into the quadratic polynomial formula. Combined with all gradient sample data of the channel, a calibration curve that uniquely represents the mapping between the calibration pressure value and the analog-to-digital conversion value of the sensing channel is accurately fitted and generated, thus completely replicating the nonlinear pressure response characteristics of the channel.

[0100] For example, by using the analog-to-digital conversion value on the horizontal axis and the calibrated pressure value on the vertical axis, and substituting the coefficients of the quadratic term, the linear term, and the constant term into the quadratic polynomial formula, the calibration curve fitted by the sensing channel is generated.

[0101] Furthermore, after the curve fitting is completed, a global error verification mechanism is activated to retrieve the verification pressure node data independent of the fitted samples. This allows for a comprehensive verification of the global adaptability, fitting accuracy, and generalization ability of the calibration curve. It also precisely identifies and eliminates anomalies such as local fitting deviations, endpoint distortion, overfitting, and underfitting. Finally, the curve verification results are output to accurately determine whether the calibration curve meets the accuracy requirements of actual applications.

[0102] In one possible implementation of this application, in specific implementation, the step of performing error verification on the calibration curve in step S1032 to obtain the curve verification result may include: S10321. Control the air pressure stabilization control unit to pressurize the finger covering cavity according to the preset intermediate verification pressure value, and in the pressure steady state stage corresponding to the intermediate verification pressure value, use the multi-channel data acquisition unit to acquire the analog-to-digital conversion sampling value corresponding to the sensing channel.

[0103] Here, the intermediate verification pressure value is a verification pressure node independent of the calibration pressure gradient sequence, used to verify the accuracy of the calibration curve. For example, the intermediate verification pressure value can be set to 5 kPa, 15 kPa, or 25 kPa. The analog-to-digital conversion sampling value is the effective sampling data of the analog-to-digital conversion signal of the sensing channel collected during the verification stage after zero-point calibration.

[0104] In this step, to verify the generalization ability and actual accuracy of the calibration curve and avoid the problem of sample overfitting, the intermediate verification pressure value that did not participate in the curve fitting calculation is selected as the verification benchmark.

[0105] Furthermore, a precise pressure adjustment command is issued to control the air pressure stabilization control unit to precisely increase the pressure to the target intermediate verification pressure value. Throughout the process, the pressure is increased at a uniform speed, and the pressure is accurately and continuously stabilized. After the air pressure, airbag deformation, and sensor output signal are completely stable without fluctuations, the multi-channel data acquisition unit is activated to accurately acquire the real-time analog-to-digital conversion sampling value corresponding to the current verification sensor channel, and simultaneously retrieves the zero-point reference value of the channel to complete the difference calibration, thus obtaining an effective analog-to-digital conversion sampling value.

[0106] S10322. Based on the analog-to-digital conversion sampling value, the test pressure value is determined using the calibration curve corresponding to the sensing channel, and based on the test pressure value, the intermediate verification pressure value, and the preset full-scale pressure value, the error result corresponding to the calibration curve corresponding to the sensing channel is determined.

[0107] Here, the test pressure value is the fitted pressure value calculated by substituting the verification sample value into the calibration curve formula; the full-scale pressure value is the maximum range pressure calibrated by the sensor. In this embodiment, the conventional full-scale pressure value can be set to 100 kPa; and the error result is the relative error between the test pressure value and the intermediate verification pressure value, which is used to determine whether the calibration curve is qualified.

[0108] In this step, the collected and calibrated analog-to-digital conversion sample values ​​are accurately substituted into the calibration curve formula that has been fitted and preliminarily verified for the channel. The fitted pressure value corresponding to the current sample value is obtained by accurately solving the formula, which is the test pressure value.

[0109] Furthermore, the preset standard intermediate verification pressure value and the sensor full scale are retrieved, and the relative full scale error between the test pressure value and the standard verification pressure value is accurately calculated according to the standardized error calculation formula. The accurate error result of the sensor channel calibration curve is obtained in a quantitative manner, providing accurate data basis for subsequent qualification judgment.

[0110] In the embodiments of this application, the expression for the standardized error calculation formula is as follows.

[0111] .

[0112] in, To test the pressure value, The intermediate pressure value is used for verification. This is the preset full-scale pressure value. This represents the error result.

[0113] S10323. Determine whether the error result is less than a preset error threshold.

[0114] Here, the preset error threshold is the criterion for determining whether the calibration is qualified. In this embodiment, the preset error threshold can be 1%FS (full scale error).

[0115] In this step, the error result obtained from the current channel solution is accurately compared with the preset error threshold, and standardized hierarchical judgment logic is executed to unify the calibration accuracy judgment standard of all sensing channels, eliminate judgment deviations, and ensure the consistency of calibration accuracy of all channels.

[0116] S10324. If the error result is less than the preset error threshold, then the curve verification result corresponding to the calibration curve of the sensing channel is determined to be passed the error verification.

[0117] In this step, if the error result of the sensing channel is determined to be less than the preset error threshold after comparison, it indicates that the calibration curve fitting accuracy of the sensing channel meets the standard, the generalization ability is good, and the actual application error is controllable. The calibration parameters of the channel are marked as qualified, and the current curve verification result is determined to be passed the error verification. All calibration coefficients and calibration curves of the channel can be formally put into use for subsequent tactile pressure analysis.

[0118] S10325. If the error result is greater than or equal to the preset error threshold, then the curve verification result is determined to be a failure of error verification.

[0119] In this step, if the error result of the sensing channel is determined to be greater than or equal to the preset error threshold after comparison, it is determined that the accuracy of the currently fitted calibration curve is substandard, there is fitting distortion or sample abnormality, and the curve verification result is determined to be unsuccessful.

[0120] Furthermore, a closed-loop rework correction mechanism is triggered to accurately locate abnormal channels. The multi-level gradient pressure application, steady-state data acquisition, sample screening, and curve fitting calculation process are restarted for the channel separately. The fitting parameters are optimized a second time, and repeated iterations are performed until the error result of the channel meets the standard, thus completely eliminating the output of unqualified calibration parameters.

[0121] S1033. In response to the curve verification result indicating that the calibration curve corresponding to the sensing channel has passed the error verification, determine the instantaneous slope of the curve corresponding to the calibration curve, and determine the instantaneous slope of the curve corresponding to the analog-to-digital conversion value below the midpoint of the range of the calibration curve as the equivalent sensitivity coefficient corresponding to the sensing channel.

[0122] Here, the instantaneous slope of the curve is the slope of the tangent at any coordinate point of the calibration curve, which characterizes the real-time sensitivity of the sensor under the corresponding deformation state; the equivalent sensitivity coefficient is the instantaneous slope corresponding to the analog-to-digital conversion value at the midpoint of the calibration curve's range, which is used as a fixed sensitivity parameter of the sensing channel for pressure conversion.

[0123] In this step, if the curve verification result determines that the calibration curve of the sensing channel is properly fitted and the accuracy meets the standard, the sensitivity coefficient calculation logic is initiated. Based on the fitted standard calibration curve, the instantaneous tangent slope of all coordinate points of the curve is calculated across the entire domain, and the curve sensitivity fluctuation range is comprehensively statistically analyzed. Subsequently, the midpoint of the standard calibration range of the sensor is precisely located, the analog-to-digital conversion value corresponding to the midpoint of the range is extracted, and the precise instantaneous slope of the curve corresponding to the coordinate point is further obtained. This stable and precise instantaneous slope is fixed as the equivalent sensitivity coefficient of the sensing channel, which is used to quantify and characterize the pressure response sensitivity characteristics of the sensing channel, and serves as the core fixed parameter for subsequent pressure analysis and conversion.

[0124] In this embodiment of the application, the instantaneous slope of the calibration curve is determined by the following formula: .

[0125] in, To calibrate the instantaneous slope of the curve corresponding to the curve, This is the analog-to-digital conversion value. and These are the coefficients of the quadratic term and the coefficients of the linear term, respectively.

[0126] S1034. Based on the zero-point reference value of the sensing channel, the zero-point offset coefficient corresponding to the sensing channel is determined using the calibration curve corresponding to the sensing channel.

[0127] Here, the zero-point offset coefficient is a channel zero-pressure offset parameter obtained by solving the zero-point reference value and calibration curve, and is used to correct the sensor zero-point drift error.

[0128] In this step, the zero-point reference value of the sensor channel in the previous calibration is retrieved, the analog-to-digital conversion numerical coordinates corresponding to the zero-point reference value are located, the zero-pressure coordinate value is substituted into the calibration curve of the channel, and the theoretical pressure offset value of the channel under zero-pressure state is calculated through reverse derivation.

[0129] Furthermore, the pressure offset value obtained from the reverse solution is fixed as the zero-point offset coefficient of the channel. This is used to quantify the inherent and unavoidable zero-point drift and initial offset error of the sensor, providing core parameter support for dynamic zero-point compensation and precise pressure correction when the dexterous robot is powered on.

[0130] S104. Based on the calibration coefficient results corresponding to each of the sensing channels, generate a tactile calibration file corresponding to the dexterous hand to be calibrated, and control each finger covering cavity to depressurize and reset.

[0131] Here, the tactile calibration file is a standardized file generated by summarizing the calibration coefficients of all sensor channels. It is divided into a readable debugging format and a microcontroller programming format, which is used by the dexterous hand control system to analyze tactile pressure signals in real time.

[0132] For example, the tactile calibration file can be saved in two forms: one for use by the host computer, where the calibration file is archived and backed up on the host computer's hard drive as a .csv or .txt file for debugging purposes, providing high readability and allowing manual viewing of calibration parameters; and another for distribution, where a binary .bin file is generated and burned onto the smart hand microcontroller acquisition board, enabling the tactile decoding program to read parameters upon power-up and complete the real-time conversion between analog-to-digital conversion values ​​and pressure.

[0133] In this step, after all five-finger sensing channels have completed independent curve fitting, full-domain error verification, and calibration accuracy has met the standards, with no unqualified channels remaining, the calibration file generation program is started. All stored channel data is traversed, and each sensing channel's unique calibration coefficients, unique channel number, sensor calibration range, fitting error parameters, and other complete set of valid information are summarized. Following a preset standardized file format and data arrangement rules, a standardized tactile calibration file adapted for reading, writing, and parsing by the dexterous hand control system is generated.

[0134] Furthermore, after the calibration files are generated, archived, and backed up, a slow pressure release command is issued to precisely control the air pressure stabilization control unit to gradually reduce the output air pressure of the five air paths, causing all finger-covering cavities to depressurize and retract at a uniform and slow speed, thus eliminating the airflow impact and deformation rebound stress caused by sudden pressure drops throughout the process.

[0135] Furthermore, after the airbag has fully returned to its initial relaxed normal pressure state and the sensor deformation has fully rebounded and reset, the fixed limit of the cavity sealing ring and the dexterous hand limit holder is released in sequence. The calibrated dexterous hand is then taken out smoothly and vertically upwards. At the same time, the temporary cache data of the equipment is automatically cleared, completing the entire fully automatic calibration operation process. The equipment then enters standby mode and can directly carry out the next batch of calibration operations.

[0136] This standardizes and documents discrete single-channel calibration parameters to meet the read / write requirements of the dexterous hand control system; slow pressure release avoids sudden pressure changes that could damage the flexible sensor, ensuring the reusability of the equipment and workpiece. Calibration files can be directly burned and used, the entire process is automated without human intervention, and after calibration, the sensor is free of wrinkles, delamination, and permanent plastic deformation, achieving non-destructive calibration.

[0137] As an example of the method described in this application: Clamping and positioning: Fix the dexterous hand, and place each of the five fingers into an independent airbag cavity, ensuring a snug fit without wrinkles or twisting; Zero-point calibration: Completely depressurize the airbag, collect the original AD values ​​of all channels at zero pressure, and set the zero-point reference; Gradient pressurization: Increment the pressure stepwise at 0 kPa, 20 kPa, 40 kPa, 60 kPa, 80 kPa, and 100 kPa, stabilizing the pressure for 8 seconds at each level; Synchronous acquisition: Synchronously acquire 320 channels of AD values ​​under steady-state conditions at each level, and generate a sample dataset after zero-point correction; Independent fitting: Fit a quadratic polynomial curve using the least squares method for each channel to solve for the complete set of calibration coefficients; Error verification: Use independent points at 5 kPa, 15 kPa, and 25 kPa for verification, and automatically rework if the error exceeds 1%FS; File generation: Summarize the qualified coefficients and generate a readable TXT / CSV file and a BIN burning file; Depressurization and reset: Slowly depressurize, remove the dexterous hand, and complete the calibration.

[0138] The bionic robot dexterous hand tactile calibration method provided in this application adopts a sealed cavity for precise positioning and clamping, and a fully enclosed flexible airbag for adaptive curved surface pressure equalization and bonding. It uses multi-level gradient air pressure steady-state control and synchronous data acquisition across all channels, and performs independent quadratic polynomial fitting and closed-loop error verification for each channel. This solves problems such as the inability of existing rigid calibration methods to adapt to irregular curved surface sensors in dexterous hands, the existence of numerous blind zones in calibration, uneven pressure distribution on sensors, easy damage to flexible sensing structures, large calibration errors, low pressure control accuracy, poor consistency in multi-channel calibration, and extremely low global calibration efficiency. It achieves standardized and precise calibration of the five fingers of the bionic robot dexterous hand simultaneously, with no blind zones across the entire curved surface, and without stress or damage. It effectively unifies the calibration benchmarks of all sensing channels, reduces the dispersion of multi-channel data, and improves the calibration accuracy and mass production calibration efficiency of flexible curved surface tactile sensors.

[0139] Please see Figure 3 , Figure 3 This is a schematic diagram of a calibration device for the tactile sense of a biomimetic robot's dexterous hand, provided as an embodiment of this application. Figure 3 As shown, the calibration device 300 includes: The clamping and positioning control module 310 is used to control the dexterous hand of the target bionic robot to be calibrated to be fixed on the positioning structure of the sealed calibration cavity, and to control each finger of the dexterous hand to be calibrated to be inserted into each finger cavity of the fully covered curved flexible airbag module, so that the flexible curved tactile sensor on the surface of each finger is in contact with the inner wall of the airbag of the finger cavity. The calibration data acquisition module 320 is used to perform zero-point calibration on the multi-channel data acquisition unit and control the air pressure stabilization control unit to pressurize each finger covering cavity step by step according to the preset calibration pressure gradient sequence. In the pressure steady state stage corresponding to each calibration pressure value in the calibration pressure gradient sequence, the multi-channel data acquisition unit is used to acquire the analog-to-digital conversion value corresponding to each sensing channel of each flexible curved surface tactile sensor. The mapping curve fitting module 330 is used to perform mapping curve fitting on the analog-to-digital conversion value and the corresponding calibration pressure value for each of the sensing channels under each calibration pressure value, so as to obtain the calibration coefficient result corresponding to the sensing channel. The calibration file generation module 340 is used to generate a tactile calibration file corresponding to the dexterous hand to be calibrated based on the calibration coefficient results corresponding to each of the sensing channels, and to control the pressure relief and reset of each finger covering cavity.

[0140] Furthermore, when the calibration data acquisition module 320 is used to perform zero-point calibration on the multi-channel data acquisition unit, the calibration data acquisition module 320 is used for: The fully enclosed curved flexible airbag module is completely depressurized, and the flexible curved tactile sensor is kept in a static, pressure-free state. In response to the fact that the fully enclosed curved flexible airbag module has been completely depressurized and the flexible curved tactile sensor is in a static and pressureless state, the original analog-to-digital conversion initial value corresponding to each sensing channel configured for each of the flexible curved tactile sensors is acquired by the multi-channel data acquisition unit. The original analog-to-digital conversion initial value is determined as the zero-point reference value for each of the sensing channels to complete the zero-point calibration of the multi-channel data acquisition unit.

[0141] Furthermore, when the calibration data acquisition module 320 acquires the analog-to-digital conversion values ​​corresponding to each sensing channel of each flexible curved surface tactile sensor during the steady-state stage corresponding to each calibrated pressure value in the calibrated pressure gradient sequence, the calibration data acquisition module 320 is used to: In the steady-state stage of pressure corresponding to each calibrated pressure value in the calibrated pressure gradient sequence, the original analog-to-digital conversion value corresponding to each sensing channel configured for each of the flexible curved surface tactile sensors is acquired by the multi-channel data acquisition unit. The difference between the original analog-to-digital conversion value and the zero-point reference value of the sensing channel is used to obtain the analog-to-digital conversion value corresponding to the sensing channel.

[0142] Furthermore, the calibration coefficient results include quadratic term coefficients, linear term coefficients, constant term coefficients, equivalent sensitivity coefficients, and zero-point offset coefficients; when the mapping curve fitting module 330 performs mapping curve fitting on the analog-to-digital conversion value corresponding to each of the sensing channels at each calibration pressure value to obtain the calibration coefficient results corresponding to that sensing channel, the mapping curve fitting module 330 is used to: For each of the aforementioned sensing channels, based on the analog-to-digital conversion value and the corresponding calibration pressure value of the sensing channel at each calibration pressure value, the least squares method is used to perform matrix operations on the preset quadratic polynomial formula to obtain the quadratic term coefficients, the linear term coefficients, and the constant term coefficients corresponding to the sensing channel. Based on the quadratic term coefficient, the linear term coefficient, the constant term coefficient, the analog-to-digital conversion value and the corresponding calibration pressure value of the sensing channel under each calibration pressure value, and the quadratic polynomial formula, the calibration curve corresponding to the sensing channel is fitted, and the calibration curve is verified for error to obtain the curve verification result. In response to the curve verification result indicating that the calibration curve corresponding to the sensing channel has passed the error verification, the instantaneous slope of the curve corresponding to the calibration curve is determined, and the instantaneous slope of the curve corresponding to the analog-to-digital conversion value below the midpoint of the range of the calibration curve is determined as the equivalent sensitivity coefficient corresponding to the sensing channel. Based on the zero-point reference value of the sensing channel, the zero-point offset coefficient corresponding to the sensing channel is determined using the calibration curve corresponding to the sensing channel.

[0143] Furthermore, when the mapping curve fitting module 330 performs error verification on the calibration curve to obtain the curve verification result, the mapping curve fitting module 330 is used to: The air pressure stabilization control unit pressurizes the finger covering cavity according to a preset intermediate verification pressure value, and during the pressure steady-state stage corresponding to the intermediate verification pressure value, the multi-channel data acquisition unit acquires the analog-to-digital conversion sampling value corresponding to the sensing channel. Based on the analog-to-digital conversion sampled value, the test pressure value is determined using the calibration curve corresponding to the sensing channel, and based on the test pressure value, the intermediate verification pressure value, and the preset full-scale pressure value, the error result corresponding to the calibration curve corresponding to the sensing channel is determined. Determine whether the error result is less than a preset error threshold; If the error result is less than the preset error threshold, then the curve verification result corresponding to the calibration curve of the sensing channel is determined to be passed the error verification. If the error result is greater than or equal to the preset error threshold, then the curve verification result is determined to be a failure of error verification.

[0144] The bionic robot dexterous hand tactile calibration device provided in this application embodiment uses a sealed cavity for precise positioning and clamping, and a fully enclosed flexible airbag for adaptive curved surface pressure equalization and bonding. It employs multi-level gradient air pressure steady-state control and synchronous data acquisition across all channels, performing independent quadratic polynomial fitting and closed-loop error verification for each channel. This solves problems such as the inability of existing rigid calibration methods to adapt to the irregular curved surface sensors of dexterous hands, the existence of numerous blind zones in calibration, uneven pressure distribution on sensors, easy damage to flexible sensing structures, large calibration errors, low pressure control accuracy, poor consistency in multi-channel calibration, and extremely low global calibration efficiency. It achieves standardized and precise calibration of the five fingers of the bionic robot dexterous hand simultaneously, with no blind zones across the entire curved surface, and without stress or damage. It effectively unifies the calibration benchmarks of all sensing channels, reduces the dispersion of multi-channel data, and improves the calibration accuracy and mass production calibration efficiency of flexible curved surface tactile sensors.

[0145] Please see Figure 4 , Figure 4This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 4 As shown, the electronic device 400 includes a processor 410, a memory 420, and a bus 430.

[0146] The memory 420 stores machine-readable instructions executable by the processor 410. When the electronic device 400 is running, the processor 410 and the memory 420 communicate via the bus 430. When the machine-readable instructions are executed by the processor 410, they can perform the operations described above. Figure 2 The steps of the bionic robot dexterity hand tactile calibration method in the illustrated method embodiment can be found in the method embodiment for specific implementation, and will not be repeated here.

[0147] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described actions. Figure 2 The steps of the bionic robot dexterity hand tactile calibration method in the illustrated method embodiment can be found in the method embodiment for specific implementation, and will not be repeated here.

[0148] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0149] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0150] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0151] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0152] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0153] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The scope of protection of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for calibrating the tactile sense of a bionic robot's dexterous hand, characterized in that, The calibration method is applied to a dexterous hand tactile calibration device, which includes a sealed calibration cavity, a fully enclosed curved flexible airbag module, a pressure stabilization and control unit, and a multi-channel data acquisition unit. The fully enclosed curved flexible airbag module is fixed inside the sealed calibration cavity, and the fully enclosed curved flexible airbag module is provided with multiple independent finger enclosing cavities. The calibration method includes: The dexterous hand of the target bionic robot to be calibrated is fixed on the positioning structure of the sealed calibration cavity, and each finger of the dexterous hand to be calibrated is controlled to extend into each finger cavity of the fully covered curved flexible airbag module, so that the flexible curved tactile sensor on the surface of each finger is in contact with the inner wall of the airbag of the finger cavity. The multi-channel data acquisition unit is zero-point calibrated, and the air pressure stabilization control unit is controlled to pressurize each finger covering cavity step by step according to a preset calibration pressure gradient sequence. In the pressure steady state stage corresponding to each calibration pressure value in the calibration pressure gradient sequence, the multi-channel data acquisition unit is used to acquire the analog-to-digital conversion value corresponding to each sensing channel of each flexible curved surface tactile sensor. For each of the aforementioned sensing channels, a mapping curve is fitted between the analog-to-digital conversion value corresponding to each of the aforementioned calibrated pressure values ​​and the corresponding calibrated pressure values ​​to obtain the calibration coefficient result corresponding to the sensing channel. Based on the calibration coefficient results corresponding to each of the sensor channels, a tactile calibration file corresponding to the dexterous hand to be calibrated is generated, and the pressure relief and reset of each finger cavity is controlled.

2. The method according to claim 1, characterized in that, The zero-point calibration of the multi-channel data acquisition unit includes: The fully enclosed curved flexible airbag module is completely depressurized, and the flexible curved tactile sensor is kept in a static, pressure-free state. In response to the fact that the fully enclosed curved flexible airbag module has been completely depressurized and the flexible curved tactile sensor is in a static and pressureless state, the original analog-to-digital conversion initial value corresponding to each sensing channel configured for each of the flexible curved tactile sensors is acquired by the multi-channel data acquisition unit. The original analog-to-digital conversion initial value is determined as the zero-point reference value for each of the sensing channels to complete the zero-point calibration of the multi-channel data acquisition unit.

3. The method according to claim 2, characterized in that, In the steady-state stage corresponding to each calibrated pressure value in the calibrated pressure gradient sequence, the multi-channel data acquisition unit acquires the analog-to-digital conversion values ​​corresponding to each sensing channel configured for each flexible curved surface tactile sensor, including: In the steady-state stage of pressure corresponding to each calibrated pressure value in the calibrated pressure gradient sequence, the original analog-to-digital conversion value corresponding to each sensing channel configured for each of the flexible curved surface tactile sensors is acquired by the multi-channel data acquisition unit. The difference between the original analog-to-digital conversion value and the zero-point reference value of the sensing channel is used to obtain the analog-to-digital conversion value corresponding to the sensing channel.

4. The method according to claim 1, characterized in that, The calibration coefficient results include quadratic term coefficients, linear term coefficients, constant term coefficients, equivalent sensitivity coefficients, and zero-point offset coefficients; for each of the aforementioned sensing channels, the process involves performing a mapping curve fitting between the analog-to-digital conversion value corresponding to each of the calibrated pressure values ​​and the corresponding calibrated pressure values ​​to obtain the calibration coefficient results for that sensing channel, including: For each of the aforementioned sensing channels, based on the analog-to-digital conversion value and the corresponding calibration pressure value of the sensing channel at each calibration pressure value, the least squares method is used to perform matrix operations on the preset quadratic polynomial formula to obtain the quadratic term coefficients, the linear term coefficients, and the constant term coefficients corresponding to the sensing channel. Based on the quadratic term coefficient, the linear term coefficient, the constant term coefficient, the analog-to-digital conversion value and the corresponding calibration pressure value of the sensing channel under each calibration pressure value, and the quadratic polynomial formula, the calibration curve corresponding to the sensing channel is fitted, and the calibration curve is verified for error to obtain the curve verification result. In response to the curve verification result indicating that the calibration curve corresponding to the sensing channel has passed the error verification, the instantaneous slope of the curve corresponding to the calibration curve is determined, and the instantaneous slope of the curve corresponding to the analog-to-digital conversion value below the midpoint of the range of the calibration curve is determined as the equivalent sensitivity coefficient corresponding to the sensing channel. Based on the zero-point reference value of the sensing channel, the zero-point offset coefficient corresponding to the sensing channel is determined using the calibration curve corresponding to the sensing channel.

5. The method according to claim 4, characterized in that, The step of verifying the error of the calibration curve to obtain the curve verification result includes: The air pressure stabilization control unit pressurizes the finger covering cavity according to a preset intermediate verification pressure value, and during the pressure steady-state stage corresponding to the intermediate verification pressure value, the multi-channel data acquisition unit acquires the analog-to-digital conversion sampling value corresponding to the sensing channel. Based on the analog-to-digital conversion sampled value, the test pressure value is determined using the calibration curve corresponding to the sensing channel, and based on the test pressure value, the intermediate verification pressure value, and the preset full-scale pressure value, the error result corresponding to the calibration curve corresponding to the sensing channel is determined. Determine whether the error result is less than a preset error threshold; If the error result is less than the preset error threshold, then the curve verification result corresponding to the calibration curve of the sensing channel is determined to be passed the error verification. If the error result is greater than or equal to the preset error threshold, then the curve verification result is determined to be a failure of error verification.

6. A calibration device for the tactile sensation of a bionic robot's dexterous hand, characterized in that, The calibration device includes: The clamping and positioning control module is used to control the dexterous hand of the target bionic robot to be calibrated to be fixed on the positioning structure of the sealed calibration cavity, and to control each finger of the dexterous hand to be calibrated to be inserted into each finger cavity of the fully covered curved flexible airbag module, so that the flexible curved tactile sensor on the surface of each finger is in contact with the inner wall of the airbag of the finger cavity. The calibration data acquisition module is used to perform zero-point calibration on the multi-channel data acquisition unit and control the air pressure stabilization control unit to pressurize each finger covering cavity step by step according to the preset calibration pressure gradient sequence. In the pressure steady state stage corresponding to each calibration pressure value in the calibration pressure gradient sequence, the multi-channel data acquisition unit is used to acquire the analog-to-digital conversion value corresponding to each sensing channel of each flexible curved surface tactile sensor. The mapping curve fitting module is used to perform mapping curve fitting on the analog-to-digital conversion value and the corresponding calibration pressure value for each of the sensing channels under each calibration pressure value, so as to obtain the calibration coefficient result corresponding to the sensing channel. The calibration file generation module is used to generate a tactile calibration file corresponding to the dexterous hand to be calibrated based on the calibration coefficient results corresponding to each of the sensing channels, and to control the depressurization and reset of each finger covering cavity.

7. A dexterous hand tactile calibration device, characterized in that, The dexterous hand tactile calibration device includes a sealed calibration cavity, a fully enclosed curved flexible airbag module, a pneumatic pressure stabilization control unit, a multi-channel data acquisition unit, and a bionic robot dexterous hand tactile calibration device applying the tactile calibration method for bionic robot dexterous hand as described in any one of claims 1 to 5; the fully enclosed curved flexible airbag module is fixed inside the sealed calibration cavity; the fully enclosed curved flexible airbag module is provided with multiple independent finger enclosing cavities; the multi-channel data acquisition unit includes multiple acquisition channels, each acquisition channel corresponding to each sensing channel configured for each flexible curved tactile sensor.

8. The dexterous hand tactile calibration device according to claim 7, characterized in that, The finger-covering cavity is made of highly elastic and highly ductile thermoplastic polyurethane elastomer or silicone flexible material. When the finger-covering cavity is inflated, it adaptively conforms to the irregular outer curved surface of each finger of the dexterous hand to be calibrated, so that the finger-covering cavity completely wraps the curved surface of each finger, so that the flexible curved tactile sensor on the surface of each finger is uniformly pressed and there are no stress concentration points.

9. The dexterous hand tactile calibration device according to claim 7, characterized in that, The sealed calibration chamber is a rigid cylindrical structure. The top of the sealed calibration chamber is provided with a positioning structure for positioning the dexterous hand to be calibrated. The sealed calibration chamber is also provided with a sealing ring and a limiting seat. The air pressure stabilization and control unit includes a miniature air pump, a precision pressure regulating valve, a pressure sensor, a pressure stabilizing air tank, and an electromagnetic control valve group.

10. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. The machine-readable instructions are executed by the processor to perform the steps of the calibration method for the tactile sense of a bionic robot dexterous hand as described in any one of claims 1 to 5.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the calibration method for the tactile sense of a bionic robot dexterous hand as described in any one of claims 1 to 5.

Images