A flexible sensor, a calibration method and a contact curved surface normal direction measurement method

Abstract

The application discloses a flexible sensor, a calibration method and a contact curved surface normal direction measurement method, and relates to the technical field of sensors.The flexible sensor comprises a flexible substrate and a strain gauge, the strain gauge is uniformly embedded on the flexible substrate in a circular array mode, and a data acquisition circuit measures the output resistance change of the strain gauge to obtain force / torque information at the contact position.The calibration method of the flexible sensor utilizes a neural network to fit the relationship between the six-dimensional force borne by the center of the hemispherical bottom surface of the flexible substrate and the resistance of the strain gauge, so that the calibration of the flexible sensor is realized;the mapping relationship between the force and the deformation of the contact point is obtained by combining finite element simulation and the neural network, and then the normal direction at the contact position is calculated by using the measured force / torque information, so that the normal perception of the unknown curved surface is realized.The application can closely adhere to the contact surface, conform to the uneven contact curved surface, realize curved surface matching and alignment, and measure the normal direction of the contact curved surface.

Description

Technical Field

[0001] This invention relates to the field of sensor technology, and in particular to a flexible sensor, a calibration method, and a method for measuring the normal direction of a contact surface. Background Technology

[0002] Currently, most existing wall-climbing robots can only climb flat surfaces or surfaces with fixed curvature. Research on wall-climbing robots adapting to varying curvature surfaces is limited. Most studies passively adapt to surface changes through structural means. For example, wall-climbing robots using magnetic adsorption can utilize track structures to achieve a certain degree of adaptability to surface changes. However, passively adapting to surfaces through mechanisms lacks predictive and active control capabilities, and the range of curvature changes that can be adapted to is relatively small.

[0003] Currently, non-contact sensors such as vision sensors and proximity sensors are widely used in various fields. These non-contact sensors achieve surface sensing capabilities through feedback control. For example, Guangdong University of Technology proposed a method combining an ultrasonic sensor array and an adsorption cavity array to actively sense the relative pose of the adsorption cavity and the surface to be adsorbed, thereby realizing adsorption and transfer motion between different surfaces. However, these sensors are difficult to integrate and suffer from visual obstruction, have high requirements for the lighting conditions of the application environment, and this method cannot achieve real-time sensing of the surface normal.

[0004] In addition, the use of flexible contact sensors that can sense force / torque to obtain the normal information of the contact surface is a major research hotspot. However, most flexible tactile sensors studied by researchers can usually only sense three-dimensional force information. How to realize six-dimensional force sensing of flexible contact sensors and then use this information to calculate the normal direction of the contact surface is a major research challenge. Summary of the Invention

[0005] This invention provides a flexible sensor, a calibration method, and a method for measuring the normal direction of a contact surface, in order to solve the technical problem in the prior art that contact sensors cannot adapt to the curvature of a surface and cannot perceive the normal direction of the contact surface in real time.

[0006] To address the above problems, the present invention provides a flexible sensor, which includes a flexible substrate and a strain gauge.

[0007] The strain gauges are uniformly distributed in a circular array on the bottom plane of the flexible substrate, and adjacent strain gauges are in different orientations to achieve a response to deformation in all directions of the flexible sensor.

[0008] It also includes a data acquisition circuit electrically connected to the strain gauge. When the flexible sensor is attached to different curved surfaces, the strain gauge bends with the curvature of the contact surface, and the resistance of the strain gauge changes. The data acquisition circuit is used to measure the resistance value R of the strain gauge of the flexible sensor. i .

[0009] Preferably, the flexible substrate is formed by casting a flexible material, and the strain gauge is embedded in the flexible substrate.

[0010] Preferably, the flexible substrate is configured as a hemispherical shape, and the diameter of the hemispherical shape is 48-52mm.

[0011] Preferably, the number of strain gauges is at least 6, and the height of the strain gauges from the hemispherical plane of the flexible substrate is 7-9 mm.

[0012] Preferably, the flexible substrate is cast from Ecoflex material.

[0013] A second objective of this invention is to provide a calibration method for a flexible sensor, employing the flexible sensor described above, wherein the calibration method includes:

[0014] Step S 100 Construct a neural network model with six inputs and six outputs;

[0015] The neural network model uses the resistance change output by the strain gauge of the flexible sensor as the input variable and the six-dimensional force at the center of the hemispherical bottom surface of the flexible substrate of the flexible sensor as the output variable.

[0016] Step S 200 :Calibrated sample dataset;

[0017] The resistance changes output by at least 1000 strain gauges and the six-dimensional force at the center of the hemispherical bottom surface of the flexible substrate are collected as corresponding input and output data to form a calibration sample dataset.

[0018] Step S 300 The neural network model is trained using the calibration sample dataset to determine the various parameters of the neural network, thereby obtaining a defined neural network model.

[0019] Step S 400 Establish the corresponding mapping relationship between the resistance change of the strain gauge and the six-dimensional force on the center of the hemispherical bottom surface of the flexible substrate, and complete the calibration of the flexible sensor.

[0020] Preferably, in step S 400In this context, establishing the mapping relationship between the resistance change of the strain gauge and the six-dimensional force acting on the center of the hemispherical bottom surface of the flexible substrate specifically includes:

[0021] Step S 401 Using a standard rigid six-dimensional force sensor as a reference sensor, the response output resistance of the strain gauge of the flexible sensor is obtained under a given force condition;

[0022] Step S 402 The six-dimensional force (F) acting at the center of the hemispherical bottom surface of the flexible substrate of the flexible sensor. x ,F y ,F z M x M y M z Using the resistance changes (R1, R2, R3, R4, R5, R6) output by the strain gauge of the flexible sensor as the input variables, different forces are applied to the flexible sensor to obtain at least 1000 sets of corresponding input variable data and output variable data as a sample training dataset.

[0023] Step S 403 A Cartesian coordinate system is established with the center of the hemispherical bottom surface of the flexible substrate of the flexible sensor as the origin. At least 1000 sets of corresponding input and output variable data are input into the neural network model for learning, thereby obtaining the six-dimensional force and strain gauge resistance R acting at the center of the hemispherical bottom surface of the flexible sensor. i Mapping relationship between them:

[0024] (F x ,F y ,F z M x M y M z )=f(R1,R2,R3,R4,R5,R6)

[0025] Wherein: F x ,F y ,F z Forces M in the x, y, and z axes respectively. x M y M z R1, R2, R3, R4, R5, and R6 represent the moments about the x, y, and z axes, respectively. R1, R2, R3, R4, R5, and R6 represent the resistance values ​​of the strain gauges under these forces. f() represents (F x ,F y ,F z M x M y M zThe mapping relationship between (R1,R2,R3,R4,R5,R6);

[0026] Step S 404 After calibrating the flexible sensor, the resistance of the strain gauge of the flexible sensor is measured, and the six-dimensional force at the center of the hemispherical bottom surface of the flexible sensor is predicted using the trained neural network model.

[0027] A third objective of this invention is to provide a method for measuring the normal direction of a contact surface, employing the flexible sensor described above. The measurement method includes:

[0028] Step T 100 When the flexible sensor comes into contact with the curved surface and deforms under force, the strain gauge resistance value R of the flexible sensor is measured by the data acquisition circuit. i ;

[0029] Step T 200 Based on the six-dimensional force acting at the center of the hemispherical bottom surface of the flexible sensor and the strain gauge resistance R... i The functional relationship (F) established between them x ,F y ,F z M x M y M z = f(R1,R2,R3,R4,R5,R6), and calculate the six-dimensional force information at the center of the hemispherical bottom surface of the flexible sensor;

[0030] Step T 300 Using the measured six-dimensional force information, the normal direction at the contact point of the contact surface is calculated.

[0031] Preferably, in step T 300 The specific calculation process for obtaining the normal direction at the contact point of the contact surface includes:

[0032] Step T 301 : Determine the resultant force acting on the hemispherical bottom surface of the flexible sensor along the normal direction and the coordinate r of the contact point H on the r-axis. H The calculation formula;

[0033] When the radius of curvature of the contact surface is greater than the radius of curvature of the flexible sensor, the contact surface is approximated by a plane α. Without applied force, it is plane α1; after applying force, the plane moves to α2. Without force, the contact point is P; after applying force F, the flexible sensor deforms, and the contact point moves to point H. Therefore:

[0034]

[0035]

[0036] r H =Rsinθ-d t cosθ=g(F x F y F z M x M y M z )

[0037] Among them, F z F is the resultant force acting on the hemispherical bottom surface of the flexible sensor in the direction normal to the surface. r For F x With F y The resultant force, r H Let H be the coordinates of point H on the r-axis, θ be the angle between the normal direction of the contact surface and the normal direction of the hemispherical bottom surface of the flexible sensor, and d be the coordinates of point H on the r-axis. n d represents the normal deformation of the flexible sensor along the contact surface. t Let E be the tangential deformation of the flexible sensor along the contact surface, and let E be the Young's modulus of the flexible sensor.

[0038] Step T 302 Using finite element simulation, different stress deformation conditions of the flexible sensor were simulated to obtain at least 1000 corresponding sets of six-dimensional forces (F) acting on the center of the hemispherical bottom surface of the flexible sensor. x F y F z M x M y M z The coordinates of the contact point H on the r-axis are r. H ;

[0039] Step T 303 Using at least 1000 sets of data as training sample datasets, a neural network is used to learn and fit the relationship:

[0040] r H =g(F x F y F z M x M y M z )

[0041] Step T 304 Using the measured six-dimensional force information, the angle θ between the normal direction of the contact surface and the normal direction of the hemispherical bottom surface of the flexible sensor is calculated, thereby realizing the measurement of the normal direction of the contact surface.

[0042] Preferably, in step T 302Specifically, the simulated deformation of the flexible sensor under different forces includes changing the degree of deformation d of the flexible sensor. n and d t .

[0043] Compared with the prior art, the present invention has significant advantages and beneficial effects, specifically reflected in the following aspects:

[0044] A flexible substrate is used to achieve a tight fit with the contact surface. An embedded strain gauge is used to sense the six-dimensional force at the contact point. The embedded strain gauge has a certain degree of compliance, which can adapt to the deformation of the flexible sensor when it is attached to the curved surface. The force / torque information at the contact point is obtained by measuring the change in its output resistance. This sensed information is then processed to obtain the normal information of the contact position. When the flexible sensor is attached to different curved surfaces, the strain gauge can bend with the curvature of the contact surface, and the resistance of the strain gauge changes accordingly.

[0045] By using a neural network to fit the relationship between the six-dimensional force at the center of the hemispherical bottom surface of the sensor and the resistance of the embedded strain gauge, the flexible sensor can be calibrated. Furthermore, by combining finite element simulation with the neural network to obtain the mapping relationship between the force and the deformation at the contact point, the normal direction at the contact point can be calculated using the measured force / torque information, thus enabling the sensing of the normal direction of an unknown curved surface.

[0046] The flexible sensor in this invention can fit closely to the contact surface and conform to the uneven contact surface. At the same time, it can realize real-time normal sensing of the contact surface and can be used to achieve surface matching and alignment, conforming to the curvature surface. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the mold required for the fabrication process of the flexible sensor in this embodiment of the invention;

[0048] Figure 2 This is a schematic diagram of a flexible six-dimensional force sensor with the ability to sense the direction of the curved surface normal in an embodiment of the present invention;

[0049] Figure 3 This is a schematic diagram illustrating the force exerted on the flexible sensor during the calculation process of the contact surface normal in an embodiment of the present invention.

[0050] Figure 4 This is a flowchart illustrating the calibration method for the flexible sensor in an embodiment of the present invention;

[0051] Figure 5 This is a flowchart illustrating the method for measuring the normal direction of the contact surface in an embodiment of the present invention.

[0052] Explanation of reference numerals in the attached figures:

[0053] 1-Flexible substrate; 2-Strain gauge. Detailed Implementation

[0054] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0055] like Figure 1-3 As shown, this embodiment of the invention provides a flexible sensor with the ability to sense the normal direction of a curved surface. The flexible sensor includes a flexible substrate 1, a strain gauge 2, and a data acquisition circuit, wherein:

[0056] The flexible substrate 1 is formed by casting a flexible material to ensure that the flexible sensor can passively conform to the contact surface and achieve a tight fit with the contact surface; at the same time, the flexible material can also protect the strain gauge 2.

[0057] The strain gauges 2 are uniformly embedded in the flexible substrate 1 in a circular array, and adjacent strain gauges 2 are in different orientations to achieve response to deformation in all directions of the flexible sensor.

[0058] The data acquisition circuit is electrically connected to strain gauge 2. When the flexible sensor is adsorbed onto different curved surfaces, strain gauge 2 can bend according to the curvature of the contact surface, and the resistance of strain gauge 2 changes. At this time, the data acquisition circuit is used to measure the resistance value R of strain gauge 2 of the flexible sensor. i .

[0059] Further explanation is needed; please refer to the following: Figure 1 As shown, before the flexible substrate 1 is cast, the strain gauge 2 is fixed in the corresponding position of the mold according to the designed layout. During the high-temperature solidification process of the flexible substrate 1 of the flexible sensor, the position of the strain gauge 2 is adjusted in time to ensure that it is in the ideal position, thereby realizing the embedding of the strain gauge 2.

[0060] Specifically, in an embodiment of the present invention, the flexible substrate 1 is formed by casting a flexible material, and the strain gauge 2 is embedded in the flexible substrate 1.

[0061] Specifically, please refer to Figure 1 As shown, in an embodiment of the present invention, the flexible substrate 1 is configured as a hemispherical shape, and the diameter of the hemispherical shape is 48-52 mm. As the most preferred embodiment of the present invention, the diameter of the hemispherical shape is 50 mm.

[0062] Therefore, the flexible sensor is a hemispherical shape with a diameter of 50mm. This ensures that the sensor fits tightly against any shape of contact surface, and the hemispherical shape can better transmit the external force on the flexible sensor to the embedded strain gauge 2, ensuring that it can sense the stress and deformation of the flexible sensor. This size can be adjusted according to the actual application.

[0063] It should be further noted that the materials used in the casting of the flexible substrate 1 include, but are not limited to, Ecoflex material. Ecoflex material is a fossil-based biodegradable and compostable plastic, which is an important basic raw material for many biodegradable and bio-based plastics.

[0064] Please see Figure 2 As shown, in an embodiment of the present invention, strain gauges 2 are embedded in a circular array on a flexible substrate 1, the number of strain gauges 2 is at least 6, and the height of the strain gauges 2 from the hemispherical plane of the flexible substrate 1 is 7-9 mm.

[0065] Therefore, the flexible sensor includes 6 or more strain gauges 2. As the best preferred embodiment of the present invention, the strain gauges 2 are selected as BHF2K-3AA model, thereby realizing the sensing of six-dimensional force.

[0066] In addition, in the embodiments of the present invention, the strain gauge 2 is about 8 mm away from the hemispherical plane of the flexible substrate 1, and adjacent strain gauges 2 are embedded in the flexible substrate 1 in different postures (parallel or perpendicular to the hemispherical plane of the flexible substrate 1), thereby ensuring that the flexible sensor can respond to the deformation caused by the force in each direction.

[0067] Please see Figure 4 As shown in the figure, this embodiment of the invention also provides a calibration method for a flexible sensor, using the flexible sensor described above. The calibration method includes:

[0068] Step S 100 Construct a neural network model with six inputs and six outputs;

[0069] The neural network model uses the resistance change output by the strain gauge 2 of the flexible sensor as the input variable and the six-dimensional force at the center of the hemispherical bottom surface of the flexible substrate 1 of the flexible sensor as the output variable.

[0070] Step S 200 :Calibrated sample dataset;

[0071] The resistance changes output by at least 1000 strain gauges 2 and the six-dimensional force at the center of the hemispherical bottom surface of the flexible substrate 1 are collected as corresponding input and output data to form a calibration sample dataset.

[0072] Step S 300 The neural network model is trained using the calibration sample dataset to determine the various parameters of the neural network, thereby obtaining a defined neural network model.

[0073] Step S 400Establish the corresponding mapping relationship between the resistance change of strain gauge 2 and the six-dimensional force on the center of the hemispherical bottom surface of flexible substrate 1, and complete the calibration of the flexible sensor.

[0074] Specifically, in the embodiments of the present invention, in step S 400 In this process, establishing the mapping relationship between the resistance change of strain gauge 2 and the six-dimensional force acting on the center of the hemispherical bottom surface of the flexible substrate 1 specifically includes:

[0075] Step S 401 Using a standard rigid six-dimensional force sensor as a reference sensor, the response output resistance of strain gauge 2 of the flexible sensor is obtained under a given force condition;

[0076] Step S 402 The six-dimensional force (F) acting at the center of the hemispherical bottom surface of the flexible substrate 1 of the flexible sensor. x ,F y ,F z M x M y M z Using the resistance changes (R1, R2, R3, R4, R5, R6) output by the strain gauge 2 of the flexible sensor as the input variables, different forces are applied to the flexible sensor to obtain at least 1000 sets of corresponding input variable data and output variable data as a sample training dataset.

[0077] Step S 403 A Cartesian coordinate system is established with the center of the hemispherical bottom surface of the flexible substrate 1 of the flexible sensor as the origin. At least 1000 sets of corresponding input and output variable data are input into the neural network model for learning, thereby obtaining the six-dimensional force and strain gauge resistance R at the center of the hemispherical bottom surface of the flexible sensor. i Mapping relationship between them:

[0078] (F x ,F y ,F z M x M y M z )=f(R1,R2,R3,R4,R5,R6)

[0079] Wherein: F x ,F y ,F z Forces M in the x, y, and z axes respectively. x M y M zR1, R2, R3, R4, R5, and R6 represent the moments about the x, y, and z axes, respectively. R1, R2, R3, R4, R5, and R6 represent the resistance values ​​of the strain gauges under these forces. f() represents (F x ,F y ,F z M x M y M z The mapping relationship between (R1,R2,R3,R4,R5,R6);

[0080] Step S 404 After calibrating the flexible sensor, the resistance of the strain gauge 2 of the flexible sensor is measured, and the six-dimensional force at the center of the hemispherical bottom surface of the flexible sensor is predicted using the trained neural network model.

[0081] It should be noted that in the embodiments of the present invention, a six-input, six-output neural network model is constructed. The neural network model includes, but is not limited to, the PSOBP neural network model. After the flexible sensor is calibrated by the above calibration method, the resistance of the strain gauge 2 of the flexible sensor is measured, and the six-dimensional force at the center of the hemispherical bottom surface of the flexible sensor can be predicted by the trained neural network model.

[0082] Please see Figure 5 As shown, an embodiment of the present invention also provides a method for measuring the normal direction of a contact surface, using the flexible sensor described above. The measurement method includes:

[0083] Step T 100 When the flexible sensor comes into contact with the curved surface and deforms under force, the strain gauge resistance value R of the flexible sensor is measured by the data acquisition circuit. i ;

[0084] Step T 200 Based on the six-dimensional force acting at the center of the hemispherical bottom surface of the flexible sensor and the strain gauge resistance R... i The functional relationship (F) established between them x ,F y ,F z M x M y M z = f(R1,R2,R3,R4,R5,R6), and calculate the six-dimensional force information at the center of the hemispherical bottom surface of the flexible sensor;

[0085] Step T 300 Using the measured six-dimensional force information, the normal direction at the contact point of the contact surface is calculated.

[0086] Therefore, the method for calculating the normal direction of the contact surface can be achieved by measuring the resistance value of the strain gauge 2 of the sensor through the data acquisition circuit, based on the established functional relationship (F). x ,F y ,F z M x M y M z The formula f(R1,R2,R3,R4,R5,R6) can be used to obtain the six-dimensional force at the center of the bottom surface of the flexible sensor hemisphere. Then, the normal direction at the contact point of the contact surface can be calculated using the measured six-dimensional force information.

[0087] The specific calculation process is as follows:

[0088] When the radius of curvature of the contact surface is greater than the radius of curvature of the flexible sensor (radius R), the contact surface is approximated as a plane (plane α1 when no force is applied, and the plane moves to α2 after force is applied). The contact point is P when no force is applied, and the contact point moves to point H after the flexible sensor deforms under force F. F can be decomposed into F1, F2, F3, F4, F5, F6, F7, F8, F9, F10, F11, F2, F1 ... x F y F z θ can be obtained by inverse solution using the following formula.

[0089]

[0090]

[0091] r H =Rsinθ-d t cosθ

[0092] Among them, F z F is the net force acting on the sensor hemispherical base in the direction normal to the bottom surface. r For F x With F y The resultant force (i.e., the net force on the horizontal surface, F) r The direction is the r-axis), r H Here, H is the coordinate value of point H on the r-axis; θ is the angle between the normal direction of the contact surface and the normal direction of the hemispherical bottom surface of the flexible sensor described in this invention; d n For the normal deformation of the flexible sensor along the contact surface (perpendicular to surfaces α1 and α2), d t E represents the tangential deformation (parallel to surfaces α1 and α2), and E is the Young's modulus of the flexible sensor.

[0093] Please see Figure 3 As shown in the embodiments of the present invention, finite element simulation is used to simulate the deformation of the flexible sensor under different forces.

[0094] Specifically, this means changing the degree of deformation d of the flexible sensor.n and d t By simulating different force conditions, a total of 1000 corresponding sets of six-dimensional forces (F) were obtained at the center of the bottom surface of the sensor hemisphere. x F y F z M x M y M z The coordinates of the contact point H on the r-axis are r. H .

[0095] Using these 1000 sets of data as the training sample dataset, and employing neural networks (including but not limited to backpropagation neural networks) to learn from them, a relational expression can be fitted:

[0096] r H =g(F x F y F z M x M y M z )

[0097] Through the above finite element simulation and neural network algorithm, fitting can be used to calculate the coordinates of point H on the r-axis using the six-dimensional force information at the center of the hemispherical bottom surface of the measured sensor.

[0098]

[0099]

[0100] r H =g(F x F y F z M x M y M z )=Rsinθ-d t cosθ

[0101] In summary, by utilizing the measured six-dimensional force information (converted to F...) z F r and r H By calculating the angle θ between the normal direction of the contact surface and the normal direction of the hemispherical bottom surface of the flexible sensor described in this invention, the normal direction of the contact surface can be measured, thus realizing the measurement of the normal direction of the contact surface.

[0102] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of this disclosure, and all such changes and modifications will fall within the scope of protection of this invention.

Claims

1. A method of measuring a direction of a normal to a contact surface, characterized by, The flexible sensor comprises a flexible substrate (1), strain gauges (2) and a data acquisition circuit electrically connected to the strain gauges (2), the strain gauges (2) are uniformly embedded on the flexible substrate (1) in a circular array, and adjacent strain gauges (2) are in different postures to realize the response to deformation of the flexible sensor in each direction. When the flexible sensor is adsorbed on different curved surfaces, the strain gauge (2) bends along the bending curvature of the contact curved surface, and the resistance of the strain gauge (2) changes, and the data acquisition circuit is used for measuring the resistance value of the strain gauge (2) of the flexible sensor ; The measurement method comprises: Step T 100 : When the flexible sensor is in contact with a curved surface and deforms under force, the strain gauge resistance value of the flexible sensor is measured by the data acquisition circuit ; Step T 200 : according to the function relationship between the six-dimensional force on the center of the hemisphere bottom surface of the flexible sensor and the resistance value of the strain sheet , the six-dimensional force information on the center of the hemisphere bottom surface of the flexible sensor is calculated.​ Step T 300 : Using the measured six-dimensional force information, the normal direction at the contact point of the contact surface is calculated; Step T 301 : Calculate the resultant force on the normal direction of the hemisphere bottom of the flexible sensor and the coordinate of the contact point H on the r-axis of the calculation formula; When the curvature radius of the contact surface is greater than the curvature radius of the flexible sensor, the contact surface adopts a plane α approximation, which is a plane when no force is applied , and moves to after the force is applied The contact point is P when no force is applied, the flexible sensor deforms after the force F is applied, the contact point moves to point H, and the following equation holds wherein, is the resultant force on the normal direction of the hemisphere bottom surface of the flexible sensor, is the resultant force on the normal direction of the hemisphere bottom surface of the flexible sensor, is the resultant force on the normal direction of the hemisphere bottom surface of the flexible sensor, is the resultant force on the normal direction of the hemisphere bottom surface of the flexible sensor, is the coordinate value of the point H on the r axis, is the angle between the normal direction of the contact surface and the normal direction of the hemisphere bottom surface of the flexible sensor, is the normal deformation of the flexible sensor along the contact surface, is the tangential deformation of the flexible sensor along the contact surface, and E is the Young's modulus of the flexible sensor. Step T 302 : using finite element simulation, simulating different force deformation conditions of the flexible sensor, obtaining at least 1000 groups of corresponding six-dimensional forces borne by the center of the hemisphere bottom surface of the flexible sensor and the coordinates of the contact point H on the r-axis ; Step T 303 : The at least 1000 sets of data are taken as training sample data sets, and a neural network is used to learn and fit a relational expression. Step T 304 : Calculate the included angle between the normal direction of the contact surface and the normal direction of the hemisphere bottom surface of the flexible sensor by using the measured six-dimensional force information , and realize the measurement of the normal direction of the contact surface.

2. The method of claim 1, wherein The flexible substrate (1) is formed by casting a flexible material.

3. The method of claim 2, wherein the contact surface normal direction is determined by: The flexible substrate (1) is provided in a hemispherical shape, and the diameter of the hemispherical shape is 48-52 mm.

4. The method of claim 1, wherein the contact surface normal direction is determined by: The number of strain gauges (2) is at least 6, and the height of the strain gauges (2) from the hemispherical plane of the flexible substrate (1) is 7-9 mm.

5. The method of claim 1-4, wherein the method further comprises: The flexible substrate (1) is cast from Ecoflex material.

6. The method of claim 1-4, wherein the method of measuring the direction of the tangent to the surface is characterized by, The calibration method of the flexible sensor comprises: Step S 100 : Constructing a six-input, six-output neural network model; The neural network model takes the resistance change output by the strain gauges (2) of the flexible sensor as the input variable, and takes the six-dimensional force on the center of the hemispherical bottom surface of the flexible substrate (1) as the output variable; Step S 200 : Calibration sample data set; At least 1000 groups of resistance changes output by the strain gauges (2) and six-dimensional forces on the center of the hemispherical bottom surface of the flexible substrate (1) are collected as corresponding input and output data to form a calibration sample data set; Step S 300 : training the neural network model by using the calibration sample data set, determining each parameter of the neural network to obtain a determined neural network model; Step S 400 : Establish the corresponding mapping relationship between the resistance change of the strain gauge (2) and the six-dimensional force on the center of the hemisphere bottom of the flexible substrate (1), and complete the calibration of the flexible sensor.

7. The method of claim 6, wherein the contact surface normal direction is determined by: In step S 400 The method further comprises the following steps of: establishing a corresponding mapping relationship between the resistance change of the strain gauge (2) and the six-dimensional force on the center of the hemisphere bottom surface of the flexible substrate (1). Step S 401 : Using a standard rigid six-dimensional force sensor as a control sensor, the response output resistance of the strain gauge (2) of the flexible sensor is obtained under a given force condition; Step S 402 : the six-dimensional force applied at the center of the hemispherical bottom surface of the flexible substrate (1) of the flexible sensor : the resistance change output by the strain gauge (2) of the flexible sensor as an output variable : as input variables, different forces are applied to the flexible sensor, and at least 1000 sets of corresponding input variable data and output variable data are obtained as sample training data sets; Step S 403 : A Cartesian coordinate system is established with the center of the hemisphere bottom surface of the flexible substrate (1) of the flexible sensor as the origin, and the at least 1000 sets of corresponding input variable data and output variable data are input into the neural network model for learning to obtain the mapping relationship between the six-dimensional force and the strain sheet resistance value at the center of the hemisphere bottom surface of the flexible sensor : Wherein: are forces in x, y, z axis direction respectively, are moments of force around x, y, z axis respectively, are resistance values of the strain gauges under the force condition respectively, f() represents the mapping relationship between and and ​ Step S 404 : After the calibration of the flexible sensor, the six-dimensional force at the center of the hemisphere bottom of the flexible sensor is obtained by measuring the resistance of the strain gauge (2) of the flexible sensor and using the trained neural network model for prediction.

8. The method of claim 1, wherein: At step T 302 Among them, simulating different stress deformation conditions of the flexible sensor specifically includes: changing the deformation degree of the flexible sensor And .

Images