CN111079333B - Deep learning sensing method of flexible touch sensor - Google Patents
Deep learning sensing method of flexible touch sensor Download PDFInfo
- Publication number
- CN111079333B CN111079333B CN201911314290.6A CN201911314290A CN111079333B CN 111079333 B CN111079333 B CN 111079333B CN 201911314290 A CN201911314290 A CN 201911314290A CN 111079333 B CN111079333 B CN 111079333B
- Authority
- CN
- China
- Prior art keywords
- sensor
- data set
- fusion
- deep learning
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000013135 deep learning Methods 0.000 title claims abstract description 18
- 238000000034 method Methods 0.000 title claims abstract description 15
- 230000004927 fusion Effects 0.000 claims abstract description 23
- 238000004088 simulation Methods 0.000 claims abstract description 18
- 230000007246 mechanism Effects 0.000 claims abstract description 17
- 238000005259 measurement Methods 0.000 claims abstract description 16
- 230000008447 perception Effects 0.000 claims abstract description 16
- 238000001514 detection method Methods 0.000 claims abstract description 9
- 239000000463 material Substances 0.000 claims description 21
- 230000006870 function Effects 0.000 claims description 19
- 238000012549 training Methods 0.000 claims description 7
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 230000006399 behavior Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 238000013139 quantization Methods 0.000 claims description 3
- 230000004044 response Effects 0.000 claims description 3
- 239000013598 vector Substances 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 2
- 238000013136 deep learning model Methods 0.000 abstract description 3
- 230000000704 physical effect Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003938 response to stress Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Force Measurement Appropriate To Specific Purposes (AREA)
Abstract
The invention discloses a deep learning sensing method of a flexible touch sensor, which comprises the following steps in sequence: establishing a sensor unit structure mechanical model; establishing a sensor array structure mechanical model; obtaining an actual measurement data set; obtaining a finite element simulation dataset; the data set fusion improves the data resolution; and combining deep learning to establish a perception mechanism model. According to the invention, through fusing the data sets, the relation between the pressure signal and the three-dimensional multi-scale geometrical dimension, the surface morphology, the physical properties and other structures of the detection object is obtained by means of a deep learning model.
Description
Technical Field
The invention relates to the field of sensing of flexible touch sensors, in particular to a deep learning sensing method of a flexible touch sensor.
Background
In the discrete numerical modeling of the perception mechanism, the numerical simulation precision based on the finite element depends on the size of the division of the finite element unit, and the contradiction between the calculation precision and the calculation efficiency can be brought. On the one hand, if a high-precision perception effect needs to be obtained, smaller and more finite element units are needed; on the other hand, more finite element units can bring the reduction of solving speed, and when the finite element units reach a certain scale, real-time numerical solving cannot be achieved by utilizing the finite element model of contact mechanics. At this point, a faster sensing mechanism approach is sought to achieve real-time prediction of sensor force signals and sensing models.
The perceptual mathematical model based on contact mechanics is a highly nonlinear model. In recent years, with the development of machine learning technology, deep learning attracts more and more attention with its excellent nonlinear fitting capability, and is widely used in semantic perception problems such as robot object recognition, object detection, and semantic segmentation. Furthermore, deep learning extends the perception based on two-dimensional images to three-dimensional space over some quantitative issues, such as object pose estimation, motion estimation, etc. The great success of these tasks demonstrates the ability to solve the quantitative estimation problem with deep learning. However, the current sensing methods such as CN106446948A only stay on the machine learning level, and the sensing mechanism research has not been performed by integrating the measured data, the finite element simulation data and the deep learning means, so as to improve the sensing precision and efficiency.
Disclosure of Invention
The invention aims to overcome the defects and shortcomings of the prior art and provides a deep learning sensing method of a flexible touch sensor, which is used for acquiring the relation between a pressure signal and structures such as three-dimensional multi-scale geometric dimensions, surface morphology, physical properties and the like of a detection object by means of a deep learning model through fusing the data sets.
The aim of the invention is achieved by the following technical scheme:
the deep learning sensing method of the flexible touch sensor comprises the following steps in sequence:
step 1), building a sensor unit structure mechanical model;
step 2), building a sensor array structure mechanical model;
step 3), obtaining an actual measurement data set;
step 4), obtaining a finite element simulation data set;
step 5), data set fusion improves data resolution;
and 6) combining deep learning to establish a perception mechanism model.
The sensor unit is a pyramid sensor unit formed by four sensitive units, and the four sensitive units are arranged at the inner part of the sensor unit in a 2x2 mode.
And the sensor unit integrates the elastic behavior model and the stress-capacitance conversion model into a sensor representative unit, performs discrete simplification and modularization on the sensor array, and constructs a response mechanism of the sensor array.
The actual measurement data set is finally established by touching an object with obvious boundary characteristics by utilizing a sensor array. The actual measurement database obtained at this time is rough and is used for training.
Further, a fine simulation data set is obtained using finite element simulation.
The data fusion is to establish a mapping relation to realize the fusion of the simulation data set and the actual measurement data set.
The establishment of the perception mechanism model comprises high-resolution pressure cloud image generation, fusion convolution operation, geometric relation reconstruction and microstructure and material attribute reconstruction.
The high-resolution pressure cloud image generation comprises the following steps:
(i) Firstly, obtaining a rough pressure cloud picture by using a linear interpolation mode from a low-resolution pressure cloud picture;
(ii) Performing downsampling self-coding operation on the low-resolution pressure cloud image by using convolution operation to obtain a series of characteristic layers;
(iii) And the characteristics of the layers are connected by adopting a U-Net mode, and finally, a high-resolution pressure cloud picture is obtained, so that higher precision is obtained.
The fusion convolution operation specifically comprises the following steps: providing a fusion convolution function for the properties of the boundary, the material and the like of the fusion object, and using the fusion convolution function in the convolution operation of each layer; let x be i Is the corresponding characteristic value, b is the offset value, N (x i ) Is x i Corresponding to the fusion convolution defined as:
x i =Ψ(M i )∑w(x j )·(x j e m j )+b
where w (·) can be regarded as a weight function, M i ={m j The corresponding adaptive matrix is the 'multiplication' operation, and the function of filtering the boundary value is achieved; in this, the function ψ (M i ) Is defined as a function related to object boundaries, microscopic appearance, material properties, etc.; thus, the function is combined with convolution operation to apply the features of the detected object to the high resolution pressure cloud reconstruction.
The geometric relationship reconstruction is specifically as follows: and carrying out feature extraction on the input three-dimensional pressure cloud image by using a graph convolution neural network in reconstruction, and then carrying out regression reconstruction on the three-dimensional model, so as to acquire the geometric shape of the detection object by using the sensor sensing signal.
The microstructure and material attribute reconstruction is specifically as follows: through actual measurement and finite element simulation, a corresponding database set of a sensor three-dimensional pressure cloud image and an object appearance microstructure, materials and the like is constructed, then the three-dimensional pressure cloud image is constructed into a 3×n matrix as an input characteristic, a multi-layer perceptron with DenseNet+ReLU as a block is adopted to conduct quantization regression training, and finally numerical vectors of the corresponding appearance microstructure and materials are output, so that the relation of material properties of the detected object is obtained through sensor perception signals.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. according to the invention, the established mechanical model of the touch sensor is utilized to carry out numerical simulation of the touch characteristic object of the sensor array, so as to obtain simulated touch data, the simulated touch data is used as an auxiliary database for collecting data of the sensor entity, and a mapping model is established to improve the accuracy of a training data set, so that a large-scale force signal and a corresponding high-resolution data set of a three-dimensional structure are constructed;
2. by utilizing the technology of convolutional neural network, constructing an end-to-end depth network to train on a data set and obtain a training model, quickly identifying the geometric and surface microstructure, material and other attributes of a three-dimensional object by utilizing force signals obtained by a sensor, and constructing a three-dimensional structural relation with the detected object.
Drawings
FIG. 1-1 is a schematic diagram of a sensor unit; fig. 1-2 are exploded views of a sensor unit.
FIG. 2 is a schematic diagram of the contact relationship between a sensor array and a surface feature of a test object.
Fig. 3 is a schematic diagram of a depth network for generating a high resolution pressure cloud.
Fig. 4 is a schematic diagram of a geometrical relationship reconstruction based on a three-dimensional pressure cloud.
Fig. 5 is a schematic diagram of a perception mechanism.
Wherein the meaning of the reference numerals is explained as follows:
1-adaptive convolution operation, 2-connection operation, 3-up acquisition operation, 4-sensor unit, 5-sensitive unit.
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but embodiments of the present invention are not limited thereto.
The invention relates to the field of sensing mechanisms of flexible touch sensors, in particular to a deep learning sensing method of a flexible touch sensor. And establishing a three-dimensional mechanical model of the contact state of the surface microstructure of the flexible touch sensor and the three-dimensional microstructure of the surface of the touch object by adopting a numerical simulation and experiment method, fusing the actual measurement data set with the finite element contact simulation data set, and acquiring the relation between the pressure signal and the three-dimensional multi-scale geometric dimension, the surface morphology, the physical attribute and other structures of the detection object by means of a neural network deep learning model. The method comprises the following steps:
step 1), building a sensor unit structure mechanical model;
step 2), building a sensor array structure mechanical model;
step 3), obtaining an actual measurement data set;
step 4), obtaining a finite element simulation data set;
step 5), data set fusion improves data resolution;
and 6) combining deep learning to establish a perception mechanism model.
The sensor is designed to form a pyramid sensor unit by four sensitive units, wherein the four sensitive units are arranged at the inner part of the sensor unit by 2x2, as shown in figures 1-1 and 1-2. The upper and lower electrode layers of each sensing unit are connected in series with adjacent units so as to realize the array connection of the sensing units. In combination with the micro-capacitance detection circuit, high resolution tactile sensing can be achieved. Furthermore, the sensor unit elastic behavior model and the stress-capacitance conversion model are integrated into a sensor representative unit, matlab programming is adopted, modularization and discretization simulation are carried out on a sensor array mechanical response mechanism, and a cooperative strain mechanism of force loads in different directions is considered. And touching an object with obvious boundary characteristics by using a sensor array, and establishing a rough actual measurement database for training. By means of a finite element tool, simulating a stress response process when the sensor array faces touch features such as different micro geometric contact morphologies (as shown in fig. 2), and the like, touch feedback data under micro boundary disturbance is obtained and used as a fine data set for strengthening and supplementing the defects of an actual measurement data set, and fusion of the simulation data set and the actual measurement data set is realized by establishing a mapping relation, namely, the physical precision of the sensor is combined with the precision of a numerical algorithm, and the detection precision and sensitivity of the sensor are improved. And constructing a model of a perception mechanism by using a depth network, wherein the model comprises high-resolution pressure cloud image generation, fusion convolution operation, geometric relation reconstruction and microstructure and material property reconstruction.
The high-resolution pressure cloud image generating step is shown in fig. 3, (i) firstly, a rough pressure cloud image is obtained from a low-resolution pressure cloud image in a linear interpolation mode; (ii) Performing downsampling self-coding operation on the low-resolution pressure cloud image by using convolution operation to obtain a series of characteristic layers; (iii) And the characteristics of the layers are connected by adopting a U-Net mode, and finally, a high-resolution pressure cloud picture is obtained, so that higher precision is obtained. A fusion convolution function is provided for fusing object boundaries, materials and other attributes and is used in convolution operation of each layer. Let x be i Is the corresponding characteristic value, b is the offset value, N (x i ) Is x i Is to the neighborhood ofThe convolution is defined as:
x i =Ψ(M i )∑w(x j )·(x j e m j )+b
where w (·) can be regarded as a weight function, M i ={m j The corresponding adaptive matrix is the 'multiplication' operation, and the function of filtering the boundary value is achieved. In this, the function ψ (M i ) May be defined as a function related to object boundaries, microscopic appearance, material properties, etc. Thus, the feature of the detected object can be applied to the reconstruction of the high resolution pressure cloud image when the function is combined with the convolution operation.
And extracting features of the input three-dimensional pressure cloud image by using a graph convolution neural network, and then carrying out regression reconstruction on the three-dimensional model, so that the geometric shape of the detection object is obtained by using the sensor sensing signal, and the reconstruction flow is shown in fig. 4.
Through actual measurement and finite element simulation, a corresponding database set of a sensor three-dimensional pressure cloud image and an object appearance microstructure, materials and the like is constructed, then the three-dimensional pressure cloud image is constructed into a 3×n matrix as an input characteristic, a multi-layer perceptron with DenseNet+ReLU as a block is adopted to conduct quantization regression training, finally numerical vectors of the corresponding appearance microstructure, materials and the like are output, and therefore the relation of material properties of the detected object is obtained through sensor perception signals, and the perception mechanism is shown in figure 5.
The above examples are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above examples, and any other changes, modifications, substitutions, combinations, and simplifications that do not depart from the spirit and principle of the present invention should be made in the equivalent manner, and the embodiments are included in the protection scope of the present invention.
Claims (4)
1. The deep learning sensing method of the flexible touch sensor is characterized by comprising the following steps in sequence:
step 1), building a sensor unit structure mechanical model;
step 2), building a sensor array structure mechanical model;
step 3), obtaining an actual measurement data set;
step 4), obtaining a finite element simulation data set;
step 5), data set fusion improves data resolution;
step 6), combining deep learning to establish a perception mechanism model;
the data set fusion is to establish a mapping relation to realize fusion of the simulation data set and the actual measurement data set;
the establishment of a perception mechanism model comprises high-resolution pressure cloud image generation, fusion convolution operation, geometric relation reconstruction and microstructure and material attribute reconstruction;
the high-resolution pressure cloud image generation comprises the following steps:
(i) Firstly, obtaining a rough pressure cloud picture by using a linear interpolation mode from a low-resolution pressure cloud picture;
(ii) Performing downsampling self-coding operation on the low-resolution pressure cloud image by using convolution operation to obtain a series of characteristic layers;
(iii) The characteristics of the layers are connected by adopting a U-Net mode, and finally, a high-resolution pressure cloud picture is obtained, so that higher precision is obtained;
the fusion convolution operation specifically comprises the following steps: providing a fusion convolution function for the boundary and material property of the fusion object, wherein the fusion convolution function is used in the convolution operation of each layer; let x be i Is the corresponding characteristic value, b is the offset value, N (x i ) Is x i Corresponding to the fusion convolution defined as:
x i =Ψ(M i )∑w(x j )·(x j ⊙m j )+b
wherein w (·) is a weight function, M i ={m j The corresponding adaptive matrix is the }, the # -is the multiplication operation, which plays a role in filtering the boundary value; in this, the function ψ (M i ) Defining as a function related to object boundaries, microscopic appearance, material properties; thus, combining the function to convolution operation applies the feature of the detected object to the high-resolution pressure cloudGraph reconstruction is carried out;
the geometric relationship reconstruction is specifically as follows: the reconstruction utilizes a graph convolution neural network to perform feature extraction on an input three-dimensional pressure cloud graph, and then performs regression reconstruction on a three-dimensional model, so that the geometric shape of a detection object is obtained by utilizing a sensor sensing signal;
the microstructure and material attribute reconstruction is specifically as follows: through actual measurement and finite element simulation, a corresponding database set of the sensor three-dimensional pressure cloud image and the microstructure and the material of the appearance of the object is constructed, then the three-dimensional pressure cloud image is constructed into a 3×n matrix as an input characteristic, a multi-layer perceptron with DenseNet+ReLU as a block is adopted to conduct quantization regression training, and finally a numerical vector of the corresponding microstructure and the material of the appearance is output, so that the relation of material properties of the detected object is obtained through sensor perception signals.
2. The method for deep learning sensing of a flexible touch sensor of claim 1, wherein the sensor unit is a pyramid sensor unit consisting of four sensing units, and the four sensing units are arranged 2x2 inside the sensor unit.
3. The method for deep learning sensing of a flexible touch sensor according to claim 1, wherein the sensor unit integrates an elastic behavior model and a stress-capacitance conversion model into a sensor representative unit, and performs discrete simplification and modularization on the sensor array to construct a response mechanism of the sensor array.
4. The method of claim 1, wherein the actual measurement data set is established by touching an object with distinct boundary features with a sensor array.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911314290.6A CN111079333B (en) | 2019-12-19 | 2019-12-19 | Deep learning sensing method of flexible touch sensor |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911314290.6A CN111079333B (en) | 2019-12-19 | 2019-12-19 | Deep learning sensing method of flexible touch sensor |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111079333A CN111079333A (en) | 2020-04-28 |
CN111079333B true CN111079333B (en) | 2024-03-12 |
Family
ID=70315541
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911314290.6A Active CN111079333B (en) | 2019-12-19 | 2019-12-19 | Deep learning sensing method of flexible touch sensor |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111079333B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111796708B (en) * | 2020-06-02 | 2023-05-26 | 南京信息工程大学 | Method for reproducing three-dimensional shape features of image on touch screen |
CN111964821A (en) * | 2020-08-05 | 2020-11-20 | 清华大学深圳国际研究生院 | Pressure touch prediction method and pressure touch prediction model for electronic skin |
US20230306261A1 (en) * | 2020-11-24 | 2023-09-28 | Max-Planck-Gesellschaft zur Förderung der Wissenschaften e. V. | Method for force inference of a sensor arrangement, methods for training networks, force inference module and sensor arrangement |
CN112802182B (en) * | 2021-01-20 | 2022-12-16 | 同济大学 | Method and system for reconstructing anthropomorphic touch object based on touch sensor |
CN114894354B (en) * | 2022-04-11 | 2023-06-13 | 汕头大学 | Pressure sensing feedback device based on surface structural color and deep learning identification method |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20030027636A (en) * | 2001-09-29 | 2003-04-07 | 홍동표 | A Sensor capable of sensing objects |
CN1539604A (en) * | 2003-11-01 | 2004-10-27 | 中国科学院合肥智能机械研究所 | Flexible touch sensor and touch information detection method |
CN110135485A (en) * | 2019-05-05 | 2019-08-16 | 浙江大学 | The object identification and localization method and system that monocular camera is merged with millimetre-wave radar |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9092737B2 (en) * | 2009-07-30 | 2015-07-28 | Northwestern University | Systems, methods, and apparatus for 3-D surface mapping, compliance mapping, and spatial registration with an array of cantilevered tactile hair or whisker sensors |
-
2019
- 2019-12-19 CN CN201911314290.6A patent/CN111079333B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20030027636A (en) * | 2001-09-29 | 2003-04-07 | 홍동표 | A Sensor capable of sensing objects |
CN1539604A (en) * | 2003-11-01 | 2004-10-27 | 中国科学院合肥智能机械研究所 | Flexible touch sensor and touch information detection method |
CN110135485A (en) * | 2019-05-05 | 2019-08-16 | 浙江大学 | The object identification and localization method and system that monocular camera is merged with millimetre-wave radar |
Non-Patent Citations (3)
Title |
---|
Multimodal Material identification through recursive tactile sensing;A. Gómez Eguíluz 等;Robotics and Autonomous Systems;20180507;第106卷;130-139 * |
基于卷积神经网络的软硬触觉感知方法研究;余乐 等;传感器与微系统;20171231;第36卷(第06期);35-37, 41 * |
电容-电阻双模式材质识别传感器设计与实验;郭小辉 等;华中科技大学学报(自然科学版);20151031;第43卷(第S1期);220-223 * |
Also Published As
Publication number | Publication date |
---|---|
CN111079333A (en) | 2020-04-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111079333B (en) | Deep learning sensing method of flexible touch sensor | |
Zhang et al. | Fingervision tactile sensor design and slip detection using convolutional lstm network | |
CN110188598B (en) | Real-time hand posture estimation method based on MobileNet-v2 | |
Narang et al. | Sim-to-real for robotic tactile sensing via physics-based simulation and learned latent projections | |
Zhang et al. | Effective estimation of contact force and torque for vision-based tactile sensors with helmholtz–hodge decomposition | |
Suresh et al. | Shapemap 3-d: Efficient shape mapping through dense touch and vision | |
Wi et al. | Virdo: Visio-tactile implicit representations of deformable objects | |
CN108594660B (en) | Working modal parameter identification method and system of time invariant structure | |
Kuppuswamy et al. | Fast model-based contact patch and pose estimation for highly deformable dense-geometry tactile sensors | |
CN111204476A (en) | Vision-touch fusion fine operation method based on reinforcement learning | |
Lee et al. | Predicting the force map of an ert-based tactile sensor using simulation and deep networks | |
Guo et al. | Inverse simulation: Reconstructing dynamic geometry of clothed humans via optimal control | |
Soter et al. | Shape reconstruction of CCD camera-based soft tactile sensors | |
Van der Merwe et al. | Integrated object deformation and contact patch estimation from visuo-tactile feedback | |
CN116911079B (en) | Self-evolution modeling method and system for incomplete model | |
CN113947119A (en) | Method for detecting human gait by using plantar pressure signals | |
Saku et al. | Spatio-temporal prediction of soil deformation in bucket excavation using machine learning | |
Du et al. | 3D contact point cloud reconstruction from vision-based tactile flow | |
Wu et al. | Example-based real-time clothing synthesis for virtual agents | |
Wang et al. | Tactile sensory response prediction and design using virtual tests | |
CN116029205A (en) | Flow field reconstruction method based on intrinsic orthogonal decomposition and deep learning fusion | |
CN100464151C (en) | Method for inspecting article surface vein and its sensor | |
Rasoulzadeh et al. | Linking early design stages with physical simulations using machine learning | |
Wang et al. | Elastic interaction of particles for robotic tactile simulation | |
Azulay et al. | Augmenting tactile simulators with real-like and zero-shot capabilities |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |