JP2023088176A - Estimation device, estimation method, estimation program, and learning model generation device - Google Patents

Abstract

To estimate a state of an imparting side for imparting a stimulus to an object using electric characteristics of the object having a conductive flexible material without using a special detection device.SOLUTION: An estimation device (1) includes: a detection unit for detecting electric characteristics between a plurality of detection points (75) predetermined in a flexible material of a glove (2) having the conductive flexible material (22) whose electric characteristics change according to a change in an imparted stimulus; and an estimation unit (5) for inputting time-series electric characteristics detected by the detection unit to a learning model that is learned so as to output gripping state information with the time-series electric characteristics as input, using the time-series electric characteristics when the stimulus is given to the flexible material and gripping state information indicating a gripping state of an article by the hand wearing the glove that gives the stimulus to the flexible material as learning data, and estimating the gripping state information indicating the gripping state of the article by the hand wearing the glove corresponding to the input time-series electric characteristics.SELECTED DRAWING: Figure 1

Description

The present disclosure relates to an estimating device, an estimating method, an estimating program, and a learning model generating device.

2. Description of the Related Art Conventionally, a shape change occurring in an object is detected, and the detection result is used to estimate the state of a person or an object that deforms the object. In terms of detecting a shape change that occurs in an object, it is difficult to detect the deformation of the object without disturbing the deformation of the object. In addition, since strain sensors used for detecting deformation of a rigid body such as metal deformation are difficult to use for articles, a special detection device is required to detect deformation of an object. For example, there is known a technique of measuring the displacement and vibration of an object with a camera, obtaining a deformed image, and extracting the amount of deformation (see, for example, Patent Document 1). Also known is a technique of measuring the pressure distribution applied to a wheelchair, a cushion, a bed, etc., using a sheet-like pressure-sensitive sensor (see, for example, Non-Patent Document 1).

WO2017/029905

https://www.nitta.co.jp/product/sensor/bpms/

However, in the aspect of detecting a shape change that occurs in an object such as an object, when detecting the amount of deformation such as displacement of an object using a camera and an image analysis method, a system including a camera and image analysis etc. is large-scale. This is not preferable because it results in an increase in the size of the apparatus. In addition, the optical method using a camera cannot measure hidden portions that are not captured by the camera. Therefore, there is room for improvement in detecting deformation of objects.

The present disclosure is an estimating device capable of estimating the state of the stimulus-providing side of the object by using the electrical characteristics of the object having a conductive flexible material without using a special detection device. An object is to provide a method, an estimation program, and a learning model generation device.

In order to achieve the above object, the first aspect is
A detection unit for detecting electrical characteristics between a plurality of predetermined detection points on the flexible material of a glove provided with a flexible material that is conductive and whose electrical characteristics change according to changes in applied stimuli. and,
Using time-series electrical characteristics when the flexible material is stimulated and grasping state information indicating a grasping state of the article by the gloved hand applying the stimulus to the flexible material as learning data, the The time-series electrical characteristics detected by the detection unit are input to a learning model trained to receive the time-series electrical characteristics and output the gripping state information, and the input time-series electrical characteristics are input. an estimating unit for estimating gripping state information indicating a gripping state of the article by the gloved hand corresponding to
is an estimating device including

A second aspect is the estimation device of the first aspect,
the electrical property is volume resistance;
The gripping state includes a state in which at least one of pressure stimulation and material stimulation is applied to the glove by the gloved hand,
The learning model is trained to output, as the grasping state information, information indicating a state in which at least one of pressure stimulation and material stimulation by the gloved hand corresponding to the detected electrical property is applied. .

A third aspect is the estimation device of the second aspect,
The pressure stimulus is a pressure stimulus generated with a movement of gripping an article with the gloved hand,
The gripping state information is information about the motion of gripping the article with the gloved hand.

A fourth aspect is the estimation device of the second aspect,
The material stimulation is material stimulation that occurs as the glove contains water,
The gripping state information is information about water content in the glove.

A fifth aspect is the estimation device according to any one aspect of the first aspect to the fourth aspect,
The grip state information includes physical state information of the person who grips the glove.

A sixth aspect is the estimating device according to any one aspect of the first to fifth aspects,
The holding state information includes hardness information about the hardness of the article to be held with the gloved hand.

A seventh aspect is the estimating device according to any one aspect of the first aspect to the sixth aspect,
It further includes an output unit that outputs the grip state estimated by the estimation unit.

An eighth aspect is the estimating device of any one aspect of the first aspect to the seventh aspect,
The glove includes a material in which conductivity is imparted to at least a part of a urethane material having a structure having at least one of fibrous and mesh-like skeletons, or having a structure in which a plurality of fine air bubbles are scattered inside.

A ninth aspect is the estimation device according to any one aspect of the first aspect to the eighth aspect,
The learning model includes a model generated by learning using a network by reservoir computing using the flexible material as a reservoir and using the reservoir.

The tenth aspect is
A computer detects an electrical characteristic between a plurality of predetermined detection points on the flexible material of the glove having a flexible material that is electrically conductive and whose electrical properties change in response to changes in the applied stimulus. ,
Using time-series electrical characteristics when the flexible material is stimulated and grasping state information indicating a grasping state of the article by the gloved hand applying the stimulus to the flexible material as learning data, the The glove corresponding to the input time-series electrical characteristics by inputting the detected time-series electrical characteristics to a learning model trained to input the time-series electrical characteristics and output the gripping state information. This is an estimation method for estimating grasping state information that indicates the grasping state of an object by a hand wearing a hand.

The eleventh aspect is
Detecting electrical characteristics between a plurality of predetermined detection points on the flexible material of a glove having a flexible material that is electrically conductive and whose electrical characteristics change in response to a change in applied stimulus. ,
Using time-series electrical characteristics when the flexible material is stimulated and grasping state information indicating a grasping state of the article by the gloved hand applying the stimulus to the flexible material as learning data, the The glove corresponding to the input time-series electrical characteristics by inputting the detected time-series electrical characteristics to a learning model trained to input the time-series electrical characteristics and output the gripping state information. An estimation program for executing a process of estimating gripping state information indicating a gripping state of an article by a hand wearing a hand.

The twelfth aspect is
A detection unit for detecting electrical characteristics between a plurality of predetermined detection points on the flexible material of a glove provided with a flexible material that is conductive and whose electrical characteristics change according to changes in applied stimuli. an acquisition unit that acquires the electrical characteristics from and gripping state information indicating a gripping state of the article by the gloved hand that stimulates the flexible material;
A grip indicating a state of gripping an article by the gloved hand that applies a stimulus to the flexible material, based on the acquisition result of the acquisition unit, taking as input time-series electrical characteristics when pressure is applied to the flexible material. a learning model generation unit that generates a learning model that outputs state information;
is a learning model generation device including

Advantageous Effects of Invention According to the present disclosure, there is an effect that the state of the application side can be estimated by using the electrical properties of an object having a conductive flexible material without using a special detection device.

It is a figure which shows the structure of the estimation apparatus which concerns on embodiment. It is a figure which shows arrangement|positioning of the electroconductive urethane which concerns on embodiment. 1 is a diagram showing a conceptual configuration of a learning model generation device according to an embodiment; FIG. It is a figure which shows the functional structure of the learning process part which concerns on embodiment. It is a figure which shows the other functional structure of the learning process part which concerns on embodiment. It is a figure which shows the electrical structure of the estimation apparatus which concerns on embodiment. It is a flow chart which shows a flow of presumed processing concerning an embodiment. FIG. 4 is an explanatory diagram for explaining a specific example of a target object according to the embodiment; It is a flow chart which shows a flow of learning data collection processing concerning an embodiment. 4 is a flowchart showing the flow of learning processing according to the embodiment; 6 is a flowchart showing a specific example of the flow of estimation processing according to the embodiment;

Hereinafter, embodiments for implementing the technology of the present disclosure will be described in detail with reference to the drawings. Components and processes having the same actions and functions are given the same reference numerals throughout the drawings, and overlapping descriptions may be omitted as appropriate. In addition, the present disclosure is not limited to the following embodiments, and can be implemented with appropriate modifications within the scope of the purpose of the present disclosure.

In the present disclosure, a person is a concept that includes at least one of a human body and an object that can stimulate an object with a physical quantity. In the following description, a concept including a person and an object will be collectively referred to as a person without distinguishing between at least one of a human body and an object. In other words, a single human body and an object, and a combination of a human body and an object are collectively referred to as a person.

First, referring to FIGS. 1 to 7, a flexible material imparted with conductivity to which the technology of the present disclosure is applied and a state estimation process for estimating the state of the flexible material on the application side using the flexible material. explain.

<Flexible material>
In the present disclosure, the term “flexible material” is a concept that includes materials that are at least partially deformable such as bending, and is a soft elastic body such as a rubber material, a structure having at least one of a fibrous and mesh-like skeleton. It contains a body and a structure in which multiple microscopic air bubbles are scattered. Examples of these structures include polymeric materials such as urethane materials. Also, in the present disclosure, a flexible material imparted with conductivity is used. The term "flexible material to which electrical conductivity is imparted" is a concept that includes materials having electrical conductivity, materials obtained by imparting electrical conductivity to a flexible material to impart electrical conductivity, and materials in which the flexible material has electrical conductivity. including. Polymer materials such as urethane are suitable for the flexible material that imparts electrical conductivity. In the following description, as an example of a flexible material to which conductivity is imparted, a member formed by blending and infiltrating (also referred to as impregnation) a conductive material into all or part of a urethane material will be referred to as "conductive urethane." will be named and explained. The conductive urethane can be formed by either blending or soaking (impregnation) of the conductive material, and can be formed by blending or soaking (impregnating) the conductive material. can be formed in combination with For example, if the conductive urethane obtained by infiltration (impregnation) has higher conductivity than the conductive urethane obtained by blending, it is preferable to form the conductive urethane by infiltration (impregnation).

Conductive urethane has the function of changing its electrical properties according to a given physical quantity. An example of a physical quantity that causes a function of changing electrical properties is a stimulus value based on a pressure value that indicates a stimulus due to pressure that deforms a structure such as bending (hereinafter referred to as pressure stimulus). In addition, the pressure stimulus includes the application of pressure to a predetermined site and the application of pressure by distribution of pressure within a predetermined range. Further, other examples of the physical quantity include a stimulus value such as a water content indicating a stimulus (hereinafter referred to as material stimulus) that changes (transforms) the properties of the material by applying moisture or the like. Conductive urethane changes its electrical properties according to given physical quantities. An example of a physical quantity representing this electrical property is an electrical resistance value. Other examples include voltage values or current values.

By giving conductivity to a flexible material having a predetermined volume, conductive urethane exhibits electrical characteristics (that is, changes in electrical resistance) according to given physical quantities. It can be regarded as a volume resistance value. In conductive urethane, the electrical paths are linked in a complicated manner, and for example, the electrical paths expand and contract according to deformation. It may also exhibit behavior in which the electrical path is temporarily disconnected, and behavior in which a different connection than before occurs. Therefore, the conductive urethane deforms according to the magnitude and distribution of stimuli (pressure stimuli and material stimuli) by given physical quantities between positions separated by a predetermined distance (for example, the positions of detection points where electrodes are arranged). It exhibits different electrical properties depending on the temperature and alteration. For this reason, the electrical characteristics change according to the magnitude and distribution of the physical quantity applied to the conductive urethane.

By using conductive urethane, it is not necessary to provide a detection point such as an electrode at a target location for deformation and deterioration. Detecting points such as electrodes may be provided at at least two arbitrary locations on either side of the location where the conductive urethane is stimulated by a physical quantity (for example, FIG. 1).

Further, more than two detection points may be used in order to improve the detection accuracy of the electrical properties of the conductive urethane. Also, the conductive urethane of the present disclosure may be formed of a conductive urethane group in which a plurality of conductive urethane pieces are arranged with the conductive urethane 22 shown in FIG. 1 as one piece of conductive urethane. In this case, the electrical characteristics may be detected for each of the plurality of conductive urethane pieces, or the electrical characteristics of the plurality of conductive urethane pieces may be synthesized and detected. When the electrical characteristics are detected for each of a plurality of conductive urethane pieces, the electrical characteristics such as the electrical resistance value can be detected for each arrangement portion (for example, detection sets #1 to #n). As another example, the detection range on the conductive urethane 22 may be divided, a detection point may be provided for each divided detection range, and the electrical characteristics may be detected for each detection range.

<Estimation device>
Next, an example of an estimating device for estimating the state of the application side with respect to the conductive urethane using the conductive urethane will be described.

FIG. 1 shows an example of the configuration of an estimation device 1 capable of executing estimation processing for estimating the state of the granting side. The estimating device 1 includes an estimating unit 5 and is connected to the object 2 so that the electrical characteristics of the conductive urethane 22 are input. The estimation device 1 estimates the state of the application side of the conductive urethane 22 contained in the object 2 . The estimating device 1 can be realized by a computer having a CPU as an executing device that executes the processing described later.

The above-described deformation and alteration of the conductive urethane occur in physical quantities given to the conductive urethane in chronological order. The physical quantity given in this time series depends on the state of the giver. Therefore, the electrical properties of the conductive urethane that change in time series correspond to the state of the physical quantity imparted to the conductive urethane. For example, when a pressure stimulus that deforms conductive urethane or a material stimulus that modifies conductive urethane is applied, the electrical properties of conductive urethane that change in time series indicate the position, distribution, and magnitude of the pressure stimulus. corresponds to the state of the side. Therefore, it is possible to estimate the state of the application side to the conductive urethane from the electrical properties of the conductive urethane that change in time series.

The estimating device 1 estimates and outputs an unknown granting-side state using a learned learning model 51 by an estimating process to be described later. This makes it possible to identify the state of the application side to the object 2 without using a special device or a large-sized device or directly measuring the deformation and deterioration of the conductive urethane 22 contained in the object. The learning model 51 is learned by inputting the state of the application side with respect to the object 2 and the electrical characteristics of the object 2 (that is, the electrical characteristics such as the electrical resistance value of the conductive urethane 22 placed on the object 2). be. Learning of the learning model 51 will be described later.

It should be noted that the conductive urethane 22 can be placed on a flexible member 21 to form the object 2 (FIG. 2). The object 2 composed of the member 21 on which the conductive urethane 22 is arranged includes an electrical property detection portion 76 . The conductive urethane 22 may be placed on at least part of the member 21, and may be placed inside or outside. In addition, the conductive urethane may be arranged so that the state of the application side to the conductive urethane can be estimated, for example, it may be arranged so as to be able to contact the person directly or indirectly, or both.

FIG. 2 shows an arrangement example of the conductive urethane 22 on the object 2. As shown in FIG. The conductive urethane 22 may be formed so as to completely fill the inside of the member 21 so that the AA cross section of the object 2 is shown as an object cross section 2-1. Further, as shown in the object cross section 2-2, the conductive urethane 22 may be formed on one side (surface side) inside the member 21, and as shown in the object cross section 2-3, the conductive urethane 22 may be formed on the member 21 A conductive urethane 22 may be formed on the other side (rear side) inside the . Furthermore, as shown in the cross section 2-4 of the object, a conductive urethane 22 may be formed in a part of the inside of the member 21. FIG. Further, as shown in the object cross section 2-5, the conductive urethane 22 may be separately arranged outside the surface side of the member 21, and as shown in the object cross section 2-6, the other side ( back side). When the conductive urethane 22 is arranged outside the member 21, the conductive urethane 22 and the member 21 may be simply laminated, or the conductive urethane 22 and the member 21 may be integrated by adhesion or the like. Even if the conductive urethane 22 is arranged outside the member 21, the flexibility of the member 21 is not hindered because the conductive urethane 22 is a conductive urethane member.

As shown in FIG. 1, the conductive urethane 22 detects the electrical properties of the conductive urethane 22 (that is, the volume resistance value, which is the electrical resistance value) by signals from at least two detection points 75 arranged at a distance. ). In the example of FIG. 1, a detection set #1 is shown for detecting electrical characteristics (time-series electrical resistance values) from signals from two detection points 75 arranged diagonally on the conductive urethane 22. there is The number and arrangement of the detection points 75 are not limited to the positions shown in FIG. good. The electrical characteristics of the conductive urethane 22 can be obtained by connecting an electrical characteristics detector 76 for detecting electrical characteristics (for example, a volume resistance value, which is an electrical resistance value) to the detection point 75 and using the output thereof.

In this embodiment, since the conductive urethane 22 is used as the sensor, for example, when a person intervenes, the sense of discomfort given to the person is extremely small compared to a conventional sensor. Therefore, it is possible to perform measurement and estimation of the state of the granting side at the same time without damaging the state of the granting side of the person during the measurement. This is an advantage over conventional sensors that separately perform measurement and state estimation on the application side, and is particularly advantageous in estimation based on long-term measurement evaluation that follows chronological changes.

The estimating unit 5 is a functional unit that is connected to the object 2 and estimates the state of the applying side using the learning model 51 based on the electrical characteristics that change according to at least one of deformation and alteration of the conductive urethane 22. be. Specifically, time-series input data 4 representing the magnitude of electrical resistance (eg, electrical resistance value) in the conductive urethane 22 is input to the estimation unit 5 . The input data 4 corresponds to the state data 3 indicating the state of the application side with respect to the object 2, for example, the state related to the behavior of the person in contact with the object 2, such as the posture and movement of the person. For example, when a person contacts the object 2, the contact is made in a predetermined state such as a posture, and the conductive urethane 22 constituting the object 2 is stimulated (at least pressure stimulation and material stimulation) in response to the state. On the other hand) is given as a physical quantity, and the electrical properties of the conductive urethane 22 change. Therefore, the electrical characteristics of the conductive urethane 22 that change in time series indicated by the input data 4 correspond to the state of the object 2, ie, the conductive urethane 22 on the application side. The estimating unit 5 also outputs output data 6 representing the state of the applying side corresponding to the electrical characteristics of the conductive urethane 22 that change in time series, as an estimation result using the learned learning model 51 .

The learning model 51 has completed learning to derive the output data 6 representing the state of the applying side from the electric resistance (input data 4) of the conductive urethane 22 that changes due to the stimulus (pressure stimulus and material stimulus) given as a physical quantity. is a model. The learning model 51 is, for example, a model that defines a trained neural network, and is expressed as a set of information on weights (strengths) of connections between nodes (neurons) that make up the neural network.

<Learning processing>
Next, learning processing for generating the learning model 51 will be described.
FIG. 3 shows a conceptual configuration of a learning model generation device that generates the learning model 51. As shown in FIG. The learning model generation device includes a learning processing unit 52 . The learning model generation device can be configured including a computer having a CPU (not shown), and is executed as a learning processing unit 52 by a learning data collection process and a learning model generation process executed by the CPU to generate a learning model 51. .

<Learning data collection processing>
In the learning data collection process, the learning processing unit 52 learns a large amount of input data 4 obtained by measuring the electrical characteristics (for example, electrical resistance) of the conductive urethane 22 in chronological order using the state data 3 representing the state of the application side as a label. Collect as data. Therefore, the learning data includes a large amount of sets of input data 4 indicating electrical characteristics and state data 3 indicating the state of the application side corresponding to the input data 4 .

Specifically, in the learning data collection process, stimulation (pressure stimulation and material stimulation) according to the state of the application side when the state of the object 2 (that is, the state of the application side with respect to the conductive urethane 22) is formed An electrical characteristic (for example, an electrical resistance value) that changes depending on the time is acquired in chronological order. Next, state data 3 is given as a label to the acquired time-series electrical characteristics (input data 4), and the set of state data 3 and input data 4 reaches a predetermined number or a predetermined time. Repeat the process until A set of the state data 3 indicating the state of the application side and the time-series electrical characteristics (input data 4) of the conductive urethane 22 acquired for each state of the application side becomes the learning data. The state data 3 in the learning data is stored in a memory (not shown) so as to be handled as output data 6 indicating the state of the giving side in which the estimation result is correct in the learning process described later.

The learning data may be associated with time-series information by adding information indicating the measurement time to each electrical resistance value (input data 4) of the conductive urethane 22 . In this case, information indicating the time of measurement may be added to a set of time-series electrical resistance values in the conductive urethane 22 to associate the time-series information with the period determined as the state of the application side.

An example of the learning data described above is shown in the following table. Table 1 is an example of a data set in which time-series electrical resistance value data (r) and state data (R) indicating the state of the application side are associated as learning data regarding the state of the application side of the conductive urethane 22. be.

It should be noted that the electrical characteristics detected by the conductive urethane 22 (temporal characteristics based on time-series electrical resistance value data) can be regarded as characteristic patterns relating to the state of the conductive urethane 22 on the application side. That is, different stimuli are applied to the conductive urethane 22 in chronological order depending on the state of the application side of the conductive urethane 22 . Therefore, it is considered that the time-series electrical characteristics within a predetermined period of time appear as characteristic electrical characteristics with respect to the state of the application side. Therefore, the pattern (for example, the distribution shape of time-series electrical resistance values in the electrical characteristics) shown in the electrical characteristics (time characteristics based on time-series electrical resistance value data) detected by the conductive urethane 22 varies depending on the state of the application side. It corresponds and functions effectively in the learning process which will be described later.

<Learning model generation processing>
Next, learning model generation processing will be described. The learning model generation device shown in FIG. 3 generates a learning model 51 using the learning data described above by a learning model generation process in the learning processing unit 52 .

FIG. 4 is a diagram showing the functional configuration of the learning processing unit 52, that is, the functional configuration of a CPU (not shown) regarding learning model generation processing executed by the learning processing unit 52. As shown in FIG. A CPU (not shown) of the learning processing unit 52 operates as functional units of the generator 54 and the computing unit 56 . The generator 54 has a function of generating an output in consideration of the sequential relationship of the electrical resistance values obtained in time series as an input.

The learning processing unit 52 uses, as learning data, the above-described input data 4 (for example, an electrical resistance value) and output data 6, which is the state data 3 indicating the state of the stimulus applied to the conductive urethane 22. A large number of sets are held in a memory (not shown).

The generator 54 includes an input layer 540, an intermediate layer 542, and an output layer 544 to form a known neural network (NN). Since the neural network itself is a known technology, detailed description is omitted, but the intermediate layer 542 includes a large number of node groups (neuron groups) having inter-node connections and feedback connections. Data from the input layer 540 is input to the intermediate layer 542 , and data resulting from the operation of the intermediate layer 542 is output to the output layer 544 .

The generator 54 is a neural network that generates generated output data 6A as data representing the state of the application side or data close to the state of the application side from the inputted input data 4 (eg, electrical resistance value). The generated output data 6A is data obtained by estimating, from the input data 4, the state of the applying side in which the conductive urethane 22 is stimulated. The generator 54 generates generated output data representing a state close to the state of the granting side from the input data 4 input in chronological order. The generator 54 learns using a large number of input data 4 so that it can generate the generated output data 6A that is close to the state of the applying side such as a person or the like to whom the stimulus is given to the object 2, that is, the conductive urethane 22. Become. In another aspect, the electrical characteristics, which are the input data 4 input in chronological order, are regarded as a pattern, and by learning the pattern, the object 2, that is, the person or the like to whom the stimulus is applied to the conductive urethane 22, can be applied to the subject. generated output data 6A close to the state of .

The calculator 56 is a calculator that compares the generated output data 6A with the output data 6 of the learning data and calculates the error of the comparison result. The learning processing unit 52 inputs the generated output data 6A and the output data 6 of the learning data to the calculator 56 . In response to this, the calculator 56 calculates the error between the generated output data 6A and the output data 6 of the learning data, and outputs a signal indicating the calculation result.

The learning processing unit 52 performs learning of the generator 54 that tunes the weight parameter of the connection between nodes based on the error calculated by the calculator 56 . Specifically, the weight parameter of the connection between the nodes of the input layer 540 and the hidden layer 542 in the generator 54, the weight parameter of the connection between the nodes in the hidden layer 542, and the node of the hidden layer 542 and the output layer 544 Each of the weight parameters of the connections between is fed back to the generator 54 using techniques such as gradient descent and error backpropagation. That is, with the output data 6 of the learning data as a target, the connections between all nodes are optimized so as to minimize the error between the generated output data 6A and the output data 6 of the learning data.

Note that the generator 54 may use a recursive neural network having a function of generating an output in consideration of the context of time-series inputs, or may use another technique.

The learning processing unit 52 generates the learning model 51 using the learning data described above by the learning model generation process. The learning model 51 is expressed as a set of information of weight parameters (weights or strengths) of connections between nodes of learning results, and is stored in a memory (not shown).

Specifically, the learning processing unit 52 executes learning model generation processing according to the following procedure. In the first learning process, input data 4 (electrical characteristics) labeled with information indicating the state of the application side, which is learning data obtained as a result of chronological measurement, is obtained. In the second learning process, a learning model 51 is generated using learning data obtained as a result of time-series measurement. That is, a set of information on weight parameters (weights or strengths) of connections between nodes is obtained as a result of learning using a large amount of learning data as described above. Then, in the third learning process, data expressed as a set of information of weight parameters (weights or strengths) of connections between nodes of learning results is stored as a learning model 51 .

Then, the estimation apparatus 1 uses the learned generator 54 (that is, data expressed as a set of information of weight parameters of connections between nodes of learning results) as the learning model 51 . If a sufficiently learned learning model 51 is used, it is also possible to identify the state of the applying side from the time-series electrical characteristics of the object 2, that is, the conductive urethane 22 (for example, the electrical resistance value characteristics that change in time series). Not impossible.

<PRC>
By the way, in the conductive urethane 22, as described above, the electrical paths are intricately linked, and the electrical paths are expanded, contracted, expanded, contracted, temporarily disconnected, and changed (deformed) such as new connections, and changes in the properties of the material. It shows behavior according to (transformation). As a result, the conductive urethane 22 behaves with different electrical properties depending on the applied stimulus (eg, pressure stimulus). This allows the conductive urethane 22 to be treated as a reservoir that stores data regarding deformation of the conductive urethane 22 . That is, the estimation device 1 can apply the conductive urethane 22 to a network model (hereinafter referred to as PRCN) called physical reservoir computing (PRC). Since PRC and PRCN are known technologies, detailed description thereof will be omitted, but PRC and PRCN are suitable for estimating information on deformation and deterioration of the conductive urethane 22 .

FIG. 5 shows an example of the functional configuration of the learning processing unit 52 to which PRCN is applied. The learning processing unit 52 applying PRCN includes an input reservoir layer 541 and an estimation layer 545 . The input reservoir layer 541 corresponds to the conductive urethane 22 included in the object 2 . That is, the learning processing unit 52 to which the PRCN is applied treats the target object 2 including the conductive urethane 22 as a reservoir for storing data on deformation and deterioration of the target object 2 including the conductive urethane 22 and learns. The conductive urethane 22 becomes an electrical characteristic (electrical resistance value) according to each of various stimuli, functions as an input layer for inputting the electrical resistance value, and stores data on deformation and alteration of the conductive urethane 22. Acts as a reservoir layer. Since the conductive urethane 22 outputs different electrical characteristics (input data 4) according to the stimulus given by the state of the applying side such as a person, the estimation layer 545 calculates the electrical resistance value of the given conductive urethane 22. It is possible to deduce the state of the unknown grantor from Therefore, the estimation layer 545 should be learned in the learning process in the learning processing unit 52 to which the PRCN is applied.

<Configuration of estimation device>
Next, an example of a specific configuration of the estimation device 1 described above will be further described. FIG. 6 shows an example of the electrical configuration of the estimation device 1. As shown in FIG. The estimating device 1 shown in FIG. 6 includes a computer as an execution device that executes processing for realizing the various functions described above. The estimating device 1 described above can be realized by causing a computer to execute a program representing each function described above.

A computer functioning as the estimation device 1 includes a computer main body 100 . The computer main body 100 includes a CPU 102 , a RAM 104 such as a volatile memory, a ROM 106 , an auxiliary storage device 108 such as a hard disk drive (HDD), and an input/output interface (I/O) 110 . These CPU 102, RAM 104, ROM 106, auxiliary storage device 108, and input/output I/O 110 are connected via a bus 112 so as to exchange data and commands with each other. Also, the input/output I/O 110 is connected to a communication unit 114 for communicating with an external device, an operation display unit 116 such as a display and a keyboard, and a detection unit 118 . The detection unit 118 has a function of acquiring input data 4 (electrical characteristics such as time-series electrical resistance values) from the object 2 including the conductive urethane 22 . That is, the detection unit 118 includes the object 2 on which the conductive urethane 22 is arranged, and can acquire the input data 4 from the electrical characteristic detection unit 76 connected to the detection point 75 in the conductive urethane 22. be. Note that the detection unit 118 may be connected via the communication unit 114 .

The auxiliary storage device 108 stores a control program 108P for causing the computer main body 100 to function as the estimation device 1 as an example of the estimation device of the present disclosure. The CPU 102 reads the control program 108P from the auxiliary storage device 108, develops it in the RAM 104, and executes processing. Thereby, the computer main body 100 that has executed the control program 108P operates as the estimation device 1. FIG.

The auxiliary storage device 108 stores a learning model 108M including the learning model 51 and data 108D including various data. The control program 108P may be provided by a recording medium such as a CD-ROM.

<Estimation processing>
Next, the estimation processing in the estimation device 1 implemented by a computer will be further described. FIG. 7 shows an example of the flow of estimation processing by the control program 108P executed by the computer main body 100. As shown in FIG. The estimation process shown in FIG. 7 is executed by the CPU 102 when the computer main body 100 is powered on. The CPU 102 reads the control program 108P from the auxiliary storage device 108, develops it in the RAM 104, and executes processing.

First, CPU 102 acquires learning model 51 by reading learning model 51 from learning model 108M in auxiliary storage device 108 and developing it in RAM 104 (step S200). Specifically, a network model (see FIG. 4 and FIG. 5), which is represented as a learning model 51 and is a connection between nodes based on weight parameters, is developed in RAM 104 to realize connections between nodes based on weight parameters. A learning model 51 is constructed.

Next, the CPU 102 acquires the unknown input data 4 (electrical characteristics) for estimating the state of the applied side due to the stimulus applied to the conductive urethane 22 in time series via the detection unit 118 (step S202). Next, the CPU 102 uses the learning model 51 to estimate the output data 6 (unknown state of the giving side) corresponding to the acquired input data 4 (step S204). Then, the CPU 102 outputs the output data 6 (state of the provision side) of the estimation result via the communication section 114 (step S206), and ends this processing routine.

Thus, according to the estimation device 1, it is possible to estimate the unknown state of the applying side from the electrical resistance value of the conductive urethane 22. FIG. Specifically, the estimating device 1 can estimate the state of the applying side of a person or the like from the input data 4 (electrical characteristics) that change according to the stimulus given to the conductive urethane 22 depending on the state of the applying side. It becomes possible. In other words, it is possible to estimate the state of a person or the like on the applying side without using a special device or a large-sized device or directly measuring the deformation of the flexible member.

<Estimation of Gripping State>
When such a conductive urethane 22 is applied to the glove G, a pressure stimulus such as partial compression of the glove G is generated according to the movement of the hand H wearing the glove G to grasp an article. Therefore, the estimating device 1 can estimate the movement of the hand H wearing the glove G to grip the article from the time-series electrical resistance value of the conductive urethane 22 . For convenience of explanation, hereinafter, the person who puts on the glove G is simply referred to as "person". Here, the "movement of gripping an item by the hand H wearing the glove G" refers to whether the person is gripping the item (whether or not the person is gripping), how much force the person is gripping the item (strength of gripping force). (strength and weakness), which finger grips the article with force (how to grip), and the like.

FIG. 8 is a diagram showing an example in which the conductive urethane 22 is applied to the glove G. As shown in FIG. That is, the glove G is the object 2 to which the stimulus is applied. Here, the glove G is not limited to one that covers the entire hand. Instead of the G, a mitten-type glove G that is not separated for each finger may be used.

In the example shown in FIG. 8, a person grips an article with a hand H wearing a glove G containing conductive urethane 22 . The fact that the glove G contains the conductive urethane 22 means that the arrangement example of the conductive urethane 22 and the member 21 (in this case, cotton or chemical fiber) that constitutes the glove G is the conductive urethane as shown in FIG. 22 and member 21 are satisfied.

Then, as shown in FIG. 8, if the electric characteristic detection unit 76 is attached to the glove G and wireless is used, the movement of the person holding the article with the hand H wearing the glove G is not hindered, and the electric characteristic detection unit 76 is detected via the communication unit 114. It is possible to connect the electrical characteristic detection unit 76 and the estimation device 1 with . Note that the electrical characteristic detection unit 76 and the estimation device 1 may be connected by wire depending on the situation.

In the example shown in FIG. 8, an electrical property detector 76 is attached to the glove G, and the electrical property detector 76 attached to the glove G detects the electrical resistance value of the conductive urethane 22 contained in the glove G. .

The type of glove G in this embodiment is not limited as long as it is worn on the hand H of a person. Also, the type of the article to be gripped by the hand H wearing the glove G is not limited as long as it can be gripped by the hand H. Examples include tennis racket grips, golf club grips, baseball bat grips, and other sports equipment grips, and treatment chairs such as handrails held by a patient sitting in a dental chair during treatment. It may be a handrail, a handrail provided on a seat of a roller coaster, or the like.

In this way, the movement of a person's hand H wearing a glove G to grip an item affects the gripping state of the item. , is an example of gripping state information indicating the gripping state of an article with a hand H wearing a glove G. FIG.

Next, learning processing for generating the learning model 51 for estimating the movement of the hand H wearing the glove G to grip the article will be described.

In the learning data collection process, the learning processing unit 52 of the learning model generation device shown in FIG. A large amount of input data 4 obtained by measuring resistance values in time series is collected as learning data.

Specifically, in the learning data collection process, the electrical characteristics (eg, electrical resistance value) of the conductive urethane 22 contained in the glove G, which changes according to the movement of the hand H wearing the glove G to grasp the article, are transferred to the glove G. It is obtained in time series from the attached electrical characteristic detection unit 76 . Next, the state data 3 is given as a label to the input data 4 which is the acquired time-series electrical characteristics, and a plurality of learning data in which the state data 3 and the input data 4 are combined are prepared.

In addition, the learning data instructs a person to make a movement to grip an article with the hand H wearing the glove G, detects the electrical resistance value at that time, and grips the article with the hand H wearing the glove G of the person. Collect movements (see Table 2). A person who makes a motion of gripping an article with a hand H wearing a glove G of such learning data may be a specific person, such as a professional tennis player, or may be a plurality of various persons.

In the following description, the electrical resistance value is used as an example of the electrical characteristics of the conductive urethane 22 contained in the glove G. However, it was mentioned above that the current value or the voltage value may be used as the electrical characteristics of the conductive urethane 22. As I said.

Table 2 shows the time-series electrical resistance data obtained from the glove G, that is, the input data 4 and the It is an example of a data set in which state data 3 (R1 to Rk) indicating the movement of gripping an article by the hand H are associated with each other.

Here, the gripping state information is, for example, information about how much force the hand H wearing the glove G is gripping the article (strength of gripping force). As for the strength of the gripping force, learning data is obtained by giving instructions to a person who makes a motion to grip an object with a hand H wearing a glove G and applying a force to grip the object in various situations according to the type of the object. collect. For example, when the article is the grip of a tennis racket, the electrical resistance value when hitting a serve (R1), the electrical resistance value when hitting a ball with a stroke (R2), and the electric resistance value when hitting a ball with a volley (R3) Collect electrical resistance values, etc. That is, it is known that, for example, a tennis expert and a beginner apply different amounts of force to grip the grip. For this reason, learning data of an expert is collected, a learning model 51 is generated, and the estimation unit 5 calculates the current gripping state of the person from the pressure stimulus for gripping the grip input to the generated learning model 51. can be estimated. Then, the estimation result is output to a terminal device such as a smart phone. Based on the output estimation results, the person practices how to put in the strength to grip the grip of the tennis racket so as to correct the difference from the training data. It is also possible to plan

In addition, when the article is a handrail of a treatment chair, the learning data includes the electrical resistance value when gripped during a normal condition (R1) and the electrical resistance value when gripped during painful treatment (R2). collect as That is, it is known that, for example, when one is tense or in pain, one grips the handrail more strongly than when one is not tense or in a normal state when one is not in pain. For this reason, learning data is collected, a learning model 51 is generated, and the estimating unit 5 estimates the current physical state of the person from the pressure stimulus to grip the handrail input to the generated learning model 51. be able to. Then, the estimation result is output to a terminal device such as a smart phone. If the output estimation result is gripping state information indicating that the person being treated feels pain, it is possible to take measures such as interrupting the treatment. In this way, the gripping state information includes physical state information such as the state of tension and the state of pain.

Moreover, the gripping state information is, for example, information about water content in the glove G. As shown in FIG. When the glove G including the conductive urethane 22 contains moisture, material stimulation occurs, and the electrical properties of the conductive urethane 22 change according to the magnitude and distribution of the material stimulation. Therefore, the estimating device 1 can estimate information about water content occurring in the glove G from the time-series electrical resistance value of the conductive urethane 22 . Here, the moisture contained in the glove G is mainly sweat generated from the hand H put in the glove G. Learning data for water content in the glove G is collected by giving instructions to a person who moves to grip an item with the hand H wearing the glove G to grip the item in various situations according to the type of the item. For example, if the article is a handrail of a treatment chair, the electrical resistance value according to the water content (R1) in a normal state, the electrical resistance value according to the water content (R2) during painful treatment, etc. are collected. That is, it is known that, for example, when one is tense or in pain, the hand holding the handrail sweats more than when one is not tense or in a normal state without pain. It is For this reason, learning data is collected, a learning model 51 is generated, and the estimating unit 5 estimates the current physical state of the person from the material stimulus of gripping the handrail input to the generated learning model 51. be able to. Then, the estimation result is output to a terminal device such as a smart phone. If the output estimation result is gripping state information indicating that the person being treated feels pain, it is possible to take measures such as interrupting the treatment. In this way, the gripping state information includes physical state information such as the state of tension and the state of pain.

The gripping state information also includes, for example, information about the hardness of the article gripped with the hand H wearing the glove G. Learning data about the hardness of an object to be gripped is collected by instructing a person to apply a force to grip an object with various hardnesses with a hand H wearing a glove G. For example, the electrical resistance value when gripping a hard object A (R1), the electrical resistance value when gripping a hard object B (R2), the electrical resistance value when gripping a soft object A (R3), and the soft object B are collected (R4), such as the electrical resistance value when the is gripped. That is, it is known that when an article is gripped with a hand H wearing a glove G, the amount of deformation of the glove G varies depending on the hardness of the article. In the case of a hard article, the amount of deformation of the glove G is greater than in the case of a soft article. Such a property is remarkable when objects having different hardnesses are gripped with the same force. For this reason, learning data is collected, a learning model 51 is generated, and the estimating unit 5 estimates the hardness of the article as gripping state information from the input pressure stimulus for gripping the article to the generated learning model 51. can do. Then, the estimation result is output to a terminal device such as a smart phone.

The gripping state information is not limited to the above-described information, and may be, for example, electrical characteristics of the conductive urethane 22 due to pressure stimulation or material stimulation of each finger of the hand H wearing the glove G. That is, the degree of bending (straightened state, fist-clenched state, etc.) of each finger of the hand H wearing the glove G is collected as learning data, the learning model 51 is generated, and the generated learning model 51 is The estimating unit 5 can estimate the degree of bending of the finger from the pressure stimulus due to the degree of bending of the finger input by . Then, the estimation result is output to a terminal device such as a smart phone. It is also possible to play sign language, rock-paper-scissors, etc. from the output estimation results.

The learning processing unit 52 can be configured including a computer including a CPU (not shown), and executes learning data collection processing and learning processing.

FIG. 9 is a diagram showing an example of the flow of learning data collection processing performed by the learning processing unit 52. As shown in FIG.

In step S100, the learning processing unit 52 instructs the person to make a motion to grip the item with the hand H wearing the glove G. In step S102, the gripping state associated with the motion to grip the item is changed by pressure stimulation. Acquire the electrical resistance value in chronological order. In the next step S104, the acquired time-series electrical resistance values are labeled with gripping states accompanied by movement and stored (see Table 2). The learning processing unit 52 makes a negative determination until a positive determination is made in step S106 until the set of the gripping state and the electrical resistance value of the conductive urethane 22 reaches a predetermined number or a predetermined predetermined time. ), and the above process is repeated. As a result, the learning processing unit 52 can acquire and store the electrical resistance value of the conductive urethane 22 in chronological order for each gripping state. A set of electrical resistance values of is used as learning data.

The learning model 51 is generated by learning processing of the learning processing unit 52 . The learning model 51 is expressed as a set of information of weight parameters (weights or strengths) of connections between nodes as a result of learning by the learning processing unit 52 .

FIG. 10 is a diagram showing an example of the flow of learning processing performed by the learning processing unit 52. As shown in FIG.

The learning processing unit 52 selects any one learning data from among the plurality of learning data shown in Table 2 in step S110. Then, time-series input data 4 is obtained from the selected learning data. In step S112, the learning processing unit 52 generates the learning model 51 using learning data obtained as a result of time-series measurement. That is, a set of information on weight parameters (weights or strengths) of connections between nodes is obtained as a result of learning using a large amount of learning data as described above. Then, in step S114, data expressed as a set of information on weight parameters (weights or strengths) of connections between nodes of learning results is stored as a learning model 51. FIG.

Note that the generator 54 may use a recursive neural network having a function of generating an output in consideration of the context of time-series inputs, or may use another technique.

Then, in the estimation device 1, the learned generator 54 (that is, data represented as a set of weight parameter information of connections between nodes of learning results) generated by the method exemplified above is used as a learning model 51. use. If a sufficiently learned learning model 51 is used, it is not impossible to identify the gripping state from time-series electrical resistance values of the glove G, that is, the conductive urethane 22 .

Note that the processing by the learning processing unit 52 is executed by, for example, an external learning model generation device (not shown) different from the estimation device 1 . The estimating device 1 includes an estimating unit 5 . The output data 6, which is information indicating the gripping state, is an example of gripping state information of the present disclosure.

Next, a grasping state estimation process using the learning model 51 according to the present embodiment will be described.

FIG. 11 shows an example of the flow of grip state estimation processing by the control program 108P executed by the computer main body 100. As shown in FIG. The estimation process shown in FIG. 11 is executed by the CPU 102 when the computer main body 100 is powered on.

First, in step S200, the learning model 51 that has been learned for grasping state estimation is read out from the learning model 108M in the auxiliary storage device 108, and is developed in the RAM 104 to acquire the learning model 51 and construct the learning model 51. do.

Next, in step S202, unknown input data 4 (electrical characteristics) to be used for estimating the grasping state of the article due to the pressure stimulus applied to the glove G is obtained in time series via the detection unit 118. FIG. Here, the unknown input data (electrical characteristics) may be input data (electrical characteristics) obtained in real time from the gripper during execution of the estimation process, or Input data (electrical characteristics) obtained in advance may be used.

Next, in step S204, the learning model 51 acquired in step S200 is used to estimate output data 6 (gripping state) corresponding to the input data 4 (electrical characteristics) acquired in step S202. Then, in the next step S206, output data 6 (gripping state) of the estimation result is output to the auxiliary storage device 108 or output via the communication unit 114, which is an example of an output unit, and this processing routine ends. Here, the estimation result output to a terminal device such as a smartphone via the communication unit 114 may be an image of a gripping state that makes it possible to recognize the difference from a skilled person.

The estimation process illustrated in FIG. 11 described above is an example of the process performed by the estimation method of the present disclosure.

As described above, according to the present disclosure, it is possible to estimate the movement of the hand H wearing the glove G to grip the article from the unknown electrical characteristics of gripping the article by the hand H wearing the glove G. It becomes possible.

Therefore, the estimating apparatus 1 can be used to determine a person's skill level and experience level by, for example, the motion of gripping an object with a hand H wearing a glove G. Therefore, the estimating apparatus 1 can be applied to the field of sports such as form check can be done.

In addition, the estimating apparatus 1 can detect the physical condition of a person by, for example, the motion of gripping an object with a hand H wearing a glove G. Therefore, the estimating apparatus 1 can be applied to the medical field, such as interrupting treatment. can.

So far, the case where the glove G is worn on a human hand has been described. By putting on the glove G containing the conductive urethane 22 on the arm of the robot, the same grasping state information as the grasping state information estimated for a human can be estimated for the robot. In particular, the estimation unit 5 is useful when estimating the hardness of an article gripped by the arm of the robot wearing the glove G described above. That is, the "hand" in this embodiment includes a robot arm as well as a human hand. Also, the glove G to be worn on the arm of the robot may be of any shape as long as it covers the tip of the arm (the part that grips the article).

When the glove G containing the conductive urethane 22 is put on the arm of the robot, first, the arm of the robot wearing the glove G is instructed to move to grasp an article, and the electric resistance value at that time is detected. Collect the movement of the human arm of the robot wearing the glove G to grasp the article (see Table 2). Since the learning data collection process is the same as the learning data collection process shown in FIG. 9, the description thereof is omitted.

In addition, the learning model generation device uses learning data in which time-series electrical resistance value data obtained from the conductive urethane 22 included in the glove G worn by the robot arm is associated with grasping state information. , to generate a learning model 51 . Since the method of generating the learning model 51 is the same as the learning process shown in FIG. 10, the description is omitted.

The estimating device 1 performs machine learning of the relationship between the feature pattern represented by the time-series electrical resistance value data of the conductive urethane 22 and the grasping state information in which the arm of the robot wearing the glove G grasps the article. By executing the estimation process shown in FIG. to estimate Also, since the estimation process is the same as the estimation process shown in FIG. 11, the description thereof is omitted.

As described above, in the present disclosure, the case of applying conductive urethane as an example of the flexible member has been described, but the flexible member is of course not limited to conductive urethane.

Moreover, the technical scope of the present disclosure is not limited to the scope described in the above embodiments. Various changes or improvements can be made to the above-described embodiments without departing from the scope of the invention, and forms with such changes or improvements are also included in the technical scope of the present disclosure.

Further, in the above embodiment, the estimation process and the learning process are implemented by a software configuration based on the processing using flowcharts. It is good also as a form which carries out.

Also, part of the estimation device, for example, a neural network such as a learning model, may be configured as a hardware circuit.

1 estimation device 2 object 3 state data 4 input data 5 estimation unit 6 output data 6A generated output data 22 conductive urethane 51 learning model 52 learning processing unit 54 generator 56 computing unit 75 detection point 76 electrical characteristic detection unit G glove H hand

Claims (12)

A detection unit for detecting electrical characteristics between a plurality of predetermined detection points on the flexible material of a glove provided with a flexible material that is conductive and whose electrical characteristics change according to changes in applied stimuli. and,
Using time-series electrical characteristics when the flexible material is stimulated and grasping state information indicating a grasping state of the article by the gloved hand applying the stimulus to the flexible material as learning data, the The time-series electrical characteristics detected by the detection unit are input to a learning model trained to receive the time-series electrical characteristics and output the gripping state information, and the input time-series electrical characteristics are input. an estimating unit for estimating gripping state information indicating a gripping state of the article by the gloved hand corresponding to
estimator including the electrical property is volume resistance;
The gripping state includes a state in which at least one of pressure stimulation and material stimulation is applied to the glove by the gloved hand,
The learning model is trained to output, as the grasping state information, information indicating a state in which at least one of pressure stimulation and material stimulation by the gloved hand corresponding to the detected electrical property is applied. The estimating device according to claim 1. The pressure stimulus is a pressure stimulus generated with a movement of gripping an article with the gloved hand,
3. The estimation device according to claim 2, wherein said gripping state information is information relating to a motion of gripping an article with said gloved hand. The material stimulation is material stimulation that occurs as the glove contains water,
The estimating device according to claim 2, wherein the gripping state information is information about water content in the glove. The estimating apparatus according to any one of claims 1 to 4, wherein the gripping state information includes physical state information of a person who grips the glove. 6. The estimation device according to any one of claims 1 to 5, wherein the gripping state information includes hardness information about hardness of an article gripped with the gloved hand. The estimating device according to any one of claims 1 to 6, further comprising an output unit that outputs the grip state estimated by the estimating unit. 1. The glove includes a material in which at least a portion of a urethane material having a structure having at least one of fibrous and mesh-like skeletons or a structure in which a plurality of fine air bubbles are scattered therein is imparted with conductivity. 8. The estimation device according to any one of claims 7 to 8. The estimation according to any one of claims 1 to 8, wherein the learning model includes a model generated by learning using a network by reservoir computing using the flexible material as a reservoir. Device. A computer detects an electrical characteristic between a plurality of predetermined detection points on the flexible material of the glove having a flexible material that is electrically conductive and whose electrical properties change in response to changes in the applied stimulus. ,
Using time-series electrical characteristics when the flexible material is stimulated and grasping state information indicating a grasping state of the article by the gloved hand applying the stimulus to the flexible material as learning data, the The glove corresponding to the input time-series electrical characteristics by inputting the detected time-series electrical characteristics to a learning model trained to input the time-series electrical characteristics and output the gripping state information. An estimation method for estimating grasping state information indicating a grasping state of an article by a hand wearing a hand. Detecting electrical characteristics between a plurality of predetermined detection points on the flexible material of a glove having a flexible material that is electrically conductive and whose electrical characteristics change in response to a change in applied stimulus. ,
Using time-series electrical characteristics when the flexible material is stimulated and grasping state information indicating a grasping state of the article by the gloved hand applying the stimulus to the flexible material as learning data, the The glove corresponding to the input time-series electrical characteristics by inputting the detected time-series electrical characteristics to a learning model trained to input the time-series electrical characteristics and output the gripping state information. An estimation program for executing a process of estimating gripping state information indicating a gripping state of an article by a hand wearing a hand. A detection unit for detecting electrical characteristics between a plurality of predetermined detection points on the flexible material of a glove provided with a flexible material that is conductive and whose electrical characteristics change according to changes in applied stimuli. an acquisition unit that acquires the electrical characteristics from and gripping state information indicating a gripping state of the article by the gloved hand that stimulates the flexible material;
A grip indicating a state of gripping an article by the gloved hand that applies a stimulus to the flexible material, based on the acquisition result of the acquisition unit, taking as input time-series electrical characteristics when pressure is applied to the flexible material. a learning model generation unit that generates a learning model that outputs state information;
A learning model generator including

Images