JP2011240468A - Contact type discriminating system for robot - Google Patents

Abstract

A contact type discrimination system for a robot capable of discriminating the type of contact with a simple configuration.
A robot contact type discrimination system according to the present invention is a contact type discrimination system that discriminates a contact type when an external force is applied by contact with a robot. The contact type discriminating system 100 is characterized by a first detection unit 110 that is mounted on the robot 1 and that detects components in three or more directions indicating the dynamic characteristics of external force, and a detection result of the first detection unit 110. A feature amount calculation unit 152 that calculates the amount, a contact type determination unit 153 that pattern-matches the feature amount and the sample feature amount acquired in advance so as to correspond to the type of contact, and determines the type of contact with the robot 1; .
[Selection] Figure 2

Description

  The present invention relates to a robot contact type discrimination system, a robot including the contact type discrimination system, and a robot contact type discrimination method.

  Conventionally, a robot capable of detecting contact with a robot has been disclosed. For example, the robots of Patent Documents 1 to 4 are configured to detect human contact with the robot by providing a so-called touch sensor on the outer surface.

  However, for example, in order to detect contact with the arm portion, it is necessary to provide a touch sensor over the entire arm portion. For this reason, the wiring arrangement becomes complicated, and the processing of the detection result from the touch sensor becomes complicated.

  Therefore, the robot of Patent Document 5 is provided with each axis torque sensor at the joint part, and detects contact with the arm part based on a change in output from each axis torque sensor.

JP 2000-326274 A JP 2008-149442 A JP 2009-45692 A JP 2006-172410 A JP 2008-229800 A

When a person comes into contact with an arm unit having a robot hand, when shaking hands, holding and pulling the hand unit, guiding the robot by holding the hand unit, handing an object to the hand unit, hitting the arm unit Cases are assumed.
However, the robot disclosed in Patent Document 5 is simply configured to detect whether or not there has been contact with the robot, and is not configured to determine the type of contact.

  An object of the present invention is made to solve such a problem, and is a robot contact type discrimination system capable of discriminating the type of contact with a simple configuration, and a robot equipped with the contact type discrimination system. And a method for discriminating the contact type of a robot.

A contact type determination system for a robot according to the present invention is a contact type determination system for determining the type of contact when an external force is applied by contact with the robot, and is a dynamic type of the external force mounted on the robot. A first detection unit that detects components in three or more axial directions indicating characteristics; a feature amount calculation unit that calculates a feature amount based on a detection result of the first detection unit; the feature amount; and the contact in advance A contact type determining unit that pattern-matches the sample feature amounts acquired so as to correspond to the type of the robot and determines the type of contact with the robot.
As described above, by using a force sensor capable of detecting components in three or more axial directions indicating the dynamic characteristics of external force as the first detection unit, it is possible to determine the type of contact with the robot with a small number of sensors. Therefore, it can contribute to simplification of the configuration of the robot. That is, the number of wirings can be reduced and the amount of calculation can be reduced.

The feature amount calculation unit calculates a horizontal component and a vertical component of a force component that are mechanical characteristics of the external force detected by the first detection unit, and uses the calculated horizontal component and the vertical component as a feature amount. The contact type determination unit uses the horizontal component and the vertical component of the force component, which are mechanical characteristics of external force, acquired in advance so as to correspond to the contact type and the feature amount, as the sample feature amount, and the feature amount. It is preferable to pattern-match the sample feature quantity and determine the type of contact with the robot.
At this time, it is preferable that the feature quantity calculation unit uses a differential component and an integral component of the horizontal component and the vertical component as a feature quantity.

Alternatively, the feature amount calculation unit calculates a component around the horizontal axis and a component around the vertical axis of the moment, which is a mechanical characteristic of the external force detected by the first detection unit, and calculates the calculated component around the horizontal axis And the component around the vertical axis as a feature amount, and the contact type determination unit acquires in advance the component around the horizontal axis of the moment, which is the mechanical characteristic of the external force, acquired so as to correspond to the type of contact and the feature amount. It is preferable that the feature amount and the sample feature amount are pattern-matched using the component around the vertical axis and the component around the vertical axis as the sample feature amount to determine the type of contact with the robot.
At this time, it is preferable that the feature amount calculation unit uses a differential component and an integral component of the component around the horizontal axis and the component around the vertical axis as the feature amount.

Alternatively, the feature quantity calculation unit uses the frequency characteristic of the external force calculated by Fourier transforming the value of the force component, which is the dynamic characteristic of the external force detected by the first detection unit, as the feature quantity, and determines the contact type The pattern matching is performed on the feature quantity and the sample feature quantity using the frequency characteristics of the external force acquired in advance so as to correspond to the type of contact and the feature quantity as the sample feature quantity, and the robot It is preferable to discriminate the type of contact.
At this time, it is preferable that the feature amount calculation unit uses the differential component and the integral component of the frequency characteristic of the external force as the feature amount.

  Alternatively, the feature amount calculation unit calculates an intersection of a straight line passing through a vector indicating the external force detected by the first detection unit and the surface of the robot or a position where the external force acts, and calculates the intersection or the calculated The position where the position is a feature amount, and the contact type determination unit is obtained in advance so as to correspond to the type of contact and the feature amount. Is preferably used as a sample feature amount, and the feature amount and the sample feature amount are pattern-matched to determine the type of contact with the robot.

  In addition, the feature amount calculation unit calculates a displacement from a trajectory of a predetermined target position in a predetermined part of the robot, uses the calculated displacement as a feature amount, and the contact type determination unit The feature amount and the sample feature amount are obtained by using, as a sample feature amount, a displacement from a trajectory of a preset target position at a predetermined part of the robot, which is acquired so as to correspond to the type of contact and the feature amount. And pattern matching to determine the type of contact with the robot.

Furthermore, a second detection unit that detects peripheral information of the robot is provided, and the feature amount calculation unit is configured to detect the position or face of a person's face in contact with the robot based on the detection result of the second detection unit. At least one of the orientation of the hand, the position of the hand or the orientation of the hand, and the contact type determination using at least one of the face position or the face orientation or the hand position or the hand orientation as a feature amount The part is acquired in advance so as to correspond to the type of contact and the feature amount, and uses at least one of a human face position or face orientation or hand position or hand orientation as a sample feature amount, and It is preferable to perform pattern matching between the feature quantity and the sample feature quantity to determine the type of contact with the robot.
At this time, it is preferable that the feature quantity calculation unit uses at least one differential component and integral component of the face position or the face orientation or the hand position or the hand orientation as a feature quantity.

  Furthermore, a third detection unit that detects human voice is provided, and the feature amount calculation unit performs voice recognition based on a detection result of the third detection unit, and determines a voice size or a word usage frequency or The content of conversation is used as a feature amount, and the contact type determination unit uses, as a sample feature amount, the volume of speech, the frequency of use of words, or the content of conversation acquired in advance so as to correspond to the type of contact and the feature amount. Preferably, the feature quantity and the sample feature quantity are pattern-matched to determine the type of contact with the robot.

  The contact type determination unit preferably performs pattern matching between the feature quantity and the sample feature quantity using an MT system, a k-NN method, or SVM.

It is preferable that a matching selection unit that selects the sample feature quantity to be subjected to the pattern matching is provided based on a detection result of any one of the first to third detection units.
A robot according to the present invention includes the above-described robot contact type discrimination system.

The contact type determination method for a robot according to the present invention is a contact type determination method for determining the type of contact when an external force is applied by contact with the robot, and is based on a position of the robot based on a position at which the external force is applied. A step of calculating a feature amount based on a detection result of a first detection unit that is disposed on a side and detects a component in three or more axial directions indicating the mechanical characteristics of the external force; the feature amount; Pattern matching the sample feature quantities acquired so as to correspond to the types, and determining the type of contact with the robot.
As described above, by using a force sensor capable of detecting components in three or more axial directions indicating the dynamic characteristics of external force as the first detection unit, it is possible to determine the type of contact with the robot with a small number of sensors. Therefore, it can contribute to simplification of the configuration of the robot. That is, the number of wirings can be reduced and the amount of calculation can be reduced.
Here, it is preferable that the step of calculating the feature amount includes a step of calculating the feature amount based on a result obtained by down-sampling the detection result of the first detection unit.

  As described above, according to the present invention, the type of contact can be determined with a simple configuration.

It is a figure which shows the contact action of a robot and a person. It is a block diagram which shows the control system of the robot by which the contact kind discrimination | determination system based on this invention is mounted. 1 is a block diagram schematically showing a contact type discrimination system according to a first embodiment of the present invention. In the contact kind discrimination | determination system of 1st Embodiment which concerns on this invention, it is a figure which shows the template produced | generated previously. It is a flowchart which shows the contact kind discrimination | determination method of 1st Embodiment which concerns on this invention. It is a block diagram which shows the different contact kind discrimination | determination system in 2nd Embodiment based on this invention. It is a block diagram which shows the contact kind discrimination | determination system in 4th Embodiment based on this invention. It is a block diagram which shows the contact kind discrimination | determination system in 5th Embodiment based on this invention. It is a block diagram which shows the contact kind discrimination | determination system in 6th Embodiment based on this invention.

The best mode for carrying out the present invention will be described below with reference to the accompanying drawings.
However, the present invention is not limited to the following embodiment. In addition, for clarity of explanation, the following description and drawings are simplified as appropriate.

<First Embodiment>
A robot contact type discrimination system, a robot provided with the contact type discrimination system, and a robot contact type discrimination method according to the first embodiment of the present invention will be described below.
The robot 1 according to the present embodiment is a so-called humanoid robot as shown in FIG. The robot 1 includes a head 2, a body part 3, an arm part 4, a leg part 5, and the like. The robot 1 is configured such that the head 2 and the body 3, the body 3 and the arm 4, and the connecting portion between the arm 4 and the leg 5 are movable. The arm portion 4 includes a hand portion 4a, a lower arm portion 4b, an upper arm portion 4c, and the like, and each joint portion is configured to be movable. The leg part 5 includes a foot part 5a, a lower leg part 5b, an upper leg part 5c, and the like, and each joint part is configured to be movable. The robot 1 is not limited to a humanoid robot.

  As shown in FIG. 2, the control system of the robot 1 includes a first detection unit 110, a second detection unit 120, a third detection unit 130, a fourth detection unit 140, a control unit 150, and a storage unit 160. The drive part 170 is provided. A robot contact type discriminating system (hereinafter, simply referred to as a contact type discriminating system) 100 can be realized by using such elements of the robot 1. That is, the contact type determination system 100 includes a first detection unit 110, a control unit 150, and a storage unit 160. The description of the second detection unit 120, the third detection unit 130, and the fourth detection unit 140 will be described in other embodiments.

  The first detection unit 110 is a force sensor that can detect components in three or more axial directions (X, Y, and Z axis directions) indicating the dynamic characteristics of external force. As the first detection unit 110 of the present embodiment, a six-axis force sensor is used. The 6-axis force sensor detects a moment around the X axis, a moment around the Y axis, and a moment around the Z axis in addition to the force components in the X, Y, and Z axes. The first detection unit 110 outputs the detected detection result (detection signal) to the control unit 150. The first detection unit 110 is a portion in which an external force acting at one place is not distributed and transmitted to the plurality of first detection parts 110 and can be detected almost entirely by the first detection part 110 at one place. Placed in. For example, the first detection unit 110 is disposed in the wrist portion, the upper arm portion, and the portion between the sole and the ankle of the robot 1.

  The control unit 150 includes a preprocessing unit 151, a feature amount calculation unit 152, and a contact type determination unit 153. A detection signal is input to the control unit 150 from the first detection unit 110. The controller 150 is an arithmetic processing unit having a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), and the like.

  The preprocessing unit 151 is a low-pass filter that filters the detection signal in order to flatten the detection signal input from the first detection unit 110. The preprocessing unit 151 outputs the flattened detection signal to the feature amount calculation unit 152.

  The feature amount calculation unit 152 calculates a feature amount based on the detection signal input from the preprocessing unit 151. For example, the feature amount calculation unit 152 calculates a horizontal component and a vertical component of a force component, which are dynamic characteristics of an external force applied by a person based on the detection signal input from the preprocessing unit 151. More specifically, as shown in FIG. 3, the feature amount calculation unit 152 replaces the force components in the three-axis directions detected by the first detection unit 110 with the world coordinate system. Then, the feature quantity calculation unit 152 uses the horizontal component of the external force (the divided component of the external force acting on the plane including the X axis and the Y axis) and the vertical based on the force component in the three axis directions replaced with the world coordinate system. A component (a divided component of an external force acting on a plane including the X axis and the Z axis) is calculated. FIG. 3 shows only the main part of the processing operation of the contact type determination system of the present embodiment.

As described above, the feature amount calculation unit 152 combines the horizontal component and the vertical component of the external force in the world coordinate system into a vector S _filtered (t), and a differential component u _diff (t) for the vector S _filtered (t) and An integral component u _integ (t) is calculated.
Here, the differential component u _diff (t) is expressed by [ _Equation 1], and the integral component u _integ (t) is expressed by [ _Equation 2].

The feature quantity calculation unit 152 uses the differential component u _diff (t) and the integral component u _integ (t) as the feature quantity, and sets the vector u as shown in [ _Equation 3].

  The feature amount calculation unit 152 performs the above-described processing intermittently while at least the detection signal is input from the preprocessing unit 151. The feature amount calculation unit 152 outputs the vector u to the contact type determination unit 153.

  The contact type determination unit 153 performs pattern matching between the feature amount and the sample feature amount acquired in advance so as to correspond to the contact type, and determines the type of contact with the robot 1. Here, the contact type determination unit 153 determines the type of contact with the robot 1 using, for example, an MT (Mahalanobis Taguchi) system.

但し、ｉ（ｉ＝１，２，・・・ｎ）は、１０ミリ秒毎にベクトルｙを算出する際の算出順番を示す。ｊ（ｊ＝１，２，・・・ｋ）は、一つの接触の種類において複数回サンプリング（例えば、握手を複数回サンプリング）する際のサンプリングの順番を示している。 That is, the contact type determination unit 153 reads the sample feature amount from the storage unit 160. The storage unit 160 stores sample feature amounts corresponding to the types of contact. For example, as a type of contact with the robot 1 to be determined, when shaking hands, when gripping and pulling the hand portion 4a, when guiding the robot 1 by gripping the hand portion 4a, when handing an object to the hand portion 4a, Assume that 4 is hit. Here, a person grasps the hand portion 4a and shakes, and the differential component and the integral component of the horizontal component and the vertical component of the force component, which are the dynamic characteristics of the external force acting at that time, are used as the sample feature amounts. Vector y is calculated. The vector y is calculated, for example, every 10 milliseconds (however, the calculation interval can be arbitrarily set), and the average value m and the standard deviation σ of the vector y at that time are calculated. Standardization is performed using the average value m, the standard deviation σ, and [Equation 4]. Therefore, the feature amount calculation unit 152 calculates the vector u every 10 milliseconds so as to correspond to the vector y.

Here, i (i = 1, 2,... N) indicates the calculation order when the vector y is calculated every 10 milliseconds. j (j = 1, 2,..., k) indicates the sampling order when sampling a plurality of times (for example, sampling a handshake a plurality of times) for one contact type.

The inverse matrix R ⁻¹ is calculated using the correlation coefficient matrix R expressed by Y _ij , [ _Equation 5], and [Equation 6] normalized using [Equation 4].

  Such a process is repeated a plurality of times for the handshake to generate the template shown in FIG. Furthermore, it repeats several times for every kind of contact, and the template shown in FIG. 4 is produced | generated for every kind of contact. The template generated in this way is stored in the storage unit 160.

The contact type determination unit 153 reads the average value m and the standard deviation σ from the storage unit 160, and standardizes the vector u using [Equation 7]. That is, the contact type determination unit 153 standardizes the calculated vector u every 10 milliseconds using all the average values m and standard deviations σ calculated in advance.

The contact type discriminating unit 153 reads the inverse function R ⁻¹ from the storage unit 160, and obtains U, U transposed matrix U ^T and [Equation 8] normalized using the inverse matrix R ⁻¹ [Equation 7]. using, it calculates the Mahalanobis distance D ^2.

At this time, the contact type determination unit 153 determines that the type of contact to which the inverse matrix R ⁻¹ belongs when the Malahanobis distance D ² is the smallest is the type in which a person actually contacts the robot 1. As described above, by using a force sensor capable of detecting components in three or more axial directions indicating the dynamic characteristics of external force as the first detection unit 110, it is possible to determine the type of contact with the robot with a small number of sensors. Therefore, it can contribute to simplification of the configuration of the robot. That is, the number of wirings can be reduced, and the amount of calculation of the control unit 150 can be reduced.

  Further, since the contact type determination unit 153 determines the type of contact using the MT system, the type of contact can be determined with high accuracy with a small amount of calculation. However, in the above embodiment, the MT system is used. However, the k-nn method (k nearest neighbors) may be used, and contact using a system having a learning function such as SVM (Support Vector Machine) may be used. The type may be determined.

  Incidentally, as shown in FIG. 3, the feature amount calculation unit 152 calculates the feature amount based on the detection signal of the first detection unit 110, and downsamples the detection signal of the first detection unit 110. After that, it is preferable to calculate the feature amount. The contact type determination unit 153 calculates the feature amount calculated after downsampling the detection signal of the first detection unit 110 in addition to the pattern matching using the feature amount calculated based on the detection signal of the first detection unit 110. And the sample feature quantity sampled so as to correspond to the feature quantity calculated after down-sampling in advance and pattern matching is performed to determine the type of contact. Thereby, even if it is a case where contact operation changes, the kind of contact can be discriminated favorably.

  In the present embodiment, the horizontal component and vertical component differential component and integral component, which are the mechanical characteristics of the external force, are used as feature quantities, but this is not restrictive. That is, since the 6-axis force sensor can detect the moment about the X-axis, the moment about the Y-axis, and the moment about the Z-axis as described above, the horizontal axis of the moment which is the dynamic characteristic of such external force Even if the differential component and the integral component of the component around the vertical axis and the component around the vertical axis are used as the feature amount, it can be implemented in substantially the same manner.

  In the present embodiment, the horizontal component and the vertical component of the force component, which are the mechanical characteristics of the external force, or the differential component and the integral component of the component around the horizontal axis of the moment and the component around the vertical axis are used as the feature amount. This is not the case. That is, the horizontal component and vertical component of the force component, which are the mechanical characteristics of the external force, or the component around the horizontal axis and the component around the vertical axis of the moment may be used as the feature amount.

  The storage unit 160 stores a program for controlling the robot 1, a template corresponding to the type of contact described above, and the like. The drive unit 170 is an electric motor, a speed reducer, or the like provided at each joint of the robot 1. The driving unit 170 is controlled based on a control signal from the control unit 150.

  Using the contact type discrimination system 100 described above, the robot contact type discrimination method of the present embodiment (hereinafter sometimes simply referred to as a contact type discrimination method) is performed as follows. In addition, the description which overlaps with description of the above-mentioned contact kind discrimination | determination system 100 is abbreviate | omitted.

First, as shown in FIG. 5, when an external force is applied to the robot 1, the first detection unit 110 arranged on the backbone side of the robot 1 from the position where the external force is applied outputs a detection signal to the control unit 150 ( S1).
Next, the preprocessing unit 151 of the control unit 150 filters the detection signal input from the first detection unit 110 and outputs the flattened detection signal to the feature amount calculation unit 152 (S2).

  Next, the feature amount calculation unit 152 calculates a horizontal component and a vertical component in the world coordinate system of the force component that is a dynamic characteristic of the external force based on the detection signal input from the preprocessing unit 151. Further, the feature amount calculation unit 152 uses the horizontal and vertical differential components and integral components of the force component, which is a dynamic characteristic of the external force, in the world coordinate system as the feature amount and outputs the feature amount to the contact type determination unit 153 (S3).

  The contact type determination unit 153 includes a differential component and an integral component of horizontal and vertical components in the world coordinate system of a force component that is a dynamic characteristic of an external force that is a feature amount input from the feature amount calculation unit 152, and a storage unit The sample feature values read from 160 are pattern-matched, and it is determined that the type of contact to which the closest sample feature value belongs is the type in which a person actually contacts the robot 1 (S4).

<Second Embodiment>
A contact type discrimination system and a contact type discrimination method according to the second embodiment of the present invention will be described below. In addition, the description which overlaps with the contact type discrimination | determination system and contact type discrimination method of 1st Embodiment is abbreviate | omitted.

  The contact type discrimination system and the contact type discrimination method according to the present embodiment calculate the frequency characteristic of the external force by Fourier transforming the value of the force component that is the dynamic characteristic of the external force detected by the first detection unit 110. The frequency characteristic of this external force is used as a feature quantity. That is, for example, when the hand is shaken, when the hand part 4a is grasped and pulled, when the robot 1 is guided by grasping the hand part 4a, when the object is handed over to the hand part 4a, when the arm part 4 is hit, the robot 1 is generated. The vibration (frequency) should be different. Therefore, the feature amount calculation unit 152 according to the present embodiment calculates the frequency characteristic of the external force that acts by contact, and uses the frequency characteristic of the external force as the feature amount. Therefore, the storage unit 160 previously performs sampling a plurality of times for each type of contact, and stores frequency characteristics at that time as sample feature values.

  The contact type determination unit 153 reads the sample feature amount from the storage unit 160, pattern-matches the feature amount that is the frequency characteristic of the external force and the sample feature amount, and the type of contact to which the closest sample feature amount belongs is It is discriminated that the person actually contacts the robot 1.

  Thereby, the contact type discrimination system and the contact type discrimination method of the present embodiment also use a force sensor capable of detecting components in three or more axial directions indicating the dynamic characteristics of the external force as the first detection unit 110. The type of contact with the robot can be identified with a small number of sensors. Therefore, it can contribute to simplification of the configuration of the robot. That is, the number of wirings can be reduced, and the amount of calculation of the control unit 150 can be reduced.

  Also in this embodiment, it is preferable that the contact type determination unit 153 determines the type of contact using the MT system. Thereby, it is possible to determine the type of contact with high accuracy with a small amount of calculation. Of course, the k-nn method may be used, and a system having a learning function such as SVM may be used. In short, a general pattern recognition technique can be used.

  Also in the present embodiment, the feature amount calculation unit 152 calculates the feature amount based on the detection signal of the first detection unit 110, and in addition to downsampling the detection signal of the first detection unit 110, the feature amount is calculated. Is preferably calculated. The contact type determination unit 153 calculates the feature amount calculated after downsampling the detection signal of the first detection unit 110 in addition to the pattern matching using the feature amount calculated based on the detection signal of the first detection unit 110. And the sample feature quantity sampled so as to correspond to the feature quantity calculated after down-sampling in advance and pattern matching is performed to determine the type of contact. Thereby, even if it is a case where contact operation changes, the kind of contact can be discriminated favorably.

In the present embodiment, the frequency characteristic of the external force is used as the feature quantity, but the differential characteristic and the integral characteristic of the frequency characteristic of the external force may be used as the feature quantity.
In this embodiment, the type of contact is determined only by the feature amount based on the frequency characteristic of the external force. However, as shown in FIG. 6, the horizontal component of the force component (and / or moment) that is the dynamic characteristic of the external force is shown. When combined with the contact type discriminating system and the contact type discriminating method of the first embodiment, wherein the differential component and the integral component of the (component around the horizontal axis) and the vertical component (component around the vertical axis) are feature quantities, It is possible to determine the type of contact with high accuracy. FIG. 6 shows only the main part of the processing operation of the contact type determination system of the present embodiment.

<Third Embodiment>
A contact type discrimination system and a contact type discrimination method according to the third embodiment of the present invention will be described below. In addition, the description which overlaps with the contact type discrimination | determination system and contact type discrimination method of 1st Embodiment is abbreviate | omitted.

The contact type discriminating system and the contact type discriminating method according to the present embodiment use, as the feature amount, an intersection between a straight line passing through a vector indicating an external force and the surface of the robot 1 or a position (coordinate) where the external force acts. That is, the feature amount calculation unit 152 is based on the detection result detected by the first detection unit 110.
A point of intersection between the straight line passing through the vector indicating the external force and the surface of the robot 1 or a coordinate at which the external force acts is calculated, and a point of intersection between the line passing through the vector indicating the external force and the surface of the robot 1 or the coordinate at which the external force acts is defined as a feature amount. Therefore, the storage unit 160 performs sampling a plurality of times in advance for each type of contact, and stores, as a sample feature amount, an intersection of a straight line passing through a vector indicating the external force at that time and a coordinate on which the surface of the robot 1 acts or external force acts is doing. Note that the calculation method disclosed in, for example, the Journal of the Robotics Society of Japan Vol22 No5, pp616-624, 2004, etc. is used as a method of calculating the intersection of the straight line passing through the vector indicating the external force and the surface of the robot 1 or the coordinate at which the external force acts. be able to.

  The contact type determination unit 153 reads the sample feature value from the storage unit 160, pattern-matches the feature value and the sample feature value, and the type of contact to which the closest sample feature value belongs is actually determined by the robot 1 It is determined that the type of contact has been made.

  Thereby, the contact type discrimination system and the contact type discrimination method of the present embodiment also use a force sensor capable of detecting components in three or more axial directions indicating the dynamic characteristics of the external force as the first detection unit 110. The type of contact with the robot can be identified with a small number of sensors. Therefore, it can contribute to simplification of the configuration of the robot. That is, the number of wirings can be reduced, and the amount of calculation of the control unit 150 can be reduced.

  Also in this embodiment, it is preferable that the contact type determination unit 153 determines the type of contact using the MT system. Thereby, the kind of contact can be discriminated with high accuracy. Of course, the k-nn method may be used, and a system having a learning function such as SVM may be used. In short, a general pattern recognition technique can be used.

  Also in the present embodiment, the feature amount calculation unit 152 calculates the feature amount based on the detection signal of the first detection unit 110, and in addition to downsampling the detection signal of the first detection unit 110, the feature amount is calculated. Is preferably calculated. The contact type determination unit 153 calculates the feature amount calculated after downsampling the detection signal of the first detection unit 110 in addition to the pattern matching using the feature amount calculated based on the detection signal of the first detection unit 110. And the sample feature quantity sampled so as to correspond to the feature quantity calculated after down-sampling in advance and pattern matching is performed to determine the type of contact. Thereby, even if it is a case where contact operation changes, the kind of contact can be discriminated favorably.

  In the present embodiment, the type of contact is determined using only the intersection of the straight line passing through the vector indicating the external force and the surface of the robot 1 or the coordinates on which the external force acts as a feature amount. However, the force component (the mechanical component of the external force) And / or moment) horizontal component (component around the horizontal axis) and vertical component (component around the vertical axis) and the differential component and integral component of the first embodiment, and the contact type discrimination system and contact according to the first embodiment When combined with the type determination method, it is possible to determine the type of contact with higher accuracy.

<Fourth embodiment>
A contact type discrimination system and a contact type discrimination method according to the fourth embodiment of the present invention will be described below. In addition, the description which overlaps with the contact type discrimination | determination system and contact type discrimination method of 1st Embodiment is abbreviate | omitted.

  As shown in FIG. 7, the contact type discrimination system 101 and the contact type discrimination method of the present embodiment include a robot in addition to the feature amount used in the contact type discrimination system and the contact type discrimination method of the first embodiment. A differential component and an integral component of displacement from a trajectory in which a preset target position at one predetermined portion is continuous are used as feature amounts. FIG. 7 shows only the main part of the processing operation of the contact type determination system of the present embodiment.

  That is, the contact type determination system 101 includes a second detection unit 120 that can detect an angle. As the second detection unit 120, an angle sensor such as a potentiometer or an encoder can be used. The second detection unit 120 is mounted on each joint portion of the robot 1. The second detection unit 120 outputs a detection signal to the control unit 150.

  The control unit 150 calculates how much displacement has occurred from the trajectory in which the hand position is preset as a predetermined part of the robot 1. That is, the feature amount calculation unit 152 of the control unit 150 determines the coordinates of the hand position of the robot 1 based on the detection result of the second detection unit 120, the length of each element of the arm unit 4, and the like (world coordinate system or base coordinate Either of the systems may be used). Here, the hand position of the hand portion 4a of the robot 1 is set in advance as a target position (target coordinates) according to the operation, and impedance control and compliance control are performed so that the target coordinates move on a continuous trajectory. The Therefore, the feature amount calculation unit 152 calculates the displacement between the coordinates of the current hand position and a preset trajectory. Further, the feature amount calculation unit 152 uses the calculated differential component and integral component of the displacement as the feature amount. Therefore, the storage unit 160 performs sampling a plurality of times for each type of contact in advance, and obtains the differential component of the displacement between the coordinates of the hand position of the robot 1 and the trajectory set in advance, and An integral component is stored as a sample feature amount. However, in the present embodiment, the differential component and the integral component of the displacement are used as the feature amount, but the displacement may be used as the feature amount.

  The contact type determination unit 153 reads the sample feature value from the storage unit 160, pattern-matches the feature value and the sample feature value, and the type of contact to which the closest sample feature value belongs is actually determined by the robot 1 It is determined that the type of contact has been made.

  As described above, the contact type determination system 101 and the contact type determination method are set in advance in a predetermined part of the robot 1 in addition to the feature amount used in the contact type determination system and the contact type determination method of the first embodiment. Since the differential component and the integral component of the displacement of the target position from the trajectory are used as the sample feature amount, it is possible to further improve the accuracy of determining the type of contact.

  Also in this embodiment, it is preferable that the contact type determination unit 153 determines the type of contact using the MT system. Thereby, the kind of contact can be discriminated with high accuracy. Of course, the k-nn method may be used, and a system having a learning function such as SVM may be used. In short, a general pattern recognition technique can be used.

  Also in the present embodiment, the feature amount calculation unit 152 calculates the feature amount based on the detection signal of the second detection unit 120, and in addition to downsampling the detection signal of the second detection unit 120, the feature amount Is preferably calculated. The contact type determination unit 153 calculates the feature amount calculated after downsampling the detection signal of the second detection unit 120 in addition to the pattern matching using the feature amount calculated based on the detection signal of the second detection unit 120. And the sample feature quantity sampled so as to correspond to the feature quantity calculated after down-sampling in advance and pattern matching is performed to determine the type of contact. Thereby, even if it is a case where contact operation changes, the kind of contact can be discriminated favorably.

<Fifth embodiment>
A contact type discrimination system and a contact type discrimination method according to the fifth embodiment of the present invention will be described below. In addition, the description which overlaps with the contact type discrimination | determination system and contact type discrimination method of 1st Embodiment is abbreviate | omitted.

  As shown in FIG. 8, the contact type discrimination system 102 and the contact type discrimination method according to the present embodiment include a human being in addition to the feature amount used in the contact type discrimination system and the contact type discrimination method according to the first embodiment. Is used as a feature amount. FIG. 8 shows only the main part of the processing operation of the contact type determination system of the present embodiment.

  That is, the contact type determination system 102 includes a third detection unit 130 that detects peripheral information of the robot 1. As the third detection unit 130, an imaging sensor such as a vision sensor (for example, a stereo camera) including a charge coupled device (CCD) or a complementary metal oxide semiconductor (CMOS) as an imaging element, or a distance measuring sensor such as a laser range finder is used. Can be used. The third detection unit 130 is disposed at a predetermined position of the robot 1 suitable for detecting a person's face and hand, and is configured to be able to swing the detection area vertically and horizontally, for example. The third detection unit 130 outputs a detection signal to the control unit 150.

  The control unit 150 calculates the face position, face direction, hand position, and hand direction of the person who contacts the robot 1. That is, the feature amount calculation unit 152 of the control unit 150 detects the face center based on the input detection signal, and uses the differential component and the integral component of the coordinates in the base coordinate system of the face center as the feature amount. At this time, the feature amount calculation unit 152 may use a differential component and an integral component of the distance from the third detection unit 130 to the face as the feature amount.

  In addition, the feature amount calculation unit 152, based on the input detection signal, outputs a unit vector in the front direction of the face in the base coordinate system and a predetermined part (for example, a hand position) of the robot 1 from the human face center. An inner product with the unit vector is calculated, and a differential component and an integral component of the inner product are used as feature amounts.

  Further, the feature amount calculation unit 152 detects the coordinates in the base coordinate system of the human hand position based on the input detection signal, and uses the differential component and the integral component of the coordinate of the hand position as the feature amount.

  Further, the feature amount calculation unit 152 is based on the normal detection unit vector of the person's palm and the predetermined center of the person's palm (for example, the hand position) in the base coordinate system based on the input detection signal. The inner product with the unit vector is calculated, and the differential component and the integral component of the inner product are used as feature amounts.

  Therefore, the storage unit 160 performs sampling a plurality of times for each type of contact in advance, and the human face position or face orientation or hand position or hand orientation differential component and integral component acquired at that time. Are stored as sample feature values.

  The contact type determination unit 153 reads the sample feature value from the storage unit 160, pattern-matches the feature value and the sample feature value, and the type of contact to which the closest sample feature value belongs is actually determined by the robot 1 It is determined that the type of contact has been made.

  As described above, the contact type determination system 102 and the contact type determination method are not only the feature amounts used in the contact type determination system and the contact type determination method of the first embodiment, but also the position of the face of the person in contact with the robot 1. Alternatively, since the differential component and the integral component of the face orientation, the hand position, or the hand orientation are used as the feature amount, the accuracy of discrimination of the contact type can be further improved.

  Also in this embodiment, it is preferable that the contact type determination unit 153 determines the type of contact using the MT system. Thereby, the kind of contact can be discriminated with high accuracy. Of course, the k-nn method may be used, and a system having a learning function such as SVM may be used. In short, a general pattern recognition technique can be used.

  Also in the present embodiment, the feature amount calculation unit 152 calculates the feature amount based on the detection signal of the third detection unit 130, and in addition to downsampling the detection signal of the third detection unit 130, the feature amount Is preferably calculated. The contact type determination unit 153 calculates the feature amount after down-sampling the detection signal of the third detection unit 130 in addition to the pattern matching using the feature amount calculated based on the detection signal of the third detection unit 130 And the sample feature quantity sampled so as to correspond to the feature quantity calculated after down-sampling in advance and pattern matching is performed to determine the type of contact. Thereby, even if it is a case where contact operation changes, the kind of contact can be discriminated favorably.

  In the present embodiment, all the differential components and integral components of the human face position, face orientation, hand position, and hand orientation are used as feature quantities. However, at least one of these differential components and integral components is used. May be used as a feature amount.

  In the present embodiment, the position of the person's face, the orientation of the face, the position of the hand, the differential component and the integral component of the orientation of the hand are used as the feature quantity, but the position of the person's face, the orientation of the face, the hand You may use a position and the direction of a hand as a feature-value.

<Sixth Embodiment>
A contact type discrimination system and a contact type discrimination method according to the sixth embodiment of the present invention will be described below. In addition, the description which overlaps with the contact type discrimination | determination system and contact type discrimination method of 1st Embodiment is abbreviate | omitted.

  As shown in FIG. 9, the contact type discrimination system 103 and the contact type discrimination method of the present embodiment include human features in addition to the feature amounts used in the contact type discrimination system and the contact type discrimination method of the first embodiment. The volume of the voice uttered or the frequency of use of words included in the voice uttered by the person or the content of the conversation is used as the feature amount. FIG. 9 shows only the main part of the processing operation of the contact type determination system of the present embodiment.

  That is, the contact type discrimination system 103 includes a fourth detection unit 140 that detects human voice. As the fourth detection unit 140, a general sound collecting microphone can be used. The fourth detection unit 140 is mounted on the robot 1. The fourth detection unit 140 outputs a detection signal to the control unit 150.

  The control unit 150 detects the volume of a voice uttered by a person, the frequency band of a voice, the use frequency of a word, the content of a conversation, and the like. That is, the feature amount calculation unit 152 of the control unit 150 detects the volume of the voice, the frequency band of the voice and the use frequency of the words, the content of the conversation, and the like based on the input detection signal, and the volume of the voice. , Voice frequency band, word usage frequency, conversation content level are defined as feature quantities. For example, you can speak with the "Hello" If a person to shake hands on the robot 1, or to speak with "please" If the person passes the object to the robot 1, when a person is to induce the robot 1 "to here. The frequency of use of these words is used as a feature amount. Further, when a person is emotional, the voice becomes louder or the frequency band of the voice becomes higher. Therefore, the storage unit 160 performs sampling a plurality of times for each type of contact in advance, and uses the voice volume, voice frequency band and word usage frequency, and conversation content acquired at that time as sample feature values. Storing. In addition, what kind of word is used as the feature amount can be arbitrarily set, and the correspondence between the loudness of the voice, the frequency band of the voice and the frequency of use of the word, the content of the conversation, and the type of contact Can also be set arbitrarily.

  The contact type determination unit 153 reads the sample feature value from the storage unit 160, pattern-matches the feature value and the sample feature value, and the type of contact to which the closest sample feature value belongs is actually determined by the robot 1 It is determined that the type of contact has been made.

  As described above, the contact type discrimination system 103 and the contact type discrimination method also include the voice volume, the voice frequency band, and the feature amount used in the contact type discrimination system and the contact type discrimination method according to the first embodiment. Since the use frequency of words, the content of conversation, and the like are feature quantities, the accuracy of discrimination of the type of contact can be further improved.

  Also in this embodiment, it is preferable that the contact type determination unit 153 determines the type of contact using the MT system. Thereby, the kind of contact can be discriminated with high accuracy. Of course, the k-nn method may be used, and a system having a learning function such as SVM may be used. In short, a general pattern recognition technique can be used.

  The contact type discrimination system, the contact type discrimination method, and the embodiment of the robot according to the present invention have been described above. However, the present invention is not limited to the above configuration, and can be changed without departing from the technical idea of the present invention. Is possible.

  In the above embodiment, a six-axis force sensor is used as the first detection unit 110. However, if a force sensor capable of detecting at least a component in the three-axis direction showing the dynamic characteristics of external force is used, the first detection unit 110 can be implemented in substantially the same manner. .

  In the above embodiment, pattern matching is executed for all templates stored in the storage unit 160. However, it is preferable to narrow down templates for pattern matching in advance. For example, the first embodiment will be described as an example. As illustrated in FIG. 2, the control unit 150 includes a matching selection unit 154. A detection signal is input from the first detection unit 110 to the matching selection unit 154. A threshold value is set in the matching selection unit 154 in advance, and the threshold value is compared with the input force component and moment. That is, since the force component and the moment are relatively small if the degree is a handshake, the matching selection unit 154 can determine that the type of contact is other than the handshake if the input force component or moment is larger than the threshold. Therefore, the matching selection unit 154 instructs the contact type determination unit 153 to omit the pattern matching with the template belonging to the handshake as the contact type when the input force component or moment is larger than the threshold value. Thus, by reducing the number of times of pattern matching, the burden on the control unit 150 can be reduced. In this case, also in the second and third embodiments, it is possible to similarly reduce the number of pattern matching using the matching selection unit 154.

  In the fourth embodiment, the matching selection unit 154 calculates the coordinates of the hand position of the robot 1 based on the detection result of the second detection unit 120, the length of each element of the arm unit 4, and the like. Then, the matching selection unit 154 compares the calculated displacement of the hand position with a predetermined trajectory and a threshold value. That is, when the robot 1 is pulled, the displacement becomes relatively large. Therefore, when the displacement is smaller than the threshold value, the matching selection unit 154 can determine that the type of contact is other than the operation of pulling the robot 1. Therefore, when the displacement is smaller than the threshold value, the matching selection unit 154 instructs the contact type determination unit 153 to omit the pattern matching with the template belonging to the operation of pulling the robot 1 as the contact type.

  In the fifth embodiment, the matching selection unit 154 determines whether or not an image with the face facing the front can be detected from the image information input from the third detection unit 130, for example. That is, since it is highly likely that an image with the face facing the front can be detected if it is a handshake, if the matching selection unit 154 cannot detect an image with the face facing the front from the input image information, the type of contact is the handshake. The contact type determination unit 153 is instructed to omit the pattern matching with the template belonging to.

  In the sixth embodiment, the matching selection unit 154 calculates the loudness of voice from, for example, the voice input from the fourth detection unit 140, and compares the calculated loudness with a threshold value. In other words, since the loudness of a person's voice is likely to be relatively small in the case of a handshake, the matching selection unit 154 determines that the template belonging to the handshake is the contact type if the calculated loudness is greater than the threshold value. The contact type determination unit 153 is instructed to omit the pattern matching.

  In the said embodiment, although it combines with other embodiment centering on 1st Embodiment, the combination is set suitably.

DESCRIPTION OF SYMBOLS 1 Robot, 2 heads, 3 torso parts 4 arm part, 4a hand part, 4b lower arm part, 4c upper arm part 5 leg part, 5a leg part, 5b lower leg part, 5c upper leg part 100 contact type discrimination system 101 contact Type discriminating system 102 Contact type discriminating system 103 Contact type discriminating system 110 First detection unit 120 Second detection unit 130 Third detection unit 140 Fourth detection unit 150 Control unit, 151 Preprocessing unit, 152 Feature quantity calculation , 153 Contact type discrimination unit, 154 Matching selection unit 160 Storage unit 170 Drive unit

Claims (17)

A contact type discrimination system for discriminating the type of contact when an external force acts by contact with a robot,
A first detector that is mounted on the robot and detects a component in three or more axial directions indicating the dynamic characteristics of the external force;
A feature amount calculation unit that calculates a feature amount based on a detection result of the first detection unit;
A contact type determination unit that pattern-matches the feature amount and a sample feature amount acquired in advance to correspond to the type of contact, and determines the type of contact with the robot;
A robot contact type discrimination system comprising: The feature amount calculation unit calculates a horizontal component and a vertical component of a force component that are mechanical characteristics of the external force detected by the first detection unit, and uses the calculated horizontal component and the vertical component as a feature amount.
The contact type determination unit uses the horizontal component and the vertical component of the force component, which are mechanical characteristics of external force, acquired in advance so as to correspond to the contact type and the feature amount, as the sample feature amount, and the feature amount. 2. The robot contact type discrimination system according to claim 1, wherein the type of contact with the robot is discriminated by pattern matching the sample feature quantity.   The robot contact type determination system according to claim 2, wherein the feature amount calculation unit uses a differential component and an integral component of the horizontal component and the vertical component as feature amounts. The feature amount calculation unit calculates a component around a horizontal axis and a component around a vertical axis of a moment, which is a mechanical characteristic of the external force detected by the first detection unit, and calculates the component around the horizontal axis and the calculated component around the horizontal axis The component around the vertical axis is used as a feature value.
The contact type discriminating unit uses a component around a horizontal axis and a component around a vertical axis, which are mechanical characteristics of external force, acquired in advance so as to correspond to the type of contact and the feature amount, as sample feature amounts. The robot contact type according to any one of claims 1 to 3, wherein the feature amount and the sample feature amount are pattern-matched to determine the type of contact with the robot. Discriminating system.   5. The robot contact type determination system according to claim 4, wherein the feature amount calculation unit uses a differential component and an integral component of the component around the horizontal axis and the component around the vertical axis as a feature amount. The feature amount calculation unit uses the frequency characteristic of the external force calculated by Fourier transforming the value of the force component, which is the dynamic characteristic of the external force detected by the first detection unit, as a feature amount,
The contact type determination unit uses the frequency characteristics of external force acquired in advance to correspond to the contact type and the feature amount as a sample feature amount, and performs pattern matching between the feature amount and the sample feature amount. The contact type determination system for a robot according to claim 1, wherein the type of contact with the robot is determined.   The robot contact type determination system according to claim 6, wherein the feature amount calculation unit uses a differential component and an integral component of a frequency characteristic of the external force as a feature amount. The feature amount calculation unit calculates an intersection of a straight line passing through the vector indicating the external force detected by the first detection unit and the surface of the robot or a position where the external force acts, and calculates the calculated intersection or the position. As feature quantity,
The contact type discriminating unit uses, as a sample feature amount, an intersection of a straight line passing through a vector indicating an external force and a position where the external force acts, which is acquired in advance so as to correspond to the contact type and the feature amount. The contact type of the robot according to claim 1, wherein the feature type and the sample feature quantity are pattern-matched to determine the type of contact with the robot. Discriminating system. The feature amount calculation unit calculates a displacement from a trajectory of a preset target position in a predetermined part of the robot, and uses the calculated displacement as a feature amount.
The contact type determination unit uses a displacement from a trajectory of a preset target position in a predetermined part of the robot, which is acquired in advance so as to correspond to the type of contact and the feature amount, as a sample feature amount, 9. The robot contact type discrimination system according to claim 2, wherein the feature quantity and the sample feature quantity are subjected to pattern matching to discriminate a type of contact with the robot. . A second detector for detecting peripheral information of the robot;
The feature amount calculation unit calculates at least one of a face position or a face direction, a hand position or a hand direction of a person in contact with the robot based on the detection result of the second detection unit, The feature amount is at least one of the position of the face or the orientation of the face, the position of the hand or the orientation of the hand,
The contact type determination unit uses, as a sample feature amount, at least one of a human face position, a face orientation, a hand position, or a hand orientation acquired in advance so as to correspond to the contact type and the feature amount. The robot contact according to any one of claims 2 to 9, wherein the feature amount and the sample feature amount are used for pattern matching to determine a type of contact with the robot. Type discrimination system.   The feature amount calculation unit uses the at least one differential component and integral component of the face position, the face orientation, the hand position, or the hand orientation as a feature amount. Robot contact type discrimination system. A third detector for detecting human voice;
The feature amount calculation unit performs voice recognition based on the detection result of the third detection unit, and uses the volume of the voice or the frequency of use of words or the content of conversation as a feature amount,
The contact type discriminating unit, which is acquired in advance so as to correspond to the type of contact and the feature amount, uses the volume of speech or the frequency of use of words or the content of conversation as a sample feature amount, and the feature amount, The robot contact type determination system according to any one of claims 2 to 11, wherein the sample feature amount is pattern-matched to determine the type of contact with the robot.   The contact type determination unit performs pattern matching between the feature quantity and the sample feature quantity using an MT system, a k-NN method, or SVM. The contact type discrimination system for a robot according to item 1.   The matching selection part which selects the said sample feature-value which carries out the said pattern matching based on the detection result of the said any one of the 1st thru | or 3rd detection part is provided, The any one of Claims 1 thru | or 13 characterized by the above-mentioned. The robot contact type discrimination system described.   A robot comprising the robot contact type discrimination system according to any one of claims 1 to 14. A contact type determination method for determining the type of contact when an external force is applied by contact with a robot,
A step of calculating a feature amount based on a detection result of a first detection unit that detects a component in three or more axial directions indicating a dynamic characteristic of the external force, which is disposed on the fundamental side of the robot from a position where the external force acts. When,
Pattern matching the feature amount and the sample feature amount acquired in advance to correspond to the type of contact, and determining the type of contact with the robot;
A method for discriminating the contact type of a robot.   The robot according to claim 16, wherein the step of calculating the feature amount includes a step of calculating a feature amount based on a result obtained by down-sampling a detection result of the first detection unit. Contact type discrimination method.

Images