JPH0244403A - Holding force controller - Google Patents

Abstract

PURPOSE:To perform appropriate holding by inferring a holding force to be generated according to a prescribed rule corresponding to the degree of an inputted sensitive feature quantity, and generating the holding force. CONSTITUTION:A fuzzy inference part 20 performs a fuzzy inference according to the prescribed rule based on information with respect to the weight, the easiness of crash, and that of sliding of an object to be held supplied from an input part 10, and finds and outputs the holding force. A sensor 41 to detect reaction generated at the time of holding the object is provided at a holding part 40, and a holding force control part 30 drives and controls the holding part 40 based on a holding force command supplied from the fuzzy inference part 20. Also, to the control part 30, a detecting reaction signal from the reaction sensor 41 is inputted, and the holding part 40 is feedback-controlled by an ordinary method so as to set detecting reaction equal to a commanded holding force. In such a way, it is possible to perform the appropriate holding.

Description

[Detailed description of the invention] Summary of the invention Fuzzy inference of grip force is made in response to sensory instruction inputs such as ``heavy'', ``easy to slip'', and ``easily crushed'', and the inferred grip force is used to hold the object. Because it can handle linguistic information input, it has the characteristics of being able to grip a wide range of objects because there are few cases and it is easy to change the gripping force.1 There is a lot of gripping information.

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a gripping force control device for gripping a target object by a gripping means.

Conventional gripping force control involves giving a target value for the reaction force that occurs when gripping a target object, and using a sensor to increase the reaction force.
Some are controlled by feedback loops. However, if only the target values of the two reaction forces are given, the attribute information of the target object is insufficient, and the object that is easily crushed may be damaged. Because of this, and because the target value must be set in advance, there is a problem that the target objects are limited.

On the other hand, there is a method in which the attributes of the target object (weight, coefficient of friction, surface hardness, etc.) are quantified and manually calculated and the required gripping force or its range within the device is calculated or determined on a case-by-case basis. There are problems such as it is difficult to quantify the attributes and the program or hardware for calculating the two gripping forces or classifying the cases is complicated.

SUMMARY OF THE INVENTION This invention enables the user to intuitively indicate the attributes of a target object.
Moreover, it is an object of the present invention to provide a gripping force control device that can perform appropriate gripping.

The gripping force control device according to the present invention includes a gripping means for gripping a target object, and at least two types of gripping means related to attributes of the target object.
Human power means 9 for inputting the degree of sensory features of a type
An inference means for inferring the gripping force that the gripping means should generate according to a predetermined rule according to the degree of the manually input sensory feature, and a control means for controlling the gripping means to generate the gripping force given by the inference means. It is characterized by having the following.

Examples of sensory features related to the attributes of a target object include weight (lightness) and crushability (difficulty in crushing).

There are slipperiness (difficulty in slipping), etc.

Typical examples of inference means include eight fuzzy inference means, including fuzzy Φ computers, fuzzy controllers, and fuzzy inference devices.

This includes what is called a fuzzy inference calculation device. Also, dedicated devices (regardless of analog type or digital type) for fuzzy inference (for example, "Nikkei Electronics", July 27, 1987,
148-152 (see Nikkei McGraw-Hill, Inc.) as well as binary type computers, processors, etc. programmed to perform fuzzy inference.

According to the present invention, it is not necessary to strictly quantify the features related to the attributes of the target object, so input is easy. Furthermore, since the amount of sensory or linguistic information is handled, there is less need to differentiate between cases, and the hardware or software is relatively simple. Furthermore, the target gripping force can be easily changed, and the type of input information is not limited to just one type, so it may be applicable to gripping a wide range of target objects.

DESCRIPTION OF THE EMBODIMENT FIG. 1 shows an embodiment of the invention. The gripping force control device is composed of a sensory instruction manual unit 10, a fuzzy inferring unit 201, a gripping force control unit 30, and a gripping unit 40.

In this embodiment, special @
Weight as quantity (lightness), ease of crushing (difficulty in crushing)
Three types are used: and slipperiness (slip resistance). These features are set, for example, by rotating dials 11, 12, and 13, respectively. The input section 10 has these dials 11 and 12.
．． 13, and a signal processing section that outputs the amount of rotation of the dials tt-ta as an analog voltage or current signal, or a digital signal. The human power of characterizing the target object can also be realized using other means, such as voice input means, slide switch input means, etc.

The fuzzy inference 20 performs fuzzy inference according to a predetermined rule based on information regarding the weight, crushability, and slipperiness of the grasped object given from the input unit 10, and calculates and outputs the gripping force. . Details of fuzzy inference will be described later.

The gripping section 40 is configured by, for example, a robot hand, and is provided with a sensor 41 that detects a reaction force generated when gripping a target object. This reaction force sensor 41 can be constituted by a strain gauge, for example.

The gripping force control section 30 drives and controls the gripping section 40 based on the gripping force command given from the fuzzy inference section 20, and includes a speed command type servo motor or a position command type servo motor. The control unit 30 also includes a reaction force sensor 4.
The detected reaction force signal from 1 is input, and the gripping section 40 is feedback-controlled in a normal manner so that the detected reaction force is equal to the commanded gripping force. Control of the gripping force control section 30 may be achieved by fuzzy control.

The sensory attributes 9 of a target object, such as weight, crushability, and slipperiness, are expressed by membership functions. An example of membership functions for these attributes is shown in T43 (A).
, (B) and (C). For example, in Figure 3 (A), for 1 weight, there are nine types of membership functions that express linguistic information: quite light, light, somewhat light, a little light, normal, a little heavy, a little heavy, 1 heavy, and quite heavy. . Similar membership functions have been determined for crushability and slipperiness, as shown in Figures 3 (B) and (C). It goes without saying that the shape of the membership function is not limited to the triangular shape shown in the figure, but any arbitrary shape can be adopted.

Regarding the gripping force to be applied to the gripping force control unit 30, the fourth
As shown in the figure, it is quite weak and 1 weak.

Nine types of membership functions are set: . . . normal 2 . . . 2 strong, and very strong.

The fuzzy inference unit 26 may be realized by dedicated analog or digital type hardware as described above, or may be realized by a binary type computer, but in any case, the fuzzy inference unit 26 shown in FIG. 3 (A
), (B), (C) and membership functions as shown in FIG. 4 are preset.

The rules for fuzzy inference in the fuzzy inference unit 20 are so-called Il', than (if, then)
These are called rules, and for example, the following rules are set.

(1r) If it is quite heavy and is prone to crushing and slipping (then), increase the gripping force considerably.

■If 1F<, is crushed<<, and is easy to slip, increase the gripping force.

■If it is light, easily crushed, and difficult to slip, reduce the gripping force.

■If it is light, easily crushed, and slips easily, abandon the first grip.

Fuzzy inference according to such rules is performed, for example, as shown in FIG.

Heavy, hard to crush, as set in the above rule ■
The goodness of fit of the weight, crushability, and slipperiness that were manually applied from the one-manpower unit IO to the membership function of ``easiness to slip'' is calculated, and the one with the smallest goodness of fit (old N operation) is selected. be done. and,
The membership function of "fairly strong" set in rule (2) is judged by the smallest degree of fitness in 1-, and the degree of gripping force indicated by diagonal lines is determined. Similarly, for the rules after the above rule (1), individual inferences are made for each rule based on the data set in the input section 10. and,
The gripping force membership functions (shaded area) obtained by these individual inferences are integrated (MAX calculation).
, for example, by determining the position of the center of gravity, a determined gripping force can be obtained. This gripping force is applied to the control section 30.

In the above embodiment, three rules (1) to (4) are shown as examples of rules, but it goes without saying that an appropriate number of these rules may be set. Also, the rules may be changed as appropriate. It goes without saying that the number of membership functions is not limited to nine for each feature, but an appropriate number can be provided and can be changed as appropriate.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the invention. FIG. 2 is a front view showing an example of the sensory instruction manual unit. Figure 3 (A), (H), and (C) are each 9
It shows an example of the membership function of weight, crushability, and slipperiness, and FIG. 4 shows an example of the membership function of gripping force. FIG. 5 shows the inference process in the fuzzy inference section. lO...Sensory instruction input unit. 20...Fuzzy reasoning section. 30...Gripping force control unit. 40...Gripping part. Applicant for the above patents) γ-ishi Electric Co., Ltd. Representative Patent attorney Kenji Ushiku (1 other person) jl!
Figure 5 ■ If) *<, and if it is hard to crush and easy to slip (↑h button), then it would be good! ) Strongly input section 10 Input from human force section IO Kari fishing force λ force section 10 Input from the fish

Claims (1)

[Scope of Claims] Gripping means for grasping a target object; input means for inputting degrees of at least two types of sensory features related to attributes of the target object; A gripping force control device comprising: inference means for deducing the gripping force that the gripping means should generate according to a predetermined rule; and control means for controlling the gripping means to generate the gripping force given by the inference means.