fruit harvesting robots in japan

4
Adv.SpaceRes. Vol.18, No. If2,pp.(1/2)181-(l/2)184, 1996 Copyr@tB1995 COSPAR 0273- 1177(95~~8~3 Printed in Great B&w. All rights reserved 02?3-1~77~6~9.~0+0.~ FRUIT HARVESTING ROBOTS IN JAPAN N. Kondo,* M. Monta* and T. Fujiura** * (Ikayumu University, l-I-1 Tsushitnu-Naka, Okayatnu, Jupun ** Shimme Universi~, 1060 Nishi-Kuwatsucho, Mutsue. Japm W have developed harvesting robots for tomato 111,petty-tomato, cucumber 121 and grape .‘3/ in Japan. These robots mainly consist of manipulators, end-effecters, visual sensors and traveling devices. These mechanisms of the robot components were developed based on the physical properties of the work objects. The robots must work automatically by themselves in greenhouses or fields, since we are considering for one operator to tend several robots in the production system. The system is modeled after Japanese agriculture which is commonly seen to produce many kinds of crops in greenhouses and in many small fields intensively. Biopr~uction in space is somewhat similar to the agricultural system in Japan, because few operators have to work in a small space. Employing robots for bioproduction in space is considered desirable in near future. The following is a description of the harvesting robots. Generally speaking, there are four types of manipulators: Cartesian coordinate manipulator, Cylindrical coordinate manipulator, Polar coordinate manipulator, and Articulated manipulator. The basic mechanism of a manipulator is defined by its degrees of freedom, type of joint, link length and offset length. These factors determine the performance of a manipulator, therefore, they are assigned by using evaluation indexes when the manipulator is designed. As the evaluation indexes, v\c have used work envelope, measure of manipulatability 14, posture diversity and so on, It is important that the work envelope must reach all vvork objects, The measure of manipulatability w is calculated as w =ddetJ( B)JT( 01) (11 vvhere J ( 8 ) is a Jacobian matrix. This index represents the easiness for positioning the open end of a manipulator, so a large index signifies that the manipulator can move efficiently. Posture diversity is an index for a redundant manipulator which has more than 7 DOE: (degrees of freedom). Such a manipulator has high flexibility and can be a multi-purpose robot; however, it has many actuators and a heavy weight. Therefore, it is difficult to control. It has a redundant space. The middle points of the manipulator can move even when the end and the base of the manipulator are fixed. The angle of the redundant space at the object is called posture diversity. Xt’hen the posture diversity is big, the manipulator can have access to the object from many directions; therefore, it is very capable of obstacle avoidance Figure I shows the manipulator for tomato harvesting. It has 7 DOF including 2 prismatic joints and may’ also be used for petty-tomato harvesting.

Upload: n-kondo

Post on 21-Jun-2016

218 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Fruit harvesting robots in Japan

Adv.SpaceRes. Vol. 18, No. If2,pp.(1/2)181-(l/2)184, 1996 Copyr@tB1995 COSPAR

0273- 1177(95~~8~3

Printed in Great B&w. All rights reserved 02?3-1~77~6~9.~0+0.~

FRUIT HARVESTING ROBOTS IN JAPAN

N. Kondo,* M. Monta* and T. Fujiura**

* (Ikayumu University, l-I-1 Tsushitnu-Naka, Okayatnu, Jupun ** Shimme Universi~, 1060 Nishi-Kuwatsucho, Mutsue. Japm

W have developed harvesting robots for tomato 111, petty-tomato, cucumber 121 and grape .‘3/ in Japan. These robots mainly consist of manipulators, end-effecters, visual sensors and traveling devices. These mechanisms of the robot components were developed based on the physical properties of the work objects. The robots must work automatically by themselves in greenhouses or fields, since we are considering for one operator to tend several robots in the production system. The system is modeled after Japanese agriculture which is commonly seen to produce many kinds of crops in greenhouses and in many small fields intensively. Biopr~uction in space is somewhat similar to the agricultural system in Japan, because few operators have to work in a small space. Employing robots for bioproduction in space is considered desirable in near future. The following is a description of the harvesting robots.

Generally speaking, there are four types of manipulators: Cartesian coordinate manipulator, Cylindrical coordinate manipulator, Polar coordinate manipulator, and Articulated manipulator. The basic mechanism of a manipulator is defined by its degrees of freedom, type of joint, link length and offset length. These factors determine the performance of a manipulator, therefore, they are assigned by using evaluation indexes when the manipulator is designed. As the evaluation indexes, v\c have used work envelope, measure of manipulatability 14, posture diversity and so on, It is important that the work envelope must reach all vvork objects, The measure of manipulatability w is calculated as

w =ddetJ( B)JT( 01) (11

vvhere J ( 8 ) is a Jacobian matrix. This index represents the easiness for positioning the open end of a manipulator, so a large index signifies that the manipulator can move efficiently. Posture diversity is an index for a redundant manipulator which has more than 7 DOE: (degrees of freedom). Such a manipulator has high flexibility and can be a multi-purpose robot; however, it has many actuators and a heavy weight. Therefore, it is difficult to control. It has a redundant space. The middle points of the manipulator can move even when the end and the base of the manipulator are fixed. The angle of the redundant space at the object is called posture diversity. Xt’hen the posture diversity is big, the manipulator can have access to the object from many directions; therefore, it is very capable of obstacle avoidance

Figure I shows the manipulator for tomato harvesting. It has 7 DOF including 2 prismatic joints and may’ also be used for petty-tomato harvesting.

Page 2: Fruit harvesting robots in Japan

(l/Z3182 N. Kmdo et nf.

11 = Omm s1 = O--2OOmm 12 =400mm sz = O--3OOmm

12 I3 =lZOmm 8, = -150-150" 14 =250mm 8, = -30-100' 15 =ZOOmm 85 =-120--O*

11 l6 = 65mm 86 = -110-110" l7 =275mm 87 =-180-180

Fig. 1. Basic mechanism of tomato harvesting manipulator.

END-EFFECTOR

The mechanism of end-effector was also developed based on the physical properties of work object. The physical properties include shape and size, dynamical properties such as cutting resistance and elasticity, optical properties, sonic properties, and electrical properties. In addition, chemical properties and biological properties were considered when necessary. The guidelines for end-effector design depend on the work objects and the tasks. Therefore, the end effector was not designed for multi-purpose but for a specific purpose.

sure sensor Cold EiIter

Optical shield

Fig.2. End-effector for tomato harvesting. Fig.3. Visual sensor using laser and PSD

As an example, tomato harvesting end-effector i5’ is shown in Figure 2. Tomato fruits grow in clusters. A cluster has several fruits which are adjacent to one another and the peduncle joints have various forms. When a farmer harvests ripe fruits one by one in the cluster, he’she can pick them off easily by bending at the joints instead of cutting. The picking force was measured in three dimensional planes before designing the end-effector. In addition, other properties for developing the end-effector were measured. The end-effector first extends the suction pad driven by a DC motor. After securing the fruit by vacuum suction, the pad pulls the fruit to separate it from others until the air pressure in the pad reaches a threshold value. The fingers then move forward keeping the absolute position of the fruit, that is, the pad is moving backward at the same speed continuously. Finally, the fingers grip the fruit and bend it. A ten Newton force was applied as the grip pressure according to the experimental result.

t?sual sensor is one of the most irn~~ant external sensors for robots, as weI1 as, human eyes. The

Page 3: Fruit harvesting robots in Japan

When the cafor of spectraI ~~~~~~~~~ of an object is different front those of others, ~~s~~~~i~~~e~ image is made by using R,Cr, B (red, green, and bhre) sjg~a~s from a cofor TV genera or by using opticali fibers. For example, tomato fruits and leaves can be disc~mi~at~d by ~orn~~~g R signal with G signal, while a ~~~~~~r fruit which has similar color with CUC~~~~ leaves and Stems can be d~~~~g~~s~$~ by interference filters in the near infrared region. Both ~~c~rn~~ fruits and stems have different reflectance as compared to its leaf reflectance in the near infrared region. The water absorption band such as 970 nm and 1170 nm, chlorophyll absorption band, and 850 nm wavelength band are especially effective for the discrimination purpose.

.~~p~op~~at~ binary ~rnag~ can be ~~btained by d~t~rrn~nat~ou of ap~r~pr~a~e t~~e~ho~d level after di$~r~m~natio~. Before r~~og~it~on, some processes may be requited for the binary image such as smoo~~~~g* ~~~~~~c?i~~~ dilatation, thin~n~~ border f~~~o~vi~g, edge d~t~~t~~~ and so on, because the image usually ~o~ta~~s both the target object and ~n~essa~y objects (plus noise). Wheten the characteristics of object are re~o~~~~ed, it is ~rn~~a~t to extract the features from the image. There are many ~ss~b~~ flouts for shape re~o~n~~on, such a-s area, gametes, Feret’s d~arnet~~, moment, fractal Dimensions ~~terse~t~~n, orientation and so on. ~n~he~~~e, textual features extracted from a gray fevel image are so~et~~~s n~~e~ar~~ snch as angular second n~o~~nt~ contrast, inverse differer~ce ~o~~nt~ correlation and so on.

Distarxce r~~eas~r~~e~t is essential for robot guidance. W e~~e~~e~t~d to measure object ~s~t~o~s by f~~~o~v~rig three methods. These methods are based on the ~r~a~ip~~ of tr~a~g~~at~o~, If two ~~~~ge~ acquired at different places are obtained, the distance from the visual sensor to objects can be measured (similar to the fiction of human eyes). This deter is called b~~~cnlar stereo vision that is first r~etbod.

If a visuat sensor is attached to the end of a manipulator, the sensor also moves to the object, when the manipulator approaches it. The distance S from visual sensor to the object is calculated by equation (21, because the number of pisels representing the object increases when the manipulator naves toward the object. This second ~eth~ has advantages of bind able to detect the object hidden by ~b~tac~~ and to utilize the number of pixels of a visual sensor effi~~~I~tly.

where Nai is the number of pixels re~r~se~t~~~ an object in the i-th image (kt or 21, and I) is the dis~n~~ traveler bf the visual sensor.

The ~~eth~~ and the binocular stereo vision method are ~~ass~fjed as passive range finder. Qn the other hand. det~~t~~g by ~ro~ect~n g light and receiving it is cakd a&~ range finder that is third method. Figure 3 shows an ~hoto-e~e~trj~ sensor :G which gets the thr~e_d~me~s~~~~aI image jn~~ud~ng distance ~~fo~rnat~o~. One laser diode emits a red light beam ~~~a~e~engt~ : 67570 nm) and the other laser diode emits an ~nfmred light beam (~vave~ength : 830 nm) in the same optical axis, The red light is ~~du~at~d with a square-wave at 5 kHz, while the infrared tight is aviated at & kHx. The two light beams are joined in the same axis using a cold filter which reffects the infrared fight and t~a~s~~ts the visible light. After both light beams are received by the same PSD, the distance from the sensor to each scanning point is cumulated bq’ the d~~~u~~t~d voltage of the infrared bearn and the color is known from the ratio of vohage of both ~~a~~” The ~n~ber of picture eicmcnts M’as 60X.&I and the field of vision was 43.2X43.2 degree.

Page 4: Fruit harvesting robots in Japan

l/2)184 N. Kmdo et al.

TRAVELING DEVICE

In the ridge, wheel type traveling device is often used, while crawler type is used in the orchard, because the robot for fruit tree in the orchard has a relatively larger manipulator. In the terraced orchard, monorail type is used sometimes. Automation of these devices has been developed and the autonomous 4 wheel device 171 has already been manufactured for real applications.

FRUIT H.4RVESTING SYSTEM WITH ROBOT

Plant training systems and cultivation methods have been changed in the effort of improving productivity and qua&y and easing farmers’ work. A farmer can work even when part of a work object is hidden in the present plant training system, because he/she has excellent arms, hands, eyes, body, legs and brain, Performance of robot components is sometimes inferior to that of human beings, Therefore, plant training systems have to be investigated again for robotic production so that a robot can work more efficiently.

REFERENCES

1. NKondo, Y. Shibano, K. Mohri, T. Fujiura and M. Monta, Request to cultivation method from tomato harvesting robot, Acta Ho~iculturae 3 I!?, 567-572( 1992).

2. NKondo, M.Monta, YShibano and S.Arima, Cucumber harvesting system by robot, Proceedings of the Food Processing Automation Conference III, 45 l-460( 1994).

3. N.Kondo, Study on grape harvesting robot, IFAC, Mathematical and Control Applications in Agriculture and Ho~iculture, Pergamon Press, Tokyo, 243-246( 199 1).

4. T. Yoshikawa, Measure of m~ipulatability of robot arm, Journal of the Robotics Society of Japan 2.(l), 63-67( 1984)(in Japanese).

5. NKondo, h4. Monta, Y Shibano, K. hlohri, J.Yamashita and T.Fujiura, Agricultural ro~ts~2):~~anipulators and fruit harvesting hands, AS.4E Paper No.923518( 1992).

6. T.Fujiura, J.Yamashita and N.Kondo, Ag~cultural robots~l):~~ision sensing system, ASAE Paper No.923517( 1992).

7. J.Yamashita, KSato, T.Fujiura, N.Kondo and TImoto, Agricultural robots{+Automatic guided vehicle for greenhouses, ASAE Paper No.923544( 1992).