Page 96 - Mechatronics for Safety, Security and Dependability in a New Era

Ch17-I044963.fm Page 80 Monday, August 7, 2006 12:42 PM
Ch17-I044963.fm
80 80 Page 80 Monday, August 7, 2006 12:42 PM

size towel. Experiment results have shown 90% success rate for the first edge to be correctly detected.
The percentage for the edge to be successfully gripped is 65%. The main failure reason is the failure to
find the best gripping angle for gripper B. Experiment results for edge tracing has shown 70% success
rate. The main reason of failure in tracing is the failure to determine the possible corner correctly. This
is due to the noise within the image(s) taken. Readjusting the threshold parameter can solve the
problem. Figure 7 shows an example of gripper A's movements (in this case Js2) during edge tracing.
This proves that our proposed method for edge tracing and corner confirmation have been successful.
0.65
0.65 A •
Tracing’s RCH-40’s RCH-40 Image processing area CCD camera
[] m 0.60 starting point position Js2
0.60
(Js2)
odinte a 0.55 I Corner conformation towel
Corner conformation
Point (by image)
Point (by image)
or B
Z-axis c 0.50 E D C Corner detected
Endoftracin
End of tracing Possible corner
0.45
0.25 0.30 0.35 0.40 0.45 0.50 0.55 At A At D At E
0.40
0.45
0.35
0.25
AtE
0.50
0.30
0.55
AtD
At A
X-axis coordinate [m]
X-axis coordinate [m]
Figure 7: Example on gripper movement pattern and corner confirmation during tracing
We have also conducted experiments for the whole process. The success rate of the whole process
without any failure stands at 50% and it took approximately 4 minutes to spread the towel. Since failure
can be detected and all the processes can be repeated over and over again, the success rate can
eventually reach 90% and over. Figure 8 shows the scene for the whole process.
(1) (2) (3) (4) (5) (6 ) (7)
turn
(10)
(8) (9) (10) (11) (12) (13) (14)
(12)
(11)
(14)
(13)
Figure 8: Scenes from towel spreading process
CONCLUSIONS
A method of towel spreading using a unique edge tracing method using a combination of infrared
sensors feedbacks and CCD camera images is presented. Manipulating deformable objects such as
clothes are indeed difficult and challenging, but with the right ideas, we have shown that clothes
manipulation can be successfully accomplished to a certain rate.
REFERENCES
[1] Hamajima K. and Kakikura M. (1998). Planning Strategy for Task of Unfolding Clothes. Proc. of
the 16* Annual Conf. of the Robotics Society of Japan, 389-390.
[2] Ono E., Kita N. and Sakane S. (1997). Unfolding a Folded Using Information of Outline with
Vision and Touch Sensor. Journal of the Robotics Society of Japan, 15:2, 113-121.
[3] Smith P.W., Nandhakumar N. and Tamadorai A.K. (1996). Vision Based Manipulation of Non Rigid
Objects. Proc. of IEEE Tnt. Conf. Automat, 4, 3191-3196.