|
TSMCSPO Based Position Control for a Hydraulic Manipulator |
|
Jie Wang, Hyun Hee Kim, Karam Dad Kallu, Min Cheol Lee(Pusan National University, Korea) |
| A new design approach of sliding mode control with terminal sliding mode and sliding perturbation observer for a hydraulic robot manipulator is described in this paper. A sliding perturbation observer can observe the unknown states and disturbance. The estimated states and perturbation are used in sliding mode for more accuracy tracking. However, it occurs some time delay when use the combined controller-sliding mode control with can drive system tracking error to origin with as faster speed than conventional one. Thus, terminal sliding mode is designed with both faster speed and precision. |
|
|
Path Planning by Multilayer Neural Network and Rapidly Exploring Random Tree |
|
Takahiro Yamashita(Kyushu Institute of Technology, Japan) |
| We propose a path generation method that combines a multilayer neural network (MLN) and a rapidly-exploring random tree (RRT), which is a path-planning method that uses random numbers. Specifically, an MLN that recalls the midpoint of the path is constructed using the data set created from RRT. The recall of the midpoint using this neural network is repeated and the path is generated. Through this method, it is possible to generate a repeatable path by using MLN at high speed. In addition, since the time required for learning is very short, relearning for environmental change is easy. |
|
|
Development of a Tiny Impact Device for Hammering Inspection |
|
Masamitsu Kurisu(Tokyo Denki University, Japan), Hiroaki Kuwahara(Corporate Manufacturing Engineering Center TOSHIBA Corporation, Japan) |
| This paper presents a development of a tiny impact device which can be mounted on a small robot for hammering inspection in a narrow place. The developed device utilizes magnetic forces of permanent magnets. It makes a movable magnet to be adsorbed to other magnet by changing the magnetic force balance and takes out an impact force generated at the moment of adsorption as a hammering force. The device can generate continuous impact with low-power consumption. Performance of the device estimated theoretically is discussed. The availability of the device is verified by a hammering test. |
|
|
Stable Grasping of Objects Using Air Pressure Sensors on a Robot Hand |
|
Dong Eon Kim, Ki Seo Kim, Jin Hyun Park, Aiing Lee, Jang Myung Lee(Pusan National University, Korea) |
| In this paper, we propose a grasping force measurement method by machine learning to control a 3-finger hand robot with attaching an air pressure sensor on its fingertip. Robotis's 3-finger robot was used for the end effector.
The method is carried out by firstly inserting and sealing the air pressure sensor at the fingertip of the finger robot, and then measure the air pressure at the time of grasping.
It can be seen that it is possible to measure the grasping force by linearizing the output value through machine learning on the measured air pressure sensing value. |
|
|
Foreground Segmentation using Convolutional Neural Network |
|
Ajmal Shahbaz, Kang-Hyun Jo(University of Ulsan, Korea) |
| This paper proposes foreground segmentation algorithm powered by the convolutional neural network. The proposed algorithm consists of two sub networks. First network is based on CNN which extracts and learn features from the input image. This network is termed as encoder. The second network based on transposed convolution expand the encoded feature map into foreground background binary mask. The proposed method is tested on baseline category of
change detection dataset. |
|
|
Automatic Rubber Attachment by a Robot Arm for Cardboard Punching Mold Fabrication |
|
Takashi Yoshimi(Shibaura Institute of Technology, Japan), Yuta Ito(Shibaura Institute of, Japan) |
| In this study, automation of rubber attachment task by a robot arm in cardboard punching mold fabrication factory is considered. This task is required to pick up a small rubber part and put it beside the blade accurately by the robot arm. We considered and proposed how to pick up the small rubber part, how to remove it from the mount, and how to put it beside the blade. The effectiveness of the proposed system, method and conditions is confirmed through experiments, and we found that the robot arm can execute this task with the almost same quality to the human worker. |
|