Grasping Motion for Small Non-Rigid Food Using Instance Semantic Segmentation
| Autor: | Tulapornpipat, Wiwat, Pimpin, Alongkorn, Chumkamon, Sakmongkon, Hayashi, Eiji |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2020 |
| Předmět: | |
| Zdroj: | Proceedings of the SICE Annual Conference 2020. :1540-1545 |
| Popis: | The importance of food automation becomes a challenge in this era since food is an essential factor for humans. In this paper, we designed the robot framework to autonomously generate grasping motion for non-rigid food objects. The system could recognize and localize objects and target regions for pick-and-place rather than only the position. Assembling a lunch box with different food required advance in both hardware and software to create an efficient process. The robot platform based on a seven-axis industrial robot arm equipped with instance segmentation based on Cascade Mask R-CNN for Japanese food. A modular end effector was designed and prototyped to facilitate soft gripper and vacuum pad on the single unit, which allows the system to handle different food objects. In the experiment, we also evaluated the performance of pick-and-place process. The system can successfully pick-and-place food into a lunch box with an outcome successful rate of 90%. The Society of Instrument and Control Engineers (SICE) Annual Conference 2020 (SICE2020), September 23-26, 2020, Chiang Mai, Thailand （新型コロナ感染拡大に伴い、現地開催中止） |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|