Contribution

Update of the SCARAB robot to sort valuable items in containers of residual waste

---

Abstract

In this paper the features of the autonomous mobile robot SCARAB are extended. SCARAB is now not only exchanging full waste containers with empty ones but also sorting out the valuable objects of the waste. For this task, a gripper was added to the robot’s end-of-arm tool. The fingers of the gripper have a Fin Ray design to robustly grasp the objects. Adaptions of the waste container allow to empty the waste onto a sorting table without additional actuators. Object detection is done with a YOLOv8 model which was initially trained with an open data set and improved with additional training data. In order to label this training data a standalone tool based on the Segment Anything Model (SAM) was developed. The paper shows the mechanical design of the gripper fingers, the adaption of the waste container as well as the design of a suitable sorting table. It is demonstrated that the waste sorting task is carried out robustly without the need of any additional expensive equipment.

Keywords: object detection, segmentation, waste sorting

How to Cite:

Reischl, D., Wenninger, J., Zwirtmayr, S. & Schröck, J., (2025) “Update of the SCARAB robot to sort valuable items in containers of residual waste”, ARW Proceedings 25(1), 7-12. doi: https://doi.org/10.34749/3061-0710.2025.1