Design und Evaluation eines automatischen Entscheidungssystems für die Greiferauswahl bei der Kommissionierung

Abstract

Grasping objects poses a challenge in the automation of order picking due to the diverse types of objects and complex real-world scenarios, necessitating the selection of an appropriate gripper for each object type. Existing gripper selection methods focus on generalized gripper selection systems able to grasp a large variety of objects in non-cluttered scenarios for handling in industrial applications. Within this paper, a knowledge-based gripper selection method for e-grocery items in cluttered scenarios is implemented as a binary decision tree. The results are validated through empirical tests, demonstrating an overall accuracy of 90.7 %. As the percentage of true negatives is 81.6 %, it is necessary to combine grasping principles to reduce the percentage of True Negatives in the future.