An Approach for Bin Picking in High-Mix Low-Volume Manufacturing
conference paper
Short production cycles and customization have led to increasing demand for flexible, adaptive solutions in industrial automation. Bin picking plays a crucial role as it can provide a significant increase in productivity and efficiency in manufacturing processes. It is the ability of a robot to choose and pick an individual object from a pile of randomly arranged ones. This field has seen ample research in recent years, with a lot of solutions presenting various uses of Artificial Intelligence techniques. However the majority of the implemented machine learning methods still lack the ability to generalize sufficiently. This is especially important in the case of high-mix low-volume situations, where the objects tend to be unique and non-uniformly shaped. These situations are commonplace in e.g. 3D printing manufacturing lines, where training a single algorithm for bin picking proposes a challenge. To tackle this we propose a modular pipeline, which splits the problem into sub-questions, each of which refers to a separate component of the picking process. This enables creation of solutions, which consist of multiple approaches combined into one. We create a first experimental implementation by addressing two parts of the modular workflow - detection and selection. This combines existing state-of-the-art bin picking techniques into a single pipeline. Preliminary results show the challenges from creating modular solutions. We argue that the benefit of such a solution is that it would enable interoperable applications. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.
Topics
TNO Identifier
998039
ISSN
21954356
ISBN
9783031574955
Publisher
Springer Science and Business Media Deutschland GmbH
Source title
Lecture Notes in Mechanical Engineering
Editor(s)
Wagner, A.
Alexopoulos, K.
Makris, S.
Alexopoulos, K.
Makris, S.
Files
To receive the publication files, please send an e-mail request to TNO Repository.