The Path to a More Efficient Circular Economy by Integrating Deep Learning into Robotic Sorting Systems

Schlosser A, Walter J, Roming L, Franke J, Reitelshöfer S (2026)

Publication Type: Conference contribution

Publication year: 2026

Journal

Lecture Notes in Mechanical Engineering Springer

Publisher: Springer Science and Business Media Deutschland GmbH

Pages Range: 334-342

Conference Proceedings Title: Lecture Notes in Mechanical Engineering

Event location: Bologna, ITA

ISBN: 9783032211538

DOI: 10.1007/978-3-032-21154-5_37

Abstract

Automated robotic sorting plays a central role in boosting material recovery and reducing environmental impact in the circular economy. Lightweight packaging waste consists of heterogeneous and visually complex materials. This presents persistent challenges for object-level detection and classification. Deep learning–based instance segmentation offers a scalable and precise approach for identifying target materials and contaminants under dynamic real-world conditions. A newly developed image processing pipeline integrates structured pre-processing with state-of-the-art segmentation models. These models are trained on synthetic and real datasets designed to reflect both idealized and challenging scenarios, including occlusions, clutter, and variable lighting. In controlled environments, the models achieve high segmentation accuracy and real-time performance. However, generalization drops significantly when applied to real data, especially for visually ambiguous objects. These results highlight the need for domain-adaptive training and realistic variation in dataset design. Integrating such segmentation pipelines into industrial systems improves classification, reliability, and operational efficiency. Continued development of these technologies is essential to meet regulatory goals, reduce resource consumption, and promote sustainable practices in recycling.

Authors with CRIS profile

Alexander Schlosser Lehrstuhl für Fertigungsautomatisierung und Produktionssystematik (FAPS) Jonas Walter Lehrstuhl für Fertigungsautomatisierung und Produktionssystematik (FAPS) Jörg Franke Lehrstuhl für Fertigungsautomatisierung und Produktionssystematik (FAPS) Sebastian Reitelshöfer Lehrstuhl für Fertigungsautomatisierung und Produktionssystematik (FAPS)

Involved external institutions

Fraunhofer Institute for Optronics, System Technologies and Image Exploitation (IOSB) DE Germany (DE)

How to cite

APA:

Schlosser, A., Walter, J., Roming, L., Franke, J., & Reitelshöfer, S. (2026). The Path to a More Efficient Circular Economy by Integrating Deep Learning into Robotic Sorting Systems. In Holger Kohl, Günther Seliger, Franz Dietrich, Giampaolo Campana (Eds.), Lecture Notes in Mechanical Engineering (pp. 334-342). Bologna, ITA: Springer Science and Business Media Deutschland GmbH.

MLA:

Schlosser, Alexander, et al. "The Path to a More Efficient Circular Economy by Integrating Deep Learning into Robotic Sorting Systems." Proceedings of the 21st Global Conference on Sustainable Manufacturing, GCSM 2025, Bologna, ITA Ed. Holger Kohl, Günther Seliger, Franz Dietrich, Giampaolo Campana, Springer Science and Business Media Deutschland GmbH, 2026. 334-342.

BibTeX: Download