EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

3052. Utilizing Graph Neural Networks for autonomous picking sequence optimization

Invited abstract in session WD-29: Optimization issues on graphs II (Contributed), stream Combinatorial Optimization.

Wednesday, 14:30-16:00
Room: 157 (building: 208)

Authors (first author is the speaker)

1. Dimitrios Tsakoumis
Industrial Management and Technology, University of Piraeus
2. Konstantinos Giannakos
University of Piraeus Research Center
3. Stathis Plitsos
Industrial Management and Technology, University of Piraeus
4. Pavlos Eirinakis
Industrial Management and Technology, University of Piraeus
5. Giulio Vivo
Centro Ricerche Fiat

Abstract

Warehouse picking is an essential part of assembly production systems. Robotic systems are often utilized to automate this process to increase efficiency and throughput. In this regard, optimizing the picking sequence of the semi-finished products and other components is significant for overall productivity improvement. In our approach, we model the picking process as a Travelling Salesman Problem (TSP), where a robotic agent needs to select the most efficient path to pick multiple different objects from various locations. We investigate different variants of TSP that arise in this setting based on the specificities of the robotic system utilized (e.g., incorporating a dual gripper for picking). Further, we explore the potential of Graph Neural Networks (GNNs) to handle the inherent complexities of such warehouse picking tasks. Although traditional methods for solving TSP have been extensively examined, GNNs have been recently proposed for solving various combinatorial optimization problems. Our choice for leveraging GNNs in this context arises from their excellence in capturing spatial relationships in graph structured data.

Keywords

Status: accepted

Back to the list of papers