10:30 Uhr
Simulation-Based Resolution of Deadlocks in Automated Guided Vehicles using Deep Reinforcement Learning
Mustafa Jelibaghu | TH Aschaffenburg
Details anzeigen
Autor:innen:
Mustafa Jelibaghu | TH Aschaffenburg
Michael Eley | TH Aschaffenburg
Alexander Palatnik | TH Aschaffenburg
This paper discusses the use of deep reinforcement learning to resolve deadlocks in material flow systems with automated guided vehicles (AGVs). The paper proposes a strategy for dealing with deadlocks based on a single Agent reinforcement learning approach (SARL). The agent will find the optimal solution strategy in real time. The proposed approach is evaluated using a material flow simulation for a real use case in industry. The effectiveness in reducing the occurrence of deadlocks as well as the number of collisions in the system is demonstrated. This study highlights the potential of deep reinforcement learning for improving the performance and efficiency of material flow systems with AGVs.
Manuscript herunterladen
11:00 Uhr
Herausforderungen für die Projektierung von Mobilen Robotern (FTS und AMR)
Maximilian Dilefeld | DUALIS GmbH IT Solution
Details anzeigen
Autor:in:
Maximilian Dilefeld | DUALIS GmbH IT Solution
This paper elaborates the importance of extensive planning when deploying complex intralogistics systems involving Automated Guided Vehicles (AGV). The deployment for such a system is divided into different planning phases. Requirements for the tools which can be used in relation to the objectives in the respective phases are pointed out. An overview for different types of planning tools is given. Important challenges for a realistic implementation of AGVs in (3D)-simulation and how they can be overcome is discussed in additional detail. A virtual commissioning approach for connecting the real Fleet Manager and other software services that implement material flow logic is presented as a promising approach for AGV planning.
Manuscript herunterladen
11:30 Uhr
Ereignisdiskrete Modellierung autonomer Transportfahrzeuge mittels Open-Source Software
Viktor Artiushenko | Otto-von-Guericke-Universität Magdeburg
Details anzeigen
Autor:innen:
Viktor Artiushenko | Otto-von-Guericke-Universität Magdeburg
Sebastian Lang | Fraunhofer-Institut für Fabrikbetrieb und Automatisierung (IFF)
Marcel Müller | Otto-von-Guericke-Universität Magdeburg
Tobias Reggelin | Otto-von-Guericke-Universität Magdeburg
This contribution addresses the need for improved methods in modelling and simulating transportation vehicles in discrete event simulation (DES), since current commercial software solutions suffer from a limited adaptability, high costs, and slow performance. We introduce an object class that enables the addition of freely moving transport resources to open-source DES libraries, including collision-free motion modelling. Applicable to all object-oriented programming languages, this class design extends the functionality of existing open-source software. A component for an intralogistics transport vehicle was developed for the Python library Salabim, and its functionality was successfully verified in a two-vehicle simulation environment.
Manuscript herunterladen