Management of the Automated guided vehicle (AGV) system for Warehouses using Multi-agent Deep Reinforcement Learning
Management of the Automated guided vehicle (AGV) system for Warehouses using Multi-agent Deep Reinforcement Learning
Abstract
In the context of Industry 4.0, smart process automation has become a foundational element in the industrial sector. The evolving demands of global trade require the reduction of human intervention in repetitive processes, coupled with the increasing integration of artificial intelligence into industrial operations. One prominent example of these advancements is the Automated Guided Vehicle System (AGVS), particularly in the domain of product transshipment. AGVS refers to a system comprising a group of vehicles that operate within a shared environment, tasked with transporting products from specific departure points to designated destinations. The efficient functioning of such a system requires sophisticated management of task allocation and resolution of access conflicts in shared areas. The technical requirements of an AGVS call for advanced algorithmic designs, particularly in terms of product allocation and conflict management. Moreover, the incorporation of machine learning techniques—specifically deep reinforcement learning—into these systems can streamline the complexity of their formalism, thereby simplifying the implementation process for system architects. The objective of this project is to develop a novel multi-agent system for managing an AGVS within a warehouse setting, utilizing deep reinforcement learning algorithms. The agents will collaborate to achieve the overarching goal of efficiently transporting all products within the warehouse. Keywords: artificial intelligence, automated guided vehicles system, deep learning, reinforcement learning, multi-agents' system
Summary
In the context of Industry 4.0, intelligent process automation has become a fundamental part of the industrial sector. The increasing demands of global trade are forcing a reduction in human intervention in repetitive processes, coupled with increased integration of artificial intelligence into industrial operations. A prominent example of these advances is the Automated Guided Vehicle System (AGVS), particularly in the area of product transshipment. AGVS refers to a system consisting of a group of vehicles operating in a shared environment, responsible for transporting products from a specific point of departure to a designated destination. The proper functioning of such a system requires sophisticated management of task allocation and the resolution of access conflicts to shared areas. The technical specifications of an AGVS require advanced algorithmic designs, particularly in terms of product allocation and conflict management. Additionally, the integration of machine learning techniques—particularly deep reinforcement learning—in these systems helps simplify the complexity of their formalism, thus facilitating the implementation process for the architects of these systems. The objective of this project is to develop a new multi-agent system for the management of an AGVS in a warehouse, using deep reinforcement learning algorithms. Agents will work collaboratively to achieve the overall goal of efficiently moving all products within the warehouse.
Artificial intelligence, automated guided vehicle system, deep learning, reinforcement learning, multi-agent system.
Strong knowledge in machine learning, deep learning, Markov decision processes, and reinforcement learning.
Thesis director : Pr. Ahmed ELHILALI ALAOUI
Thesis co-supervisor : Pr. Mohamed RHAZZAF
Please send the application before September 30, 2024 to:
Cedoc.admission@ueuromed.org & a.elhilali-alaoui@ueuromed.org