نوع مقاله : مقاله پژوهشی (کاربردی)
عنوان مقاله English
نویسندگان English
Background: Significant advancements have been made in the development of processing tasks across various industries to enhance productivity, reduce costs, and improve flexibility. However, in the manufacturing industry, resource constraints remain a persistent challenge, often manifesting as precedence relationships among factors. Therefore, task scheduling requires greater precision and adaptability. To address this issue, this study presents a scheduling algorithm that utilizes a multi-agent system to solve the Flexible Job Shop Scheduling Problem (FJSP) in a dynamic input environment. Method: The resource precedence environment is modeled as a decentralized partially observable Markov decision process (Dec-POMDP), where tasks act as independent agents and select an available collaborative robot (cobot) based on their current observations. To overcome the challenge of managing the high-dimensional operational space in concurrent multi-task scheduling, an architecture is designed within a multi-agent reinforcement learning (MARL) system with intrinsic motivation. This model simultaneously considers both coordinated behavior among agents and individual agent performance. Results: The novelty of this research lies in the application of an intrinsic motivation-based multi-agent system for problem-solving and optimizing overall job execution time. Compared to well-known methods such as rule-based approaches, metaheuristics, and cooperative strategies, the proposed model demonstrates superior performance. The results from the case study highlight the flexibility and training stability of this model, suggesting that it can serve as an effective solution for addressing task scheduling challenges in various industries.
کلیدواژهها English