Self-Optimizing Multi-Agent Manufacturing Systems Using Hybrid Optimization and Intelligent Control Techniques

Abstract

The increasing complexity and dynamic nature of modern manufacturing systems, driven by the Industry 4.0 and emerging Industry 5.0 paradigms, require intelligent, adaptive, and decentralized control frameworks. This paper proposes a novel Self-Optimizing Multi-Agent Manufacturing System (SOMAMS) that integrates multi-agent systems (MAS), hybrid optimization techniques, and intelligent control strategies into a unified architecture. The proposed framework adopts a layered design that combines supervisory coordination, hybrid Genetic Algorithm-Particle Swarm Optimization (GA-PSO), and reinforcement learning-based control to enable real-time, autonomous decision-making in distributed manufacturing environments. The system is formulated as a multi-objective optimization problem aiming to minimize makespan, energy consumption, tardiness, and operational cost under practical manufacturing constraints. The hybrid GA-PSO algorithm enhances the speed of convergence and the quality of the solution by effectively balancing exploration and exploitation, while reinforcement learning enables agents to dynamically adapt to environmental changes. The decentralized agent-based structure supports scalable coordination among product, machine, and resource agents, facilitating self-organization and resilience. Simulation results demonstrate the effectiveness of the proposed framework in achieving rapid convergence, generating Pareto-like trade-offs between conflicting objectives, and maintaining balanced machine utilization. Additional analyzes, including scheduling visualization, learning performance, and agent interaction patterns, validate the system’s ability to optimize production efficiency while ensuring adaptability and robustness. The integration of cyber-physical systems and digital twin concepts further enables real-time monitoring and continuous self-optimization. In general, the proposed approach provides a comprehensive solution for next-generation smart manufacturing systems, offering significant improvements in flexibility, efficiency, and sustainability. The framework also establishes a foundation for future advancements in deep reinforcement learning and agentic artificial intelligence for fully autonomous industrial systems.