Intellectual Property India Publishes Patent Application for Graph Neural Network-Based Collaborative Multi-Agent Cyber-Physical System For Intelligent Manufacturing Process Optimization

MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202641109022 A) filed by S Singaravelan; Dr. J. Veneeswari; R. Srinivasanbalu; D. Mohanapriya; R. Madhu Keerthana; J. Yamuna Bee; S. Saranya; V. G. Saranyavaishalini; K. Vignesh; G. Akiladevi; P. Sathya; and Dr. Aravind B on September 10, 2026, for Graph Neural Network-Based Collaborative Multi-Agent Cyber-Physical System For Intelligent Manufacturing Process Optimization.

Inventors include S Singaravelan; Dr. J. Veneeswari; R. Srinivasanbalu; D. Mohanapriya; R. Madhu Keerthana; J. Yamuna Bee; S. Saranya; V. G. Saranyavaishalini; K. Vignesh; G. Akiladevi; P. Sathya; and Dr. Aravind B.

The application for the patent was published on September 18, 2026, under issue no. 38/2026.

Abstract: The present invention relates to an intelligent manufacturing optimization system based on a Graph Neural Network (GNN)-enabled collaborative Multi-Agent Reinforcement Learning (MARL) framework integrated with a Cyber-Physical System (CPS). The system comprises a plurality of manufacturing entities including machines, robotic units, automated guided vehicles, workstations, sensors and production resources, wherein each manufacturing entity is represented as a node and relationships between the manufacturing entities are represented as edges to form a dynamic manufacturing graph. A plurality of intelligent agents associated with the manufacturing entities acquire real-time operational parameters from the physical manufacturing environment through sensors and industrial communication networks. A graph neural network processes the dynamic manufacturing graph to generate relational and contextual representations of the manufacturing entities. A collaborative multi-agent reinforcement learning module utilizes the graph representations to determine coordinated actions including production scheduling, machine selection, resource allocation, job sequencing and material transportation. A multi-objective reward mechanism evaluates production throughput, machine utilization, processing time, waiting time, energy consumption and production quality. The cyber-physical system continuously updates the graph and agent states based on real-time manufacturing conditions, enabling adaptive decision-making. The proposed invention thereby provides intelligent, distributed and collaborative optimization of manufacturing processes while improving production efficiency, resource utilization and operational adaptability.

Disclaimer: Curated by HT Syndication.