A distributed physical architecture and data-based scheduling method for smart factory based on intelligent agents. (October 2022)

• Record Type: Journal Article
• Authors: Gu, Wenbin; Liu, Siqi; Zhang, Zequn; Li, Yuxin
• Is Part Of: Journal of manufacturing systems. Volume 65(2022)
• Page Start: 785
• Page End: 801
• Publication Date: 2022-10
• Subjects: Smart factory -- Personalized customization -- Intelligent agent -- Data-based scheduling algorithm -- Real-time scheduling mechanism; Manufacturing processes -- Periodicals; Production engineering -- Data processing -- Periodicals; Robots, Industrial -- Periodicals; Production, Technique de la -- Informatique -- Périodiques; Robots industriels -- Périodiques; Electronic journals; 670.42
• DOI: 10.1016/j.jmsy.2022.11.006
• Languages: English
• ISSNs: 0278-6125

Abstract

Abstract: With the increasing demands for personalized customization, the characteristics of order gradually show dynamics and diversity, which bring new challenges to the production mode of traditional factory. In the context of industry 4.0, the emergence of advanced technologies has made the vision of transforming manufacturing system into smart factory stronger and stronger. However, the establishment of smart factory based on the original automatic workshop needs to consider two problems: one is how to redesign the physical architecture of the factory, and another is how to improve the scheduling performance of the factory. Therefore, this paper proposes a distributed physical architecture of smart factory based on intelligent agents (DPASF-IA). First, it is divided into multiple units according to different functions, namely processing unit, transportation unit, storage unit and decision-making unit. Then, the intelligent agent (IA) is designed for the building of these heterogeneous units. Moreover, in order to improve the scheduling performance of smart factory, a data-based scheduling algorithm is proposed based on reinforcement learning which realizes the high-quality management and improves the performance of scheduling. Meanwhile, combing with the proposed architecture and scheduling algorithm, a real-time scheduling mechanism is also proposed. Finally, a prototype system experimental platform for performance evaluation of scheduling is built according to the proposed DPASF-IA. The experimental results show that the proposed method can realize real-time scheduling and has good scheduling performance compared with other scheduling methods. Highlights: This paper proposes a distributed physical architecture of smart factory based on intelligent agents (DPSAF-IA) according to different functions, namely processing unit, transportation unit, storage unit and decision-making unit. An intelligent agent (IA) is designed for the building of these heterogeneous which can realize three unified functions: state perception and information exchanging; real-time scheduling; controlling physical equipment. In order to improve the scheduling performance of smart factory, a data-based scheduling algorithm is proposed based on reinforcement learning which consists of five modules. Combing with the DPASF-IA and data-based scheduling algorithm, a real-time scheduling mechanism is also proposed. Experimental results show that the DPASF-IA can use the data-based scheduling algorithm to realize real-time scheduling, and has good scheduling performance compared with other scheduling methods.