Autonomy and viability in agent -based manufacturing systems
Abstract
Based on agent theory and organization, a new model of manufacturing systems termed agent-based manufacturing systems (ABMS) is developed. In an ABMS, agents are autonomous and working collaboratively with each other through the regulation of structured dialogues (task application protocols, TAP). Agents are connected by communication links. The autonomy functions of an agent are demonstrated by four internal functions: internal resource management, a reflexivity mechanism, a goal adjustment mechanism, and collaboration management. The four functions assist an agent to interact with other agents and with the environment (modeled by tasks) and to adjust itself to situate in the dynamic environment. Agents are similar to living systems. Based on the concept of living systems, autonomy and viability of agents are proposed and defined as new measures for manufacturing systems. A theory of agents' coexistence and parasitism in ABMS is developed to identify the performance characteristics of ABMS. The definition of viability is derived both at the agent level and at the system level. This research also develops a parallel simulator called Teamwork Integration Evaluation/Agent (TIE/Agent) for evaluating ABMS. TIE/Agent is programmed on a Paragon supercomputer. Four compute nodes represent an agent. The information transmission between agents is realized by the Message Passing Interface (MPI) communication library. Experiments are preformed in a case study format on TIE/Agent. The experiments demonstrate scenarios for applying TIE/Agent. They also validate TIE/Agent as a stable simulator. The results in the case-study experiments indicate that processing time of the tasks, rewards of the tasks, and centralization/distribution of resources are three significant factors to the system viability, whereas interarrival time of the tasks and cost of keeping agents functional are two insignificant factors. In addition, the experiments indicate that TIE/Agent. is an effective tool to study the relationships (1) between autonomy factors (including reflexivity mechanism, goal adjustment mechanism, and error recovery) and system viability; and (2) between the number of agents and the system viability. TIE/Agent can also help establish outsourcing policies for an ABMS. In summary, this research is performed to build a new study platform for future manufacturing systems that are highly distributed, autonomous, and collaborative. Models, measures, a simulator, and case studies are developed to investigate this new paradigm of manufacturing systems.
Degree
Ph.D.
Advisors
Nof, Purdue University.
Subject Area
Industrial engineering|Systems design|Management
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.