Agent-Based Manufacturing

As explained in the Foreword the idea of an international programme on Manufacturing Science was originally initiated in 1989 by Professor H. Yoshikawa of Japan, the then Professor of Precision Engineering at the University of Tokyo. His 1989 paper states

The intelligent manufacturing system takes the intelligent activities in manufacturing and uses them to better harmonise human beings and intelligent machines... integrating the entire corporation from marketing through the design, production and distribution, in a flexible manner, which improves productivity.

Future manufacturing, to stay competitive, should possess the capabilities to permit much faster response to changes in the business and technology environment. Companies should be able to transform their operations, organisation and technology at much shorter notice than what is prevalent today. A promising structure, in this respect, would be organic in nature and would resemble a conglomerate of distributed and autonomous units. The units would have the ability to self-determine their actions while communicating and cooperating with others to steer the manufacturing system towards meeting the challenges placed upon the organisation. This paper examines the fractal, bionic and holonic concepts that advocate such organic structures. The concepts are first explained and the application in manufacturing noted. Comparison of these concepts with respect to their design, operation and self-organising abilities is undertaken. This comparison reveals a rich insight into the similarities and differences among them that can be fruitfully exploited to fill any shortcomings in one or another concept.

For most of the 20th century, a central theme of industrial and academic research in manufacturing has been the development of flexible systems. In this chapter we discuss the notion of manufacturing flexibility and, in particular, the efforts toward achieving flexibility through advanced automation. After years of research into these flexible manufacturing systems it has become evident that the existing technologies alone are not the solution; effective system integration and coordination are also necessary. The second part of this chapter looks at recent efforts to achieve system integration and coordination through emerging technologies such as multi-agent systems and holonic manufacturing systems.

Properly architected holonic systems can enhance the ability of each set of players in the industrial automation and control market to deliver added value by encapsulating, reusing and deploying their specialized intellectual property at succeedingly higher levels of integration. Such an architecture expands on the HMS concept of cooperation domains to include both low-level control (LLC) and high-level control (HLC) domains. LLC refers to normal, non-holonic control and automation functions, while HLC refers to the integration of these functions into holons through the use of software agent technology. Function blocks, as defined in the International Electrotechnical Commission (IEC) 61499 series of standards, can be used for encapsulation, reuse, distribution and integration of both LLC and HLC functions, while HLC functionality can be standardized as defined by the Foundation for Intelligent Physical Agents (FIPA).

This paper documents the fact that the research in the two areas of holonic manufacturing systems and multi-agent systems contains substantial overlap. Many technologies and results achieved in the multi-agent area can be applied with advantage in the holonic field. This is especially true when talking about standardisation, which should enable interoperability of systems. The FIPA international consortium has introduced a systematic approach to development and maintenance of specifications and standards for the multi-agent domain. The principles of the FIPA Abstract Architecture are briefly described. Two examples of applications of FIPA standards in different areas within the scope of holonic systems (control, and production planning and supply chain management) are presented. FIPA standards have been recognised as a suitable candidate for ensuring complete interoperability in the holonic field.

An abstract model of holonic systems is presented using concepts of cooperating knowledge-based systems. The model is shown to be fault tolerant and to handle failures as locally as is possible. The model is sufficiently abstract to enable theoretical properties to be derived. In particular it is shown that if a task is reported as being completed correctly, then this is indeed the case. Tasks that are not explicitly aborted by the system are guaranteed to terminate successfully.

This paper investigates a computational model for an abstract HMS operational scenario with a view to outlining rules that should be implemented in a holonic system to ensure a correct dynamic behaviour, particularly during operational disturbances. The model considers operations at the substructure levels of holons. It is claimed that without such a model the system behaviour will be unreliable, potentially causing not only wasted production but also uncontrolled executions. Three features considered under the computational model are execution consistency, execution termination in a finite time and operational stability. A failed action must be rolled back to a recoverable point for possible redoing, execution must follow a sequence of rules, termination must be guaranteed, and stability must be ensured. The paper argues against the implementation of a holonic system without such behavioural features.

This chapter investigates the function block-based Holonic Manufacturing System (HMS/FB) architecture and specifies the HMS operational model by blending concepts from the HMS/FB architecture and Co-operating Knowledge Based Systems (CKBS). The HMS operational model addresses holonic interaction issues viz. negotiation and co-operation, during task distribution, scheduling, execution and execution control. We envisage two levels of sophistication of the HMS operational model: (1) a basic operational model, supporting only the minimum functionality of the HMS, and (2) a fault-tolerant operational model, supporting advanced operational strategies such as fault tolerance. We have used motor assembly testbed requirements, provided by the Japanese HMS consortium to illustrate the basic operational model, and we have generalised two holonic shopfloor layouts, namely, the DaimlerChrysler’s engine assembly testbed and Rockwell Automation’s proposed simulation testbed for elaborating the fault-tolerant operational model. Although the manufacturing requirements used for this study are necessarily abstract, the ideas discussed in this chapter can provide a useful abstract foundation for many HMS implementations.

Diagnosis is an important function of a holonic manufacturing system if the desired levels of stability, adaptability and flexibility are to be achieved. Our research agenda is to study holonic behaviours (such as diagnosis and control) through the incorporation of these behaviours into operational industrial systems. Given the lack of fielded holonic solutions in industry, we are currently constrained to use conventional systems in our work. In this paper we describe the development of a holonic diagnostic capability for a PLC-controlled vehicle assembly line. A novel model-based strategy is used for diagnosis. Because of the constraints imposed on model formation in this environment, a two-phase approach consisting of off-line fault space generation and online fault space analysis is used. The fault space analysis utilises heuristics to achieve the desired performance levels (diagnosis in less than 60 seconds and success rates of greater than 90%). Areas for further research in holonic diagnosis are identified.

In this chapter, desirable features of Holonic Manufacturing System (HMS) development and implementation environments are extracted, and HMS-Shell and JDPS are explained, as examples that provide these features. HMS-Shell is a GUI-based design tool for HMSs based on the Cooperation Domain (CD) that encapsulates cooperation partners and cooperation protocols. JDPS is a HMS implementation and execution environment based on reliable broadcast messaging that achieves complete location and multiplication transparency of entities. These features of HMS-Shell and JDPS make application programs independent of cooperation mechanisms, and enable easy dynamic reconfiguration and incremental development of HMSs while systems are in operation.

Transport agents, or holonic automated guided vehicles (H-AGVs), are important sub-systems of agent-based manufacturing systems such as holonic manufacturing systems (HMS). In this chapter concepts for the development of transport agents are presented, including taxonomy, agent models, architectures, realisations and the concept of a prototyping suite for the development of transport agents. Related work began at the start of the HMS project and has continued within several other projects.

Holonic features such as cooperative and autonomous behaviour can ensure the flexibility and robustness of a manufacturing system. When a holonic system is to be implemented, several important issues have to be resolved: How to describe the task to be completed? How to decompose a task for different production devices? How to optimise the workflow and co-ordinate the production?

Our holonic shot-blasting system is a demonstration, where two separate robots work together to complete surface treatment tasks. A 3D design model of a product, with selected surface treatment operations, is the starting point of holonic shot-blasting. The workpiece (the product) must be localised, with an optical sensor in the robot cell, before any treatment can proceed. The treatment plan is transformed into TCP paths. The final decision of the active robot depends on the kinematics of the robot, on the tool in hand, on the workload and on the free space in the cell. In our system, the decision mechanism is a blackboard and robots are active and willing to complete tasks.

Holonic manufacturing systems support a more plug-and-play approach to configuring and operating manufacturing processes, and thereby address increasing efforts to meet the needs for market responsiveness and mass customised products. This chapter is primarily concerned with the control approaches associated with holonic manufacturing systems. It addresses three key issues. Firstly, a clear business rationale for a holonic approach to manufacturing control is outlined. Secondly, a number of the key developments in holonic control systems are summarised. Finally a number of outstanding issues for the design and implementation of holonic control systems are highlighted.