The OOPSLA98 Mid-Year Workshop on Distributed Object Technology for Manufacturing will provide a forum for experienced practitioners to share their insights on the application of object technology to the problems of distributed manufacturing. This workshop will also enable participants to collaborate on a vision for the future solutions that object technology might offer to manufacturing enterprises.

• An analysis of key challenges that must be addressed by the Object Technology (OT) community to meet the needs of manufacturing enterprises;
• A definition of Object Oriented (OO) domain models and processes for the manufacturing domain;
• Requirements specification of OO tools and environments for supporting concurrent engineering in distributed manufacturing;
• Considerations on the use of advanced OO techniques and technologies for distributed enterprise integration (e.g. Application Frameworks, Software Agents);
• Recommendations to industry groups (particularly the OMG Manufacturing DTF) on manufacturing industry needs and priorities.

Additional information is available in the accompanying background document compiled from the workshop proposal of the organizers.

Manufacturing is the set of activities that contribute to the product life-cycle from initial development through after-sale support. The manufacturing domain focuses on the management of factories, resources and processes that are involved in transforming materials into salable products.

The labels Computer Integrated Manufacturing (CIM) and Manufacturing Execution Systems (MES) are often applied to the software and hardware systems that support the core processes of manufacturing.

Object Technology has slowly been gaining acceptance in manufacturing applications. The early emphasis focused on using object-oriented development environments (e.g. C++, Smalltalk, Java) to create customized applications.

As software systems continue to grow in size and complexity, the importance of software reuse and maintainability increases. While a large number of general methods and notations have been developed for creating and reusing designs and architectures, domain specific techniques and methodologies are usually not available.

There is a need in the manufacturing domain to agree upon a domain model (including product and process models), a set of domain specific tools for concurrent engineering, and distributed architectures for factory integration. Object Orientation is an enabling technology for these goals.

The workshop will consist of multiple focused sessions spread over three days. The first day will focus on the state of the art in object technology solutions for manufacturing. The second and third day will concentrate on future trends and industry needs.

Each day will begin with a series of brief presentations selected from the position papers of workshop participants. The remainder of each session will consist of small group breakout sessions to address specific topics. The emphasis of the workshop will produce concrete deliverables that can become the basis for future work. The workshop will conclude with an overview plenary session to share results and coalesce outcomes.

Individuals interested in participating in this workshop should submit a short Position Paper (1-3 pages) defining experiences with the use of object technology for manufacturing. Participants may also provide supporting documentation on their topic such as references to documentation on their activities, an architecture diagram or visualization of the manufacturing system, recommendations for potential breakout sessions. Some participants will be asked to deliver a brief presentation of their manufacturing activity.

Please send submissions along with your name, address, voice phone, fax phone, e-mail, and affiliations to Prof. Giuseppe Menga at the e-mail address specified below.

Position papers may be in HTML, Acrobat, MSWord, or postscript formats (in order of preference). You are encouraged to provide URLs for information that supplements your submission.

Participant submissions are due before June 30th.

Paper Title

Authors affiliations and address

Introduction

Section 1

Section 2

.

Conclusion

References

Authors biography

Domain Modeling for Manufacturing:

• Description of product models for the exchange of information among those involved in the product life cycle;
• Description of process models to monitor and control distributed and interrelated manufacturing processes;
• Specifying and implementing business objects for manufacturing;

• Object oriented tools to support concurrent engineering;
• Tools for domain analysis and modeling;
• Integrating systems development with business process engineering;

• Using Agent technology and mobile code for distributed computing;
• Frameworks, Design Patterns, and Component Development for developing distributed architectures;
• Using distributed object infrastructures for manufacturing (CORBA, DCOM, Internet)

• Object Oriented solutions for distributed production planning and scheduling;
• Applying object technology to algorithmic control logic, sequencing, and routing;

Giuseppe Menga

Dip. di Automatica e Informatica

Politecnico di Torino

Corso Duca degli Abruzzi 24

10129 Torino, Italy

Phone/Fax: +39-11-564-7012 / +39-11-564-7099

menga@polito.it

Giuseppe Menga has been professor and has held the chair of Automatic Controls at the Polytechnic of Turin since 1981. For the last ten years, his academic and professional activity has been focused on modeling and control of computer-integrated manufacturing systems, and he has pioneered the use of object-oriented technology in this field since 1985.

His most significant publications can be found in the IEEE Transactions on Automatic Control, the IEEE Transactions on Robotics and Automation, Communications of the ACM, Journal of Object-Oriented Programming. He has contributed to several international books. He was the general chairman of the 1992 IEEE Robotics and Automation International Conference held in Nice (France).

Outside the academic area, his most significant achievements have been the design of the flexible manufacturing system (FMS) for Borg-Wagner in York, Pennsylvania and the conception of G++, a commercial software environment for the development of large object-oriented concurrent and distributed software applications specifically for the Computer Integrated Manufacturing domain.

Terence Lammers

BCAG Information Systems

P.O. Box 3707, MS 6C-KM

Seattle, WA 98124

Phone/Fax: 206.234.7657 / 206.234.2303

terence.l.lammers@boeing.com

Terry Lammers has over 6 years experience in domain engineering and system development in the factory automation domain. This experience includes work on 9 projects in the domain, all of which have used object technology. His major domain engineering projects are:

• The OO domain analysis of data collection,
• The design and delivery of components from the data collection domain for a shop business management system,
• Analysis, design and delivery of components for a graphic user interface framework for the same shop system,
• Analysis of a shop floor control domain model,
• Analysis of the product data management domain and prototyping of this domain,
• The definition of an OO domain engineering process,
• The analysis, design and delivery of a framework for the machine monitoring domain.
• Development of a chemical process monitoring application using the framework.
• Development of an equipment monitoring application using the framework.

Lammers current areas of expertise are:

• OOT, OO analysis, design and programming;
• Managing OO software development projects;
• OO domain analysis and engineering;
• System engineering, including software and hardware architectures;
• Technical software application architectures;
• Software engineering methods, practices and tools.

All attendees of the Midyear Workshops should remember to register before June 12 for reduced hotel rates. More information can be found at the following URL:

http://www.acm.org/sigplan/oopsla/oopsla98/midyear/log.html< / P>

In this paper we show how we have applied, in a current project, mobile agent technology to an industrial process in the field of textile manufacturing. The project involves several companies, each of which is a Europe-wide representative of one of the key phases of the textile production process (e.g. spinning, weaving, dyeing, finishing and ready made clothing). The main contribution of this paper is the description of the agent-oriented application framework which integrates their business enterprise information systems, to create a supply and distribution chain in the volatile (seasonal) environment of top fashion mens and womens wool garment manufacturing.

1. Introduction

For a long time research in Computer Integrated Manufacturing focused on finding new architectures for building complex distributed control systems for Factory Automation and Integration.

In this context Object Oriented Technology plays an important role, since it offers techniques for creating powerful reference models and scalable architectures. Design Patterns (see [1] for a survey) and Application Frameworks (see [2] for a survey) are two new techniques emerged from the Object-Oriented world. Design Patterns are a tool for describing solutions to common problems in a specific application domain. Frameworks are a technology that enables software reuse by providing a set of reusable and extendible components for a specific application domain.

Recent changes in the business environment of Computer Integrated Manufacturing (CIM) led to an increasing need to integrate process chains beyond company boundaries to customers and suppliers, in order to create network wide integrated distributed Virtual Factories (VFs) [3].

A Virtual factory is a specific example of a distributed system which supports highly specialized, concurrent and distributed task-planning and decision making, integrating enterprise-wide business functions with Flexible Manufacturing Systems (FMS).

In order to ensure coherence in problem-solving and decision making and to guarantee that end-users and application designers use a VF effectively, a technology which allows geographically dispersed activities to be flexibly combined and coordinated is necessary.

The goal of this paper is to show how we have applied Agent Technology (for a survey see [5]) and Code Mobility (for a survey see [4] in the development of a communication and coordination infrastructure between distributed information systems. The infrastructure integrates the production planning and marketing activities of several European textile manufacturing factories involved in a current ESPRIT project.

2. Production and Supply Electronic Chain

The project we describe will create a supply and distribution chain in the volatile (seasonal) environment of high fashion mens/womens wool garment manufacturing. The cycle of the textile process is made up of long chain of operations (spinning, weaving, dyeing, finishing and ready made or made-to-measure clothing), operation usually performed by different firms which cooperation is based on traditional schemes and methods. This organization makes difficult implementation of a continuous real-time adaptation of production planning, which constitutes a serious handicap, especially in the worsted and woolen industry, whose high priced goods, with high percentage of fancy colours and patterns are manufactured.

The textile market is characterized by rapid changes of fashion, style and look. This in turn requires deliveries within a period of weeks. This factor compels the operators in the different branches in the cycle to look for ways of establishing closer integration and creates the need for speed in communication.

The market requirement is therefore for an interconnecting network where product data coupled with a strongly integrated scheduling of the production between retailers, the garment producer and its suppliers, commission transformers and customers, reduces the time between securing an order and delivering the product to a minimum.

 

Our contribution to the project has provided the architectural infrastructure that allows the different local factories to automate the process of cooperation and it consists of three parts: distributed Product Data Management, distributed planning and scheduling system, and an interactive access system to the supply chain through the Web. Issues such as heterogeneity, scalability, reconfigurability, easy accessibility, performance and security were all considered when designing the framework.

 

The distributed Product Data Management (PDM) system will be accessible to all of the partners. Upstream (wool-spinning) the chain will respond mostly to the exchange of technical data with the objective of facilitating commercial transactions and identifying the product better. Product quality and technical aspects, quality and environmental (eco labelling) data will by-pass the manufacturer (in other words it will no longer be under the sole control of the manufacturer) and will flow to weavers and spinners directly. The PDM system will allow garment product data right to be tracked up to the wool data.

 

In the center of the chain (spinning-weaving-manufacturing) the emphasis will be mostly on order-material management aspects. This part of the chain will stress and demonstrate the viability of distributed planning and it will be mostly responsible for the quick response capabilities expected from the system.

 

The framework for interactive access to the supply chain through the Web will raise the inter-factory relationship from the current file transfer level, established by private networks with added value services, to the level of direct interaction between partners offering each partner the advantage of operating directly on the others facilities.

3. The Framework

The supply-chain infrastructure has been designed as an organization of cooperating agents, i.e. integrated software units of design that can be flexibly composed with other similar units to build complex systems.

Interactions among agents are established dynamically according to the dependencies among their functionality. The same capability may be provided by different agents and the same agent may provide several capabilities.

Agents can cooperate since they share the same communication language and a common vocabulary, which contains words appropriate to common application areas and whose meaning is defined in a shared ontology [6].

In our framework, agents communicate using code mobility as well: the content of a message is an entire program with a complexity which can range from simple procedures to objects with state and behavior (mobile agents). Code mobility is achieved using the IBM Aglets framework (see [7]) based on the Java language. Java provides a syntax for the transformation of the source code into a form (the script), which can be transmitted over the network and interpreted by a standard program, called virtual machine. Mobile agents have a behavior, a state and a location. They are able to migrate from one static agent to another one.

Our agent-based framework consists in a set of static and mobile agents organized in three functional levels:

• Supply Chain Level. A set of static agents ( Factory Agents) operate the physical factories which are part of the supply chain. They manage product, business, and information models and integrate the factorys business functions (order management) with its Enterprise Resource Planning (ERP) system. They reside on computers tightly coupled to the factories they operate and provide the execution environment for the mobile agents which migrate from other Factory Agents.

• Factory Level. A physical factory is usually organized in a hierarchy of control modules with different responsibilities allocated to each level [8]. At the highest level the Facility represents one complete production system and performs long-term, strategic planning. It communicates orders to a shop floor module which schedules production lots (a lot is an administrative entity indicating a group of identical pieces to be manufactured together) and coordinates the factorys FMS.

• User Level. The users access the services provided by the VF through a Web browser, which hosts mobile agents called Interface Agents. They help and guide the users in retrieving and using services and information through the supply chain. They interact with the Factory Agents in order to post the customers orders.

The separation into three functional layers increases the flexibility of the multi-agent organization and allows the designer to exploit the appropriate technology for the maintenance of the system. New users can access the distributed system through new Interface Agents, new collaboration protocols can be shared by the Factory Agents, and new physical factories can become part of the supply chain and offer new services.

1. Conclusion

Collaborating autonomous agents are becoming an increasingly attractive alternative to traditional hierarchical control architectures, offering better support for robustness, scalability, reconfigurability, and reusability, especially in systems where problem solving and decision making must be distributed. This is particularly true for Virtual Factories.

In this paper, we have shown how a supply chain integration infrastructure can be implemented using object-oriented technology, software agent technology, and code mobility.

 

2. References

Due to advances in networking technology, increasingly heterogeneous distributed computing resources are becoming available on the Internet. This proliferation has introduced new requirements in the design of distributed systems.

• Scalabiliy. The critical scaling issue is not the size of the applications to be developed or the number of applications that can be interconnected, but rather the flexibility of the distributed system to the introduction of new components or new interconnections.
• Interoperability. The distributed system should be able to interoperate with applications that were unknown or nonexistent prior to the creation of the system.

• Adaptability. A distributed system continually evolves: it should be able to rapidly adapt to new environmental conditions and requirements.

Effective use of the Internet by end-users and application designers calls for technology which allows distributed resources to be flexibly combined, and their activities coordinated. Object Oriented (OO) Technology has proved to be an extremely powerful software development technique which promotes modularity, composability, and reusability of complex systems. Unfortunately it has turned out to be excessively brittle with respect to the specific requirements of distributed systems (Guerraoui 1996), especially for the development of reusable application frameworks. This is due to the fact that in OO programming the interactions between two components are specified and implemented as operations on one of the two: the more likely the interactions are to be changed in future applications, the less reusable is the component that express the joint behavior (Mili et al. 1995). There is the need for a computational model which increases the adaptability of the reusable components of a framework in terms of the interactions with other components.

Several attempts to extend the traditional object model have been proposed in literature, with the goal of enhancing its flexibility in the development of distributed applications: Interaction Protocols (Bokowski 1996), Contracts (Holland 1992), Actors (Agha et al. 1992) The distributed object model underlying OMG CORBA (Vinoski 1997) and Software Agents (for a survey see (Riecken 1994)).

Software Agents are integrated systems that incorporate major capabilities drawn from several research areas: Artificial Intelligence, Databases, Programming Languages, and Theory of Computing. Usually, software agents are conceived as "one-off" systems built to investigate a single application. A new trend in Distributed Artificial Intelligence (DAI) considers software agents as software units of design, that may be customized and composed with other similar units to build complex systems.

The agent model presented is viewed as an extension of the traditional object model. In particular, the agent model will be described in terms of Agent State, Agent Behavior, Agent Interface, and Agent Identity.

• Agent State: Agents, like objects, have an internal state, which reflects their knowledge. However, this knowledge may be partially specified or based on default assumptions (Conrad, Saake, and Tuerker 1997). During its life span, an agent may extend or refine its knowledge, by performing inferences on its past and present states, or by acquiring information from the outside environment.

• Agent Behavior: Agents have a behavior which is determined by the set of admissible operations on their internal state. As in the RISC model, the internal operations of our agents are represented by first class objects, called actions. In contrast to traditional objects, whose behavior is specified in a procedural way inside a method definition, the behavior of our agents may be specified inside actions in a declarative way (e.g. by rules, constraints, goals) as well.

• Agent Interface: The agents have a message-based interface, which is defined in terms of the set of messages that the agent can interpret. The agents adopt the declarative approach to communication, which is based on the Speech Act theory (Searle 1969).

• Agent Identity: An agent's identity is that property of the agent which distinguishes it from all other agents in the same system. An agent's identity is represented by an identifier, such that no two distinct agents in the system are allowed to have the same identifier.

The agent application framework. builds on the agent model defined in previous section. Following the classification defined in (Brugali, Menga, and Aarsten 1997), the framework consists of elemental, basic design and domain dependent components.

• Elemental components determine the language in which the agent architecture is written. In the early stages of the frameworks life span (Brugali, Menga, and Aarsten 1997) they are conceived as white-box components. As the framework evolves, concrete implementations are provided, examples being communication modules which are to be specialized to embed specific protocols, knowledge representations modules, and reasoning modules.

• Basic design components are agent architectures which integrate elemental components of the framework. They differ from each other according to their internal structure.

• The other components are domain dependent and consist of concrete implementations of agents with specific functionalities.

A number of researchers are reexamining OO's basic potential, that the object paradigm can give the programmer tremendous flexibility (Guerraoui 1996). In some application domains (e.g. open distributed systems) the object model has turned out to be too brittle.

Several researchers are exploiting ideas from other disciplines in order to enhance the flexibility and usability of the object model in the development of complex systems. Agent technology is an attempt to accommodate basic OO concepts (e.g. abstraction, modularity) and advanced Artificial Intelligence techniques (e.g. reasoning, learning). The aim is to provide the programmer with a basic unit of design (the Agent), which enhances software modularity, maintainability, and reuse. Through the development of this framework, we are pursuing this idea further, by raising the level of reuse from single agent components to entire architectures (multi-agent systems) and up to the development of application frameworks.

Jeff Sutherland was organizer of OOPSLA'97 Workshop on Business Object Design and Implementation III. Here are some conclusions from the last year Workshop on Business Object Design and Implementation (see http://jeffsutherland.org/oopsla97/oo97final.html).

The variety of papers presented and the high level of expertise at the workshop led to new insights on several important issues:

Significant work has been done on the Resource-Event-Agent (REA) accounting model since the last workshop. The REA model is viewed as the core model for state-of-the-art accounting systems, and the appropriate pattern for automation of any exchange of resource in a business object system. Geerts and McCarthy discovered a new and better object design for REA implementation based on work presented by Fowler in last year's workshop [Fowl96]. Haugen is building an application using the REA pattern. SAP is moving towards it, as are other institutions.

Building application frameworks that enable horizontal reuse across domains and products is a challenging problem. Choudhry and Patel built upon the experience of Hertha et al [HBPP96] and others to design a domain framework for customer services.

Critical issues are business modeling and segmentation of business models into stable fragments, and public release of models useful across domains which are usually considered proprietary if developed by a commercial institution. An organization like the OMG is the best place to assemble the required expertise in a public forum to solicit and evaluate domain frameworks.

Enterprise architectures based on business objects are evolving rapidly. There is increasing integration of Business Process Engineering with implementation of a Business Object Architecture which enables reuse of more than implementation artifacts.

Marshall provided an excellent report on implementing a Java BOA in practice [Mars97]. The approach and implementation of a DBOA model presented by Kitty Hung is to achieve the of right granularity of Business Object reuse not only on code level but rather the reuse of business expertise and business knowledge. The Dynamic Systems Development Method is to provide a project management approach coupling with the Business Object Architecture in response to the dynamic change of business climate.

The software architecture on the factory floor currently consists of three vertical levels: enterprise resource planning (ERP), manufacturing execution system (MES) and Controls. The functional boundaries of these vertical systems are not defined, instead the real time response requirement is the differentiation of each of the system. In AMIS  Agile Manufacturing Information System, all elements of the factory floor  devices, resources, people, machines, jobs, orders, etc. Are represented by autonomous agents.

The system is self-organized into a virtual factory (and even enterprise) using the concept of emergent behavior. The dimension and architecture of the virtual factory depends on the Job (or order) and its complexity. The life span of the virtual factory depends on the job; the job is the "glue" that forms and keeps the virtual factory together.

The ultimate goal of the project is to develop a very flexible, sustainable, intelligent, evolving information system that can rapidly and seamlessly adapt to new environments, modify its own behavior, re-configure itself to a new architecture as a result of addition or removal of resources and changes in production.

The system PROMOS - PROduct Model of Odense Shipyard, has been devised to integrate the diverse CAD systems used by the shipyard. The result is an object oriented, 3-D enabling tool and is being successfully used in making strategic decisions during the design phase. The strategy decisions are then converted downstream into tactical process and schedule plans in a seamless manner.

The goals that prompted the development of the PROMOS system and the resulting architecture will be discussed. Following this, there will be a summary of the applications that are already implemented and those that are in the various stages of development.

Focussing on the steel hull, the key application is the planning of the build-strategy which is done interactively in 3-D. The paper will describe how this is then used in the down-stream process chain in a seamless manner.

Finally, the plans to increase the scope of integration using CORBA will be explained.

+-----------------------------------------------------------------------+
| Dan Dragomirescu                                                      |
+-----------------------------------+-----------------------------------+
| Politecnico di Torino             | phone:    +39-11-5647090          |
| Dip. Automatica e Informatica     | fax:      +39-11-5647099          |
| corso Duca degli Abruzzi, 24      | e-mail:   dand@athena.polito.it   |
| 10129 Torino (Italy)              |                                   |
+-----------------------------------+-----------------------------------+