PROPOSAL BAA 95-37 Volume I

Title: Design Space Colonization

Duration: three (3) years

Business Types: Educational

Prime Contractor: Stanford University

Contact: Mark Cutkosky
Center for Design Research
560 Panama Street
Stanford, California 94305-2232
voice: (415) 725-1588
fax: (415) 725-8475
e-mail: cutkosky@cdr.stanford.edu

Administrative Contact: Ruth Kaempf
Senior Contract Officer
Sponsored Projects Office
857 Serra St., Rm. 260
Stanford University
Stanford, CA 94305-4125
voice: (415) 723-4740
fax: (415) 723-1654
e-mail: as.rmk@forsythe.stanford.edu

Subcontractors: Option-1: Pipeline Systems Analysis

(B)Executive Summary

Political, economic and technological forces are changing the landscape of engineering. There is an increasing need for organizations to form joint design and manufacturing teams that collaborate for the life of a project and then disperse. These teams need to quickly locate, evaluate and make effective use of the best resources (tools, facilities, people) available, wherever they may be found. In this proposal we concentrate on a subset of this problem that is most critical in the early stages of developing a product, when new ideas, processes, components and materials are being explored and prototypes are developed. We propose solutions that will enhance the ability of teams to exploit novel tools and processes for improving product performance and reducing costs, and to document their explorations for others to follow. In this way we believe that the slow cycle of exploration, maturation and widespread adoption that typically accompanies significant advances in materials and processing capabilities can be compressed from decades to years.

Innovative Ideas

Design Space Colonization (DSC) is the component of the Palo Alto Collaborative Engineering (PACE) effort, consisting of CDR, EIT, KSL, and Lockheed-Martin, that addresses the ability of teams of engineers to explore systematically the design space of complex products. By analogy with space exploration, designers explore the design space to find optimal or novel solutions. As they explore, they colonize the space with examples and organize and map it with models and documentation for others to follow. We propose to provide revolutionary computer support for these two major functions of design exploration: design mapping/navigation and population.

Design navigation and mapping concern decision and coordination support for systematic group exploration of alternatives, trade-offs, conflicts, etc. It is achieved by providing organized documentation of the design, the design process and the provision of active help in tracking, mapping, and generally navigating through a team-generated design space. This functionality is imperative for a large project involving exchanges among engineers in various disciplines over an extended period, as characterizes the design of many defense systems. Based on our experiences with MadeFast and other distributed design exercises, we argue that a formal distributed design process is essential for virtual design teams to undertake such projects and to receive support from computational tools for generating and organizing design documentation.

Design space population concerns the generation and analysis of novel design alternatives and choices. There is a substantial body of research and technology available on this general topic. Our focus will be on an important gap in this technology: enabling teams of engineers to take advantage of the very large design spaces made possible by novel manufacturing processes and materials such as the solid free-form fabrication (SFF) manufacturing processes that have recently been developed at Stanford and a few other institutions. These processes have an exciting potential for realizing designs that have never before been feasible to produce but only if engineering teams can easily find out about them, evaluate them and learn how to use them effectively.

Our general approach will be to use: (1) structured design methods, drawing upon previous work with Redux [Petrie et al. 1994] and DesignRoadmap [Park 1995] on formal design process representation, coordination and mapping and (2) an Agent-Based Engineering (ABE) approach in which capabilities, design rules and simulation code associated with novel processes are loaded on-demand into the designers' working environment. The ABE approach builds upon results of the SHARE [Toye et al. 1994] and SHADE (SHAred Dependency Engineering) [McGuire et al. 1993] projects to develop generic ontologies, protocols and APIs for integrating informal documentation (as captured with tools like PENS [Hong et al. 1995]) and formal results (as produced by the proposed Constraint Manger agent), and ensures that the tools developed by different communities can inter-operate.

Comparison to current Approaches

When in doubt
Make it stout, 
Out of things 
You know about
                (D. G. Ullman, The Mechanical Design Process, McGraw Hill,1992)

In large development projects design changes typically flow through a bureaucracy of change orders and management structures, and ultimately from one engineer to another. Mistakes are made and opportunities lost along the way. Coordination is recognized as the crucial bottleneck of large defense projects today [Forney 1995]. Novel and rapidly developed designs can only be achieved by small co-located teams. Projects involving multiple organizations have little chance of bringing in a good design, on time and under budget.

Given these difficulties, it is understandable that engineering teams are hesitant to explore novel and somewhat risky alternatives to tried and true materials and processes; hence the truism quoted at the top of this section. Deficiencies in the design process provide too many opportunities that unforeseen problems will necessitate expensive last-minute corrections. Moreover, with projects running behind schedule, there is simply not enough time to thoroughly explore the space of design alternatives.

As a consequence, new technologies, manufacturing methods and materials are slow (on the order of 25 years) to be exploited effectively. Engineers treat composites as "black aluminum" and don't learn how to use new manufacturing processes, such as shape deposition, to full advantage. This behavior persists despite ample evidence that systematic, aggressive design space exploration works. For example, the Toyota design process, which is generally recognized as "best in class" compared to competitors in Japan and around the world, involves substantially more exploration of design alternatives and a "least commitment" approach in which many options remain viable late in the design process. The end result is that superior quality designs are produced with about half as many engineers as the competition, in less time and with lower ultimate production costs [Ward et al. 1995].

However, the point to remember in considering this example is that Toyota is only able to pursue this approach when working with familiar processes and when collaborating with long-term partners such as Nippondenso. Thus it is the antithesis of the image of virtual design teams formed in response to sudden opportunities, seeking out and exploiting novel materials and processes.

Our experience from MadeFast and similar long-distance design exercises in our graduate design curriculum makes it clear that the difficulties associated with coordinating and managing design efforts, and with ensuring systematic design space exploration, are exacerbated by the limited human interactions associated with distributed design teams. The proposed research on Design Space Colonization is aimed directly at overcoming these problems by providing a combination of (1) a formal distributed design methodology supported by novel tools for coordination, decision maintenance and rationale capture and (2) direct agent-based methods for helping designers to discover, evaluate, experiment with, and become familiar with new tools and manufacturing processes.

Expected Impact

What if ... we could eliminate waste and inefficiency with direct peer-to-peer communications of changes and interactions? - We could develop more complex systems and artifacts than are currently possible.

What if ... we could provide bookkeeping and coordination of design changes and their rationales? - Designers could explore a wider range of design alternatives and ultimately produce better designs.

What if ... we could help designers explore unfamiliar, cutting-edge manufacturing processes and materials? - Radically different, cheaper and better designs could emerge decades earlier than is possible now.

We believe that the technology proposed here and by other MADE contractors has the potential to create a revolution similar to that of the first generation of computer systems and to that of the Internet. Technology for the coordinated distribution of tasks will change the nature of development of large projects.

In particular, (1) the design mapping and navigation technologies proposed will allow development and maintenance of complex products to dramatically improve in cost, time, and complexity; (2) the agent-based approach for exploring and experimenting with new processes has the potential to reduce the typical lengthy exploration, maturation and adoption cycle from a few decades to a few years. We should start to see the design of parts and integrated assemblies that literally have never before been contemplated.

The proposed work covers a narrow, but deep slice of the general design space colonization problem. Although we will focus on a few manufacturing processes and facilities, our emphasis will be on the development of languages, methodologies, protocols and agent-based-engineering (ABE) methods that allow us to extrapolate from our work to include additional tools and services for design coordination, process simulation, analysis and documentation. By the same token, it allows us to integrate relevant tools and services developed by others in the context of distributed design exercises proposed both for the current work and as a component of large projects such as the Simulation Based Design (SBD), AIMS and AM3 projects, in which we are a subcontractor.

The proposed technologies have the potential to be widely used. The ABE approach offers an extensible open approach to the problem of wrapping and integrating distributed, heterogeneous tools. This strongly mitigates the burden of injecting new technology into established design processes. The engineers can use existing tools, but will have added functionality.

We observe also that this approach does not require that all players invest in heavy-weight design environments, but it provides a way to make specific capabilities of powerful analysis and modeling tools accessible to small teams and it provides vendors of such tools with a mechanism to make them selectively shareable. We believe that these qualities are essential for broad acceptance by the engineering community, to become de facto standards instead of being relegated to a few, costly proprietary CAD systems.

A few CAD vendors have also begun to realize that the future may involve completely unbundling their systems and providing "nuggets" of capability on a "pay-per-view" basis. This kind of line-item accountability is increasingly desired. The technology described in this proposal, along with complementary work by other MADE contractors, is an essential component of this change in business paradigm. By continuing to work with systems companies (particularly SGI and SUN) and CAD vendors (particularly Concentra) we expect this technology to appear in commercial systems.

More generally, the technology is likely to be commercialized as proprietary libraries and tools, as well as CAD augmentations. PENS [Hong et al. 1995] is indicative of a variety of documentation tools that will appear in the commercial marketplace. One example is a tool that would browse design documentation using the formal language (AEDNL) proposed here to help designers find alternatives for current issues and tasks; an approach different from the indexing methods of current case-based reasoning attempts to reuse designs.

Process and metrics

In this work we will be collaborating closely with partners at Lockheed-Martin, (using MECE as an example notebook environment and using the AIT missile design program as an application domain), EIT (using and contributing to their proposed Virtual Design Workspace, and KSL (using and contributing to their efforts on formal methods for agent-based engineering). We will also work intimately with Prof. Prinz's Rapid Prototyping Laboratory (RPL) at Stanford to explore how best to access, evaluate and incorporate process models of their layered material deposition process into the engineering teams' design environment. Our work in this area will draw upon the agent-based engineering notebook technology that we have developed under SHARE and as a collaborator in SHADE, and on our previous work on modeling, tracking and coordinating agent-based design processes with Redux [Petrie et al. 1994].

Although we will collaborate most closely with the groups just mentioned, there are a number of other complementary techniques that have promise for exploring and managing very large design spaces. We discuss specific collaboration plans with other groups promoting some of these techniques in Section (K) Evaluation Metrics.

Our proposed metrics are defined more fully in Section (K) and consist of 1) measurements of factors that produce time and cost savings by increasing the efficiency of distributed design processes, 2) factors that measure the manufacturability and serviceability of the designs that are developed, 3) the use of novel manufacturing materials and processes, and 4) increases in the number and range of design alternatives explored.

Anticipated project deliverables

As detailed in Section (E) Deliverables and Products, we will deliver agent-based engineering (ABE) tools and services that allow distributed engineers to map and navigate in a changing shared design space while populating it with design alternatives based on novel processes and materials. We will deliver three major technologies: 1) our Process Open Description System (PODS) will help users to discover and evaluate new processes and materials in an integrated agent-based environment; 2) ProcessLink will coordinate the distributed exploration of the design space by users and their CAD tools; 3) DesignJournal will provide distributed users with lightweight informal documentation authoring and critiquing tools that encourage creative exploration while providing enough structure for integration with PODS and ProcessLink.

Visionary system description

The basic vision behind our work is described in Sections (D) and (I). We envision an industrial community that is able to respond to opportunities to produce small quantities of defense-related electromechanical systems with unique specifications, and for which the opportunities depend on being able to deliver the systems quickly and at low cost. The community will utilize a combination of pre-existing teaming agreements and core technology that allows design teams to organize around a distributed design process model and methodology. The specific contributions of our work toward this vision are (1) the development of ProcessLink technology that utlizes the distributed design process model to provide support for team coordination and mangement, (2) the development of an agent-based approach to support exploration and evaluation of novel processing and materials opportunities and (3) the development of a DesignJournal that supports creative design in the context of (1) and (2). Our approach is specifically designed to be interoperable with additional components needed to realize this vision. Complementary technologies from PACE and other MADE contractors will be integrated as detailed in "Comparison with ongoing research" in Section (J) Technical Rationale.

Demonstration program

In Section (H) Statement of Work, we lay out a program in which we provide ARPA with a demonstration of technologies every six months. At the end of the first year, we propose to hold a technology transfer workshop together with PACE and other MADE contractors for the defense community. An initial integration of PACE technologies will be demonstrated and explained in design exercises that all attendees particpate in. The exercises will both transfer technology and provide user feedback.

At the end of the second year, we will conduct a Concept Feasibility Demonstration (CFD) with PACE and other MADE contractors with more robust and more tightly integrated versions of all prototype software. This will be a simulated rapid design exercise.

In the third year, we plan to participate in a MadeFast-like project, MadeFast Done Right, with the adoption of formal design process management tools for coordination and documentation. As an option, we will provide the project coordination and integration functions we performed in the original MadeFast exercise. These demonstration are described in more detail in Section (E) Deliverables and Products.

In Section (H) Statement of Work, we lay out a program in which once a year, we provide ARPA with a demonstration of technologies integrated with those of at least one other MADE contractor, such as EIT, KSL, or Lockheed. Also, at the end of the first year, with at least one of these contractors, we will hold an ARPA workshop to demonstrate and explain the technologies. These workshops will be for the purpose of technology transfer, with attendees from the defense community.

Finally, we plan to participate in a MadeFast-like project ,"MadeFast Done Right", with the adoption of formal design process management tools for coordination and documentation in the third year. As an option, we will provide the project coordination and integration functions we performed in the original MadeFast exercise.


(C) Innovative Claims

The basic idea of this proposal is Design Space Colonization, a combination of novel technologies that will allow distributed teams of designers to explore systematically the design space of complex products. We focus our attention particularly on enabling design teams to explore the greatly expanded design spaces made possible by new materials and processing techniques such as the layered shape deposition process under development at Stanford. Our approach consists of providing computational support for two critical functions needed for systematic exploration of new design spaces: design space navigation/mapping and design space population.

Design navigation and mapping concerns decision and coordination support for systematic group exploration of alternatives, trade-offs, conflicts, etc. It is achieved by providing organized documentation of the design and the design process and the provision of active help in tracking, mapping, and generally navigating through a team-generated design space, based upon the design process models of DesignRoadmap [Park 1995] and Redux [Petrie et al. 1995] .

Design space population concerns the generation and analysis of novel design alternatives and choices. It is achieved by providing a combination of mechanisms by which designers' agents can find, evaluate and incorporate the characteristics of novel processes directly into the CAD environment and tools that support creative exploration of concepts and alternatives, while maintaining linkage to the structure imposed by a distributed design process model.

Expected Impact

Although the work focuses on a particular aspect of the design space exploration problem, i.e., the exploration and population of greatly expanded design spaces that arise from novel materials and processes, it is carefully defined so that it can immediately be extrapolated to more general problems in coordinating the progress of virtual design teams working on complex products. We will validate these extensions through distributed design exercises proposed both for the current work and as a participant in Simulation Based Design (SBD), AIMS and AM3 and Airborne Laser (ABL) projects. We are working with Lockheed-Martin, Hughes , and Rockwell on the use of the proposed technologies in the development of missile systems. We are already involved in discussions with CAD vendors with regard to the commercialization of these technologies. PSA Inc. will license PENS technology from Stanford University to create and market commercial versions of PENS for popular computer platforms. With the anticipated market penetration, the PENS offering, developed as core MADE technology, will greatly augment the effectiveness of team collaboration in design.

The design mapping and navigation technologies proposed will allow development and maintenance of complex products to dramatically improve in cost, time, and complexity. Equally important, we believe that these technologies are essential for geographically distributed design teams to function efficiently. The agent-based approach for exploring and experimenting with new processes has the potential to greatly compress the lengthy infancy that typically accompanies significant advances in materials and processing. We should start to see the design of parts and integrated assemblies that literally have never before been contemplated.

More generally, the technology proposed here and in collaboration with other MADE contractors has the potential to revolutionize that way that the design of complex systems is done. Our experiences with MadeFast and other distributed design exercises convince us that this technology is essential for the vision of virtual design "tag teams" to become a reality.


(D) Vision

Defense systems are slow to use novel manufacturing technologies and materials. The proposed Process Open Description System (PODS) and DesignJournal technologies directly address this problem. The ability to discover, evaluate, and become experienced with new materials and processes within months rather than years or decades is our target -- an order of magnitude faster adoption of processes and materials for defense systems.

But the discovery and exploitation of new technologies , especially by distributed teams, can only be accomplished with technologies for the systematic exploration of the corresponding design spaces. The major impediment now to rapid design of complex systems such as missile seekers is coordination[Forney 1995] . This problem becomes more important when engineers are distributed temporally and spatially. The ProcessLink framework provides a critical solution to the coordination of tasks and integration of heterogeneous software over the Internet. The design mapping and navigation technologies proposed for ProcessLink will allow development and maintenance of complex products to improve in cost, time, and complexity Our target is an order of magnitude reduction in time and cost for products of equivalent complexity.

What is proposed here is not just a set of technologies: it is a new design and development methodology . Today changes often follow a complicated paperwork path up levels of managers before they cross over organizational barriers. ProcessLink will enable more direct peer-to-peer communications among design engineers whether synchronous or asynchronous. This will be a departure from the current method of change control by management hierarchies. Furthermore, the communications will all be semi-structured and will be organized automatically as design documents to be used by follow-on engineers in the same project or for redesign in future projects.

The PODS and ProcessLink technologies proposed are delivered via agent-based technologies that span the Internet and can be used with any legacy software. This Agent-Based Engineering (ABE) approach will become ubiquitous as valuable services (constraint solving, project management, geometric analysis, component location, finite element analysis, etc.) become available either for free or as commercial services over the Internet.

The technologies proposed provide the basis for a new way of doing business. They will need to be accompanied by new economic and legal structures outside the scope of this proposal. But the requirements for tag-team design make the adoption of these technologies, and similar ones, inevitable.


(E) Deliverables and Products

We will deliver open Agent-Based Engineering (ABE) technologies specifically designed to generate/evaluate new design alternatives based on new processes and materials and allow a distributed team to explore this new space in a systematic manner. The Agent-Based Engineering (ABE) set of agent protocols and conventions will be based on existing KQML and CORBA work, but specialized for design and engineering applications. We will deliver open Agent-Based Engineering (ABE) technologies specifically designed to generate/evaluate new processes and perform design space mapping/navigation functions in a distributed dynamic environment across heterogeneous software and platforms for engineering design

Process Open Description System (PODS) (Figure I-2) will allow designers to discover new processes and CAD tools to import, dynamically, design rules that will constrain design options to those consistent with the process being explored. PODS agents will exchange process geometry and material features between agents. Simulation code will be imported with Java applets or similar HTTP-based facilities to allow evaluation not possible with only geometry features. The initial target is the Shape Deposition process being developed at Stanford's Rapid Prototyping Laboratory headed by Prof. Fritz Prinz. PODS will consist of the following components:

The ProcessLink framework (Figure I-1) provides for the integration, coordination, and project management of distributed interacting CAD tools and services in a large project. Specifically, this framework implements a structured design methodology based upon the Redux and DesignRoadmap models and provides a design space structure , as well as mapping and navigation services for distributed design. This framework consists a reusable protocol and generic domain-independent services including:

We will also deliver DesignJournal tools that permit design information to become self-documenting and self organizing:

We will be performing the technology transfer functions described in the Section (M) Technology Transition. Of special note, a patent has been applied for PENS and commercialization is underway as described immediately below. Also, the ProcessLink framework as described is being incorporated into the SBD architecture under contract. And funds have been allocated by SIMA to sponsor a graduate student for a year to transfer portions of ProcessLink to Hughes Aircraft in Tucson. ProcessLink has also been partially sponsored by Concentra, a CAD vendor, which will continue to follow the research to see what might be incorporated into their CAD framework.

In addition, we propose two MadeFast-like projects as well as a special Workshop to be held in conjunction with other MADE contractors and to be used for technology transfer to participants.

Commercialization of PENS

Pipeline Systems Analysis (PSA), a research development consultancy whose principals were intimately involved with PENS development and testing, propose to further develop and deliver PENS for commercial distribution on all popular computer platforms (i.e. MS Windows, Macintosh, UNIX, Newton PDA). As an option to this Design Space Colonization proposal, PSA, as subcontractor, will license the current PENS technology from Stanford, transform it into a commercial software product, and market it for wide-spread distribution to the internet community. PSA is also in dialog with private and venture capital investors to fully fund and capitalize on this opportunity to market the PENS product. As part of this optional subcontract, PSA will also collaborate with the DSC project team so that a robust platform for PENS development within DSC and the commercial version of PENS can readily incorporate the assisting agents technology being developed. Specifically, PSA will address a list of user-requested features and issues in the commercial version of PENS.

(F) Schedules and Milestones

6 months - Technology Demonstrations

PODS - demonstrate front end design agent, process broker, RPL process agent communicating, controlled from within CAD environment (most likely ProEngineer + MECE).

ProcessLink - demonstrate working Redux' agent with registration of domain agents and Constraint Manger working with single constraint solver.

12 months - PACE Workshop for Technology Transfer

PODS - demonstrate a rough prototype for downloadable light-weight simulators and process interface integrated with EIT virtual design space environment together with refined RPL agent and preliminary IMTL agent from within CAD environment.

ProcessLink - Demonstrate Redux' rationale capture and active coordination, Constraint Manager with at least two constraint solvers, and initial version of Project Manager with task planning. Integration - demonstrate the above in conjunction with Lockheed's MECE, and EIT's Virtual Workspace.

18 months - Technology Demonstrations

PODS - demonstrate a prototype for design environment with access to novel processes. This will be a proof-of-concept version with a severely restricted range of available geometries and materials variations.

ProcessLink - demonstrate project planning and revision due to design changes with Project Manager, using Redux' and Constraint Manager services, in conjunction with MECE.

24 months - Concept Feasibility Demonstration (CFD):

Perform a simulated rapid design exercise in which PACE and other MADE contractors design, over the Internet, a missile component (e.g., structure for supporting and orienting optical components of a missile seeker) using PODS + ProcessLink framework and create prototypes of their designs. Document the design processes and analyze the organization of the process and artifact design for redesign and retrieval.

30 months - Integrated Demonstration

Demonstrate tested more robust versions of all software integrated with that of other PACE members and other MADE contractors.

30-36 months -Exercise

Perform a rapid design exercise in which several geographically distributed MADE contractors design and prototype novel subsystems (e.g., optics & controls) of seekers that occupy dramatically different corners of the design space, drawing substantially on the documentation and pointers to tools and resources developed for the 24 months demonstration, showing reuse of elements of that design. This project is intended to be larger than MadeFast: a MadeFast Done Right resulting in a usable missile design. Optionally, CDR will act as project manager of the exercise.


(G) Proprietary Claims

Our intent is to share with the government and the R&D community everything necessary for the successful adoption of the technology developed in this project. Moreover, we believe the most effective dissemination strategy is to commercialize the results, thereby ensuring that they will have a lasting impact. Therefore we intend to reserve all proprietary rights to the full extent permitted by law, including rights in technical data, in computer software, in know-how, and in prototype systems, and including all commercial rights, intellectual property rights under patent or copyright law, data rights, or other proprietary rights that may be retained.

Notwithstanding this retention of rights, all data developed under the proposed contract will be delivered to the United States Government with appropriate controls to ensure confidentiality of data volunteered by participants. Specifically, the U.S. Government will be granted the right to use the software developed under the proposed contract for the purposes and time period of the funded project. Copies of certain software developed at private expense may also be supplied to the U.S. Government under this contract for limited purposes.

To the extent that the software developed and delivered under this effort depends upon these software products, such products will be integrated into the delivered software system and licensed without charge to the government for use solely as part of the delivered system. At the government's request, the developers will grant other members of the research community similar no-cost licenses to the necessary company-owned software systems for use with the delivered software, for purposes of evaluation of that software. Software developed at private expense and supplied under the proposed contract shall be maintained in confidence by the U. S. Government and shall be restricted to use for government purposes only.

Preexisting tools and models used in developing the delivered system can be further developed at the developers' own expense independently of the proposed contract for example, by improving the software, extending the modeling environment, and developing advanced implementations. The government's use of systems, software, and data purchased or produced at private expense in performing the contract shall have no effect on any developer's proprietary rights in such systems, software, and data.


(H) Statement of Work

We propose to develop the items specified in Section (E) Deliverables and Products . The major task areas are: Process Open Description System (PODS), ProcessLink, DesignJournal, Assessment, and Integration.

PODS and DesignJournal together provide a mixed formal and informal approach to facilitate the exploration and population of design spaces made possible by novel materials and processes. ProcessLink complements these with tools and methodology that allow distributed teams to function efficiently and explore alternatives and trade-offs systematically. The Assessment task provides input for the Metrics in Section (K) and guides the refinement of the tools, languages and methodologies. The Integration task includes distributed design exercises that feature the products that we develop working together with products of other MADE contractors.

Software, data, research and experimental results developed under the MADE program will be shared with all MADE participants to the maximum extent possible. Our tools and services will be available via the Internet as public agents. PENS will be made a commercial product. DEliza will be added to a commercial CAD tool. Documentation and software will be available on the WWW, with demonstration interfaces. The results of design exercises will be available on the WWW.

Task area 1: PODS

Responsible personnel: Mark Cutkosky, Charles Petrie, George Toye.

0-12 months (PODS1) Begin definition of process capabilities and characteristics in cooperation with the Rapid Prototyping Lab (RPL) of F. Prinz and the IMTL at Sandia Labs. Develop preliminary ontologies and protocols for describing manufacturing process capabilities, including constraints and geometry features, in consultation with the Stanford Knowledge Systems Lab (KSL). Develop exemplar of lightweight process/feature simulator. Develop "front-end" design and process agents to exchange process and design information and a process broker to locate appropriate candidate process.

12-24 months (PODS2) Develop language and downloadable process simulation capabilities for RPL and Sandia, together with a third process (to be determined). Extend basic representation, communication, encapsulation capabilities to address tools/services from other MADE contractors. The likely first candidate is Rockwell's Design Sheet for analyzing and plotting effects of controlling material density, hardness, etc. as a function of spatial parameters.

24-36 months (PODS3) Continue to refine and develop PODS environment, language and agents in collaboration with KSL, RPL and Sandia. Continue incorporation of external tools/services for analysis, visualization. Explore Virtual Design Workspace (VWD) from EIT for visualizing/sharing the enlarged design parameter space as a complement to the use of DesignSheet.

Task area 2: ProcessLink

Responsible personnel: Mark Cutkosky, Charles Petrie

0-12 months (ProcessLink1) Develop prototype Constraint Manager broker running a constraint problem solver of local construction for a demonstration problem. Develop design for extended Redux dependencies and message types for project management and implement a core subset.
Develop agent-based version of DesignRoadmap and project planning tool that can be used to analyze and map the feature-based task relationships in an engineering project, with assistance of KSL in providing general modeling tools.
Develop a preliminary version of Agent Electronic Design Notebook Language (AEDNL), in consultation with EIT, Lockheed, and KSL , to be used by: Redux', DesignJournal (see task below), EIT's Virtual Design Workspace (VWD), and KSL's device modeling tools.

12-24 months (ProcessLink2) Refine the Constraint Manager and demonstrate it with systems provided externally.
Extend Redux for project management, including the use of constraints to detect over-use of consumable resources and project total resource costs. (These may be demonstrated with a human project manager.)
Develop first version of formal methodology for distributed design, and a semi-formal Quality Function Deployment [Clausing ] agent that works in concert with formal tools such as Redux' and DesignRoadmap and with external tools such as ITI's RAPPID.
Explore project management using the constraint manager and Redux'. A demonstration will show how design changes impact the process plan and schedule, and how the ProcessLink agents are used together to make incremental changes in the schedule. The demonstration will involve design decisions made by different engineers not co-located and will result in non-intuitive changes in the process plan.

24-36 months (ProcessLink3) Continue to extend and refine all ProcessLink components. Demonstrate exchange of EPL messages and AEDNL files among ProcessLink agents and MECE (Lockheed) and GEM (Rockwell) software and possible other MADE contractor products. Prepare for integrated demonstration to show utility of ProcessLink for distributed project management.

Task area 3: DesignJournal

Responsible personnel: George Toye, Larry Leifer

0-12 months (DesignJournal1) Develop activities and process tagging capability with respective description templates in PENS. Coordinate terminology with MECE and preliminary AEDNL (see ProcessLink). Develop preliminary agent that helps designer reorganize notes.

12-24 months (DesignJournal2) Refine PENS and develop and test DesignEliza agent that accesses PENS information as source for real-time design process inquiries. Develop filters to translate PENS note records to/from AEDNL to allow formation of community notebook webs with access to integrated ABE services.

24-36 months (DesignJournal3) Continue to test and refine PENS and DesignEliza and to explore integration with other notebook and design tools and with the EIT Virtual Design Workspace.

Task area 4: Assessment

Responsible personnel: George Toye, Larry Leifer, Mark Cutkosky

0-12 months (Assessment1) Evaluate PENS both in Stanford's E210 design curriculum and in joint experiments with Lockheed-Martin and EIT (see Management Plan).

12-24 months (Assessment2) We will evaluate products of DesignJournal both in Stanford's ME210 design curriculum and in joint experiments with Lockheed-Martin and EIT.
Conduct focused trials of ProcessLink and PODS with students and defense company engineers for artificial projects selected by Stanford researchers. Perform user studies on the efficacy and performance (range of geometries & materials explored, number of design/manufacturing errors introduced) as compared to using conventional CAD and project planning tools.
Evaluate the results of task Integration1.

24-36 months (Assessment 3) Continue evaluations of DesignJournal, as well as any products from other MADE contractors deemed ready for introduction into the E210 curriculum.
Continue focused trials of ProcessLink and PODS.
Evaluate the results of task Integration2.

Task area 5: Integration

Responsible personnel: George Toye, Larry Leifer, Mark Cutkosky, Charles Petrie

0-12 months (Integration1) In collaboration with Stanford RPL and Sandia Labs, demonstrate PODS let designer select preliminary IMTL and RPL agents and load process descriptions into CAD environment (probably using ProEngineer).
We will collaborate with industrial MADE partners in the E210 project and assessment environment (see (L) Management Plan).
Hold the PACE Workshop for technology transfer and integration as described in Section (F) Schedule and Milestones.

12-24 months (Integration2) An advanced CFD of PODS , ProcessLink, and MECE and/or GEM will be held and demonstrated with PACE and other MADE contractors as described in Section (F) Schedule and Milestones.
Collaborate with industrial MADE partners in the E210 project and assessment environment (see (L) Management Plan).

24-36 months (Integration3) We will participate on an integrated exercise, MadeFast done right , as described in Section (F) Schedule and Milestones.
Collaborate with industrial MADE partners in the E210 proejct and assessment environment (see (L) Management Plan).

Optional technology transition tasks

(0-36 months) Option1: Building on the concepts developed in SHARE, PSA Inc. will license DesignJournal technology to create commercial versions of PENS. PSA will market PENS and its add-on extensions. PSA will also collaborate via this optional subcontract with us in developing new software platforms for PENS research prototypes.

(24-36 months) Option2: As an option, we will manage MadeFast done right. This will require additional personnel to coordinate and integrate tools and people (see Sections L and II-C).


(I) Visionary System Description

Our vision of distributed teams that can quickly explore a shared design space is illustrated in Figure I-3. As a particular example, we envision an industrial community that is able to respond to opportunities to produce small quantities of defense-related electromechanical systems with unique specifications, and for which the opportunities depend on being able to deliver the systems quickly and at low cost.

An example opportunity might consist of a request for 1000 modified AIT missiles with the ability to look within 5 degrees of the direction of flight so that the target need not always be tracked from the side but only if they can be delivered within a year and at roughly one half the cost of the original versions with redesign will need to be largely complete in 2 months. Taking advantage of some pre-existing teaming agreements, a core design team is composed on the fly, using engineers at a couple of major defense contractors as well as some small systems houses with expertise in such areas as rapid prototyping or digital controls. Part of the pre-existing agreement is the adoption of the common Electronic Project Language (EPL) and a secure network for communications. All tools are equipped with ABE compliant APIs that send EPL messages via either KQML or CORBA using encryption and electronic signatures for authentication. A set of ProcessLink generic service agents is generated for the project. All tools register, via their APIs as agents with Redux'.

As part of the pre-existing partnership agreement, it is understood that a formal Distributed Design Methodology (DDM), based upon formal task analysis, will be followed for managing the project. The basic static task analysis is based on the DesignRoadmap model (with contributions by Eppinger [Eppinger et al. 1992] ) and driven by the dynamic Redux model for recording history and using it for coordination. A working model of tasks, subtasks, responsibilities, and authority is developed based upon analysis and the agents registered. A project plan is generated by the Project Manager, tied to the inputs and outputs of the tasks.

This approach increases startup overhead but dramatically reduces confusion and uncertainty about who is responsible for what, by when, and subject to which other activities. Perhaps more importantly, it provides a methodology and structure in which powerful project management and documentation tools can be employed to drastically reduce runtime confusion, thrashing, and waste. Actions, decisions, task completions, conflicts are tracked both to provide automatic notification of potential problems and opportunities and to improve the quality of the documentation providing a structure for indexing it.

At the working level, much of this support is invisible. The designers observe that a structured methodology is being followed and a formal project planning (the Project Manager agent) tool is in place. Occasionally the system notifies them of opportunities to revisit previous ideas or asks them to re-evaluate a previous result in light of new information. The engineers work with a variety of CAD simulation and analysis tools, all of which are connected to their engineering notebook programs (such as MECE and PENS). Following a couple of face-to-face meetings, most of the communication is remote, asynchronous and in the context of the tools that the engineers regularly use.

The project begins with a formal description of the new mission requirements. These are entered using a Requirements template in the distributed engineering notebook environment, which is composed of a collection of interacting tools including PENS, MECE that can interpret the EPL. In the spirit of agile design, it is understood that the requirements are subject to modification as new information and results are uncovered as a result of design activities - including further communications with customer. The requirements are compared to the on-line documentation regarding existing AIT designs and several related products (e.g., THAAD).

A key item in reducing weight and manufacturing costs will be to make the new optical system compact and with as few moving parts as possible. Analysis of the "traditional" five (5) mirror system leads to its disqualification because adding forward direction will lead to an even more expensive window and mirrors.

Previous design documentation is searched with software, possibly commercial products, using the AEDNL format, for issues containing the keywords optics, compact and elimination of moving parts. One seeker design had a window in the nose but was only good for missiles that could be guided by radar until it was close enough to the target that the heat from the window itself no longer obscures the target heat image. Another design was not for a missile seeker but an underwater vehicle that corrected for a spherical shape and the water refraction by using materials refraction.

Neither design uncovered applied directly but they lead to two workable alternatives. One is to use the "traditional" side window but to augment this with a nose window to be used at sufficiently close range. Initial thermal analysis shows this would allow the missile to fly with an effective behavior consistent with the requirements. The government agent that generated the requirement is contacted, remotely plays a simulation, and approves a modification of the requirement on the fly.

Another option to be considered is a strongly refractive side window with a fiber optic cable replacing mirrors altogether. It is determined that this would meet the original requirements directly and could result in the desired cost and weight savings. However, no manufacturing process in use at the contractor can produce the design. The PartNet system is searched, using the KQML-compliant CAD agents in the ProcessLink framework, and a subcontractor found that can supplied the novel optical elements, including a compound lens.

Early on, it is determined that a significant cost savings could be obtained if some of the structural elements were manufactured by a process such as die casting or molding, instead of CNC machining. The success of this approach hinges on obtaining comparatively inexpensive molds. The features are generated and fed to a PODS design agent. Its search for prototyping processes uncovers facilities at Stanford, ALCOA and Carnegie-Mellon University that can do sprayed-metal forms for making die-cast metal or chopped fiber composite parts at about 1/10 the cost of machined tooling. The sprayed metal forms will not last as long as conventional tooling and do not achieve as good a surface finish, but this is not a concern in the present application.

The designers now need to evaluate is process and to start designing for it. The decision is made to investigate sprayed metal facilities, while pursuing conventional milled aluminum parts in parallel as a back up This process involves a series of queries and responses among the designers' PODS design agents, a broker agent and the processes' process capability agents.

A request from the PODS design agent of an engineering team is sent to a broker to obtain the information needed to generate design rules. In contrast to the design rules developed for processes such as 6 micron VLSI design and for services like MOSIS, these design rules must be obtained on-the-fly for each new process. Figure I-2 shows an extract from such a dialogue. The designers' PODS design agent asks the process capabilities agent about the basic process type, process constraints and feature set. The requests are formulated in a standard agent communication language and refer to standard ontologies of features, process constraints and process characteristics for defining the terminology used. The ontologies may, in turn, refer to standards such as the PDES/STEP standard for geometric features and product data modeling. The dialogue need not be entirely autonomous; the main point is to load a set of design rules, features and simulation nuggets into the design environment with as little custom effort as possible.

The process agent replies with a standard feature set that refers to PDES/STEP part 43 and a set of simulation procedures that can be used to evaluate the manufacturing process for designs composed of those features. The simulation procedures again refer to a standard ontology that defines their scope. In this example, they are implemented as applets in hot JAVA. With a set of structured examples to draw from, and a set of design rules and process modeling capabilities loaded into the design environment, the process of exploring what the process can do becomes much less tedious and error prone. In particular, the designers soon discover that by using a layered manufacturing process, there is no cost penalty associated with making molds that are riddled with cooling passages to minimize thermal distortion and prolong mold life.

As it happens, the substitution of molded chopped-fiber parts for aluminum has reduced the total part count and weight for the missile seeker assembly. The adoption of a new, commercial microprocessor-based controller further trims the weight so that the overall weight constraint is (temporarily) no longer active. Redux' notifies the power supply team of this and reminds them that they may want to revisit a surplus battery and power system that they earlier been attracted to because it was very robust and inexpensive, but had rejected as too heavy. They revert to this alternative, saving cost, reusing inventory, while remaining within the weight budget. The Project Management agent notes that this design change affects the project plan: delivery of lithium batteries must be canceled and a new power regulator must be designed, somewhat lengthening the project.

The design project is now slightly behind schedule, but because of the aggressive exploration of alternatives and the incorporation of manufacturing issues and constraints into the design process, the development of the manufacturing plans and actual production proceed very rapidly. Furthermore, incremental revision of the project plan by the Project Management agent ensures that there are no surprises and the delay is acceptable.

The design project follows the practice of gradually narrowing the design space so that production of critical tooling can begin even as the design of subsystems is still being completed. Notifications of gradually tightening tolerances on space, weight, power consumption, bandwidth, heat dissipation etc. are propagated by the appropriate constraint solvers, managed by the Constraint Manager agent, and the error and weight budgets are updated by the Project Manager.

As the teams work, each new result, conflict and question is automatically tagged both to the step or activity in the overall design process that it is associated with and to the elements of the design artifact that it addresses. The resulting design documentation is better structured and more amenable to automatic navigation that it would be if crafted "by hand" as in MadeFast. Future designs can be evaluated against the existing rationale, ensuring that no design elements will be superfluous or otherwise without a valid reason for existence. In fact, for a later design for which expense was no object, the rejection of the lithium battery was flagged as unwarranted. But that's another story.


(J) Technical Rationale

Background

New materials and processes expand the parameter space of feasible designs by removing constraints and increasing the number of design variables that can be controlled. However, designers are slow to capitalize on the opportunities that new materials and processes afford. Part of the problem is that without examples to draw from, people lack the confidence and intuition needed to explore the newly expanded design space and use it effectively. Moreover, when people are hesitant to explore new technologies the experience base grows slowly. Our argument is that industry, and the defense industry in particular, can no longer afford to wait years and sometimes decades to make effective use of the opportunities provided by new materials and manufacturing technologies.

As an example of what typically happens, consider the adoption of composites in aircraft, automobiles and bicycles. The first attempts were sometimes labeled "black aluminum" replacing metal parts with graphite, without significantly changing the part geometry or loading. Inappropriately used composites often performed poorly, making other designers hesitant to try them. Only recently, decades after the development of composites as a competitive technology, have engineers in these industries learned to fully exploit their characteristics. The new composite designs look dramatically different and perform far more reliably than their predecessors.

A new opportunity

The current proposal is partly motivated by the development of novel layered rapid-prototyping processes at Stanford and other institutions. These processes include RPL Shape Deposition [RPL], MD*, microcasting [Amon 1994] and laser sintering and are collectively described as solid free-form fabrication (SFF) processes. They remove many of the constraints associated with shaping parts by conventional means (e.g., machining) and then assembling them. With these processes it is just as easy to create a ship inside a sealed bottle, building the assembly layer-by-layer, as it is to make each item separately. From a more practical standpoint, SFF processes make it possible to build high-performance structural parts with embedded circuits, sensors, etc. However, the idea of shaping integrated assemblies in a single process is so new that we have difficulty envisioning the possibilities. After all, every complex product that we can think of has been assembled from discrete parts!

SFF processes also permit engineers to control material composition and microstructure throughout a part. Performance of systems and components can be significantly improved through this kind of custom tailoring of material and shape. To illustrate this point, consider that high performance aircraft frequently contain elements with high stiffness to weight ratios such as carbon fiber reinforced composites (CFRC). However, a shortcoming of composites is their lower toughness when compared to engineering alloys. Tough materials such as metals tend to be heavy, whereas light materials like carbon are brittle. Carbon fiber reinforced composite parts consequently involve a compromise between density and toughness. Additional constraints are imposed by the need avoid features like holes and notches which give rise to stress concentrations. Trade-offs like this are characteristic of almost any design scenario. Breakthroughs are possible as soon as one or more constraints can be removed and the design space expanded accordingly.

The key bottleneck with SFF processes is not the characteristics or limitations of the processes themselves (although these admittedly are not fully understood) but the ability of engineering teams to come to grips with many-dimensional design space that they provide. While numerous analysis tools are available to determine strength, fatigue life, and a range of other mechanical properties, few methods exist to help designers during the creation of structures with custom tailored material properties. The present proposal attempts to create an agent-based design environment to aid and guide engineers step by step in exploring performance and manufacturing constraints. The interactive and incremental nature of the proposed design and fabrication method is expected to lead to novel designs with previously unmatched performance.

In summary, novel materials and processes provide us with newly expanded design spaces containing large uncharted and unpopulated regions. The proposed research on design space colonization is aimed at providing a combination of services and design tools that will enable teams of engineers to explore these regions rapidly, to populate them with examples, and to organize and document their explorations with maps and models for future use.

Approach and rationale

Our approach for achieving this goal has two thrusts: (1) systematic, coordinated design space exploration and documentation and (2) direct methods for exploring and populating large design spaces associated with novel processes.

1. Design navigation and mapping

The design of complex electromechanical products such as seekers is necessarily a team effort. Thus it is essential to provide support for interdisciplinary engineering teams systematically to explore the design spaces involved in such products, to populate them with examples and to document their explorations. The vision in this case is analogous to a well organized scouting party that combs the features of a new landscape and carefully documents its findings with maps and charts. When regions become well understood they can be populated with design examples (colonies) that serve as guideposts for future explorations.

There is evidence that extensive, systematic exploration of the design space associated with a new product can lead to faster product development cycles as well as better designs. In particular, it appears that this is the approach taken by Toyota Motor Corporation, both internally and when working with long-term trusted suppliers such as Nippondenso [Ward et al. 1995]. Importantly, the engineering design process of Toyota Motor Co. is recognized as "best in class," resulting in high-quality products brought to market in significantly less time and with fewer people than its Japanese and American competitors.

For example, if Nippondenso is supplying an exhaust system to Toyota they are expected at the early design stages to come up with not one or two, but from ten to fifty designs that occupy distinctly different regions of the design space and to provide an informed discussion of the tradeoffs among them. Gradually, the design space is winnowed to a few promising prototypes. In contrast to the prevailing wisdom in the U.S., Toyota does not use co-located teams. Moreover, Toyota engineers spend less time communicating directly with suppliers than their counterparts at other automobile companies. They do, however, maintain records of "lessons learned" while exploring a wide range of alternatives [Ward et al. 1995].

The lessons to be taken from the Toyota example are these:

* systematic, aggressive design space exploration works;

* it can be accomplished with large teams, and when working with external partners and suppliers;

* it requires careful documentation of alternatives explored and lessons learned.

However, it is important to recognize that Toyota can only achieve this kind of design exploration with teams that have a long history of collaboration - often as much as 10 years. Newer, and less trusted suppliers than Nippondenso are given much stricter instructions about what to design and are not expected to participate in design space colonization. Also, it is important to remember that the Toyota process is used only with well understood processes and on evolutionary designs.

The challenge that we address is whether we can obtain similar benefits when virtual teams are assembled for a one-time project and when using novel processes that greatly expand the design space beyond their previous experience. We argue that just as structured programming has become necessary for complex projects, the methodology of structured design is vital for distributed design "tag teams". Not only do structured methodologies facilitate team design and better documentation, they also provide a basis for computer support of team-design. When structure is introduced, it becomes much easier for software to interpret and add value to the process. The key difficulty to date has been the lack of design process models that can be used for distributed design.

Our work under SHARE has provided design process models that can be used as a basis for structuring design spaces. These models provide a tasked-based methodology for analyzing engineering design processes. The research has also provided initial implementations of these models on which we propose to build a suite of tools to support distributed engineering. Specifically, the proposed work on design navigation and mapping will build upon our previous work with Redux' and DesignRoadmap to provide CAE tools so that teams of designers and engineers can provide maps of the space to each other and so that some automatic navigation "beacons" are provided. This work also provides a basis for integrating informal design documentation

Redux [Petrie et al. 1995] is an instance of the design rationale approach to concurrent engineering, but is fundamentally novel. Redux provides a small and ubiquitous design decision ontology based on a model that assumes designers are using depth-first search in parallel to satisfy loosely connected objectives or tasks. Redux supplies both structure for design space mapping and active navigation functions.

This search-based model provides the semantics for novel automatic change management services that reduce the necessity for domain-specific expressions of interest. An important example, from the complete description in [Petrie et al. 1995], is notification of loss of Pareto optimality [Doyle 1995] of the incremental global design. An important part of design is backtracking because of global constraint conflicts. The Redux' agent automatically constructs a backtracking decision rationale and notifies the designer if later changes by other designers in the shared design space make that backtracking unnecessary. Without such a notification, a designer may pursue unknowingly a poor search path. In general, Redux can be used to eliminate thrashing and signal distributed engineers when changes by others open up opportunities to use favored but previously rejected design alternatives. Redux can also be used to provide other basic distributed design coordination functions.

Moreover, Redux performs these novel functions with a minimal model. Not all possible interactions are captured. The interactions that are have domain-independent defined semantics rather than being ad hoc. And finally, the interactions result directly from the required task-based analysis of the process and from the methodology, reducing the burden on the engineer to specify interactions manually.

We will build on the Redux model with the ABE approach to provide an extended ProcessLink [Park et al 1994] framework that can be used to integrate legacy CAD software and implement a Distributed Design Methodology (DDM) Figure I-1 in the Appendix shows the ProcessLink framework and typical messages. This framework will include a new Electronic Project Language (EPL), an ABSML-based documentation language (AEDNL), CAD tool-specific wrappers, and a set of generic engineering design agents including: the basic Redux' coordination agent, a Constraint Manger capable of supporting diverse constraint solvers, and a Project Manager capable of not only process planning but also propagating the effects of design changes to the plan. These services provide strong "navigational aids" for distributed designers, as well as automatic design rationale.

The Redux ontology and semantics will be used to develop a fundamental design documentation language to be used by design documentation tools, such as Lockheed's MECE, and advanced collaboration and display frameworks such as EIT's Virtual Design Workspace. And this structure will be used to provide users with an overall map of the design and the status of the design process.

The task-based design process analysis model of DesignRoadmap [Park 1995] provides the basis for the project plan and the "wrapping" of design tools with ABE-compliant APIs. This new advanced process model will allow analysis of the task interconnections at various levels of abstraction and can be used to develop the most efficient input/output paths between component agents in an engineering development system. This analysis is now done by hand. An example is the interaction of structural, optical, thermal, control, and sensor analysis subsystems in the "kill vehicle" part of anti-missile missile design. DesignRoadmap [Park 1995] can be used to elucidate the feature dependencies among tasks and facilitate the conversion of these systems into ProcessLink cooperating agents. Similarly, it can also be used for design project process planning.

The proposed work provides structure for the integration and coordination of engineering design software. All software services, including legacy systems, are to be integrated by using ABE-compliant API wrappers. The Redux and DesignRoadmap models provide a structured basis for these wrappers, over and above the minimal commitment of agent communication protocols such as KQML.

The Redux model offers an incremental approach to implementation in that an application can be modeled at a very high level of abstraction. Increasingly fine-grained models provide more structure and power as desired. But a robust, "first-cut" can always be made at a high level of abstraction. Finally, the ABE approach allows Redux' facilitators to be used in a hierarchical, federated fashion, to ensure scalability. The overall result is a new methodology and toolset that could dramatically change engineering practice for large distributed projects.

Finally, we note that this approach facilitates the exploration of various alternatives concurrently by different engineering teams. In particular, the ProcessLink Constraint Manger agent will allow designers to simultaneously explore sets of consistent parameter values for global consistency, even hypothetically. We will tie this together with the extension of Redux' to consider multiple designs simultaneously.

The focus thus far has been on provisions for coordinating design teams and helping them to understand "where they are" and "where they have been" in terms of a design project. Informal creative exploration is the selection of "where do we go from here?" as one is exploring - i.e. discovering new processes/technology. Design research studies have shown that informal creative exploration of design spaces is stifled by rigid structure. Clearly, a rich design environment must support both informal and formal design information sharing. While many of our tools help the formal aspects of design, and a smaller set deal with the informal, few are effective in bridging them. Yet, the fluidity by which designers are able to transition between the informal and the formal, is central to an efficient design process.

Our development of PENS [Hong et al 1995] addresses designers' need for informal information recording and sharing. The success of PENS is traceable to its usage model, as a personal notebook and a place to reflect upon design activities. From observations of PENS use, we see that in reflection, designers reorganize their personal notes as previously recorded information is re-accessed, enriched and updated. These reflection-based activities present opportunities to transform the amorphous informal into semi-structured formal information granules.

To summarize this section, the Toyota example suggests that the adoption of a systematic process in which the set of design alternatives is first expanded and then gradually narrowed lends itself to geographically distributed design teams with limited communications bandwidth. The key idea behind our approach is to support this kind of communal design space exploration with tools that allow engineering teams to better coordinate their activities and to see where they, and others before them, have been and to integrate these with tools for creative exploration. As discussed below, these include means which designers can quickly learn about the fundamental characteristics of processes, incorporate them into their working CAD environments and experiment with them. This allows designers to quickly populate newly expanded design spaces with examples for themselves and others.

2. Design space population

Design space population is the generation and analysis of novel design alternatives and choices. There is a substantial body of research and technology available on this general topic. Our focus will be on an important gap in this technology: enabling teams of engineers to take advantage of the very large design spaces made possible by novel manufacturing processes and materials such as the solid free-form fabrication (SFF) manufacturing processes that have recently been developed at Stanford and a few other institutions. These processes have an exciting potential for realizing designs that have never before been feasible to produce but only if engineering teams can easily find out about them, evaluate them and learn how to use them effectively.

Specific methods that we will use include:

i. Direct incorporation of processing constraints into the design environment to generate Design Rules on Demand.

ii. Ontologies and engineering models that impart local order in a universally accessible form.

iii. Semi-automated design space exploration.

iv. Populating the space with organized examples

* Direct incorporation of processing constraints into the design environment (i.e., finding processing capabilities on the Internet; obtaining process models and constraints expressed with respect to standard ontologies of features, constraints, actions, effects; incorporating the constraints and capabilities directly into the design and simulation tools used by designers for incrementally evaluating the design.)

The basic technical approach is illustrated in Figure I-2 in the Appendix section. The designer has used a broker or facilitator agent to find several potentially useful rapid prototyping processes that meet state objectives. At the instant portrayed in this figure, a dialogue is taking place between the designer's Design Agent and the Process Capabilities agent of a Solid Free-Form Fabrication (SFF) process.

The results of this dialogue are a set of Design Rules, in the same sense that design rules were developed to make MOSIS and other VLSI processes feasible, but with the important difference that these design rules must be generated on-demand because we do not know a priori which process or facility will be used at one side of the dialogue, and we do not know what CAD environment will be used at the other. It therefore is necessary to provide for a standard and systematic way of generating design rules on the fly. We believe that this approach, in which the design rules and characteristics are obtained on-demand from the processing facility where they are generated and kept up to date, is essential for scaling the concept to address a cornucopia of possible processes and services that will ultimately be offered over the Internet.

The essence of the exchange, shown as a pseudo-KQML message at the bottom of the figure, is "What can you tell me about your process and features, in terms of the terminology in the following standard ontologies {list}?" The "standard ontologies" include descriptions of features, process constraints and these in turn refer to other lower-level ontologies and standards such as PDES/STEP Parts 41-43. Features may be extended to include allowable material properties, such as density, alloy concentration, and microstructure. The exchange is performed using open standards (KQML/CORBA/HTTP) allowing distributed designers to import process-specific geometric features, constraints and process characteristics as applied to an emerging design. The effect to be achieved is that of dynamically generated design rules that mitigate the difficulty of designing

* Ontologies and engineering models that impart local order in a universally accessible form.
The ability to incorporate processing models and constraints into the design environment hinges upon being able to obtain those models in a standard form. This work requires the formalization of novel manufacturing processes and materials. It is claimed [Prinz] that processes such as SFF are inherently more amenable to formalization than conventional processes. To the extent that formal models of a new process or material exist it should be easier for teams of human and computational agents to determine how well a design "maps" to these new technologies. Will assist Prinz and others (Sandia) in this effort.

Our emphasis will not be on basic research into the process physics, but rather on helping the groups that are doing that research to put their models into a standard form that can be incorporated into design environments. In this effort we will continue to collaborate with our colleagues at KSL to develop useful ontologies and protocols for representing and acquiring process information. Our efforts will be more application-specific than those at KSL, focusing specifically on the need to describe components, processes, features and constraints associated with specific processes such as SFF and facilities such as those at Stanford, Sandia and CMU.

Specifically, we propose to build a library of ontologies describing the shape deposition process and a collection of simulation software that can be used to evaluate it in different situations. These are to be used by a Shape Deposition Process Capabilities Agent (SD-PCA), an exemplar of an open set of Process Capabilities Agents (PCAs). PCAs are located by, and communicate process capabilities to, a Technology Facilitation Broker (TFB) upon request for technologies by a Design Technologies Agent (DTA) on the user's side. The SD-PCA is one example of a process agent that can register with the TFB. In turn, the DTA is an example of a domain-specific agent that can request technology from a public TFB, or network of TFBs. In many cases, the characteristics of a process are expressed as procedures and numerical simulations. In this case, we propose to make process simulation "nuggets" be loadable into the designers' CAD environment using technologies such as Hot Java [Java] or a similar lightweight code open transfer mechanism. We call this collection of functions the Process Open Description (PODS).

* Semi-automated design space exploration

Even with appropriate process models and constraints at their fingertips, designers can only scratch the surface of many-dimensional design problems associated with continuously controlling shape, material composition and material properties throughout a structure, if they have to do it manually.

Techniques for automated or semi-automated search of large design spaces are therefore needed. One approach is to draw upon research in the fields of shape optimization and configuration design. Promising techniques for shape and configuration optimization include simulated annealing [Reddy and Cagan 1993], genetic algorithms [Gage 1995, Conru 1994], A teams [Talukdar et al. 1991] and other numerical methods suited for exploring large, non-convex search spaces. Where such methods are applicable, we will work closely with F. Prinz's group to adapt them to the design of structures with tailored material properties.

The focus of our work will not be on algorithm development, but on providing agent-based support and access to such methods from within the design environment. The approach will be similar to the approach we have used in building a multi-agent system for cable harness routing that uses a combination of fast kinematics and potential-based methods in conjunction with a genetic algorithm for exploring different harness topologies [Conru]. In particular, genetic agents can make dramatic mutations that quickly explore the design space and come up with configurations that people do not anticipate. However, a feature of the multi-agent approach and standard protocols is they allow human intervention, treating the human input as another agent.

* Populating the space with examples

In the absence of formal models, examples are helpful. Each example is an existence proof for a certain region of the design space analogous to a colonial outpost in a spatial wilderness. However, the usefulness of examples is entirely a function of how well they are organized and documented so that they can easily be retrieved by various searching methods, and interpreted from different perspectives (e.g., different disciplines such as optics or mechanics and from the standpoint of different stages of the product life-cycle). We believe that the work described in the previous section on formal methods for design coordination, decision maintenance and rationale capture, along with the tools for creating rich personal and group webs being developed by our colleagues at Enterprise Integration Technologies, will provide the necessary structure for helping designers to locate relevant examples and understand where they reside in the overall design space. In this area of work we will also continue our long collaboration with researchers at NASA AMES to employ and extend documentation navigation technologies such as those developed for Dedal [Baudin et al. 1992].

Comparison with other ongoing research

The proposed Design Space Colonization work is not comprehensive. It does not include component location, device modeling, 3-D envisioning, argumentation, negotiation, tradeoff analysis, or domain-specific coordination strategies.. We will work with all other contractors to integrate services and the ABE approach provides a sound open method for doing so We do make the claim that the PODS and ProcessLink technologies proposed here are fundamental to tag-team design and constitute unique research duplicated nowhere else.

This proposal is closely tied to parallel efforts at EIT, KSL and the Lockheed-Martin AI center that support other aspects of the Palo Alto Collaborative Engineering (PACE) environment.

EIT: The Virtual Design Workspaces (VDW) will be closely tied to our work, especially through the development of AEDNL. We expect privileged display environments for documents conforming to this ABSML-based documentation language. The VDW is also necessary for informal communications in a distributed design environment.

KSL: The Collaborative Device Modeling Environment (CDME) will also complement this proposal by providing a fundamental basis describing device models. In particular, the

Compositional Modeling Interchange Language (CMIL) can be used as a sub language of AEDNL to precisely describe devices and their features.

Lockheed-Martin: The Context Integrated Design (CID) system connects to our technologies most strongly via the MECE component, which will share information with ProcessLink through AEDNL file s and EPL messages. The ProcessLink Redux' agent will also provide fundamental coordination technology for the CID coordination functions.

The Design Space Colonization effort is also allied with other approaches to achieve more systematic exploration of the design spaces associated with complex products. Two in particular are the set-based design methodology proposed by Ward [Ward et al. 1995] and adopted by ITI for their proposed RAPPID work . We are in continuing discussion with ITI on how to marry our complementary technologies, especially for support of negotiation and tradeoffs in design.

The Rockwell Sciences Center distributed design project management tool GEM (Graphical Enterprise Model) [Rockwell] will be used on airborne laser project to support multiple facets of the process (multiple views from different perspectives). We propose to subcontract with Rockwell to support that effort and integrate ProcessLink with GEM. We also would like to work with Rockwell to make DesignSheet an ABE agent. We recognize that it is a challenging example because the preferred mode of interaction is with a user interface closely tied to the underlying kernel. It may be possible for the kernel to run remotely as an agent and use Hot Java or similar technology for a responsive, platform-independent front-end.

We expect to make use of the PartNet system and to integrate its use with PODS and ProcessLink agents. We are also aware of ongoing efforts in PDES/STEP community to define standard for process modeling and description. We will certainly use as much of that standard as appropriate and believe that we can help extend it. We will also work to make our technologies CORBA-compliant. We are also interested in furthering collaboration with NASA AMES on Dedal and successors to Dedal. Will work with EIT to continue to convert Dedal and its successors into a useful Web Librarian.


(K) Evaluation metrics

Evaluation will occur continuously during the project, both within controlled academic settings and through design exercises, as described in Section (F) Schedule and Milestones, with major defense contractors to determine the efficacy of the deliverables.

The benefit of the deliverables to ARPA will be in terms of the time and money saved on large projects. Thus we need metrics of the following anticipated benefits of our work:

* ABE agents allow engineers to find and exchange design information faster, speeding up the design process;

* PODS allow more design alternatives to be explored with more manufacturing processes and materials, increasing the quality of the design and reducing the manufacturing cost and time for the product;

* ProcessLink coordinates the process, eliminating design thrashing, and mismatched connections between tasks, provides better project management and version control, and catches missed opportunities to improve the design of the artifact, all resulting in a shortened process, fewer mistakes, and a better-designed product.

These savings in time and money are difficult to measure in large projects and especially hard to quantify as reductions against a control project. However, each of the bullets above describes the causes of the effected time and money savings. We propose to measure these causes, to support evidence of time and cost savings.

One important measure will the be increased number of design alternatives considered. Studies such as [Ward et al. 1995] have shown that this measure can be translated into person-years saved on major projects. We propose to measure additional design alternatives considered in the design exercises. We will do this by appealing to previous studies of the average number of alternatives evaluated and then measuring the number considered using DSC technology. Then, using studies such as [Ward et al. 1995; Eppinger et al. 1991], we will translate this into time and money saved.

Another measure will be the use of novel manufacturing materials and processes. We will determine, from instrumented DSC technology use and user interviews, what novel processes and materials were used in the design because of DSC. Based upon manufacturing and life-cycle costs for similar products manufactured with conventional processes and materials, we will provide estimates of time and cost savings due to the use of the novel processes and materials. We will also use the DFM analysis tools of Ishii [Ishii et al. 1994] to evaluate the manufacturability and serviceability of designs produced in the design exercises to address the question of cost effectiveness.

The third important measure will be the efficiency of the design process. We will use previous studies of design and development projects to determine typical costs in time an d money of selected inefficiencies in such projects: thrashing, mistakes, and missed opportunities where the information to make a better decision was known but forgotten. We will then analyse the EPL messages sent during the design exercises to determine a measure of thrashing and mistakes eliminated and opportunities noticed by ProcessLink that were used in the design. Again, using prior management studies, we will assign cost and time savings based on this analysis.

Tactics

One big advantage to adopting a structured design process methodology and utilizing design process management tools such as Design Roadmap, GEM and other tools (e.g, PERT and SADT tools) is that the process becomes amenable to instrumentation, documentation, and formal analysis. There is a rich literature on process analysis and a number of numerical and symbolic analysis methods that can be applied (e.g., [Eppinger et al. 1991]). For example, one can examine the process structure for evidence of "process deficiencies' [Park 1995] such as process bottlenecks, confusion of parallel and alternative sequences, execessive propagation of low-level interactionsand dependencies to higher levels, excessive iterations, etc.

(One difficulty is that most project planning tools are inherently not well suited to use with virtual teams exploring novel design alternatives. As explained in the Techical Rationale section, we believe that Design Roadmap overcomes the limitations of earlier approachesby providing richer semantics than previous representatons and by allowing dynamic modification of the design process model at multiple levels of detail, with automatic consistency checking as one moves to higher and lower levels of abstraction.)

Beyond looking for structural deficiencies in a design process, one can use a design process model as generated in Design Roadmap [Park 1995] and/or GEM (from Rockwell) as the basis for dynamic simulations [Levitt] and analyses of probable opportunity costs associated with delayed actions or decisions [Eppinger et al 1991].

We will apply these analyses to the design process models that we generate both in the E210 curriculum and design exercises at Stanford (see also, Assessment subsection in Statement of Work) and to collaborative exercises with other MADE and DOD contractors (see Management Plan and Statement of Work).

Evaluations will include:

* user surveys regarding ease of use, problems, etc.

* data on frequency of use of each tool and technology

* average information access and publication times

* average design iteration cycle time

* extent of reuse of previous design information (number of visits, and incorporations of previous information into current design project web).

* number of alternatives explored

* amount of electronic documents generated

* percentage of the documents that are structured (e.g, with labels corresponding to design process or artifact model) as opposed to informal (e.g, Hypermail).

This detailed analysis will be corelated with time and money savings using relevant management studies in the literature.


(L) Management Plan

The Center for Design Research is composed of regular faculty , research associates, and graduate students employed as research assistants. The PIs for the DSC project are Mark Cutkosky and Larry Leifer. Senior research associates Charles Petrie and George Toye help in implementation and in managing the research assistants on the tasks as specified in Section Statement of Work.

In this work we will be collaborating closely with partners at Lockheed-Martin (using MECE as an example notebook environment and using the AIT missile design program as an application domain), EIT (using and contributing to their proposed Virtual Design Workspace, and KSL (using and contributing to their efforts on formal methods for agent-based engineering). We will also work intimately with Prof. Prinz's laboratory at Stanford to explore how best to access, evaluate and incorporate process models of their layered material deposition process into the engineering teams' design environment. Our work in this area will draw upon the agent-based engineering notebook technology that we have developed under SHARE [Toye et al. 1994] and as a collaborator in SHADE [McGuire et al. 1993] and on our previous work under SHARE on modeling, tracking and coordinating agent-based design processes with Redux <[Petrie et al. 1995]. Our working practice, as demonstrated in SHARE and MadeFast, not only consists of regular meetings and seminars, but also sharing students. In the case of KSL, we regularly fund a student at CDR to work at KSL on topics of mutual interest and then use the expertise and technology at CDR. In the case of Lockheed-Martinand EIT, these companies regularly employ our students on consulting jobs, summer projects, and as permanent employees after graduation, providing a rich interchange between these comapanies and CDR.

PIs and Research Associates meet with project research assistants on a weekly basis to evaluate the direction and progress of the projects and ensure that they conform to the schedule in Section Schedule and Milestones and the deliverables conform to those in Section Deliverables and Products. We routinely make presentations and demonstrations of on-going work to ARPA and defense contractors more often than every six months.

In addition to our project exercises with other contractors, we will solicit electromechanical design projects in E210, sponsored by our defense industry partners in MADE, AM3, SBD and ABL programs (i.e., TI, Hughes, Rockwell, EIT and Lockheed-Martin). As is standard policy in Stanford's graduate design curriculum, E210, the project sponsors will designate a technical liaison who interacts closely with the design team, conducts design reviews, answers questions about the client preferences and participates in the negotiation of requirements. Liaisons interact by reviewing the project web for their project and communicating via a combination of mail, electronic notebook evironments (e.g., PENS), videoconferencing, telephone, etc. We will ask that the technical liaisons be drawn from same pool as people participating with us in demonstrations and distributed design exercises for MADE AM3, SBD, AIMS and ABL programs. We have enjoyed this kind of dual research/E210 project interaction with G. Bouchard from Lockheed-Martin in the past and it has resulted both in excellent E210 projects as well as rich two-way transfer of technology and ideas. We have also observed with other companies that those clients that most successfully embrace the SHARE/E210 environment also enjoy the best interactions with their design teams.

CDR is willing to be assigned to an integrator of a prime contractor chosen by ARPA. In MadeFast, CDR clearly demonstrated its capability to play an integrating and coordinating role in such exercises. We will continue to find opportunities for utilizing the tools and technologies of other ARPA MADE contractors in the context of distributed collaborative design.

As an option, we offer our services as the prime contractor for a Madefast Done Right exercise in the third year. For this option, we would hire a full time staff member to become the project leader as well as an additional research assistant. It is structurally important that the prime for such projects have complete technical and budgetary control based on our experience with MadeFast. Such a project should be run more like a real defense development project with performance-based payment. As the prime in such exercises, we would set deadlines for performance of tasks. Unless it could be shown that failure to complete on time was due to a delayed critical input from another exercise participant, subcontractors would be penalized for late work. All subcontractors would participate in the development of a project plan and be required to notify the prime of any deviations with explanations. The prime would have the authority to revise the plan as required with agreement from the subcontractors.


(M) Technology Transition Plan

The proposed work covers a narrow, but deep slice of general the design space colonization problem. Although we will focus on a few manufacturing processes and facilities, our emphasis will be on the development of languages, methodologies, protocols and agent-based-engineering (ABE) methods that allow us to extrapolate from our work to include additional tools and services for design coordination, process simlulation, analysis and documentation. By the same token, it allows us to integrate relevant tools and services developed by others in the context of distributed design exercises proposed both for the current work and as a component of large projects such as the Simulation Based Design (SBD), AIMS and AM3 projects, in which we are a subcontractor.

A few CAD vendors have also begun to realize that the future may involve completely unbundling their systems and providing "nuggets" of capability on a "pay-per-view" basis. This kind of line-item accountability is increasingly desired. The technology described in this proposal, along with complementary work by other MADE contractors, is an essential component of this change in business paradigm. By continuing to work with systems companies (particularly SGI and SUN) and CAD vendors (particularly Concentra) we expect this technology to appear in commercial systems.

More generally, the technology is likely to be commercialized as proprietary libraries and tools, as well as CAD augmentations. PENS [Hong et al. 1995] is indicative of a variety of documentation tools that will appear in the commercial marketplace. One example is a tool that would browse design documentation using the formal language (AEDNL) proposed here to help designers find alternatives for current issues and tasks; an approach different from the indexing methods of current case-based reasoning attempts to reuse designs.

In Section (H) Statement of Work, we lay out a program in which once a year, we provide ARPA with a demonstration of technologies integrated with those of at least one other MADE contractor, such as EIT, KSL, or Lockheed. Also, at the end of the first year, with at least one of these contractors, we will hold an ARPA workshop to demonstrate and explain the technologies. These workshops will be for the purpose of technology transfer, with attendees from the defense community.

Finally, we plan to participate in a MadeFast-like project ("MadeFast done right" with the adoption of formal design process management tools for coordination and documentation) at the end of two years. As an option, we will to provide the project coordination and integration functions we performed in the original MadeFast exercise.


(N) Facilities

Facility: Center for Design Research

Location: Building 02-560, Stanford University

Purpose: The Center for Design Research, located on the Stanford Campus, is dedicated to facilitating individual creativity, understanding the team design process, and developing advanced tools - methods that promote superior design and manufacturing of products. It is being used to create state-of-the-art product prototypes that range from dextrous robotic fingers to MADEFAST missle seekers. User-transparent computer -based instrumentation allow study of how people use design and collaboration tools, access design information databases and synthesize their thoughts for communication within the computer environment. There is strong emphasis on coordination and collaboration, human factors, product simulation, rapid prototyping, agent based design support, and manufacturing for the full product life cycle. The approach is incremental and integrative with special attention to visualization (human and machine), modeling (analytic and physical) and communication between design, manufacturing and management processes.

Systems:
Sun Sparc based workstations (10)
Silicon Graphics Indigo 2 workstations (2)
DEC MIPS based workstations (2)
Apple Macintosh 680x0 based workstations (6)
Apple Macintosh PowerPC based workstations (3)
Apple laptop computers (10)
80x86 based workstations (6)

All the above systems are connected to SUNet, the campus network which provides access to the world wide internet. Multimedia peripherals and software support internet based video and audio conferencing.

Languages:
Fortran, C, C++, Lisp, Assembler, Prolog, C, Pascal, TCL/TK, Perl

Major Engineering Packages:
PATRAN, NASTRAN, ANSYS, SDRC IDEAS, Concentra/Wisdom, Mathmatica, Working Model


(O) Experience

Work proposed in this document builds directly upon previous accomplishments by the principal investigators and key personel. Accordingly, that work has been referenced extensively in the context of new work proposed. We refer the reader to section J - Technical Rationale for the extended, context appropriate presentation of this material. The following few paragraphs are the short form summations of that work.

As a primary participant in the MADEFAST demonstration, this research group has obtained unique insight into both the capabilities and deficiencies of currently available tools and methodologies used to support distributed tag-team design. The research proposed here is directed at major technical shortcomings we observed first hand in MADEFAST: the inefficiencies associated with coordination and management of design space exploration amongst team members distributed in time and space.

Toye, Cutkosky and Leifer's knowledge and experience, as practicing mechatronics design engineers and instructors, provide solid technical grounding for assessing the value, effectiveness and impact of the technology being developed in support of mechatronics design.

The proposed project builds on a substantial heritage of research in areas such as AI and enginering, concurrent engineering, enterprise integration, CSCW (Computer-Supported Cooperative Work) pioneered at Stanford University and affiliated industrial research laboratories. Examples of directly relevant research contributions from this research team include SHARE, SHADE, DEDAL, First-Link, Next-Link, ACaPS, PENS.

The objective of the ARPA SISTO sponsored SHARE project is to provide the enabling technologies that will support design engineers by allowing them to create and access helpful information via the internet. These technologies were tested by co-located and distributed design teams working on industry sponsored design projects to assess their usability, effectiveness, and impact on the design process.

The ARPA SHADE (SHAred Dependency Engineering) project led by Lockheed-MartinAI Center is concerned with the information sharing aspect of concurrent engineering, It's objective is to develop and demonstrate a flexible infrastructure to support dynamic, knowledge-based, machine-mediated collaboration between disparate engineering tools.

DEDAL, developed collaboratively with researchers at NASA Ames Research, is an intelligent tool which uses device model representations for indexing and multimedia information retreival about the device being designed. It incorporates methologies for incremental, realtime modeling and indexing of design information as it is generated during the design process so that it can be available for retrieval and reuse.

First-Link is a prototype computer-support system for concurrent design of aircraft cable harnesses. An agent-based framework was developed to coordinate and manage the multiple perspectives and complex set of dependencies in the cable design process.

Next-Link, a continuation of the First-Link project, is an exploration into the principles of coordination that will enable computational agents to facilitate distributed design and engineering. A goal is to reduce time and cost associated with information flow in large design organizations by better facilitating direct peer-to-peer communication about design and engineering changes.

A collaborative experimental effort, the Agile Cable Production Service (ACaPS) project demonstrated the advanced cable production capabilities at Lockheed's Missles and Space Division as the first competitive capability to offer as a service assessible over the internet. ACaPS has shown that the NII can provide a powerful tool for devense conversion by enabling manufacturers to reach new markets, and by enhancing competitiveness through agility.

Personal Electronic Notebook with Sharing (PENS) is a recent development that addresses designer's personal needs for note recording and information reuse. Its agility in helping designers organize information nuggets nurtures reflective thought. Its operational simplicity promotes informal information sharing via group notebook publication on the WWW.


(P) Key Personel

Omitted.

(Q) Qualifications

Qualifications: Mark R. Cutkosky

Mark R. Cutkosky is the Associate Chair for Design and Manufacturing in the Mechanical Engineering Department and a co-director of the Stanford Integrated Manufacturing Association at Stanford University, Stanford, California. He joined Stanford in 1985, after working for several years in the Robotics Institute at Carnegie-Mellon University and as a design engineer at ALCOA, in Pittsburgh, Pennsylvania. Dr. Cutkosky research activities include computational support for concurrent product and process design and dextrous manipulation with robotic hands and tactile sensing. He is an NSF Presidential Young Investigator, an Anderson Faculty Scholar and the Charles M. Pigott Associate Professor at Stanford. He is a member of ASME, IEEE, SME and Sigma Xi.

Education:

1/85 Ph.D., Carnegie-Mellon University, Pittsburgh, PA.

5/82 M.E., Carnegie-Mellon University, Pittsburgh, PA

5/78 B.S., University of Rochester, Rochester, NY

Honors and Awards:

9/94 Appointed Charles M. Pigott Associate Professor

9/94 Best Paper award, 1994 ASME Database Symposium, (co-authors G. Olsen, J.M. Tenenbaum).

5/93 Outstanding Paper award, 1993 IEEE International Conf. on Robotics and Automation

(co-author J. Hyde).

9/89 Appointed Anderson Faculty Scholar at Stanford

2/86 National Science Foundation Presidential Young Investigator Award.

9/81 Phillip M. McKenna Foundation Fellowship.

Selected publications:

M. R. Cutkosky and Jay M. Tenenbaum, "A Methodology and Computational Framework for Concurrent Product and Process Design," Mechanism and Machine Theory, Vol. 25, No. 3, April,1990, pp. 365-381.

M. R. Cutkosky and J. M. Tenenbaum, "Providing Computational Support for Concurrent Engineering," the International Journal of Systems Automation: Research and Applications, Vol. 1, No. 3, pp. 239-261, 1991.

M. R. Cutkosky, D. Brown and J. M. Tenenbaum, "Working With Multiple Representations in a Concurrent Design System," ASME Journal of Mechanical Design, Vol. 114, No. 3, 1992, pp. 515-524.

K. Ramani, A. Miller and M. R. Cutkosky: "A New Approach to the Forming of Thermoplastic-Matrix Continuous-Fiber Composites. Part I: Process and Machine." Journal of Thermoplastic Composite Materials, Vol.5, July 1992, pp. 184-201.

M. Cutkosky, R. Engelmore, R. Fikes, T. Gruber, M. Genesereth, W. Mark, J. Tenenbaum and W. Weber, "PACT: An Experiment in Integrating Concurrent Engineering Systems," IEEE Computer , special issue on Computer Support for Concurrent Engineering, January 1993, pp. 28-37.

S. Kambhampati, M. R. Cutkosky, J. M. Tenenbaum and S-H Lee, "Integrating General Purpose Planners and Specialized Reasoners: Case Study of a Hybrid Planning Architecture," IEEE Transactions on Systems, Man, and Cybernetics,, Vol. 23, No. 6, November/December, 1993, pp. 1503-1518.

H. Park, M. R. Cutkosky, A.B. Conru and S-H. Lee, "An Agent-Based Approach to Concurrent Cable Harness Design," Artificial Intelligence for Engineering Design, Analysis and Manufacturing, Vol. 8, March, 1994, pp. 45-61.

G. R. Olsen, M.R. Cutkosky and J. M. Tenenbaum and T. R. Gruber, "Collaborative Engineering Based on Knowledge Sharing Agreements,"Concurrent Engineering Research and Applications,, Vol. 3, No. 2, 1995, pp. 145-159.

Other Major Support:

ARPA MADE SHARE project, ONR Tactile Sensing, Sandia Intelligent Agents

Qualifications: Larry J. Leifer

Professor Leifer's formal academic training was obtained at Stanford University. He holds a Bachelor of Science degree in Mechanical Engineering (1962) and a Master of Science degree in Product Design (1963). His Biomedical Engineering doctoral thesis (1969) dealt with the electrophysiology and control of voluntary human movement. From 1969 to 1972 he worked on human information processing at the NASA Ames Research Center. This work continued through 1973 as a NASA research exchange fellow to the MIT Man-Vehicle Laboratory. From 1973 to 1976 Dr. Leifer was an Assistant Professor of Biomedical Systems Analysis at the Swiss Federal Institute of Technology in Zurich. His work dealt with the neuromuscular control of posture, functional electrical stimulation of hearing, and measurement of motor axon conduction velocity distributions.

A member of the Stanford faculty since 1976, he now teaches the industrial project-based Graduate Automation and Machine Design course series, directs the industrial Design Affiliates Program and manages the Joint Design Research Seminar.

His research is based in the Center for Design Research (CDR), a member laboratory in the Stanford Institute for Manufacturing and Automation (SIMA). As CDR's founding Director (1984) he endeavors to develop basic design theory and methodology through the application of knowledge-based engineering technology to a wide range of industrial machine design problems. Special interest projects include: a), development of an electronic design notebook for "design knowledge capture"; b), development of telerobotic assistants for medical, rehabilitation, industrial and space applications; c), development of concurrent product and process design software for machine and injection mold design; and d), development of conceptual design analysis tools.

Selected publications

Leifer, L.J., "On the Nature of Design and an Environment for Design." In Rouse, W.B., and Boff, K.R., editors, System Design: Behavioral Perspectives on Designers, Tools, and Organizations, North-Holland, 1987, pp.199-210

Edwards, L., Kessler, W., and Leifer, L., The Cutplane: a Tool for Interactive Solid Modeling, ACM-SIGCHI Bulletin, October, 1988, pp. 72-78

Lakin, F., Wambaugh, J., Leifer, L., Cannon, D., and Sivard, C., The Electronic Design Notebook: performing medium and processing medium, The Visual Computer, Springer International, Vol.5, No.4, August, 1989, pp. 214-226

Hammel, J., Hall, K., Lees, D., Leifer, L., Van der Loos, M., Perkash, I., and Crigler, R., Clinical Evaluation of a Desktop Robotic Assistant, Journal of Rehabilitation Research and Development, Vol.26, No.3, July, 1989, pp. 1-16.

G. Toye, M.R. Cutkosky, L.J. Leifer, J.M. Tenenbaum and J. Glicksman, "SHARE: A Methodology and Environment for Collaborative Product Developement," The International Journal of Intelligent and Cooperative Information Systems , vol.3, no.2, June 1994, p. 129-53.

Baya, V.; Gevins, J.; Baudin, C.; Mabogunje, A.; Toye, G.; Leifer, L.; "An Experimental Study of Design Information Reuse" , In Proceedings of the 4th International Conference on Design Theory and Methodology, ASME, Sept. 13-16, Scottsdale, Arizona, pp 141-147, 1992.

L.J. Leifer, C. Baudin, J. Gevins, G. Toye, V. Baya, A. Mabogunje, "A tool and protocol for developing reusable design knowledge," Proceedings of First AIAA Aerospace Design Conference, 1992

Other Major Support:

ARPA MADE SHARE project, SIMA ICM project, NSF Synthesis project, NASA GCDK project

Qualifications: George Toye

Education

'83-'89 Ph.D. Mechanical Engineering (minor: Electrical Engineering), Stanford Univ.

'80-'81 M.S. Mechanical Engineering (Automatic Control), Univ of California at Berkeley

'77-'80 B.S. Mechanical Engineering (Energy Conversion), Univ of California at Berkeley

Professional Experience:

Associate Director:/Research Associate Stanford Center for Design Research, Stanford, CA, '89-

Administer, direct and coordinate research projects in mechanical engineering - mechatronics design research and design education at CDR. Other research interests include fault tolerance and high reliability systems design and analysis.

Biomedical Engineer Veterans Administration Rehabilitation R&D center, Palo Alto, CA, '85-'87

Developed full system level fault tolerance technology for mobile robots that assist and serve quadriplegic patients.

Engineer Systems Control, Inc., Palo Alto, CA, '81-'84

Analyzed and optimized operation of BWR/PWR nucluear power plants, pulp and paper mills using first principle based simulation model and controls. Designed first ever fault tolerant digital feedwater flow control system for use in nuclear power plants.

Additional Information:

California Board of Registration certified Professional Engineer in Mechanical Engineering

Chairman of U.C. Berkeley Campus Safety Committee (1979-1981). Member of ASME, IEEE, ASEE, NSPE. Editor of CSPE Bay Area Engineering Newsletter, a publication of the Golden Gate Chapter of California Society of Professional Engineers (1990-1991). Active volunteer for MATHCOUNTS (1986-1995)

Recent Publications:

Hong, J, Toye G., Leifer, L. Personal Electronic Notebook with Sharing, submitted to the Fourth IEEE Workshop on Enabling Technologies, April, 1995.

Hong, J, Toye G., Leifer, L. Using the WWW for a Team-Based Engineering Design Class, Electronic Proceedings of the 2nd WWW Conference, Chicago, IL, October, 1994

G. Toye, M.R. Cutkosky, L.J. Leifer, J.M. Tenenbaum, J. Glicksman, "SHARE: A methodology and environment for collaborative product development," Int. Journal of Intelligent and Cooperative Information Systems, Vol 3, No 2, pp129-153, 1994

G. Toye, L.J. Leifer, "Helenic fault tolerance for robots," Computers Elect. Engng, Elsevier Science Ltd., Vol. 20, No. 6, pp 479-497, 1994

Toye, G.; Isaksen, L., Mukherjee, S., Carmichael, L.; "Demand control of commercial buildings: dynamic modeling experimental design" , In Proceedings of 1992 EPRI Advanced Computer Technology Conference, Scottsdale, Arizona, December, 1992.

Baya, V.; Gevins, J.; Baudin, C.; Mabogunje, A.; Toye, G.; Leifer, L.; "An Experimental Study of Design Information Reuse" , In Proceedings of the 4th International Conference on Design Theory and Methodology, ASME, Sept. 13-16, Scottsdale, Arizona, pp 141-147, 1992.

L.J. Leifer, C. Baudin, J. Gevins, G. Toye, V. Baya, A. Mabogunje, "A tool and protocol for developing reusable design knowledge," Proceedings of First AIAA Aerospace Design Conference, 1992

G. Toye, L.J. Leifer, "Failure Modes and Effects Recognition in TMA Design," Proceedings of SPIE's 1990 International Symp. on Optical and Optoelectronic Applied Science and Engineering, July 1990

Other Major Support:

ARPA MADE SHARE project, SIMA ICM project, NSF Synthesis project, JPL ICM-MIDAS project

Qualifications: Charles J. Petrie

Education

He has a B.S. in Mathematics and a M.S. and Ph.D. in Computer Science from the University of Texas at Austin in Austin, Texas.

Employment History

Stanford University Center for Design Research, 1993. Research Associate.
Research interests: coordination of distributed design.

Microelectronics & Computer Technology Corp. (MCC), 1984 to 1993. Sr. MTS and Project Leader. Research interests: hybrid inference architectures, truth maintenance, distributed design and planning.

Sperry Corporation, 1978 to 1984. Systems Analyst and Manager. Performance analysis, seismic processing benchmarks, and custom software development.

Solar Research Inc., Austin, 1974 to 1978. Engineer Programmer.

U.S. Eighth Army. MP in Korea from 1970 to 1971.

Professional Activities

He has served and continues as a reviewer, guest editor, and/or program committee member for IEEE, AIEDAM, CERA, AI, AAAI, IJCAI, ECAI, and CAIA. He has four times been been a visiting scientist at the University of Kaiserslautern and the DFKI. Dr. Petrie was the Program Chair of the 1992 ICEIMT series of the international workshops and conference documented in the MIT Press book he edited: "Enterprise Integration Modeling". He also chaired the 1992 AAAI Workshop on AI in Enterprise Integration, and continues to organize workshops on concurrent engineering. The last was a 1994 workshop on distributed scheduling held at the IEEE Conference on Artificial Intelligence Applications. He was the invited speaker at the workshop on "Conflict Management" at AID-94, the conference on AI in design. He was a keynote speaker at the 1995 German conference on knowledge-based systems, XPS-95. He is the General Chair of the 1996 IEEE WETICE to be held in June at Stanford.

Selected Publications:

``Context Maintenance'', Proc. AAAI-91.

Decision Revision,'' Proc. AAAI-92.

Enterprise Integration Modeling, (ICEIMT) editor, MIT Press, October, 1992.

``A Minimalist Model for Coordination'', AAAI-92 Workshop on Design Rationale. Also in Enterprise Integration Modeling.

``The Redux' Server,'' Proc. Internat. Conf. on Intelligent and Cooperative Information Systems (ICICIS), Rotterdam, May, 1993. Abstract / Report

``Design Space Navigation as a Collaborative Aid,'' Proc. AI in Design: 3rd Internat. Conf., pp. 611-623, Lausanne, August, 1994. Abstract / Report

``Using Pareto Optimality to Coordinate Distributed Agents,'' with T. Webster and M. Cutkosky, AIEDAM, to be published, 1995. Abstract / Report

Other Major Support:

ARPA MADE SHARE project


(R) Other Proposals

Omitted

(S) Bibliography

[Amon 1994] C. Amon et al., "Modeling Novel Manufacturing Processes," Manufacturing Science and Engineering, PED v 68-2,ASME,. pp. 535-546, , 1994.

[Baudin et at. 1992] Baudin, C., Gevins, J., Baya, V., and Mabogunje, A., "Dedal: Using Domain Concepts to Index Engineering Design Information," Proceedings of the 14th Annual Conference of the Cognitive Science Society, August, 1992.

[Clausing ] Hauser, J. R., Clausing, D., The House of Quality, Harvard Business Review, May/June 1988, pp 63-73

[Conru 1994] A. Conru, "A Genetic Approach to the Cable Harness Routing Problem," Proceedings of the IEEE World Congress on Computation Intelligence, 1994.

[Doyle 1985] J. Doyle, "Reasoned Assumptions and Pareto Optimality," Proc. of the 9th IJCAI , pp. 87-90, 1985.

[Eppinger et al. 1991] Eppinger, S. D., D. E. Whitney, R. P. Smith and D. A. Gebala 1991. Organizing the Tasks in Complex Design Projects. Computer-Aided Cooperative Product Development. D. Sriram and R. Logcher, ed. New York: Springer-Verlag. 229-252.

[Gage 1995] P. Gage, "New Approaches to Optimization in Aerospace Conceptual Design," Ph.D. Dissertation, Stanford University, 1995, also published as NASA Contractor Report 196695.

[Gebala and Eppinger 1991] D.A. Gebala S.D. Eppinger, "Methods for Analyzing Design Procedures," in Design Theory and Methodology, ASME Vol. DE-31, 1991, pp. 227-232.

[Forney] Paul Forney informal presentation on the Lockheed-Martin AIT/THAD Missile Seeker Programs, August 11, 1995.

[Hong et al. 1995] Hong, J, Toye G., Leifer, L. Personal Electronic Notebook with Sharing, In Proceedings of Fourth IEEE Workshop on Enabling Technologies, April, 1995.

[Ishii et al. 1994]

[Java] Sun Hot Java, WWW URL http://java.sun.com/.

[Olsen et al. 1995] G. R. Olsen, M.R. Cutkosky and J. M. Tenenbaum and T. R. Gruber, "Collaborative Engineering Based on Knowledge Sharing Agreements,"Concurrent Engineering Research and Applications,, Vol. 3, No. 2, 1995, pp. 145-159.

[Park 1995] H. Park, Modeling of Collaborative Design Processes for Agent-Assisted Product Design, Ph.D. Dissertation, Stanford University, March 1995.

[Park et al. 1994] H. Park, M. R. Cutkosky, A.B. Conru and S-H. Lee, "An Agent-Based Approach to Concurrent Cable Harness Design," Artificial Intelligence for Engineering Design, Analysis and Manufacturing, (AIEDAM) Vol. 8, March, 1994, pp. 45-61.

[Petrie et al. 1994] C. Petrie, M. Cutkosky and H. Park, "Design Space Navigation," Proceedings of the Third International Conference on AI in Design, pp. 611-623, August, 1994, Lausanne, Switzerland.

[Petrie et al. 1995] C. Petrie, T. Webster and M. R. Cutkosky, "Using Pareto Optimality to Coordinate Distributed Agents," to appear in Artificial Intelligence for Engineering Design, Analysis and Manufacturing, (AIEDAM special issue on conflict management, 1995.

[Reddy and Cagan 1993] G.M. Reddy and J.Cagan, "Optimally Directed Truss Topology Generation Using Shape Annealing," DE-Vol. 651, Advances in Design Automation, ASME 1993, pp. 749-759.

[Rockwell] Rockwell Science Center, Systems Design and Modeling, Project 809 (GEM) FY1995 Final Report, September 1995. (In Preparation.)

[RPL] Rapid Prototyping Laboratory WWW URL http://www-rpl.stanford.edu/ .

[Weis et al. 1992] L.E. Weiss, F.B. Prinz, D.A. Adams and D.P. Sieworeck, "Thermal Spray Deposition," Journal of Thermal Spray Technology, Vol 1. No. 3, 1992, pp. 231-237.

[McGuire et al 1993] J.G. McGuire, D.R. Kuokka, J.C. Weber, J.M.Tenenbaum, T.R. Gruber, and G.R. Olsen, "SHADE: Technology for Knowledge-Based Collaborative Engineering," Concurrent Engineering Research and Applications (CERA), 1(3):137-146.

[Talukdar et al. 1991] Talukdar, S. N., P. S. deSouza, R. Quadrel and V. C. Ramesh, "A-Teams: Multi-Agent Organizations for Distributed Iteration," EPRI/NSF Workshop on Application of Advanced Mathematics to Power Systems, , 1991.

[Toye et al. 1994] G. Toye, M.R. Cutkosky, L.J. Leifer, J.M. Tenenbaum and J. Glicksman, "SHARE: A Methodology and Environment for Collaborative Product Developement," The International Journal of Intelligent and Cooperative Information Systems , vol.3, no.2, June 1994, p. 129-53.

[Ward et al. 1995] A. Ward et al., "The Second Toyota Paradox: How Delaying Decisions Can Make Better Cars Faster," Sloan Management Review, Spring 1995, Vol. 36, No. 3, pp. 43-61.

APPENDIX:

1. Acronyms

2. Figure I-1:

3. Figure I-2:

4. Figure I-3:

4. Letter of Support: SGI

5. Letter of Commitment: PSA


Acronyms

ABE Agent Based Engineering

ABL Airborne Laser

ABSML A Better Structured Markup Language

ACaPS Agile Cable Production Service

AEDNL ABSML based Engineering Design Notebook Language

AI artificial intelligence

AIMS Agile Infrastructure for Manufacturing Systems

AIT Atmospheric Intercept Technology program

AM3 Affordable Multi-Missle Manufacturing

APCL Agent Process Communication Language

API Application Programming Interface

ARPA Advance Research Project Agency

BPI Boost Phase Intercept

CAD Computer Aided Design

CAE Computer Aided Engineering

CAM Computer Aided Manufacturing

CDR Center for Design Research (Stanford Univ)

CFD Concept Feasibility Demonstration

CFRC Carbon Fiber Reinforced Composites

CM Constraint Manager

CMU Carnegie Melon University

CNC Computer Numerical Control

CODE COoperative Design Exploration

CORBA Common Object Request Broker Architecture

CPM Critical Path Management

CSCW Computer Supported Cooperative Work

DDE Design Documentation Environment

DDM Distributed Design Methodology

DEliza DesignEliza

DOD Department of Defense

DSC Design Space Colonization

DTA Design Technologies Agent

EIT Enterprise Integration Technologies, Inc.

EPL Electronic Project Language

GEM Graphical Enterprise Model

HTTP Hyper-Text Transport Protocol

IMTL Integrated Manufacturing Technology Laboratory, Sandia

ITI Industrial Technology Institute

KKV Kinetic Kill Vehicle

KQML Knowledge Query and Manipulation Language

KSL Knowledge Systems Laboratory

MADE Manufacturing Automation and Design Engineering

MD* Metal Deposition

MECE Multimedia Engineering Collaborative Environment

MOSIS Metal Oxide Semiconductor Integrated Systems

NMD National Missle Defense

PI Principal Investigator

PCA Process Capabilities Agent

PDA Personal Digital Assistant

PDES Product Data Exchange Standard

PENS Personal Electronic Notebook with Sharing

PERT Project Evaluation and Review Technique

PM Project Manager

PODS Process Open Description System

QFD Quality Function Deployment

PSA Pipeline Systems Analysis

RPL Rapid Prototyping Laboratory

SADT Structured Analysis and Design Technique

SBD Simulation Based Design

SD-PCA Shape Deposition Proces Capabilities Agent

SGI Silicon Graphics Inc.

SHADE Shared Dependency Engineering

SIMA Stanford Integrated Manufacturing Association

SSF Solid Free-form Fabrication

TFB Technology Facilitation Broker

THAAD Theatre High Altitude Airborne Defense

TI Terminal Intercept

TMD Theatre Missle Defense

VLSI Very Large Scale Integration

WWW World Wide Web


Figure I-1: ProcessLink Framework with example EPL message types and CAD software.


Figure I-2: An agent dialogue for generating Design Rules on Demand for a new manufacturing process. In this case, the processes under consideration include layered shape deposition processes and a flexible machining and welding cell. The dialogue includes definitions of process capabilities, features and constraints, referred to shared ontologies and standards. Process simulation capabilities are also transported via Hot Java.


Figure I-3: Illustration of the Design Space Colonization concept relative to missile design. Explorers of the space collect, record and analyze design options. New technologies offer portals to new design spaces.