AN INTEGRATED FRAMEWORK TO SUPPORT DISTRIBUTED CAD

AN INTEGRATED FRAMEWORK TO SUPPORT DISTRIBUTED CAD OVER THE INTERNET Amar Gupta, Sanjeev Vadhavkar, Feniosky Peña-Mora, Jason Yeung Massachusetts Institute of Technology 77 Massachusetts Avenue, Rm E40-194a Cambridge, MA 02139 {agupta, vada, feniosky, jyeung}@mit.edu design issues are gathered and evaluated, the initial design assumptions are refined and a better understanding of the user requirements is achieved. The decentralization of operations adds to the complexity of the design process due to differentials of time, space and organizational dimensions. Design teams invent, design, analyze, refine, simulate, build, test, and document the artifact. For efficient handling of these tasks, they must create, distribute, interpret, and assimilate nearly overwhelming amounts of information, and to do so the most important single element is the rationale of the ongoing work. Designers work in a relay-team environment, in which they must transfer information through space and time, and rapidly exchange and assimilate information to ensure success. Hence, the design project consists to a large extent of information recording, management, processing and communication, especially so in concurrent or simultaneous engineering ventures. Both large and small design teams face this issue. In a large team, each individual owns a smaller share of the information, but there is a much larger requirement on effective communication; in a small team, communication is easier, but each individual is responsible for a much larger chunk of information. In view of the above, one seeks several types of functionality from new computational tools, especially in the following areas: • Creation, Capture and Representation of Design Rationale: All design begins with creation: an unpredictable process that is highly dependent on each individual designer. Since each designer is creative under different conditions, capturing this free flow of information is a great challenge. At present, designers typically enter information in design notebooks, characterized by severe limitations and rigidity with respect to information entry and interrupt, and take time away from the design process. Given these facts, there are several needs for information creation and capture that need to be addressed. First, in order to stimulate and cross-pollinate ideas in a collaborative environment, the ideas and concepts of designers should be shared as soon as possible. Second, the design rationale information should be captured in a form that is useful for more formal downstream design processes, and that allows

Abstract Collaborative settings of multi-disciplinary teams separated across geographical and temporal boundaries call for an integrated framework to support intelligent and distributed CAD. This paper presents a framework to improve the ability to represent, capture, and reuse design rationale by using: computer-supported design rationale model (DRIMER) to capture design rationale; collaborative tools for handling team interactions over the Internet; and case-based reasoning mechanisms for organizing and analyzing design artifacts and rationale.

Introduction The next generation of engineering support tools for CAD/CAE systems will rely heavily on collaborative software across local networks and/or the Internet. At the application level, the vision for the next generation includes integrated and distributed CAD/CAE systems with product and process models, coupled with design rationale information that is shared, reused and merged with other design processes as part of the collaborative design endeavor. Knowledge about the design of previous artifacts and the process by which a design is realized are of great importance to designers [MSS97]. Design rationale information can enable improved understanding of designs and simplify the problems of design revisions and reuse. Recording the rationale behind design decisions can greatly facilitate collaboration among design team members. Traditional techniques and computer tools have relied on geometrical representations of the design in the form of drawings and specifications. However, tools and methodologies to support design decisions and the underlying reasoning, i.e., design rationale, are still at an early stage of evolution. Design processes in complex engineering systems can be construed as collaborative-iterative decisionmaking activities that involve significant interactions among various specialists. Each of these activities is organized to conceive the idea, prepare the description, determine the plan by which resources are converted, and produce end results. As the design evolves and more data about the ideal solution and

1

•

•

•

be provided to allow the growth and evolution of design rationale within the repository. • Information sharing and networking: The design rationale documentation provides the central link between the creations of the design team and the outside world. Therefore, the documentation must describe the designed product and address requirements in detail without ambiguity. When it comes to documenting and creating presentations during an ongoing design process, there is always a scramble to collect and compile data. The problem space increases when we take into account the current global nature of mechanical design where designers work beyond geographic boundaries and come from differing work cultures and work settings. Hence, it becomes especially important to record not only the design rationale information in design repositories, but also to provide an interface for sharing this vital information among all the designers. This ability will have a direct impact on the effectiveness of communication between design teams. A framework supporting intelligent and distributed CAD/CAE can provide significant benefits especially in the following areas: • Reduce design cycle through task coordination; • Evaluate more design alternatives by providing faster turn-around; • Design optimization by recording and reusing design rationale; • Increase product performance by efficient search of larger design space; • Reduce design rework through inclusion of manufacturing constraints and user requirements; • Provide design representation for effective reuse at a later stage; and • Reduce design cycle through ease of information retrieval.

comparison with other concepts and evaluations against specifications. Above all, the method utilized for capturing design rationale information should require minimal effort from the designer. The computer tools processing the design rationale information should be transparent thereby reducing the amount of time required by the designer to enter relevant information into the computer. Storage of information in design repositories about the design of previous artifacts and the process by which the design was realized: For proper use and reuse of design rationale information created in the first stage, it is necessary to store and index this information in design repositories. Merely providing access to schematics of artifacts is inadequate for this purpose; such information would lack the depth to convey adequate understanding of the design artifact and the underlying design process. The design repositories would be designed to provide a central resource through which, designers separated across geographical and time boundaries, could access case studies which are related to a problem of interest (for example, design of similar artifacts, design of artifacts with similar intents, similar design problems, similar design decisions), and abstract information which could be applied to the problem in hand. Search of design rationale information: Quick and easy access to background information is critical at all stages of the design process. Designers spend a great deal of time searching for design rationale. Due to the incompatibilities in data presentation for design rationale information, this systematic task becomes very time consuming. A streamlined approach is needed to provide a designer with timely and accurate access to the design repositories during the design process. Tools are needed within a collaborative design system with the ability to define a seamless search space of design rationale information and technical and organizational constraints for the creation of the product. Growth and evolution within the context of design repositories: For proper use of the design repositories, it is necessary to adopt an evolutionary growth rate. To utilize the corporate knowledge and design rationale information gained during a design process, it is necessary to update the design repositories at every stage of design. The repositories should be designed to handle and mimic the evolutionary design process. Additional access mechanisms need to

Framework for Distributed Design In order to improve the ability to represent, capture, and reuse design rationale in collaborative settings, the following focus areas have been identified: 1. Creation of design rationale model and tools: New methods and computer-supported representations to capture, maintain, and reuse records of the basis for the decisions made by designers. The design rationale model will enable the active capture of design rationale information and support the integration of analytical results into the design rationales. In a collaborative environment, designers are drawn

2

from different domains to work together. They usually have some common knowledge or a shared perspective of the design process. In collaborative design involving geographically and time separated multi-disciplinary teams this common knowledge serves as their only link to understanding and appreciating individual professional perspectives. In order to achieve efficient understanding at minimum costs, the approach would be to use existing knowledge to define a map from the designer’s rationale to the designer’s knowledge. Specifically, the following objectives will be addressed: provide a representation that stores, integrates and manages the various types of design knowledge; provide a representation that models and represents information such as requirements, design rationale, versions, etc. 2. Integration of the design rationale model with case-based reasoning principles: The models and tools conceived for capturing and storing design rationale and design process information will be combined with case-based reasoning principles. Specifically, this will involve easing the access to and acquisition of

3.

design rationale information in design repositories by providing knowledge-based mechanisms. Providing a collaborative framework for the overall architecture: The overall architecture will be geared to support the access to design repositories by distributed design teams over the Internet. The collaborative framework could include among others: multi-media support, overall performance on different types of networks, performance versus bandwidth, and integration with existing CAD/CAE tools. Specifically, the following objectives will be addressed: provide a shared environment in which engineers can explore space of alternative designs and communicate their designs in a uniform manner to the design repositories; provide each designer with a private working space where the individual design capabilities can be greatly enhanced by reusing design cases from the design repositories; manage the design processes by providing adequate communication and coordination capabilities in order to improve the quality as well as productivity of the entire collaborative design effort.

Figure 1: DRIMER (Design Recommendation and Intent Model Extended to Reusability)

3

Design Recommendation and Intent Extended to Reusability (DRIMER)

concrete experience in domain. By capturing the past experience, DRIMER offers a mechanism to leverage these cases effectively. DRIMER can be used for supporting the capture, modular structuring and effective access of design rationale information needed for effective collaborative design. By capturing the evolution of a particular system or its individual components in a form understandable by the computer, DRIMER taps the computer's resources to record, represent, and index the design process underlying the created product. DRIMER provides the pivotal framework for building comprehensive software environments that integrate design rationale much more tightly with the design process itself and helps in the design decision-making process. Design rationale capture process is non-obtrusive and occurs as a natural step in the design process in a collaborative setting.

Model

The design rationale model we wish to utilize for the capture and retrieval of design rationale information is called Design Recommendation and Intent Model Extended for Reusability (DRIMER) [PV96] [PV97] [PVD01] and is shown in Figure 1. DRIMER uses the Object-Oriented Modeling Technique (OMT) [Real91] as a representation language. The designer, after collaborating with other designers, presents project proposals based on design intent. The design intent refers to the objectives of the design project, the constraints involved, the functions considered or the goals of the project. The designer can present a number of different proposals satisfying common design intent. A given proposal may consist of sub-proposals allowing for a representation of the decomposition approach taken in a design process to reduce complexity. A proposal may react to an existing proposal by either supporting, contradicting or changing the ideas put in the existing proposal, thus representing the argumentative nature of collaborative design. A project proposal includes the designer's recommendation and the justification of why that particular proposal is recommended representing a relationship between a solution (i.e., artifact) and the problem (i.e., intent). The introduction of design intents and recommendations allows for a representation of the evolutionary nature of design in which problems, solutions and justifications are discovered as the design process proceeds. An artifact has both behavioral and structural properties. The artifact comprises the whole system as well as the individual components in the system. Justification explains why the recommendation satisfies the proposed design intent. In DRIMER, the artifact component in a design context represents a sub-component of the various design cases in a design repository. The design repository acts as a resource through which the designer can access various design cases that are similar (similar artifact or proposal or intent) to the current design situation. The recommendation introduces or modifies the cases in a design repository. DRIMER allows for the explicit capture of design rationale during the design of complex electro-mechanical parts. If this design rationale is not captured explicitly, it may be lost over time. This loss deprives the maintenance teams of critical design information, and makes it difficult to motivate strategic design choices to other groups within a project or organization [Schmidt95]. Identifying, documenting and reusing useful design cases requires

Collaborative Backbone for Distributed Design Teams The design, manufacturing, and repair of artifacts require collaboration among several specialists. The design expertise of these specialists contributes in a large way to the organizational knowledge, and the availability and communication of such knowledge is among the key competitive factors in the manufacturing sector. Furthermore, the decentralization of operations and the differentials of time, space, and organizational dimensions among several firms add to the complexity of the design process. Traditionally, this problem has been addressed at two levels: relocating the entire design team to a central design facility where they can interact and design the product or run multiple iterations by passing the design documents between the design team members. Both methods result in a sub-optimal use of human expertise and organizational resources. Decision making process is typically long and the implicit design rationale information during the design process is never retained for reuse. Hence, the main reason to provide a collaboration backbone is to create “virtual” design teams by coalescing distributed design teams without the constraint of collocation [HPS95]. Computer based collaboration encompasses a variety of complex systems. However, the computerbased collaboration problem can be decomposed into three layers of service: collocation, cooperation, and coordination [PHVB00]. This model projects the necessary physical meeting elements into a general requirement list. These requirements are based on the fact that physical meetings require three key components in order to exist:

4

1.

A physical meeting room in which the participant can meet (collocation); 2. A common language and a shared understanding of materials to be presented in the meeting (cooperation); 3. An agenda and an individual or set of individuals that ensures the agenda is maintained and the group is focused on resolving the issues outlined in the agenda (coordination). Collaborative and cooperative software environments are commonly referred to in the literature as “groupware”. “Groupware” has been loosely defined as computer-based collaborative systems designed to support multiple users engaged in a common task, thereby providing an interface to a shared environment [EGR91]. There are a number of commercial and academic conferencing systems that can be broadly classified into: videoconferencing systems (for example, AT&T’s RAPPAPORT), shared document managers (for example, Lotus Notes) and shared workspace managers (for example, MIT DICE project [SL91]). Current versions of “groupware” systems on the market satisfy only some of the requirements (collocation, cooperation, coordination) of physical design meetings. In addition, most of the “groupware” systems fail to provide active computer support for design rationale capture and dissemination during the decision making process between distributed design teams. Hence, a collaborative backbone is required for supporting distributed design teams, which enables multimedia collaboration among group members as well as provide active computer support for design rationale.

Case Based Reasoning Case-based reasoning systems are knowledge-based systems that store information about situations in their memory. Figure 2 from [Kol93] shows various relationships between primary processes of casebased reasoning systems. The primary process in any type of case-based reasoning system starts with case retrieval. In order to ensure that poor solutions are not repeated along with the good ones, the case-based reasoning system evaluates the solution. In the problem-solving role, the case-based reasoning system proposes initial solution by extracting the solution from a retrieved case. This is followed by adaptation, the process of fixing an old solution to fit a new solution, and criticism, the process of critiquing the new solution before the actual implementation. In the interpretive role, after proposing initial solutions, the case-based reasoning system justifies the solution proposed by creating an argument in favor of the proposed solution. Figure 2 shows the various recursive processes. Case-based reasoning systems suggest a model of reasoning that incorporates problem solving, understanding, learning and integration with historical cases [Kol93] [Veca93]. The premises underlying this model are as follows: • Historical design cases can be reused to a significant advantage. • Design descriptions are typically inadequate and unstructured, requiring further understanding. Hence, re-interpretation of the design problem is a necessary prerequisite to learning, which in turn helps the reuse of the past design problems. • Proposed solution needs to be adapted, as the historical design case is not exactly similar to the new case. For example, even with closely matching intents, the final design may be completely different depending on the context. • Learning process occurs as a natural consequence of reasoning. • Feedback and feedback analysis, along with reasoning, are integral parts of the understanding/learning/reuse cycle. The combination of reasoning and learning behavior in a case-based reasoning system and its ability to hold experience-acquired associative knowledge [Veca93] provide a strong mechanism to leverage design experience from the past. In general, the reuse of case-based reasoning principles with design cases from the design repository serves three principal benefits: • Suggests proven and tested design cases to the problems faced by the user.

RETRIEVE

PROPOSE INI. SOL.

ADAPT

JUSTIFY

CRITICIZE

EVALUATE

STORE

Figure 2: Case-Based Reasoning Cycle

5

•

[Gamma94] Gamma, E., Helm, R., Johnson, R. and Vlissides, J. “Design Patterns,” Addison-Wesley Publishing Company, Reading, Massachusetts, 1994. [GBS96] Goel, A. Bhatta, S. and Stroulia, E. “KRITIK: An Early Case-Based Design System,” in Issues in Case-Based Design, Maher, M. and Pu, P. (eds), Hillsdale, NJ. [HPS95] Hussein, K., Pena-Mora, F. and Sriram, D. “CAIRO: A System for Facilitating Communication in a Distributed Collaborative Engineering Environment,” In Proceedings of the Fourth IEEE Workshop on Enabling Technologies: Infrastructure for Collaborative Enterprises. p. 174-183. April 1995. [Kol93] Kolodner, J. “Case Based Reasoning,” Morgan-Kaufmann, 1993. [MSS97] Murdock, J., Szykman, S. and Sriram, D. “An Information Modeling Framework to Support Design Databases and Repositories,” in Proceedings of the DETC’97, 1997 ASME Design Engineering Technical Conferences, Sacramento, CA. [PV96] Pena-Mora, F. and Vadhavkar, S. “Design Rationale and Design Patterns in Reusable Software Design,” in Gero, J. editor, Artificial Intelligence in Design ’96, Kluwer Academic Publishers, London, England, 1996. [PV97] Pena-Mora, F. and Vadhavkar, S. “Augmenting Design Patterns with Design Rationale,” Artificial Intelligence in Engineering Design and Manufacturing, 11(2), 1997. [PVD01] Pena-Mora, F., Vadhavkar, S. and Siva Kumar, D. “Component-Based Software Development for Integrated Construction Management Software Applications,” to appear in Artificial Intelligence in Engineering Design and Manufacturing (AI EDAM), 2001. [PHVB00] Pena-Mora, F., Hussein, K., Vadhavkar, S. and Benjamin, K. “CAIRO: A Concurrent Engineering Environment for Virtual Design Teams,” in Artificial Intelligence in Engineering, Vol. 14 (2000), Issue 3, pp 203-219. [Real91] Rumbaugh, J., Blaha, M., Premerlani, W., Eddy, F. and Lorensen, W. “Object-Oriented Modeling and Design,” Prentice-Hall Inc., Englewood Cliffs, NJ, 1991. [Schmidt95] Schmidt, D. “Experience using Design Patterns to Develop Reusable Object-Oriented Communication Software,” Communications of the ACM, 38:65-74, October 1995. [SL91] Sriram, D. and Logcher, R. “The MIT DICE Project,” Computer, 26(1): 64-65, January 1993. [Veca93] Veloso, M. and Carbonell, J. “Automatic case generation, storage and retrieval in PRODIGY,” in Minton, S., editor, Machine Learning Methods for Planning and Scheduling, Morgan-Kaufmann, 1993.

Provides a context for understanding or assessing the design. • Provides explicit justifications during the design decision-making process. In general, designers tend to interleave the abovementioned processes according to the need of time. For example, in problem solving situations, design cases from the repository can be used to solve problems whereas in interpretive situations, the same cases can be used for criticism, justification and evaluation of the solution.

Conclusion and Outlook The paper presents a design rationale based framework to assist geographically distributed designers to collaborate remotely. The proposed framework allows for computer support to multiple designers in the area of active design rationale capture for reusability. The approach is based on abstracting the design as well as the process involved in designing an artifact and reusing them to design newer artifacts and systems. In addition, the approach envisions a paradigm shift from the specify-buildthen-maintain life cycle assumed in the past to one of component-based design. In summarizing the framework, the contributions can be divided into two parts. First, a model for representing design rationale has been developed. This model (i.e., DRIMER) provides primitives for representing design knowledge in terms of the reasoning process used by the designers to generate an artifact that satisfies their design intents. Second, the framework combines DRIMER with intelligent search mechanisms, to offer active assistance to designers in reusing artifacts and systems. Although the framework emphasizes the importance of documenting the design process, instead of laying extra burden on the programmers, it assists them by providing active computer assistance in recording the key design decisions. A prototype based on the framework highlighted in this paper is under development. The prototype will be tested in an industrial setting with a normal working environment where there are time pressures and organizational constraints preventing widespread reuse of design rationale information.

References [Alex79] Alexander, C. “The Timeless Way of Building,” Oxford University Press, NY, 1979. [EGR91] Ellis, C., Gibbs, S. and Rein, J. “Groupware: Some issues and experiences,” Communications of the ACM, 34(1): 38-58, 1991.

6