David C Brown

Analyzing complex scientific data, eg, graphs and images, often requires comparison of features: regions on graphs, visual aspects of images and related metadata, some features being relatively more important. The notion of similarity for comparison is typically distance between data objects which could be expressed as distance between features. We refer to distance based on each feature as a component. Weights of components representing relative importance of features could be learned using distance function learning ...

Having offered our own definition of engineering design, we now go on to study design as an activity – that is, the process of design. In this chapter, we review some of the established models of the design process. Then we go on to describe more recent articulations, some of which are rooted in the AI-based ideas mentioned earlier. Parts of the discussion may seem vague or abstract because we are trying to describe a complex process by breaking it down into smaller, more detailed pieces, but we are not going to produce a detailed cookbook that must be followed in order to complete a design. We are simply trying to picture, in words and diagrams, what is going on in our head when we are doing design. Dissecting the Design Process We start by looking back at the questions we asked (and others that we might have but did not) in our ladder-design exercise. In so doing, we find that we can decompose or break down the process into a sequence of steps by extracting and naming some of those steps. For example, when we ask, “How much weight should a safe ladder support?” and “For what purposes is the ladder to be used?,” we are clarifying the client's requirements .

Design expert systems can be characterized by the type of compiled knowledge used. Compiled knowledge is efficient for design problem-solving with a bounded set of design problems, but often leads to failure in slightly different design problem situations. Design decomposition knowledge is often compiled into design systems. Design specification changes often impose problem decomposition changes. Generating decomposition knowledge for a slightly different design problem is difficult.

Computational design creativity is hard to study, and until fairly recently it has received very little attention. Mostly the focus has been on extreme non-routine cases. But there are hard sub-problems and others ways of moving towards creative systems that are worth considering. This paper presents three of the alternatives, discussing one in more depth: i.e., to look at what changes can be made to routine design systems in order to produce more creative outputs. This focuses on working “upwards” towards creativity, examining smaller, ingredient decisions that make a difference to the result. As the amount of creativity displayed by a design is a judgment made by some person or group, it should be possible to investigate the degree of impact of changes to routine design mechanisms. This will contribute to our understanding of less “extreme” reasoning that leads to judgments of increased creativity: i.e., the foundation on which other methods rest.

Design Expert Systems can be built using many small, cooperating, limited function expert systems called Single Function Agents (SiFAs). Using this approach we will be able to investigate and discover primitive problem-solving and interaction patterns, specific for multiagent design systems, and should gain a deeper understanding of the types of knowledge involved. This paper presents some categories of conflicts that have been studied using the SiFA approach, and makes a brief presentation of the SINE implementation of SiFAs.

Scientific data is often analyzed in the context of domain-specific problems, for example, failure diagnostics, predictive analysis, and computational estimation. These problems can be solved using approaches such as mathematical models or heuristic methods. In this paper we compare a heuristic approach based on mining stored data with a mathematical approach based on applying state-of-the-art formulae to solve an estimation problem. The goal is to estimate results of scientific experiments given their input conditions. We present a comparative study based on sample space, time complexity, and data storage with respect to a real application in materials science. Performance evaluation with real materials science data is also presented, taking into account accuracy and efficiency. We find that both approaches have their pros and cons in computational estimation. Similar arguments can be applied to other scientific problems such as failure diagnostics and predictive analysis. In the e...

this paper we showed that information displays can be viewed as a design problem, and how, based on this, the presentation process can be modeled as a design process. Functional design is an important part of the design of presentations. Interesting problems are raised in the domain of representing and reasoning about the functions of designed objects, especially static functions. We have listed a set of interesting research topics raised by our approach to presentation. Further research will concentrate on these topics.

The work of Margaret Boden (1990; 1994) is familiar to everyone involved in the field of Computational Creativity. Her work, although at times philosophical, opened up new areas of research about creativity. However, some (Haase, 1995; Ram et al., 1995) have criticized the lack of detail in her models of creativity. Making a general model more detailed can remove some of the subjectivity; allow more options for a model to be tested; and, of interest to this workshop, move closer to models that concern designing. More recently, Wiggins (2006) has continued to refine his formal framework around Boden’s ideas about creativity as well as show that her model, even if lacking detail, was useful and revealing. Specifically he introduces a key distinction between R, a “rule set” that constrains the possible conceptual spaces, and T, a rule set that allows traversal of a space.

Requirements, both functional and non-functional, are the driving force behind the many decisions required to develop a software system. These decisions, alternative solutions, and the reasons behind them, can be captured in the rationale for the software system. The rationale contains the arguments for and against each alternative solution, which in turn relate to the functional and nonfunctional requirements. When this information is captured in the rationale for a system, it provides an alternative mechanism for tracing the impact of functional and non-functional requirements on the software system. In this paper we describe how the Software Engineering Using RATionale system (SEURAT) supports requirements traceability by incorporating functional and non-functional requirements into the argumentation.

Customer requirements are the foundation upon which a software system is built. These requirements, derived from the customers’ needs and desires, are used to both guide the development of the system and to determine if the completed system is what the customer requested. Because of its importance, requirement specification has become a research area known as Requirements Engineering (RE) both in Software Engineering [Zave, 1997; Nuseibeh & Easterbrook, 2000] and Systems Engineering [Dorfman, 1990; Chandrasekaran & Kaindl, 1996].

Design problems typically have a very large number of problem states, many of which cannot be anticipated at the onset of the design. Some design problem states are characterized by as many as hundreds of parameters. Given these amounts of uncertainty and information, AI design systems faced with learning tasks cannot know from the beginning what needs to be learned, and whether these needs will remain the same. In this abstract we describe how LEAD (Learning Expectations in Agent-based Design), a multiagent system for parametric and configuration design, addresses these challenges in design learning.

This paper describes research in progress whose goal is to synthesize design method-ologies for rapid product development in multidisciplinary design situations. The potential outcome is superior design methodologies that facilitate integration and collaboration between different disciplines, conduct design tasks concurrently, and apply to a wide range of design problems, thus reducing costs and time-to-market. The approach is based on simulating the design process using a multi-agent system that mimics the behavior of the design team. The multi-agent system activates the pieces of design knowledge when they become applicable. The use of knowledge by agents is recorded by tracing the steps that the agents have taken during a design project. Many traces are generated by solving a large number of design projects that differ in their requirements. A set of design methodologies is constructed by using clustering and abstraction techniques to generalize the traces generated. These method...

Software maintenance has long been one of the most difficult and expensive phases of the software life-cycle. Maintenance is especially difficult for large-scale systems. The more code involved, the larger the chance that there may be unexpected interactions that may cause problems when updates and corrections are made during maintenance. The large number of developers who were probably involved at various points in the system’s creation means that it is likely to be difficult to answer questions about the intent behind the design and implementation decisions. The designer’s, or developer’s, intent can be captured as their Design Rationale. Unlike standard design documentation, which is a description of the final design, Design Rationale (DR) offers more: not only the decisions, but also the reasons behind each decision, including its justification, other alternatives considered, and argumentation leading to the decision.

t is generally acknowledged that the only way to actually determine the quality of interactive systems is to perform a usability evaluation. This proves especially true for multimedia systems, where using concurrent media generates further usability problems. We aim to develop effective evaluation methods to meet industry demand and address concerns about the lack of cost-effective methods. These issues prevent most companies from performing usability evaluation, resulting in poorly designed and unusable software. Among the various usability evaluation approaches, usability inspection methods are gaining popularity because they cost less than traditional lab-based usability evaluations. These methods involve expert evaluators only, who inspect the application and, based on their knowledge, provide judgments about the usability of the different application elements. (Examples of inspection methods include heuristic evaluation, cognitive walkthrough, guideline review, and formal usabi...

In this paper we are concerned with generating parametric design systems based on cases from existing design systems. Issues that distinguish this particular application of CBR in design from those commonly seen are highlighted. In summary: we are trying to reuse past knowledge of the design process rather than design results; Adaptation could be non-trivial and crucial to accomplishing this goal; The reuse is carried on in a decomposition hierarchy with subcase reassembly involved; Various other kinds of knowledge such as design object and decomposition knowledge play an important part throughout each stage of the CBR process. The problem decomposition part has been tackled by Liu (1993) in which a system called KDD (Knowledge-based Design Decomposition) is shown to be able to produce a hierarchy of subproblems for a given design problem. The decomposition KDD achieves is based on object knowledge and heuristics, with a goal of minimizing design-time backtracking. The output of the...

It is common wisdom that people should be given tasks that computers can’t do well, and computers should be given tasks that people can’t do well. So in design computing why are we attempting to study computational design creativity? The main answer is that the field (like many others) progresses by tackling simpler problems first and moving towards harder ones. Routine parametric design and design checking were starting points, moving gradually to Configuration and most recently to harder problems such as distributed/collaborative design and to creative design: moving from routine to non-routine [Brown 1996]. One goal has always been to build working systems, while another is to learn more about the knowledge and reasoning used for each type of design activity studied. Computational design creativity is hard to study, and until fairly recently it has received very little attention, even though it is widely held to be very important both from intellectual and economic points of view...

Software maintenance has long been one of the most difficult and expensive phases of the software life-cycle. Maintenance is especially difficult for large-scale systems. The more code involved, the larger the chance that there may be unexpected interactions that may cause problems when updates and corrections are made during maintenance. The large number of developers who were probably involved at various points in the system’s creation means that it is likely to be difficult to answer questions about the intent behind the design and implementation decisions. The designer’s, or developer’s, intent can be captured as their Design Rationale. Unlike standard design documentation, which is a description of the final design, Design Rationale (DR) offers more: not only the decisions, but also the reasons behind each decision, including its justification, other alternatives considered, and argumentation leading to the decision. To drive and evaluate our research into using rationale for s...

This research originates in the work started several years ago at Worcester Polytechnic Insti-tute dedicated to the investigation, modelling and evaluation of multi-agent based design. Themain thrust behind our approach was the idea of finding the elementary patterns of agent prob-lem-solving and interaction in design tasks. To achieve this goal we introduced and definedthe concept of Single Function Agents (SiFAs) [Dunskus et al, 1995]. SiFAs are agents spe-cialized to perform one single function during the design process. Some typical functionswould be to select, evaluate, and provide critique. The SiFA concept is generic and agents canbe instantiated for different, particular design domains. This introduces the other two impor-tant characteristics of SiFAs: the unique target on which they operate and the unique point ofview from which they perform their task. In parametric design the targets are represented bydesign parameters. For example, a Selector (function) can be specialize...

The current methodologies for multi-disciplinary product design are based on compromising between different disciplines rather than integrating them. These methodologies do not use a systematic and holistic approach to the problem of multi-disciplinary design and thus are piecemeal rather than comprehensive. This paper presents a new approach to producing design methodologies for integration of the different disciplines in the design process. A multi-agent system has been developed that designs a 2-DOF robot arm by incorporating five proposed strategies for integration between disciplines. Design methodologies are extracted by tracking the system and generalizing the traces that are produced. The results show that the trace of the system provides invaluable information on how to improve the design process.

Abstract Recommender systems suggest resources to users based on collaborative filtering techniques, typically by exploiting correlations between individual user ratings of the resources they are interested in. Tags are a new form of metadata increasingly used in social bookmarking sites by users to annotate bookmarked resources. Our goal is to harness the implicit knowledge contained in these tags to improve the quality of recommendations to users. We use both tag and resource-interest knowledge in our user-based collaborative ...

We have now defined, at least tentatively, what we mean by design, and we have described several views of the process of design. In so doing, we have identified some of the ways that problem-solving strategies could be employed in the design process. Now we turn to the task of trying to outline an organizational structure or taxonomy for design. According to the dictionary, a taxonomy is the result of the “study of the general principles of scientific classification” and a classification , at least in the natural sciences, is the “orderly classification of plants and animals according to their presumed natural relationships.” Similarly here, a taxonomy of design might (1) allow us to classify design problems according to certain characteristics; and (2) facilitate the organization of the knowledge, representation, and reasoning schemes that would be useful in modeling different kinds of design. Are such taxonomies important? Why? One viewpoint is that a scientific theory of design cannot be developed without such a taxonomy. Another is that such taxonomies allow us to compare design methods and design tools – especially those reflecting the newer computer-aided technologies the ideas of which are reflected in this discussion. Our own viewpoint is stated somewhat differently in that we would stress any increase in our ability to understand and model the thought processes involved in design as being the best reason for developing such taxonomies. As we work toward this objective, we will simultaneously increase our abilities to compare various design tools and to develop and explore new design methods. We note also that, as with our characterizations of the design process, various taxonomies and classifications have been proposed, and there is certainly some overlap of ideas among them. After we review these various taxonomies, in an order roughly corresponding to their chronological development, we will try to analyze in a principled way what we have learned from so describing the thought processes of design.

A variety of approaches to the intelligent analysis of complex manufacturing data have been described in the literature. A comparative analysis of these systems reveals underlying similarities in their functional organizations. This leads to the development of a general functional model for intelligent data analysis (IDA).

The authors presents an approach to expert systems for mechanical design called design refinement, which addresses a subset of design activity by using a hierarchy of conceptual specialists that solve the design problems in a distributed manner, top-down, choosing from sets of ...

Our goal is the detailed study of the interactions, conflicts and conflict resolutions that are possible with a multi-agent design system. We are concerned with the" primitives" of knowledge and reasoning. We have stripped down each agent to an extremely simple form, in order to closely study the multi-agent system's behavior, the functionality required, the interactions that are possible (and necessary), the amount of and types of knowledge needed, and the conflicts that occur. In this way we can build a better understanding of ...

The main goal of this research is to discover the knowledge structures, control strategy, and problem-solving behavior required to determine how two objects best fit together. This sort of reasoning arises in many contexts and can involve any combination of objects, making it difficult to formalize. We therefore restrict the problem space to a particular class of objects. Each object is composed of a base, with features such as pegs, blocks, and holes of various shapes and sizes projecting inward and outward from its surfaces. A fit occurs when surfaces from any two objects can be brought close together by inserting the projections on either surface into the holes on the other. The degree of fit is determined by how tightly the projections fit into the holes and by how close the surfaces lie as a result.We recognize four stages in this reasoning process. The Grouping and Orientation stages select a particular juxtaposition between two objects that could lead to a fit. The Grouping stage identifies relevant features on an object's surface and forms a feature group. The Orientation stage selects two feature groups that appear to be compatible, based on the relative locations of features within them. If necessary, one object may be rotated to bring the feature group surfaces into opposition, or to line up the features.The Matching and Confirmation stages test the fit proposed by the two preceding stages. The Matching stage examines the surfaces region by region, matching features that are roughly complementary in shape and in corresponding positions. Only regions with features are considered. The Confirmation stage examines each feature pair in detail, analyzing the size, shape, and orientation of each member to determine how well they fit. If all the feature pairs are confirmed, the fit itself is confirmed.This decomposition into stages reduces the search space of potential fits. The Grouping stage narrows the focus from all features to the set of features relevant to a fit. The Orientation stage selects a single orientation. The Matching stage confines the search for mating features to localized regions on the object's surface. The Confirmation stage is thus able to examine features one pair at a time.Much of this reasoning is qualitative. Qualitative reasoning involves the analysis of how systems make the transition between discrete, qualitatively different states as their parameters reach certain critical values (Bobrow, 1985). Typically only ordinal relationships among values are considered. We treat knowledge about fit as a qualitative state. As new knowledge about fit is discovered, different and increasingly specific types of reasoning are applied. Eventually the knowledge of fit becomes sufficiently complete to indicate the specific measurements needed to test fit quantitatively.We have developed an implementation that currently covers the Matching and Confirmation stages. Input consists of a pair of objects that have already been Grouped and Oriented, as shown in the figure. In this example, the Matching procedures pair the features from corresponding corners since these features match in location and are of complementary shape. The Confirmation procedures examine each of these four pairs in turn. F1A is found to be smaller than F1B, and this is confirmed as a loose fit. In cross-section F2A fits into F2B, but since F2A is too long the fit is not confirmed. The radius of F3A exceeds that of F3B, and again fit is not confirmed. The radius and length of F4A are found to be smaller than those of F4B, resulting in a loose fit.We have identified several types of knowledge and control strategies that are important for reasoning about fit. We are currently working on representations for storing knowledge about the geometric and spatial relationships that accumulates as this reasoning proceeds. We are also interested in incorporating non-geometric properties, such as object functionality, into our model. Such extensions provide new avenues for reasoning about fit, and facilitate the exploration of such important areas as inferring function from structure (Stanfill, 1983) and the compilation of routine design knowledge (Brown, 1985).

%. % Note that these bibtex entries have no labels. You must add unique labels. % to incorporate these into your database. A design flaw in bibtex. % Wonder what the rationale is? %. @InProceedings{,. Author={Sue Abu-Hakima},. Title={Rationale: a tool that reasons explicitly for the purpose of explanation},. Booktitle={Fourth International Conference on Data Engineering},. Publisher={National Research Council},. Year={1988}. }. @techreport{,. Author={Guillermo Arango and Eric Schoen},. Title={Using Product Models to Compose Rationales},. ...

Fixtures are designed to accurately locate and secure a part during machining operations such that the part can be manufactured to design specifications. To reduce the design costs associated with fixturing, various computer-aided fixture design (CAFD) methods have been developed through the years to assist the fixture designer. One approach is to use a case-based reasoning (CBR) approach where relevant