Georgia Institute of TechnologyDesign & Intelligence Laboratory
Printable Version of this PageHome PageRecent ChangesSearchSign In

Research


DIL has pioneered research on:
(1) Computational theories of creative design, design by analogy, and case-based design,
(2) Teleological models of physical systems (called Structure-Behavior-Function models) and their use in case-based design and design by analogy,
(3) Teleological models of software agents (called Task-Method-Knowledge models) and their use in meta-reasoning for failure-driven and task-directed agent self-adaptation,
(4) Interactive case-based techniques and tools for aiding engineering, architectural and interface design, and
(5) Visual analogies in problem solving, diagram understanding and design.

Current projects take these theories, techniques and tools to real problems.

Current Projects

Meta-Reasoning in Classification Learning: The AN Project. The goal is to develop autonomous classification agents that automatically repair their classification knowledge when their predictions turn out to be incorrect over time, e.g., incorrect stock market predictions. Research issues include the representation and organization of the classification knowledge to support self-diagnosis and self-repair, and probing the world to collect additional information. Project members include Joshua Jones and Ajay Chaudhari. Josh is writing his Ph.D. dissertation based on this work.

Learning Mental Models of Complex Systems: The ACT Project. The goal is to develop an interactive learning environment for modeling complex natural systems in middle school science. The research issues are what are the mental models of middle school children, and how might we enable learning of deeper models of complex systems such as aquaria. Project members include Swaroop Vattam and Vivek Menon. This is a joint project with Dr. Spencer Rugaber, Prof. Cindy Hmelo-Silver (Rutgers University) and Prof. Rebecca Jordan (Rutgers University).

Story Construction in Intelligence Analysis: The STAB Project. The goal is to develop an interactive assistant that can construct a coherent story from large amounts of intelligence data. Research issues include representation and organization of story plots, retrieval of story plots, and composition of new plots. We are also intergating the STAB system with an interactive information visualization tool called Jigsaw developed by the Information Interfaces group. Project members include Avik Sinharoy and Adity Dokania. This is a joint project with Prof. John Stasko and Prof. Anita Raja (UNCC) under the Southeastern Regional Visual Analytics Center.

Self-Adaptive Game-Playing Agents: The GAIA Project. The goal is to develop autonomous game-playing agents that they can adapt themselves as the game designers change the rules of the game. The research issues are how might an autonomous software agent reflect on its decisions and actions (e.g., attacking an city in the interactive strategy game called FreeCiv) and repair its decision-making knowledge, and how might the agent assist game designers in making adaptations to designs of game-playing software agents. Project members include Joshua Jones, Chris Parnin, Sameer Indrapu and Lee Martie. This is a joint project with Dr. Spencer Rugaber.

Biologically Inspired Engineering Design: The DANTE Project. The goal is to develop an interactive environment for fostering biologically inspired design by enabling analogical transfer of biological design patterns. The research issues are how do teams of engineers and biologists generate new design ideas, how may be capture knowledge of biological design patterns on a computer, and how might we use this knowledge to facilitate innovative design. Project members include Swaroop Vattam and Michael Helms. This is a joint project with Prof. Jeannette Yen under the Center for Biologically Inspired Design.

Visual Cognition in Autism. The goal is to develop interactive tools for visual communication with autistic children. The research questions are what kinds of multimodal information processing may explain autistic behavior and what is the nature of visual cognition in autism? Project members include Maithilee Kunda.

Recent Projects

Diagram Understanding and Knowledge Acquisition by Analogy: The Archytas Project. The goal was to develop an autonomous computer system that can understand design drawings, such as the drawings in a patent database, and acquire teleological models of designs by transferring and adapting the teleological model of a known drawing. Research issues included representation of drawings and their teleological models in an abstraction hierarchy, analogical mapping between the input drawing and the known drawings, and analogical transfer of the teleological model from a known drawing to the input drawing. Patrick Yaner wrote his 2007 Ph.D. dissertation based on this work. The current project on biologically inspired design uses some of the design representations developed in the Archytas project.

Visual Analogy in Problem Solving: The Galatea Project. The goal was to develop a cognitively plausible computational technique for visual analogies, and in particular, for analogical transfer of problem-solving procedures depicted in an animation. Research issues included visual representation of problem-solving procedures, dynamic analogical mapping, and cognitive modeling of subject data. Jim Davies wrote his 2004 Ph.D. dissertation based on this work. The current project on visual cognition in autism uses some of the visual representations used in the Galatea project.

Proactive and Retrospective Self-Adaptation in Intelligent Agents: The REM Project. Bill Murdock wrote his 2001 Ph.D. dissertation based on this project. The Self-Adaptive Game-Playing Project evolves from the REM project.

Some of DIL's past research is described here on the AAAI AI Topics website, and here in the Encyclopedia of Computer Science and Technology.



Last modified 26 May 2009 at 4:11 pm by mhelms3