The research objective of this award is to assess how creative a particular design concept is, and, more specifically, to determine what attributes lead to a more (or less) creative design concept. To achieve this objective, the research uses credit assignment approaches derived for multiagent coordination to evaluate the impact of various attributes on the creativity of the full design concept. In particular, the research will provide clear guidance on what the criteria should be to evaluate the creativity of design concept variants generated in an automated design environment.
If successful, the results of this research will help identify the most creative design concepts early in the design process. The intellectual merit of this award is in its blending state of the art automated concept generation techniques with new agent coordination strategies to allow the emergence of creativity as a combination of attributes (e.g., agents). Quantifying the impact of specific attributes (local impact) on the creativity of a concept (global impact) is a key contribution. The broader impact includes far-reaching effects on engineering product design. A method that generally captures the creative potential of a design, a-priori to market introduction, could revolutionize how industry designs and develops new products, and provide an advantage to United States industry. This research will engage both the multiagent and engineering design communities. Workshops will be held at the flagship conferences in each field. In addition, the PIs will leverage an NSF funded effort that created an Engineering Virtual Organization to both enter and receive data to assess concepts. Furthermore, the impact of methods that enhance and bring forward the creativity of students will be tested through an automated design competition at the end of the project to engage the students. Finally, students in the introductory and senior design courses will be engaged to experiment with the new method.
The goal of this project was to define a automated approach to product design that would guarantee a set of innovative concepts. The outcome is an approach to automated design that identifies, quantifies, and enhances the creativity of automatically generated concepts. Specifically, we identify, functionally model, and cluster innovative concepts, create a measure of creativity, and leverage multiagent coordination techniques to determine the impact of functional elements on the creativity of concepts. Our extended research plan included developing multiple creativity measures (global utilities), and producing an adaptive agent framework where agents (e.g., functional elements) are evaluated based on their contribution to a product’s creativity (local utilities). This framework leads to an autonomous agent system that chooses design solutions that optimize the creativity measures on both a local and global level. This connection between local and global utilities is one of the fundamental research problems in agent coordination and has impact on applications as diverse as robot coordination, air traffic management, nano device control and automated concept generation. By leveraging this new and rich research area, we can not only predict the level of creativity of a concept based on the elements from which it is composed, but also automatically generate new concepts that score highly on multiple creativity measures. Three specific deliverables from this project are: A new repository for innovative products (RIP); In-class validation of new creativity metrics; and A methodology to propagate creativity metrics back to functional elements of the design based on multiagent coordination research.
