Large Language Model Agents Enabled Generative Design of Fluidic Computation Interfaces

This is the authors' abstract. We don't add key points for this paper.

The creation of interactive devices is a major area of interest. However, traditional design tools in this field often require a significant learning curve and may not effectively support creative ideation. This study explores the use of fluidic computation interfaces as a case study to examine the potential of enhancing design tools for physical devices with Large Language Model (LLM) agents. With LLM agents, the Generative Design Tool (GDT) can understand the capabilities and limitations of new devices, suggest diverse, insightful, and practical application scenarios, and recommend designs that are technically and contextually appropriate. Additionally, it generates the necessary design parameters for the traditional components of the design tool to visualize results and create files for fabrication.