AI-Infused Soft Fluidic Tactile Sensing

摘要

Soft tactile sensors can transform how robots interact with humans and their intricate environments, making such encounters more efficient and safely intimate. However, existing technologies are complex and costly, which has constrained their widespread application. Therefore, there is an urgent need to embrace more straightforward solutions utilizing traditional off-the-shelf sensors. In this context, we integrate fluid-solid interaction with artificial intelligence to develop a soft tactile interface crafted from elastomer-encapsulated fluid. Instead of relying on electronic circuitry, we utilize fluid pressure to transmit tactile information to available pressure sensors. We have devised efficient machine learning algorithms to infer touch position and intensity accurately. In its simplest form, we fabricated a 1D sensor with a linear channel, demonstrating its ability to accurately estimate touch position and force along a straight line using pressure readings solely from its two ends. Additionally, this system served as a test platform to study how geometrical parameters and different fluid mediums (such as air, water, and oil) impact the sensing capabilities of the developed sensor. We envision expanding this simple approach to create a cost-effective, electronics-free sensing front, providing skins for robots designed to be aware of their environment.