CSE - Machine-Learning Assisted Multi-Modal Robotic Skin
Machine learning has enabled electronic devices like gloves and skins to track hand movements and perform tasks such as gesture and object recognition. Despite their utility, these devices are typically bulky and struggle to adapt to the natural contours of the body. Moreover, existing signal-processing approaches depend heavily on large datasets of labeled information to identify individual tasks for every user. In contrast, printed sensors, such as resistive and capacitive types, are lightweight, highly conformal, and easily customizable into various shapes, making them ideal for diverse applications. This project seeks to integrate printed sensors with an unsupervised meta-learning framework to enable user-independent and data-efficient recognition of various hand movements. Leveraging advanced machine learning, a single printed sensor can track movements across multiple finger joints simultaneously, ensuring simplicity, flexibility, and low computational demands.
What makes the Internet of Matter Lab uniquely vibrant is its fusion of Human-computer Interaction, digital fabrication, circuit design, and machine learning to prototype systems that operate in harmony with the natural world. From biodegradable electronics to AI-enabled sensing systems, we work across disciplines to build computing platforms that disappear gracefully when no longer needed. Collaboration is at the heart of our work. We partner closely with researchers in HCI, robotics, sustainability science, and electronic engineering, forming a high-energy, interdisciplinary environment where ideas flow freely and innovation thrives.