[CS] Romain Nith Candidacy Exam/Jun 17, 2025

[via cs]

This is an announcement of Romain Nith's Candidacy Exam.
===============================================
Candidate: Romain Nith

Date: Tuesday, June 17, 2025

Time:  1 pm CST

Remote Location: https://urldefense.com/v3/__https://uchicago.zoom.us/j/99261951766?pwd=aWkrbXE4QVlmbjRXeDNwbmZVL3pOZz09__;!!BpyFHLRN4TMTrA!8hpDAP0-FVyk5LSevRov1oEboYn1t1t2BJd8PQyfUP3kaXsNkxGtzAq4Z5BFlYQTQu1aikrpM6oQKWJdOUQm$

Location: JCL 298

Title: Embodied Assistance: general and adaptable physical assistance via muscle stimulation

Abstract: Over the last decades, electrical Muscle Stimulation (EMS) has emerged as a promising method for delivering physical assistance directly through the body. However, I argue that the state-of-the-art EMS-based assistive systems remain limited in scope and impact. Most implementations rely on predefined stimulation patterns, which are not adaptable and fail to generalize beyond their designed, task-specific contexts. While these systems can move the body, they typically offer assistance only for foreground tasks—those the user is actively attending to—overlooking the potential of EMS to assist in the background. Moreover, when EMS is used to impart physical knowledge, it often relies on static demonstrations, limiting its effectiveness for learning dynamic or personalized motor sequences. To tackle these shortcomings, we propose a new approach to Embodied Assistance via EMS that addresses these limitations through adaptive, intelligent control strategies. First, by integrating multimodal AI reasoning, we propose a generalizable physical assistance system for muscle stimulation across diverse physical tasks and user contexts. We then further demonstrate how such systems can augment human cognitive capacity in multitasks. Finally, we show that by adapting the EMS strategies according to the user’s skill level can significantly outperform static demonstrations in helping learn motor sequences—leading to faster acquisition and better retention.

Advisors: Pedro Lopes

Committee Members:Pedro Lopes, Ken Nakagaki, Craig Schulz, Jun Rekimoto