GRACE: Generalizing Robot-Assisted Caregiving with User Functionality Embeddings

  • 2025-01-29 18:55:07
  • Ziang Liu, Yuanchen Ju, Yu Da, Tom Silver, Pranav N. Thakkar, Jenna Li, Justin Guo, Katherine Dimitropoulou, Tapomayukh Bhattacharjee
  • 0

Abstract

Robot caregiving should be personalized to meet the diverse needs of carerecipients -- assisting with tasks as needed, while taking user agency inaction into account. In physical tasks such as handover, bathing, dressing, andrehabilitation, a key aspect of this diversity is the functional range ofmotion (fROM), which can vary significantly between individuals. In this work,we learn to predict personalized fROM as a way to generalize robotdecision-making in a wide range of caregiving tasks. We propose a noveldata-driven method for predicting personalized fROM using functional assessmentscores from occupational therapy. We develop a neural model that learns toembed functional assessment scores into a latent representation of the user'sphysical function. The model is trained using motion capture data collectedfrom users with emulated mobility limitations. After training, the modelpredicts personalized fROM for new users without motion capture. Throughsimulated experiments and a real-robot user study, we show that thepersonalized fROM predictions from our model enable the robot to providepersonalized and effective assistance while improving the user's agency inaction. See our website for more visualizations:https://emprise.cs.cornell.edu/grace/.

 

Quick Read (beta)

loading the full paper ...