A Multi-Layer Sim-to-Real Framework for Gaze-Driven Assistive Neck Exoskeletons
A new AI framework uses eye gaze data from VR to control a physical neck exoskeleton, restoring movement for patients.
A research team led by Daniel S. Brown from UC Berkeley and UT Austin has published a novel framework for developing an AI-powered, gaze-controlled neck exoskeleton. The system is designed to assist individuals with dropped head syndrome, a debilitating condition caused by neurological diseases like ALS or Parkinson's that weakens neck muscles. The core innovation is a multi-layer 'sim-to-real' pipeline that uses virtual reality (VR) to safely and efficiently collect the crucial training data—paired eye gaze and head movement—from healthy participants. This data trains machine learning models to predict a user's intended head movement based solely on where they are looking.
The framework then rigorously evaluates potential control algorithms across three decreasing levels of abstraction: in simulation, in VR with a human in the loop, and finally on the physical exoskeleton hardware. This staged approach acts as a filter, rejecting poor-performing controllers early and cost-effectively. The research successfully identified two novel gaze-driven models that performed strongly when deployed on the real device. A key finding was that no single controller was universally best, underscoring the necessity for personalization in assistive technology. This work, accepted for presentation at the prestigious IEEE ICRA 2026 conference, provides a validated blueprint for accelerating the development of safe and intuitive wearable robots.
- Uses VR to collect training data for AI models that predict head movement from eye gaze alone.
- Implements a novel 3-stage sim-to-real pipeline (simulation, VR, physical) to filter and identify the best controllers.
- Proves no single AI controller works for everyone, highlighting the critical need for personalized assistive tech.
Why It Matters
This accelerates the development of intuitive assistive devices that could restore independence for thousands with neuromuscular diseases.