A Gaze-Driven Robotic System for Post-Stroke Active Ankle Rehabilitation Training.

Area of Science:

• Rehabilitation Engineering
• Human-Computer Interaction
• Robotics

Background:

• Effective lower-limb rehabilitation requires intuitive control channels for robotic devices.
• Existing methods may lack precision, predictability, or safety for clinical application.

Purpose of the Study:

• To develop and evaluate a clinic-friendly, binocular gaze-driven control system for a two-degree-of-freedom ankle robot.
• To assess the system's accuracy, latency, safety, and usability in healthy adults and a pilot post-stroke cohort.

Main Methods:

• A binocular gaze-tracking system mapped pupil centers to screen quadrants for discrete commands.
• Low-order regression with ArUco-guided homography and affine correction enabled precise gaze mapping.
• A dwell/occupancy rule triggered jerk-limited robot trajectories under cascaded control with safety limits.

Main Results:

• Achieved 99.94% selection accuracy with a 157.6 ms median end-to-end delay in healthy adults.
• Demonstrated robustness to head tremor, with no incorrect decisions at high severity.
• Showcased precise passive tracking (RMSE ≤0.224°) and safe torque profiles.
• Reported high usability, low workload, and minimal visual fatigue in human factors outcomes.

Conclusions:

• Gaze control is a practical, hands-free primary channel for seated ankle rehabilitation.
• The developed system is suitable for integration into clinical workflows.
• This technology holds promise for improving lower-limb robotic therapy outcomes.