Design and computational evaluation of an eye gaze wheelchair retrofit with temporal command supervision
Year: 2026
Type: preprint
Abstract: Eye gaze offers an alternative control channel for powered mobility, but reliable operation requires both deliberate command selection and timely withdrawal of motion when input becomes invalid. This study presents iMove, a wheelchair retrofit comprising a commercial eye tracker, a tablet interface, an Arduino Uno, and two servomotors that mechanically actuate the wheelchair joystick. We combined preliminary functional assessment with executable controller replay, fault injection, component ablation, Monte Carlo sensitivity analysis, and a kinematic stopping model. Preliminary testing yielded eight successful command selections in ten attempts, corresponding to 80.0% success with a 95% Wilson interval of 49.0–94.3%. The baseline controller activated on target entry and retained its motion command after target exit or interrupted communication. We therefore developed and computationally evaluated a temporal supervisor with a 0.60s motion dwell, a 0.15s input-freshness threshold, and a 0.30s receiver timeout. In deterministic fault scenarios, the supervisor neutralized the command 0.12s after input freezing and 0.24s after host or serial-link interruption; the baseline remained active throughout the 3.5s post-fault window. Ablation demonstrated that receiver timeout alone could not detect stale input accompanied by continuing messages. At an assumed speed of 0.50 m/s, a linear-deceleration model using the preliminary 3–4s stopping interval produced a conditional stopping envelope of 0.78–1.13m. The findings establish complementary roles for selection dwell, input freshness, receiver expiry, and braking characterization in gaze-controlled mobility. Physical implementation of the supervisor and evaluation with intended users remain necessary to establish operational performance.
Source: Research Square
Institution: Kuwait College of Science and Technology
Language: English
Subfield: Human-Computer Interaction
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7224745709