Development of adaptive gaze-based human-machine interface for enhancing assisted daily living

The efficacy of gaze-based assistive technologies for individuals with severe motor impairments is often limited by a dual challenge: the cognitive inefficiency of rigid interfaces and the practical inaccessibility of costly hardware. This thesis presents the design, implementation, and validation of an integrated assistive system engineered to solve this dual problem. The core of the system is an adaptive, intention-based interface that utilizes a machine learning model to interpret pupillometry in real-time. By classifying the user's data into 'decision-making' and 'focus' events, this method replaces static dwell-time with a control mechanism that responds dynamically to user’s intend, creating a more fluid and intuitive user experience. This intelligent software is paired with an accessible hardware solution: a low-cost, 3D-printable robotic controller. This mechanism retrofits standard powered wheelchairs through the mechanical actuation of the existing joystick, making advanced gaze-based navigation widely deployable without requiring extensive modifications or specialized installation. The system's functionality is further extended to semi-autonomous robotic manipulation by integrating an object detection layer, which offloads the profound cognitive load of the task by allowing users to execute complex pick-and-place actions with a single gaze command. The performance of this integrated system was evaluated through a series of user studies focusing on realistic navigation and manipulation tasks. The evaluation employed a mixed-methods approach, analyzing objective performance metrics, such as task completion time and command frequency, alongside subjective feedback on usability and workload. The results provide strong evidence for the success of this approach, showing that the adaptive interface was significantly more efficient and was subjectively preferred by participants over traditional dwell-based methods. The overall outcomes demonstrate that by concurrently addressing the fundamental challenges of usability and accessibility, this work provides a powerful and practical framework for enhancing assisted daily living, offering a validated pathway towards more adaptive and affordable assistive technologies.