Shared Autonomy for an Intuitive Human-Robot Interface Assisting the Elderly in Activities of Daily Living

Abstract

This thesis investigates shared autonomy as a key approach for intuitive human robot interaction to support elderly individuals in performing activities of daily living (ADL). The aim of the study was to find out which of the proposed control modes assists best for grasping and manipulation tasks. After reviewing state of the art shared autonomy methods across various robotic interfaces, two shared autonomy approaches were developed, an active constraint (AC) and a weighted sum (WS) based framework. Several subsystems were integrated into a unified system combining motion and hand pose intent detection. Therefore, inertial measurement units and surface electromyography were utilized respectively, as well as DLR grippers, DLR Fingertac for vibration feedback and the DLR light weight robot. A user study with 12 participants (age of 27:3 ± 6:8 years) was conducted, evaluating three different modes, fully teleoperation (Teleop), AC and WS. Objective metrics were measured, such as task completion time, positioning accuracy, and traveled distances. Moreover, participants completed subjective measures using NASA Task Load Index and the Virtual Embodiment Questionnaire. Objective and subjective results show that AC is the best suited form of shared autonomy assistance. AC significantly reduced mental and physical demand, especially in the demanding task compared to Teleop and AC. These findings highlight the advantages of AC, especially for the target group of elderly users. Future work could include hardware and software improvements, further refinement of indexing and control robustness, and a focused user study with elderly participants. Overall, this thesis demonstrates the potential of shared autonomy to enhance robotic assistance in ADL for aging populations.