Reducing caregiver burden with robotics: Limb manipulation assistance for caregivers in bed bathing

Human caregivers often perform physically demanding tasks during assistive activities, including manipulating or repositioning the arms or legs of care recipients. These repetitive and strenuous actions place caregivers at risk of significant muscle fatigue in the back, shoulders, and upper limbs, increasing the possibility of musculoskeletal injuries. To address the challenge, this paper presents a robotic framework designed to assist with limb manipulation, thus mitigating the physical strain on caregivers and improving the overall caregiving experience. For effective collaboration with caregivers, our framework integrates a multi-modal perception module that provides continuous and occlusion-resistant sensing of both the environment and caregiver actions. Caregiver comfort is modeled from the physical load of eight caregiving demonstrators, quantified through postural analysis with the Rapid Entire Body Assessment (REBA) and muscle activity measured via electromyography (EMG). This model is then used with Bayesian Optimization to determine optimal limb positions. A limb manipulation controller is designed to manipulate the limb to the optimal configuration. Additionally, a navigation strategy enables the robot to safely navigate through the environment and move to the optimal assistance location. In this paper, we evaluate the potential of the proposed framework in an assistive bathing scenario for bed-bound individuals. We conduct a within-subject study with 20 participants of varying heights and weights. To evaluate the efficacy of our framework, we collect REBA scores and EMG values and a questionnaire rating acceptance, capability, ease of hygiene task, fluency, physical comfort, and perceived safety. Participants complete the bathing task under four conditions: (a) Caregiver - no robotic assistance. (b) Macro - the robot selects the optimal posture that maximizes the participant's comfort across actions. (c) Micro - the robot dynamically selects action-wise postures that maximize comfort while minimizing the travel distance of the joints in the target limb. (d) Random - the robot adopts a feasible posture at random limb configuration. The experimental results indicate that robotic assistance with ergonomic optimization, particularly dynamic optimization, improves caregiver comfort and increases their preference for robotic support. Overall, this work shows the effectiveness and potential to enhance caregiver experience during bathing assistance. A supplementary video is available at https://youtu.be/R8EeZ8AjyVk .