Bodies Uncovered: Learning to Manipulate Real Blankets Around People via Physics Simulations

While robots present an opportunity to provide physical assistance older adults and people with mobility impairments in bed, frequently rest bed blankets that cover the majority of their body. To for many daily self-care tasks, such as bathing, dressing, or ambulating, a caregiver must first uncover from part person’s In this work, we introduce formulation robotic bedding manipulation around which robot uncovers blanket target body while ensuring human remains covered. We compare two approaches optimizing policies grasp release points body: 1) reinforcement learning 2) self-supervised optimization generate training data. trained conducted evaluations these physics simulation environments consist deformable cloth mesh covering simulated lying supine on bed. addition, transfer simulation-trained real mobile manipulator demonstrate it can parts manikin Source code is available online $^{3}$ .