Clothing assistance is a basic and important assistance activity in the daily life of the elderly and disabled people. However, robotic clothing assistance is still considered an open problem by most robotics researchers. Design of a robust framework involves reliable cloth state estimation in real-time and a motor skills learning framework that can detect and adapt to various failure scenarios. Existing studies do not explicitly handle these challenges and perform point-to-point motion planning in an offline manner for clothing assistance [1],[2]. An alternate approach is to formulate robotic clothing assistance as a reinforcement learning problem wherein the robot learns to recover from failure scenarios and adapt to new settings from experience.
展开▼