Learning to Collaborate from Simulation for Robot-Assisted Dressing
This work addresses the problem of assisting individuals with impairments in dressing tasks, though it is incremental as it builds on existing simulation and DRL methods.
The study tackled robot-assisted dressing by using deep reinforcement learning and haptic feedback control to train collaborative policies in simulation, achieving successful dressing tasks on a hospital gown and T-shirt and enabling a real PR2 robot to dress a humanoid robot's arm.
We investigated the application of haptic feedback control and deep reinforcement learning (DRL) to robot-assisted dressing. Our method uses DRL to simultaneously train human and robot control policies as separate neural networks using physics simulations. In addition, we modeled variations in human impairments relevant to dressing, including unilateral muscle weakness, involuntary arm motion, and limited range of motion. Our approach resulted in control policies that successfully collaborate in a variety of simulated dressing tasks involving a hospital gown and a T-shirt. In addition, our approach resulted in policies trained in simulation that enabled a real PR2 robot to dress the arm of a humanoid robot with a hospital gown. We found that training policies for specific impairments dramatically improved performance; that controller execution speed could be scaled after training to reduce the robot's speed without steep reductions in performance; that curriculum learning could be used to lower applied forces; and that multi-modal sensing, including a simulated capacitive sensor, improved performance.