Imitation Learning for Variable Speed Object Manipulation

To operate in a real-world environment, robots have several requirements including environmental adaptability. Moreover, the desired success rate for the completion of tasks must be achieved. In this regard, end-to-end learning for autonomous operation is currently being investigated. However, the issue of operating speed has not been investigated in detail. Therefore, in this paper, we propose a method for generating variable operating speeds while adapting to perturbations in the environment. When the work speed changes, there is a nonlinear relationship between the operating speed and force (e.g., inertial and frictional forces). However, the proposed method can be adapted to nonlinearities by utilizing minimal motion data. We experimentally evaluated the proposed method for erasing a line using an eraser fixed to the tip of a robot. Furthermore, the proposed method enables a robot to perform a task faster than a human operator.