Accurate Robotic Pouring for Serving Drinks
This work addresses a specific task in robotic cooking, offering incremental improvements in accuracy for practical applications like serving drinks.
The paper tackled the problem of robotic pouring for serving drinks by developing a system that uses recurrent neural networks to generate angular velocities for the source container, achieving an error as low as 4 milliliters when tested on unseen containers.
Pouring is the second most frequently executed motion in cooking scenarios. In this work, we present our system of accurate pouring that generates the angular velocities of the source container using recurrent neural networks. We collected demonstrations of human pouring water. We made a physical system on which the velocities of the source container were generated at each time step and executed by a motor. We tested our system on pouring water from containers that are not used for training and achieved an error of as low as 4 milliliters. We also used the system to pour oil and syrup. The accuracy achieved with oil is slightly lower than but comparable with that of water.