Manipulating liquid is widely required for many tasks, especially in cooking. A common way to address this is extruding viscous liquid from a squeeze bottle. In this work, our goal is to create a sauce plating robot, which requires precise control of the thickness of squeezed liquids on a surface. Different liquids demand different manipulation policies. We command the robot to tilt the container and monitor the liquid response using a force sensor to identify liquid properties. Based on the liquid properties, we predict the liquid behavior with fixed squeezing motions in a data-driven way and calculate the required drawing speed for the desired stroke size. This open-loop system works effectively even without sensor feedback. Our experiments demonstrate accurate stroke size control across different liquids and fill levels. We show that understanding liquid properties can facilitate effective liquid manipulation. More importantly, our dish garnishing robot has a wide range of applications and holds significant commercialization potential.
We propose a two-phase approach to control the thickness and volume-per-length of the dispensed liquid stream from a squeezing bottle. In the exploration phase, we perturb the bottle to gather liquid property information. This information allows us to predict how the flow rate changes over time with our consistent squeezing motion and how the liquid forms strokes. In the dispensing phase, we utilize these predictions to guide the drawing speed, achieving control over stream thickness and volume-per-length.
We collected data with 20 liquids to train the MLPs and tested our approach on 5 other liquids.
We collected data with 20 liquids to train the MLPs and tested our approach on 5 other liquids. The liquids cover a wide range of viscosity, ranging from 1cP (like water) to 70,000 cP (like toothpaste).
The table below shows the Flow Rate Curve (FC) and Stacking Pattern Curve (SC) mean prediction error of testing liquids in various fill levels across different approaches.
We use a 6DOF robot arm (UR5e by Universal Robotics) equipped with a 2-fingered gripper (WSG50 by Weiss Robotics) for our experiments. A 6-axis F/T sensor (NRS-6050-D80 by Nordbo Robotics) with 1000 Hz sampling rate is attached to the robot wrist. To ensure the stability of the bottle while squeezing, a specially designed bottle holder is attached to the robot. All the liquids used in the experiments are contained within a 16 oz squeezing bottle.
We task the robot with squeezing streams of 10cm in length at various fill levels of the testing liquids to achieve thickness of 5, 10, 15, and 20 mm. Our approach demonstrates comparable performance in both thickness accuracy and consistency control when compared to PP, outperforming other baseline methods.
@inproceedings{huang2024kitchen,
title={Kitchen Artist: Precise Control of Liquid Dispensing for Gourmet Plating},
author={Huang, Hung-Jui and Xiang, Jingyi and Yuan, Wenzhen},
booktitle={2024 IEEE International Conference on Robotics and Automation (ICRA)},
pages={13933--13939},
year={2024},
organization={IEEE}
}