Stir to Pour: Efficient Calibration of Liquid Properties for Pouring Actions

Abstract

Humans use simple probing actions to develop intuition about the physical behavior of common objects. Such intuition is particularly useful for adaptive estimation of favorable manipulation strategies of those objects in novel contexts. For example, observing the effect of tilt on a transparent bottle containing an unknown liquid provides clues on how the liquid might be poured. It is desirable to equip general-purpose robotic systems with this capability because it is inevitable that they will encounter novel objects and scenarios. In this paper, we teach a robot to use a simple, specified probing strategystirring with a stick-to reduce spillage when pouring unknown liquids. In the probing step, we continuously observe the effects of a real robot stirring a liquid, while simultaneously tuning the parameters to a model (simulator) until the two outputs are in agreement. We obtain optimal simulation parameters, characterizing the unknown liquid, via a Bayesian Optimizer that minimizes the discrepancy between real and simulated outcomes. Then, we optimize the pouring policy conditioning on the optimal simulation parameters determined via stirring. We show that using stirring as a probing strategy result in reduced spillage for three qualitatively different liquids when executed on a UR10 Robot, compared to probing via pouring. Finally, we provide quantitative insights into the reason for stirring being a suitable calibration task for pouring -a step towards automatic discovery of probing strategies.

How to Select and Use Tools? : Active Perception of Target Objects Using Multimodal Deep Learning

“…The robot discerned soil characteristics like moisture and weight by stirring the soil with a stick, then used image and force sensor data to calculate appropriate arm trajectories for scooping according to those characteristics. Saito et al and Lopez-Guevara et al realized a robot capable of pouring liquid from a container [22], [23], and allowed the robot to determine the features of the liquid like amount and viscosity from sensor data while shaking or stirring the container. Schenck et al realized a robot to dump a granular object to deform into a desired shape [24].…”

Cooking Robot Mastery: Learning Online Motion Generation with Active Perception

“…Some previous research has focused on the perception of characteristics using multimodality [21,22], however they only allow robots to handle stable targets that do not considerably change their state dur-ing motions. Some studies have focused on deformable targets such as liquids, however they had to control the robots to conduct pre-defined exploring motions in the beginning to perceive the object characteristics before starting the target motions [23,24,25]. We hypothesize that the difficulty occurs because the robots cannot perceive characteristics and generate motion quickly and efficiently enough for handling real-time changes.…”

Generating Parametric BRDFs from Natural Language Descriptions

“…Several methods in machine learning aim to mimic human intuition about complex physical systems (Spelke et al, 1992;Bramley et al, 2018) and develop approximate inference methods (Lake et al, 2016;Tenenbaum, 2018;Ullman et al, 2018) for downstream applications such as system identification in robotics (Lopez-Guevara et al, 2017). There is some evidence that LLMs require 'coaxing' to elicit reasoning natural to humans .…”