Understanding Dynamic Tactile Sensing for Liquid Property Estimation

Abstract: Humans perceive the world by interacting with objects, which often happens in a dynamic way. For example, a human would shake a bottle to guess its content. However, it remains a challenge for robots to understand many dynamic signals during contact well. This paper investigates dynamic tactile sensing by tackling the task of estimating liquid properties. We propose a new way of thinking about dynamic tactile sensing: by building a light-weighted data-driven model based on the simplified physical principle. The liquid in a bottle will oscillate after a perturbation. We propose a simple physics-inspired model to explain this oscillation and use a high-resolution tactile sensor GelSight to sense it. Specifically, the viscosity and the height of the liquid determine the decay rate and frequency of the oscillation. We then train a Gaussian Process Regression model on a small amount of the real data to estimate the liquid properties. Experiments show that our model can classify three different liquids with 100% accuracy. The model can estimate volume with high precision and even estimate the concentration of sugar-water solution. It is data-efficient and can easily generalize to other liquids and bottles. Our work posed a physically-inspired understanding of the correlation between dynamic tactile signals and the dynamic performance of the liquid. Our approach creates a good balance between simplicity, accuracy, and generality. It will help robots to better perceive liquids in different environments such as kitchens, food factories, and pharmaceutical factories.

Paper link: https://arxiv.org/abs/2205.08771

  title={Understanding Dynamic Tactile Sensing for Liquid Property Estimation},
  author={Huang, Hung-Jui and Guo, Xiaofeng and Yuan, Wenzhen},
  journal={arXiv preprint arXiv:2205.08771},

Liquid classification

We try to classify three common liquids, water, oil, and detergent. Their viscosity are different, leading to different oscillation decay rate Lambda, which can be perfectly extracted by our method.
We achieve 100% classification accuracy for water, oil, and detergent, even without fixing their weight and amount. In addition, the model can be trained with only 24 data points, showing our method is data efficient.

Sugar water concentration regression

We target the liquid with smaller differences, the sugar water with different concentration and height. We train a Gaussian Process Regression model using lambda and omega as input and regress on the sugar concentration. We achieve a concentration regression Mean Average Error of 15.3%, and log viscosity prediction MAE of 0.16.

Liquid height regression

We use the same dataset and method to regress on the liquid height. We achieve a height regression MAE of 0.56mm, which is near the limit of the rulers’ precision.

Liquid generalization

We trained on sugar water data and test the model on other liquids.

Container generalization

Our method can be generalized to different containers. Such as a large cuboid container and a cylinder bottle.