Tobias Schubert

Liquid Pouring

In this project we challenge the problem of pouring an unknown liquid. Robotic assistants in a home environment are expected to perform various complex tasks involving liquids. This could range from pouring drinks into cups or accurately measuring out a liquid for baking or cooking. The latter task is more difficult in that it requires pouring an exact amount of liquid. In order to complete these tasks, it is necessary to detect and track the liquid level during a pour to determine when to stop. Widely available sensors such as RGB-D cameras, provide noisy liquid level measurements which depend on the transparency of the liquid and its index of refraction. We developed a novel approach to track the liquid level using an RGB-D camera. It is able to determine when to stop pouring, making it possible to fill a cup to a specific, user- defined height. Additionally, we look at classifying the liquid as either transparent or opaque, in the event that no prior knowledge is available.