Can a robot be a better waiter than a human? To answer this question, Dr. Luis Figueredo, Senior Scientist in Prof. Sami Haddadin's team, mounts a robotic arm from the Munich-based robotics provider Franka Emika on a table and connects it to a computer. The robot's hand grabs a glass filled to the top with water, lifts it up and swings it back and forth without spilling any liquid. "And it does it faster and more safely than a human," says the scientist from the Munich Institute of Robotics and Machine Intelligence (MIRMI) at TUM.

The team fed the robot with formulas from algebra that are several hundred years old. It is based on a Moroccan tea tray, which is constructed according to the principle of a spherical pendulum. Together with doctoral student Riddhiman Laha and master's student Rafael I. Cabral Muchacho, Figueredo embeds the dynamics of a spherical pendulum in the robot's control program, which also means that the movements that the robot performs are restricted by the basic rules of geometry.

The mathematics of the pendulum

In addition, he and his team integrate the correct angles, speeds and accelerations into the model. "If you understand how a pendulum moves, know how it works, it's suddenly very simple," says Figueredo.

Spill-not-science' is a complex field. "Most previous approaches have mainly focused on limiting acceleration in order to get a grip on spillover. Or they have looked at the dynamics of liquids to calculate how these substances behave and predict trajectories," explains Figueredo: "This takes at least a few minutes, if not hours, with uncertain results."

Use in care and for the transportation of hazardous liquids

In practice, the scientists initially envisage using the innovative robotic support for elderly people and those in need of care. "However, industry involved in the transportation of hazardous biological and chemical substances is also likely to be interested in such a solution," says Figueredo. Another critical point is safety: ideally, a robot should be able to recognize dangerous situations. "We need better perception for this," explains scientist Figueredo.

Sensors would then enable the machine not only to recognize humans, but also to predict their movements. This is the only way to ultimately rule out a collision with the robot. Up to now, the robot has worked with 'tactile sensors' as a safety mechanism. In the current spill-not mode, the robot arm notices the collision and retracts immediately, but also pays attention to the liquid.