Rahul Somvanshi
01
A robot at TU Wien has been trained to clean sinks by observing human movements—picking up on techniques like scrubbing speed, force, and angle from just a few demonstrations.
Photo Source- TU Wien TV (Youtube)
02
Equipped with a sensor-loaded sponge and tracking markers, the robot learns intricate cleaning methods that go beyond basic motions to master complex surfaces.
03
How do you teach a robot to scrub like a pro? This system uses machine learning to convert a few human demos into data that can teach movement details down to the angle and force.
04
With only the front edge of a sink as its practice zone, this robot can adapt to clean entire washbasins—what’s next for this tech?
05
TU Wien’s robot isn’t limited to cleaning; its data processing could support other surface tasks like sanding wood, polishing car paint, or welding metal.
Photo Source- TU Wien
06
How does it keep getting smarter? The robot uses statistical data and neural networks to learn “motion primitives”—basic moves that it combines for full-surface cleaning.
07
The project, awarded “Best Application Paper” at IROS 2024, beat over 3,500 entries for its innovative approach to machine learning for complex physical tasks.
Photo Source- Ecole polytechnique (CC BY-SA 2.0)
08
Imagine workshops where robots learn locally but share their knowledge through federated learning, a method TU Wien hopes will push surface automation to new levels.
Photo Source- Ars Electronica (CC BY-NC-ND 2.0)
09
This robot can not only clean but also adjust its grip and force to suit different shapes—making it ideal for tasks that involve both flat and curved surfaces.