The Technical University of Vienna (TU Wien) has made a robot learn to clean by observing humans . Developed at the Institute of Automation and Control at TU Wien, the device promises to change the way automated cleaning is perceived, applying an advanced combination of artificial intelligence and machine learning .
A “cleaning robot” that learns by seeing
Rather than relying on rigid algorithms, the robot trains itself by watching a person clean the rim of a sink . Using a sponge equipped with force sensors and tracking markers, the machine records every movement and adjusts pressure and speed in real time. This process generates a vast data set that is used to train neural networks, optimizing the “motion primitives” that determine the robot’s actions.
The demonstration data is recorded with an instrumented tool. Source: TU Wien
The application of this technology is not limited to cleaning. According to Professor Andreas Kugi, the robot could be adapted for tasks such as sanding, polishing or applying adhesives to various industrial surfaces . This approach allows the robot to imitate and adapt its movements to different geometries and materials.
The potential of this innovation became evident during the IROS 2024 congress in Abu Dhabi, where it received the ” Best Application Paper Award “. In the press release, the university also shared the original study .
Furthermore, the TU Wien team plans a future where these robots collaborate and share knowledge with each other. Thanks to “federated learning,” machines could collectively acquire and improve skills without compromising data specific to each work environment. This would open the door to a network of robots that optimize tasks in carpentry, metalworking, and other workshops.
Beyond cleaning sinks
Researcher Christian Hartl-Nesic points out that while the task of capturing the geometry of a sink may seem simple, teaching the robot precise movements is a complex challenge. However, tests in Vienna have shown that the robot can adapt its cleaning techniques even to curved and variable surfaces.
PhD student Christoph Unger, meanwhile, emphasises the importance of flexibility in industrial applications, where demands are constantly changing.
Follow us on social media and don’t miss any of our posts!
YouTube LinkedIn Facebook Instagram X TikTok
Source and photo: Vienna University of Technology