Innovative System Learns to Clean Surfaces Without Damage
Who wouldn’t want a robot system that learns to tackle the unglamorous task of cleaning a dirty sink in seconds without damaging the porcelain surface? Researchers from Vienna have successfully created such an AI-powered robotic system. However, their approach promises much more than just a flexible cleaning robot, according to scientists from the Vienna University of Technology (TU Wien).
Among the 3,500 submissions for the recent prestigious IROS robotics conference held in Abu Dhabi, the work of researchers from the Institute for Automation and Control Engineering (ACIN) at TU Wien was selected for a “Best Application Paper Award,” recognizing it as one of the year’s most important innovations. This accolade demonstrates that researchers in Vienna are making “useful advancements” in this highly dynamic scientific field, explained Andreas Kugi and Christian Hartl-Nesic from ACIN.
The AI system learns more or less directly from human actions. The tool used is a “instrumented sponge,” a standard household item that researchers have modified by drilling holes in it and adding force-torque sensors. This allows for detailed tracking of how much force is applied at various points and curvatures while handling the sponge against the porcelain sink.
Additionally, the movements of the leading hand are precisely recorded in 3D using an optical tracking system. This combination yields hundreds to thousands of measurement data points per second. The high resolution enables the AI data processing to quickly understand the requirements after just a few seconds of observation. “We can extract a very good understanding of the process,” Hartl-Nesic stated.
The learning process involves a mathematical-statistical approach that breaks down similarities in movements. This allows the demonstrated actions to be technically translated into a manageable number of parameters and ultimately replicated. “For each new edge, the system can use the statistical description to calculate a meaningful tool movement,” the scientist explained.
During demonstrations, it was shown that the robot could replicate the sink-cleaning task after minimal demonstration—a human merely wiped the front edge of the sink for a few seconds. The cleaning of the rest of the geometrically complex object could then proceed autonomously. In the future, the robot will also be designed to check its own performance after completing the task and make adjustments if necessary.
However, this assistant won’t be available for household use anytime soon. Currently, there are no high-tech robotic arms or advanced systems available for commercial purchase. Kugi noted, “There is still much fundamental research to be done, but the essential components are in place.”
The researchers envision a future where robots serve as assistants, which is particularly important for Austria, a country with many small and medium-sized enterprises (SMEs) in the manufacturing sector, especially in light of the growing shortage of skilled labor. Kugi, who is also the scientific director of the Austrian Institute of Technology (AIT), emphasized that such systems would “not replace people but significantly support them.” Potential applications include any area where various surfaces need to be sanded, polished, painted, glued, or assembled.
Now is the time for companies to consider how their production will look in 10 to 15 years. “Intelligent assistance systems with semi-autonomous functions” will be part of that future. It is also conceivable to collaboratively improve such approaches across multiple companies and applications. “Private data—such as the specific shape of a workpiece—would remain confidential, but important learned principles could be shared to enhance the capabilities of all robots,” stated representatives from TU Wien.
