Researchers at Vienna University of Technology (TU Wien) have pursued an alternative method. Instead of hard-coding rules, they employed a teaching approach: a human demonstrates the cleaning process to the robot several times, using a specially prepared sponge to scrub the sink's edge. Observing this, the robot learns to mimic the cleaning process and can then adapt to objects of different shapes. This innovative approach was presented at IROS 2024 in Abu Dhabi, a leading global robotics conference.
Beyond just cleaning: A robotic approach to surface treatment
Cleaning is just one example of surface treatment tasks that share similarities with other industrial processes like sanding, polishing, and painting. "While capturing a washbasin's shape with cameras is straightforward, teaching the robot how to adjust movements, speed, angle, and pressure is much more complex," says Prof. Andreas Kugi from TU Wien's Automation and Control Institute.
As Christian Hartl-Nesic, head of TU Wien's Industrial Robotics group, explains, people naturally learn these nuances through observation and experience. "In an apprenticeship, a mentor might guide someone, saying 'apply more pressure here.' We aimed to develop a way for the robot to learn similarly."
To achieve this, the team created a specialized sponge with force sensors and tracking markers. This tool allowed humans to clean only the front edge of the sink repeatedly, producing extensive data that enabled the robot to interpret and understand proper cleaning techniques.
Learning through data-driven modeling
The TU Wien team leveraged a unique data processing approach that synthesizes several machine-learning methods. The system first processes the demonstration data statistically, training a neural network on specific movement patterns, or "motion primitives." This framework then allows the robot arm to maneuver the sponge optimally across complex surfaces.
Even though the robot only observes cleaning a single sink edge, it learns to apply the technique to entire sinks and other intricate surfaces. "The robot adapts the sponge's grip and pressure based on surface contours, applying more force in curved areas and less on flat surfaces," says PhD student Christoph Unger of the Industrial Robotics group.
A collaborative future for workshop robots
The team's advancements extend beyond cleaning, with potential applications in diverse industries, including sanding wood, polishing auto bodies, or welding sheet metal. Eventually, robots equipped with this learning algorithm could operate on mobile platforms, serving as flexible assistants across various workshops.
Looking ahead, TU Wien envisions a network of robots that can share knowledge. "Imagine multiple workshops using self-learning robots for tasks like sanding and painting. Each robot could learn locally but share critical insights with others, advancing collective intelligence while preserving private data," explains Prof. Kugi. This concept, known as "federated learning," would enable workshops to benefit from a shared pool of robotic experience, enhancing the robots' capabilities.
Research Report:ProSIP: Probabilistic Surface Interaction Primitives for Learning of Robotic Cleaning of Edges
Related Links
Vienna University of Technology
All about the robots on Earth and beyond!
| Subscribe Free To Our Daily Newsletters |
| Subscribe Free To Our Daily Newsletters |
