TUM Graduates Innovate Robots Designed for Handling Flexible Textiles

The founders of the start-up sewts are especially intrigued by materials with high tensile strength and low compressive strength. Anisotropic materials – the term applied to materials with properties that vary according to the direction of measurement – have long been a challenge for gripper robots. “Textiles are flexible,” explains sewts co-founder Alexander Bley. “They change their shape when I hold them up.” When freshly laundered towels or sheets from an industrial laundry are prepared for delivery to a hotel, a few steps are necessary. This includes taking the laundry out of the basket and placing it lengthways on the conveyor belt that feeds the folding machine. Industrial robots from the start-up, founded almost five years ago, can now do this. A gripper arm pulls a laundry item out of the container and drops it onto a conveyor belt. A few meters further on, a second robot arm grips the piece of laundry at a corner, clamps it on one side, and hands it over to the so-called sliding robot, which stretches the textile widthwise. The stretched item is transferred to the folding machine within a few seconds.

First, simulate, then train AI algorithms with synthetic data

During their studies at TUM, the founders laid the foundations for their later innovations. At the Chair of Carbon Composites, Alexander Bley investigated the properties of technical textiles, which behave differently depending on the direction of loading. In a “draping simulation”, he developed an idea for handling flexible materials. Co-founder Till Rickert worked on autonomous driving at the Chair of Automotive Engineering. An essential part of the research involves using synthetic training data to train AI algorithms. And Tim Doerks, who, like Bley, comes from the Carbon Composites department, specializes in building and testing prototypes. All of this has flowed into the innovations of their startup.

The technical challenge for the first application was for the robot to learn how to grip a laundry item and hand it over in a suitable form. Depending on how a towel is laid out, for example, it can take on many different shapes. It can also be striped or checkered, white or colored. The founders of sewts play tricks to teach the robot these countless variations. They simulate the various shapes a towel can take in the computer and generate their own (artificial) training data for the system. Instead of photographing or filming a towel from all sides and in all shapes and styles and feeding the system with these data, the computer generates the images itself and provides them to the robot as training data. “We use synthetic data to train our AI algorithms,” says Bley, who, like co-founders Tim Doerks and Till Rickert, completed his master’s degree in mechanical engineering at TUM.

Boost via XPLORE, XPRENEURS, EXIST, and the Initiative for Industrial Innovators

This case study is not the only thing keeping the three founders busy, however. In the XPLORE program of the UnternehmerTUM innovation center, in which start-up teams develop their business ideas, the future entrepreneurs have already come up with around 70 possible applications, from vehicle seat covers and awnings to cable harnesses. This is because a key aspect of XPLORE was to identify products that fit particularly well in individual markets. For the first application, the XPRENEURS incubation program focused on identifying industrial laundry customers, assessing the size of the German and worldwide markets and identifying the main barriers to entry. But the real green light came with the approval from the Federal Ministry of Economics for an EXIST grant, which supports start-ups for a year. Meanwhile, the start-up also secured a small six-figure sum as additional support via the Initiative for Industrial Innovators, in which UnternehmerTUM and the European Investment Fund were also involved. This laid the foundation for major financing rounds.

It was the right move: with the systems sold to date, the company, which employs just under 30 people, has already made millions in sales. And the potential is far from fully tapped: Alexander Bley estimates 5,000 to 6,000 folding machines are sold annually in Europe and the USA. “You could simply introduce our robot there,” says the young entrepreneur.

First the industrial laundry – now the returns business

The second major case study is now on the agenda: returns processing in online fashion retail. A pilot project has tested the new robot system with items returned to a major German retailer. The essential difference compared to the industrial laundry is the much wider range of garments. T-shirts, long-sleeved shirts, with and without buttons or V-necks, and trousers with zippers. The AI system will have to learn about all of them to prepare garments for the folding machine. The new venture capital, totaling seven million euros from 2023, has come at the right time to improve the technology, enter international markets and launch further use cases. Or – as Alexander Bley says: “To let the verticals play out, one after another.”