EN
Country: EN (United States)
Select country/region

Language

  • English
stage-teaser-component-pace-start-up-compass-2023-image-webx

SICK Start-up Compass

In SICK’s Start-up Arena, Govinda Kempermann is trying to get an unruly cable under control. The cable connects a 3D camera with a graphics processor – and mustn’t interfere with the sensor, which is also part of the new setup. But the cable is stubborn and keeps twisting back in the other direction. Finally, Kempermann clamps it onto a metal brace.

“Here, we’re working on a system that can elegantly and efficiently solve multiple logistics problems,” says Josef Baak. The 49-year-old, together with systems engineer Kempermann (38) and software developer Jan Gerber (29), makes up an in-house start-up at SICK. They chose the name Compass, which stands for Cognitive Machine Perception – a Scalable Solution. The three of them work in a former production hall at SICK headquarters. The room is divided by plywood walls, and the ceiling panels have been torn out; there’s plenty of room to think. One of the problems they’re trying to solve is jammed baggage carousels at airports. Time and time again, backpacks, umbrellas, and baby carriages create blockages. “For airports, this is a real problem. First of all, the transport system can be damaged; furthermore, passengers can get impatient if they have to wait too long on their baggage,” says Jan Gerber.

Neural networks deliver decision
The day before, the three of them didn’t get back from Amsterdam until late at night. They spoke with the operators of Schiphol Airport about how their system could be integrated into baggage transport and what the operators were looking for in order to avoid blockages. Kempermann places a backpack under the two cameras with the stubborn cable, while Gerber starts up the system on his laptop. The backpack activates the sensor, and the 3D cameras, installed across from the sensor, send images to the graphics processor. “In addition to a normal photograph, our sensor generates what is known as a depth image, which depicts the distances in the room,” Kempermann explains. Using the two images, the graphics processor from US-based company Nvidia creates a point cloud that depicts the backpack in three-dimensional space. “That’s the basic information that we then feed into the neural nets,” says Gerber.

They plan to add more neural nets. After all, it takes three or four consecutive nets just to detect the backpack; then another net decides whether to transport the backpack or discharge it. The advantage of this solution: flexibility for the customers. They can define the categories that the AI – since that’s what the neural nets form – uses to make its decisions. Or they can use the factory presets as much as possible. “Most likely, the customer will need to keep adjusting the categories,” says Gerber. For example, when new backpacks are brought to market that are especially prone to creating jams.

Training the system with data
For the next step, the start-up needs data: photos of baggage that should ideally be discharged. This data is used to train the system. Baak, Kempermann and Gerber plan to gather photographs at the airports in Paris and Amsterdam in the near future. But before that, they have an important meeting – with the SICK Executive Board. “A week from today, we’ll have the chance to pitch our idea. Then they’ll decide whether or not to give us a budget,” says Baak.
For a successful pitch, they have to demonstrate that there’s a concrete, real-world application for their idea: improving baggage transport at airports. But that’s not all – in order to offer longterm prospects, the product has to be scalable, i.e., it has to have potential applications beyond airports.

Baggage conveyor with suitcases

Pocket-sized 3D camera
Josef Baak opens up a presentation on his laptop. One slide shows a huge pile of parcels at a logistics center, dumped there with no rhyme or reason. Beside it is a photo of a parcel that has been damaged and sloppily taped back together. This already hints at the larger market that the start-up hopes to penetrate: parcel shippers want to know at any given time where their parcels are – and if any are lost along the way. Multiple 3D cameras installed at key positions in a parcel-sorting center could make a significant difference.

And when a package is delivered but arrives damaged, customer claims have to be as fast and easy as possible. “Our 3D camera solution can be flexibly and affordably applied in all these areas,” says Baak. He takes a Visionary T-300 3D camera out of his pants pocket. A small box, only about nine centimeters long, four centimeters tall and three centimeters wide, can hold two cameras. “The only reason we can do that here is because we have a product development team that is working on the sensor right now, backing us up,” says Baak. What’s new about the Visionary T-300: the computer is now separate. Thanks to a collaboration with Nvidia, this makes it possible to offer customers flexible and tailor-made solutions.

And flexibility is always called for, particularly in the start-up business, no matter whether it’s in development or marketing. Originally, Baak, Kempermann and Gerber wanted to use the 3D camera in a parcel-handling robot; and they may still get the chance. In the meantime, the three entrepreneurs have successfully made their pitch. Without missing a beat, they switch gears and soon land at the airport.