While companies like Amazon and Google are racing to develop advanced warehouse robots, a small German company believes it has an advantage that these companies don’t: Their robots can see.

Magazino, established in 2014 and currently backed by Siemens, has created a warehouse robot with advanced computer vision called TORU that can accurately identify and pick items off a shelf, “store them in their little back pack, and bring them to a sorting machine,” said Frederick Brantner, CEO and cofounder of the company.

At the moment, most warehouse robots, like Amazon’s Kiva, can move entire pallets or shelves­­, but don’t have cameras. “They only drive to a fixed point,” said Brantner. “Kiva looks at QR codes on the ground, and then basically lifts the entire shelf. That doesn’t have to be very precise, because each shelf looks the same, you just lift it up from bottom and you can transport.”

While Brantner said that Kiva is really good at transporting entire shelves, it will run into a problem “the moment you want to once handle something like a shoebox, or something that’s not always in the same spot.” So human workers are still needed move around from shelf to shelf, picking items and moving them in the warehouse. “You still need humans to do the picking,” said Brantner. Also, sometimes the barcode might be missing, or an object isn’t placed properly in a shelf–so you need to have a robot with computer vision.

SEE: 6 ways the robot revolution will transform the future of work (TechRepublic)

And whereas a robot like Rethink Robotics’ Baxter has arms that can make very precise movements allow it to perform different kinds of tasks, it is still a stationary machine. The fact that TORU is highly specialized, said Brantner, allows it to excel in the one task it’s designed for. “We looked at a specific problem and constructed the robot around it,” said Brantner.

So why is this an advantage? Brantner has done the math­­–and a lot of time and money is wasted when humans need to move around the warehouse selecting objects. Germany spends $1.7 billion each year on manual picking, he said. While TORU is still slightly slower at picking than a human worker, in picks-per-hour, “we are heading to reach similar numbers soon,” said Brantner. And, a warehouse robot is able to work longer shifts.

SEE: Why China is scooping up robots from Rethink Robotics to solve its manufacturing problem (TechRepublic)

According to Brantner, costs for one picker are nearly $34,000. TORU, which costs $113,000 per unit, pays this cost off in about three years. “The more it works, the quicker it pays off this cost,” said Brantner. “So if it works two shifts per work day, the robot pays off in even less than two years.”

Google has also recently announced plans to develop warehouse robots. Brantner said that their plan “highlights how difficult the picking of items is, and the importance of computer vision. We’re proud to say that we solved this already.”

SEE: Amazon, robots and the near-future rise of the automated warehouse (TechRepublic)

Brantner sees TORU as being the next step in the evolution of warehouse robots, offering a machine solution to a task previously only handled by humans. Still, he doesn’t see that it will entirely replace the pallet-moving robots–instead, it could complement them, working together as part of the warehouse ecosystem.

And while TORU is great at handling books, boxes, phone books, etc. it cannot handle all objects yet. Still, Brantner believes, it is an important move forward in warehouse robotics. “Complex tasks can be done now by robots because they have a seeing-eye and intelligence to make their own decisions on the spot,” he said.

Also see…