Artificial Intelligence

"Hi, I'm Toru, the intelligent warehouse robot"

A start-up called Magazino has built an intelligent warehouse robot. The goal is to actually replace human labor.

29 Feb. 2016 Jochen Fuchs
lager-roboter-1
A start-up called Magazino has built an intelligent warehouse robot. The goal is to actually replace human labor.

Not far from a dodgy drinking hole along a heavily traveled road in Munich stands a long commercial complex – directly on the railroad tracks. A spartan covered sidewalk passes by a small range of stores, some more than a little bizarre. The storefronts seem almost disconnected from their location, even if there's no doubting they were conceived as retail spaces. In the inside of the building, long, bare corridors pass by bone-dry offices and companies, leading to a startup that itself seems thoroughly out-of-place in this environment: logistic robotics specialist Magazino, which is currently waging a frontal assault on classic 'intralogistics,' as the system of internal logistics within a company's fulfillment center is known.

lager-roboter-2
The lounge area at Magazino. (Photo: Jochen G. Fuchs)

The loft-like office that awaits inside is awash in light; one passes by an elegant, human-sized Magazino logo on the way through the office's lobby area, the gaze gliding over steel-and-wood furniture designed by Magazino's in-house designer and produced by the company itself. From the casual chat corner to the espresso bar/kitchen, the atmosphere is very much a blend of Scandinavian design, Stockholm start-up flair and a Silicon Valley attitude. Only the din of the drill press on a workbench, humming 3D printers and a metal workshop in one corner of the giant room gives away that this is "only" a mechanical engineering startup from Munich. Although that's a bit of an understatement: like any classic startup from the digital business, Magazino has gathered a cadre of developers, designers, and marketers – as well as robotics experts and mechanical engineers.

lager-roboter-3
The natural home of mechanical engineers and the birthplace of the robots: the workshop. (Photo: Jochen G. Fuchs)

Tidy up no more: How a dream can become an innovation-driven company

Back when he was a student, Frederik Brantner would sometimes visit a colleague in her drugstore. He was always especially fascinated with its drug dispensing machine: simply press a button on the computer and the machine then spat out a box of medicine. "I thought to myself: I could use something like that at home. I just take all of the things scattered about in the chaos that is my life, throw them in a drawer and it'll tidy them up for me. And if I need something, I simply use my 'cover flow' and select my battery charging device and then it's smart enough to hand out the rechargeable batteries to me."

The idea that germinated in his head in that summer of 2011 never quite left Brantner. He organized a "creative thinking weekend" and received strong feedback from those present. Mechanical Engineer Lukas Zanger climbed on board as a co-founder – or as Brantner formulates it with a smile: "He didn't run away fast enough, and suddenly he was a part of it." The founding team was later joined by IT specialist Nikolas Engelhard — and Magazino was born.

lager-roboter-5
The Magazino founders, from left to right: Nikolas Engelhard, Lukas Zanger and Frederik Brantner.(Photo: Magazino)

After an initial rejection, in early 2013 Magazino received an 'Exist' startup stipend from the German Federal Ministry for Economic Affairs and Energy. And then things started moving quickly. A limited liability company was founded in 2014. After a month of simply coming in and getting to work, the first employee — Kai, now the director of electronics — was the first to receive a work contract. A second followed — industrial designer Joachim, who had also designed furniture for Magazin — as did a Ph.D. robotics expert and software developer named Ulrich Klang. All three are still part of the company today.

Magazino later also received the so-called 'Flügge' grant from the Bavarian Economic Ministry, a seed financing round through two angel investors and the Hightech-Gründer-Fonds (HTGF) and then something unprecedented: Siemens offered to buy out the business angels and the HTGF, and now holds a 49.9 percent share of the young robotics firm. Magazino had successfully exited the HTGF faster than any other startup in history.

Venture Capital? Thanks, but no thanks...

As a result, Magazino now enjoys a luxury that many venture capital-financed startups can only dream of. Brantner describes their current financial cushion with tangible satisfaction and calm: "We were granted a very large amount of latitude from Siemens that will allow us in the coming years to work on a product without worrying about other follow-on financing. This, beyond its technical expertise and readiness to provide support in a variety of areas, was one of the major factors that spoke for Siemens. With venture capitalists, we'd be chasing every year from one round of financing to the next, that wouldn't help us. First I'm doing the Series A, and suddenly I'm already focused on the Series B. And I constantly have to be holding up the developers, that won't work. In the setup phase in particular we have to get the technological basis in place quickly."

lager-roboter-5
Maru the drugstore robot uses a rotating set of shelves to hold its mixed inventory of medicines. (Photo: Jochen G. Fuchs)

At first it appeared that Magazino truly wanted to realize Brantner's dream of intelligent domestic robots, ready to spit out small items on demand. The first product it delivered, the "Maru," is a clunky round box that looks a bit like an oversized telephone booth. Maru is — what else — a medicine delivery machine for druggists. Yet it differs from others of its ilk in one important way: it is not filled like a classic soda vending machine, with each medicine in its own fixed spot. For Maru, different products are placed next to one another. Put simply, you fill up the machine by dumping in a box of unsorted medicines. The machine doesn't simply deliver whatever is in a preordained spot like a candy machine, but rather seeks out the right medicine from its 'belly.'

lager-roboter-6
The medicine is then dispensed to the vending slot. (Photo: Jochen G. Fuchs)

To infinity and (far) beyond: Magazino crosses the borders of previous robotics

At this point, Magazino was beginning to address a problem known in expert circles as Moravec's Paradox: "It is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility." The paradox was formulated in this way by Canadian trans-humanist and robotics specialist Hans Moravec in the 80s, among others. Magazino gives its robots perception and the possibility of adjusting independently to changes. "In everything we do, we're pushing past the borders of what the automation industry is currently able to do," Brantner emphasizes.

Druggist's-assistant Maru is no longer being produced. Entering that hard-fought market would have been too service intensive and expensive, Magazino determined during the pilot phase. The last Maru robot was reconfigured into a creative 'standing meeting' spot in Magazino's Munich premises. Yet the project was hardly a flop — the technology at play in Maru is now in Toru, the mobile picking robot that is working for the logistics and mail order business instead of drugstores.

lager-roboter-7
Toru picks up a book: camera-equipped gripper arm in action. (Photo: Magazino)

Robotics in intralogistics: rotating shelves and dumb robots

Standard industrial robots, including those used in the logistics field, have been engineered to perform a series of repetitive, preprogrammed work and movement steps. They dutifully perform those processes millions of times in a row without issue — until something runs counter to plan. If a carton to be filled slips out of place on a conveyor belt, then the robot is helpless and unable to proceed.

Most logistics robots responsible for picking and commissioning merchandise to be shipped — meaning fetching it from a shelf location and adding it to an order — are precisely this kind of dumb robot, such as shuttles and lift systems. Beyond this, they typically only bring containers with the product inside to human workers; the robots deployed at Amazon's offices in the US and Poland can't reach onto a shelf or a carrier and pick out individual objects. Those robots cannot replace humans, who can be deployed flexibly within the warehouse.

"Things are changing here at breakneck speeds. Robotics is seeing major breakthroughs — even if we're still quite far away of being able to do what people can," Brantner explains. He adds: "Toru behaves like a person. He should be able to do exactly the same things as a person, including multitasking. That's very important because it will free the retailer from having to decide whether to automate its warehouse or not. There will instead be the choice to use one single robot for specific tasks. In this way the robot will be able to do what it's good at and what is difficult for the human, and the human will handle the things the robot can't do." Toru can do exactly the same thing as a human worker: target and retrieve specific merchandise from storage and then transport it.

lager-roboter-8
Toru moving along the shelves. (Photo: Magazino)

Toru, the intelligent picking robot

Magazino is offering a series-produced logistics robot capable of handling 'picking,' meaning the identification, retrieval and transport of a specific warehoused item. Toru handles these tasks autonomously and uses computer vision algorithms and an integrated camera to move independently through the open spaces of a warehouse. It receives its picking assignments from a connected ERM system. Toru is outfitted with a rules set that allows it to make decisions based on the situation and plan its actions accordingly. At the start of its deployment, Toru traverses its assigned warehouse area, forming a 2D map of the space, including the position of the warehoused merchandise, and compares it with the received ERM data.

lager-roboter-9
(Photo: Jochen G. Fuchs)

The robot is capable of seeking out and gripping individual square objects from a shelf or storage container on a shelf. The current pilot model then brings the picked items to the commissioning station; the robot model to be delivered to pilot customer Sigloch will be outfitted with an integrated commissioning shelf that allows the robot to pick several objects before handing them over. Toru is slower than humans, at least for right now. Working with a batch size of one, meaning one single piece of warehoused merchandise per pick as is common in the B2C mail-order business, humans can achieve 120 picks per hour on average. Toru currently manages 80 picks, but upgrades are slated to bring that number up to 120 by May. One issue limiting its performance is the maximum possible controllable speed of the gripper arm. Another is Toru's movement speed, currently capped at one meter per second. That speed cap is just one of the many safety mechanisms and behavioral rules integrated into Toru to ensure that it can work safely around humans and other machines — after all, there are a variety of regulations at both the legal and insurance level to be accounted for.

For this reason Sigloch, a major book distribution firm, is only using Toru as 20 to 30 percent of its workforce. Toru picks technical and scientific books, which unlike the rest of its catalog tend to be ordered alone. The 20-30 percent degree of automation at play there could be raised to 100 percent, but for economic reasons is not currently planned by the pilot customer. Roughly 20 to 30 percent of an average Amazon distribution center with mixed warehouse inventory can currently be handled using automated technology, Brantner says. The rest is outside of the realm of Toru's gripper mechanism, either because it is not rectangular or because it is too heavy. If a mail-order firm working solely with rectangular warehouse inventory — a B2C bookstore, or a shoe store — wanted to work with robots, then the picking process could be upgraded to 100 percent automated right now.

Breaking the 'invest heavily in automation' model: robots as Hardware-as-a-Service

The benefits of automation, and in particular the benefits of robots in logistics centers, are clear, Brantner feels: "I am brought the single object and not an entire carrier. That cuts down on the need for personnel. In this way I can introduce radically different business models, perhaps with an entirely different picking schedule, such as at night. Or I can just pay per pick."

Right now, when an online retailer opts to upgrade to an automated warehouse, it's always an all-or-nothing decision. There's no middle ground. If the retailer wants to buy a conveyor belt system, it needs to commit fully to automation of the affected process steps — and whatever sum of investment is involved. Autonomous, mobile robots like Toru open up greater flexibility, Brantner says: "We offer customers an entirely different way to enter into the market than other automation makers. We can propose to the customer: 'How about we start with one robot — and if you're satisfied, you can add in another one and simply pay per pick.' We don't need to get started with a multi-million project at a big corporation, but rather can send in one robot to get started and show off how well it performs."

The decisive cost factors for mail-order retailers looking to work with a robot are: pick costs, pick errors, speed and flexibility. A robot's pick costs are calculated using a simple 'price/performance' formula: the cost per pick is determined by whether the robot should be used on one, two or three shifts — and whether it needs to be amortized out over two or three years, for example.

Magazino also intends to work with a flexible billing model: instead of buying each robot, the startup will provide the hardware to mail order retailers and logistics companies and charge by the pick. This will allow retailers to scale up quickly without requiring massive investment sums.

lager-roboter-10
Toru the picking robot in the hallway of the Magazino office during a test run. (Photo: Jochen G. Fuchs)

Man and machine: safety, jobs and societal impacts

No robots, not even Toru, can replace people right now. In the coming years, robots and humans will continue to co-exist in the warehouse. This opens up a variety of questions: Is this safe for the human? Will humans disappear sooner or later from the distribution center?

"One of the biggest working areas is occupational safety involving robots. When it comes to a barrier at a parking lot that might fall on my head, you expect more of the person. For a robot driving autonomously through a warehouse and encountering people, it's the robot that is subject to countless rules, some of them very old. The speed at which the robot is currently allowed to move, not to mention thousands of other rules, makes it quite unlikely that a person will be injured."

And the jobs? Asked about the potential savings for Magazino's first pilot customer, Brantner answers: "We'll be implementing two robots at first. They'll be performing a high number of picks and will quickly pay for themselves." In concrete terms it follows that a Toru robot could potentially completely replace a human worker. Translated into jobs, this might mean that a warehouse with 2,000 employees might be able to cut 20 percent, or 400 positions through automation. Asked whether the increased productivity through automation will eliminate jobs, Brantner says: "Our robots help boost productivity in the warehouse, and thus the overall stability of the jobs there. Because after all, competitiveness is a necessary factor in creating a high number of jobs. Only those sites engaging in technological progress head on will be able to compete on the international stage. In the long term, robots will be able to engage in more and more tasks.

Yet we're unlikely to see warehouses empty of humans anytime soon. Automotive production has grown more efficient through robots, and that has allowed production to stay in Germany. And there are many, many new jobs in machining being created, to the point that we're now the global leader. There are studies that say that robots create more jobs than they eliminate. There are also studies that say things are moving so fast we can't keep up as a society."

Artificial Intelligence Disruptive TechnologiesStartups CEBIT RSS Feed