news    views    talk    learn    |    about    contribute     republish     crowdfunding     archives     events
past episodes   |  about

Robots Podcast #226: Toru Robots, with Dr. Moritz Tenorth

         


interview by
January 21, 2017

totus_cube

In this episode, Ron Vanderkley spoke to Dr Moritz Tenorth, head of software development at Magazino, a Munich-based startup developing mobile pick-and-place robots for item-specific logistics. They discussed his work on the Toru robot and what it means to the warehouse industry today and in the future.

Dr Moritz Tenorth

Dr. Moritz Tenorth
Dr. Moritz Tenorth

After obtaining his PhD in robotics from TU Munich Dr Moritz Tenorth spent several years as a post-doc and freelance robotics consultant, with stays at the CMU Robotics Institute and the ATR labs in Japan. His research focuses on knowledge representation methods that can help autonomous robots make smarter decisions, and on methods for task coordination in robotics.

 Links


Transcript


Jana Witt: TORU with robots, the podcast for news and views on robotics.
 

 

 

Hi and welcome to the robots podcast. Today, we are visiting German startup company Magazino, based in Munich. At Magazino teams are developing perception controlled mobile robots for use in warehouses or manufacturing. Their current model called TORU robot consists of a mobile base, a removable shelf and a retractable and rotatable column with a gripper system. The robot can identify objects using 2D and 3D cameras. It can grasp a variety of rectangular objects from shoe boxes to lexicons, store the object on its shelf and delivery them to where they’re needed. TORU’s main strength is its flexibility compared to other systems, which usually require a warehouse to be custom-built or at least adapted. With Magazino’s solution, robots could be used in existing warehouses alongside human workers. Our interviewer Ron spoke to Dr. Moritz Tenorth, Head of Software Development at Magazino about the TORU robot series and what it means to the warehouse industry today and in the future.
Ron: Good morning Moritz. If I can get you to first introduce yourself to the podcast listeners.
Dr. Tenorth: Okay. Good morning. My name is Moritz Tenorth. I’m heading the software team at Magazino.
Ron: Could you give us a little bit about your background.
Dr. Tenorth:  Before joining Magazino I spend several years in research. I did my PhD at the Technical University in Munich on a topic related to knowledge for presentation for robots. It was more on the combination of AI and robotics. Then I spend some years as a postdoc and project manager for European Project, namely a project that investigated how task instructions from the internet could be conformed into a format that is executable on robots. For example, cooking recipes, how could robots understand them? How could they execute them in the real world?
Ron: Great. Okay, firstly can you give us a little bit of background about Magazino?
Dr. Tenorth: 

 

 

Magazino’s a startup company located in Munich, Germany. We were founded in 2014 and our goal is to develop robots for intralogistics. What is special about these robots is that they can pick individual items and fetch them from the shelves, from their storage locations, to the drop-off location and bring them to the humans. Our main product at the moment is TORU. It’s a mobile pick and place robot that is able to fetch items from shelves, and operate in warehouses that have originally been made for humans. So far we are focusing, with TORU, on cuboid objects. Magazino has grown quite a lot in the last year. At the moment we are at about 15 employees.
Ron: Great. Can you give an idea of the problem that you’re overcoming?
Dr. Tenorth: The problem in many intra-logistic scenarios is that the classical automation doesn’t make sense for many customers. A classic automated warehousing system is a very expensive, very inflexible thing. But it don’t always operates at the level of boxes, of crates, of pallets. It cannot fetch the individual items, and the sizes and dimensions are normally pretty predefined at design time.
 

 

 

 

The problem for many customers is they have very dynamic business, especially newcomers, and they can’t really plan ahead for several years. As a consequence, the classical automation doesn’t make sense for them and it’s normally still humans that walk along rows of shelves and pick items and orders. When you order something in an online shop, basically the order is sent to a human that walks through the warehouse with a little wagon, picks the item and brings them to some drop-off location. On the one hand that’s pretty costly. On the other hand it’s often difficult to find enough humans that would like to do this job, especially in some areas here in Germany or in surrounding countries. That’s where TORU comes into the game. TORU is able to operate in the same environment as the human does, so you don’t have to modify the environment a lot. It can navigate through the environment, determine its location and find the objects, pick individual items and bring them out of the warehouse.
Ron: I see. I see. What has the journey been like coming up with a product like this?
Dr. Tenorth: Well yes, it’s a pretty complex product and I think it’s at the edge of becoming possible. Robotics research I think has come to a point where perception has made lots of progress for identifying these objects, but also robot control and task coordination research has developed solutions that are now applicable and that allow us to build products for things that have only been research demonstrators in the earlier years.
 

 

 

 

 

 

 

One big problem for us is, of course I mean nobody has built such a kind of robot before. Now, some obvious ways of putting robot arm onto mobile platform, but these are not really well adapted to the problem that we have to solve so we have to cover the full range of space that a human can operate in. We have to store the picked items at some places and so on. We had to basically build the robot from scratch ourselves, design it ourselves, construct it mechanically, electrically and also design the software. This very fast co-development of hardware and software, of course, produces a lot of challenges. We follow a very iterative process, so we have new hardware revisions every few months. From a software point of view, of course, it’s a challenge to keep up with these very fast hardware changes and to design software in a way that it’s able to run on all of these different hardware revisions and still fulfil the job.
Ron: I see. Who and what are your competitors out in the real world now?
Dr. Tenorth: 

 

 

 

There are, of course, other companies that develop automation for warehouses. We don’t really see the classical warehouse automation as much of a competitor because they solve different problems in a way. Recently there have been some startups that started to work on these kinds of problems. One of the best known is Fetch Robotics in the Silicon Valley. Another one is IAM Robotics, in Pittsburgh I think, that also develop robots for intralogistics and that partly also address the picking problem. They have a different design, a different approach, so they rather use more classical robot arm on a mobile platform where we have very special kind of hardware for picking the items.
Ron: Right. When you’re talking about the hardware, what type of hardware are we talking about? Sensors? Drive mechanism? Navigation? Grippers? 3D camera? 2D stereo? What kind of array that makes it efficient?
Dr. Tenorth: 

 

 

A very important aspect is, of course, perception so we’re using different kinds of 2D and 3D cameras and always look at the requirements a certain problem has. For example, for a picking cuboid-shaped objects we’ve developed a special detection mechanism using a 2D camera and a cross laser.
The drive for our robot is a very classical differential drive. It’s a bit non-standard because it’s off-center and because the robot is pretty large compared to its environment. The robot is one meter, twenty long, 60 centimeters wide. We have to drive pretty close to obstacles which makes it challenging from a robotics point of view.
In terms of navigation, we are using lasers for safety, for ensuring safety in collaboration with humans and also for localizing and the robot in the environment and for navigation and we’re also using other kinds of sensors such as distance sensors for detecting, for example, shelves and other kinds of obstacles.
 

 

The gripper TORU uses is an in-house development. It’s a specialized gripper that is able to pick cuboid-shaped objects from stacks and can do this pretty fast and pretty reliably. If one of our pallet customers is shipping books and grasping books, it’s non-trivial things so they can flap open, they can have holes on the top. You can’t really use suction or two-finger grippers for grasping them, so the specialized gripping device is able to handle books, but also other kinds of boxes, shoe boxes and so on, pretty fast and pretty reliably.
Ron: We’ve just briefly talked about the hardware. What about the software? What type of software? What type of algorithms would you tend to employ on problems like this?
Dr. Tenorth: 

 

 

Well, there’s, of course, a whole bunch of algorithms on all levels from motion control to localization, navigation, perception and so on. As a software architecture, we are using ROS. From our point of view that also helped us a lot. It helped us with finding people that are already trained on the job for onboarding people. It made it much easier to get people on board and to have this massive growth and survive the massive growth of last year. Also, from a systems engineering point of view we are pretty happy with that. Many of the functional components, the algorithms have been developed in-house because we found that if you move away from the mainstream robotics areas then often the algorithms that come out of the box as open source can’t be used as is but have to be adapted or you have to build your own algorithms that are adapted for your problem.
Ron: Okay. How does this effect the current warehouse job? Are we transitioning out of the pickers? Are we still having workers still on-site or it’s a transition to servicing the robots themselves? How do you feel that it’s going or will be going?
Dr. Tenorth: I think TORU won’t really replace human workers. Workers will do something else and will also continue to do picking. Robots are not at the stage where they can pick each and every object from every position. There are some aspects of the human jobs that are very difficult to automate still. For example, quality control of the items, checking that all items that are sent to a customer are still in very good shape. That’s something that’s very difficult for robots still, but it’s very simple and also not very unergonomic for humans.
 

 

What the robot can do is, for example, drive around the warehouse, pick from places that are not very pleasant for humans like very down on the floor or pretty high up, and do these kinds of tasks. For many of our customers the problem is not so much that they would like to replace the human workforce, but that they can’t scale their business at the moment because they don’t really find enough workers. They would like to compliment them with robots that do more the easy, easy from a robotics point of view not easy for humans necessarily, and dull tasks, and that humans can take over the more interesting and those tasks that actually still require humans.
Ron: Does that mean that there are also energy savings? Your current robots, do they require the ambient lighter? In other words, does a warehouse need to be fully illuminated for the tasks?
Dr. Tenorth: No. The robots has its own light so it can basically illuminate the shelf and the shelf compartments, picking from actually woodwork better without ambient light or, for example, without sunlight or something. It can handle that but you could also switch it off, yes.
Ron: Yes. No, that sounds like a saving in itself.
Dr. Tenorth: Yeah definitely. Another saving is that you can operate at times where the humans are not there, so pre-pick things in the night and use the off-working hours.
Ron: What do you think the future of warehouse technology will be? Is it going to get to a point where the situation is fully automated in the next 10 or 20 years, with Q&A, et cetera, part of that?
Dr. Tenorth: 

 

We do expect that many more of the picking tasks, transporting tasks for complete box and pallets and so on to be automated a lot. We think it will still be a mix of the classical automation for both areas where it makes sense. If you have predictable demand, if you have stable requirements, similar kinds of objects and a high throughput, then a classical automation system is often a good choice. But for the areas where you don’t have this, for Ecommerce or personalized production, we think that robots will take over more and more the intralogistic tasks as well. We would expect that there are still objects that are difficult for robots to handle, like very fragile, large, difficult to recognize objects, and that there will still be tasks that will have to be done by humans.
 

 

 

I would expect that logistics is one of the fields where robotics will really take off in the next years. It’s kind of in between production and the open world, so compared to an assembly line you still have more freedom, you have more uncertainty and you can’t really use the classical robot, industrial robot, and classical programming for handling these kinds of topics. Compared to a household or the outer world, you still have more structure that you could exploit. You have numbered shelves. You have trained people. You have many very regular structures in the warehouses. That’s, of course, very good for robots to be applied. I would expect that this is one of the domains where robotics will really take off in the next years.
Ron: Just to the side, is your company also looking at the capability of doing audits in warehouse from a robotics point of view or is that still a human task?
Dr. Tenorth: You mean like inventory?
Ron: Yes, yes.
Dr. Tenorth:  Yes, actually that’s another potential saving aspect. The robot often have to do a kind of inventory for themselves in order to find the objects more easily. By scanning a compartment we, of course, also scan those objects that are not the one we’re about to pick. We do have pretty good inventory information and can also improve the data that is describing the objects, which is often pretty bad in today’s databases. Companies don’t really have good information of how large the object actually are, what their shape is and so on. If you can automate these measurements by the robots, that’s very valuable information for computing the sizes of shipping boxes and so on.
Ron: Finally, what is the job prospects in robotics, as far as you can see? Is this a growing area for companies like the one that you work for? Are there jobs?
Dr. Tenorth: Jobs, yes. We are constantly hiring. We are searching for people in many areas, from research and development to QA, deployment, both electrical engineering, mechanical engineering, software. I think there are many other startup companies that are founded or some larger companies start to work on autonomous robots. Especially in logistics, many of the larger forklift or material handling companies also started departments working on similar aspects, so I think the job prospects from an applicant’s point of view are very good.
Ron: Okay Moritz, I would like to finish by thanking you for your time and we’d like to talk to you in the future as the technology improves.
Dr. Tenorth: Yeah. Thanks a lot for the interview. Yeah, I’d be happy to catch up in the future.
Jana Witt: That’s the end of this episode. We hope you enjoyed our visit to Germany today. You can check out lots more information and all our past episodes at robohub.org. We’ll be back with another episode in two weeks time. Until then, goodbye.
TORU with robots, the podcast for news and views on robotics.

 

All audio interviews are transcribed and edited for clarity with great care, however, we cannot assume responsibility for their accuracy.


comments powered by Disqus


follow Robots Podcast:

Recent episodes: