Advanced robotics and AI in industrial manufacturing- Episode 1 transcript

Chris Pennington: Hi everyone. Welcome back to the Siemens Digital Transformation Podcast series. I’m your host, Chris Pennington, global industry marketing leader for industrial machinery at Siemens Digital Industries software. This is the first episode of 2025 and will shift into the topic of advanced robotics in industrial manufacturing.

As usual, I’m joined by our industry expert Rahul Garg, Vice President for industrial machinery vertical software strategy at Siemens Digital Industries software. In addition, we’ll be speaking with four more experienced experts knowledgeable in automation and robotics. Thanks for joining us today, gentlemen. Could you each give us a brief introduction?

Maximilian Metzner: Hello, my name is Maximilian. I’m a robotics expert in our Siemens factory in Erlangen and I’m also leading the module autonomous manufacturing systems for the plant network in Siemens Digital Industries.

Etienne Ferre: Hello, my name is Etienne Ferre. I’m managing the research and development at Kineo, a specialist in detection, flexible cable simulation, and software industry.

Jens Meckel: My name is Jens Meckel. I’m head of business line advanced automation and robotics and responsible for the product semantic robot library as well as Pick AI.

Chris Pennington: It’s great to have such a wealth of experience in the room. I can see that we’re going to have some interesting conversations ahead. I’d like to begin with a question to Rahul and Max. Over the last few years, we’ve seen a revolution and transformation in the field of automation. How have you seen recent advancements in automation impacting industrial manufacturing?

Rahul Garg: Thanks Chris. It’s great to be on this topic of advanced robotics and manufacturing. When I look at what’s been going on over the last couple of years, one of the biggest transformations that’s come about is the deduction of robotics, obviously, but more in the context of AI as well. I think the introductions of AI and robotics and automation, we are now getting into a juncture of some new innovations that, frankly speaking, were not there as much in the past.

The ability to automate and drive a more adoption of robots and automation is gonna become a bigger reality.  One of the main areas that I see this impacting is when you think about automation. Typically speaking, it’s for high volume production, whereas by introducing robots we add a lot more flexibility into the production process.  bringing that that together, we will be able to go towards a more flexible production process of catering to a more personalized needs of production processes as well.

Maximilian Metzner:  Yes. This is actually something we’ve experienced first-hand, especially picking up what Rahul said, combining automation with flexibility. This is, in my opinion, a paradigm shift that’s taken place over the last years and this is something I think that enables automation in the 1st place to happen in many industries right now this has something to do with the advent of cobots.

The ability to more closely integrate automated systems with manual systems so you don’t just have the binary choice between a fully automated system which is just not feasible for many industries or many production sites, and a manual system, you can have the best of both basically. But with it came the lowering of entry barrier.

Today, automation systems and robot systems are relatively easy to use. You don’t have to be an expert before you start. You can start simple and then grow your experience while you’re already doing value adding stuff.  Obviously AI is a big topic because it allows us to enhance the capability or the level of autonomy of automation or robot systems and basically enabling them to do more tasks more flexibly that they might have not been able to do. I don’t know five years ago if they did at a much higher price than what is possible today.

Rahul Garg: I think one of the things you just mentioned in a very brief word Max was around simulation and digital twins. I think that’s another very important capability that’s gonna drive much more rapid adoption of advanced robotics and advanced automation in into the manufacturing processes as we see here.

Chris Pennington:  Rahul actually mentioned artificial intelligence there, which of course is pervasive in in every meeting or every show that you go to these days. Max, could you maybe talk about the exact capabilities that you’re referencing when you talk about AI and automation?

Maximilian Metzner: Obviously I’m coming from a certain viewpoint here, looking at our factories and our automation journey. It offers a lot of different skills and capabilities, and some might be more relevant to some industries. But for example, if you’re looking at electronics manufacturing, you will have basically the vast majority of tasks being assembly tasks and in assembly the process relatively easy.

If you basically find the part that needs to be assembled, you pick it, you go over to the assembly where you would want to assemble it, and then you place it. This is where AI can, for example, help you with finding the part, identifying the part, and determining how you can grasp it, which is something that’s very natural to a human operator. You don’t have to program anything; you just give him a box of the parts that you want assembled and they will be able to work with it for a robot that’s a completely different task.

Back in the day, you would have to come up with a feeding mechanism or presort parts which then again is a manual effort. Or whatever you had to do to give it to the robot in an extremely defined way. The robot didn’t have to do any thinking, basically.  today that, for example, has changed. So, we see a lot of bin picking systems that are now out in the market.

It’s also a technology that we very much use in production today because when you automate something that was done manually before most of the time you will have to do some sort of bin picking to be able to just grab the parts the next step would be to insert or assemble the part itself. Here the human also makes use of their integrated skill set so to say.

No human is able to usually assemble stuff without the parts touching before, and that’s also not required. But for robots, we kind of always assume that to be the case; but it doesn’t have to be. Most cobots at least that you can buy today have some sort of sensor equipment outfitted.  Making use of that by for example AI algorithms helps you to get a yeah, more human like haptic tactile interaction feedback that allows you to also automate assembly tasks that you wouldn’t have done three years ago.

Chris Pennington:  Yeah, that’s a great use of AI. If we look at Erlangen, you’re obviously making a lot more use of software around automation than you have previously. Could you give us some ideas on the sort of gains that you’ve experienced from implementing software to support automation capabilities?

Maximilian Metzner: Yes. One big factor which I think is also true for a lot of the industry is most of the time you don’t start on a greenfield, you have an existing production which often times is manual and you want to do some automation here. Software for planning is absolutely crucial because you have a running process if you just guess how you could do the automation and then try to implement it.

You will definitely limit your production while you’re integrating, which could then probably take months to do, which is something you really cannot afford and which will definitely eat up your automation business case.  Here for example, it doesn’t really matter if you do like classical automation or advanced automation. You have to get the basics right.

The robot has to be able to reach everything and do the motion in between the things. You have to know how long it will take and you have to know how the entire system behavior will change. After the automation you still meet your goals of output, throughput, time, and all that.  this for example is basically impossible to do without planning and simulation software that we used literally on every system we integrate.

Another thing is obviously with AI you also get a certain loss of determinism. A classic automation system will basically do the same thing in the same way again and again and again. This is not true for automations or for AI based automation systems. Depending, for example, the part you want to pick its orientation, the robot will do a different motion and you have to make sure that for every possible motion.

That the robot needs to be doing it doesn’t crash into anything, or at least the plans around it or something like that to make sure that your design will be able to clear most of the bin during bin picking. This is something you cannot really do without software and planning and so on.

The last thing maybe I want to touch on is not just getting the basics and the concept right, but also the implementation itself. If you need to integrate for example brownfield equipment, if you need to integrate machines that you already have that you will not just purchase a second time just for the sake of automating them, it is crucial to also test beforehand how does the signal handling work. How do you integrate them? What is the concept?

Here again simulation, virtual commissioning, and robot offline programming can give you an edge to really get into a functioning automation system in a brownfield environment very quickly within a matter of days, worst case maybe months of fine tuning and trial and error. In a productive production line, which is basically the worst-case scenario and will probably stop many people from even trying automation.

Rahul Garg: Max, one of the things that will be helpful for everyone to know is what does Siemens build at the Erlangen factory? Maybe you can just introduce the factory a little bit for people who do not know what we do here.

Maximilian Metzner: The Erlangen factory is basically a power electronics and control electronics factory. We do electronics production, but it has very little to do with consumer electronics or your smartphone that you probably have or anything like that. But it’s electronics that actually drive motion. Control machines could be the driver for a conveyor belt. It could be for a machine tool system access. It could be for a harbor crane, or it could be for a pipeline pump.

Basically, the spread or the variety of products is huge, and it doesn’t boil down to software variants, but actually everything is in hardware. One thing we always have to keep in mind is the flexibility that we need to produce what the customer actually wants, because we are a made to order business and we produce based on a customer order. We have to be able to meet that and also make sure that we can produce economically or with a good efficiency and because electronics production is a very sought-after market.

Rahul Garg: To your last few points, you obviously have to be very flexible in your production process being able to meet varying customer needs and demands. Could you give us a sense of how many different types of products do we produce in a year through this factory?

Maximilian Metzner: Yes. So, we’re talking about a number of variants, more than 1000 of which are hardware variant. We don’t produce the same hardware twice and then flash a different software on it. Actually, it’s a different type of PCB and that basically makes it so hard to automate while keeping up the flexibility that you usually would get from manual labor.

Rahul Garg: That’s great. And keeping that view of the flexibility that you need to be
driving in your facilities was that one of the key drivers to start adopting the robots and cobots in the factory?

Maximilian Metzner: Yes, obviously that was one of the major enablers to get started in the first place. If I can give you a little bit of history, if you would have visited our site eight years ago, you wouldn’t have seen a single robot or any automation in like the final assembly or the through whole technology part of our production simply because we needed the flexibility and back in the days. Automation just couldn’t provide that at a reasonable cost. It was never really an option for us.

As I said, there were basically two major enablers that we had that really helped us get started in automation. One was the elimination of that hard barrier between automated and manual. Basically, the advent of cobots and easy to use type of automation. And the second one was then another booster maybe about five years ago: the integration of AI into automation. It helped you automate things that were basically only possible for the human before that.

Rahul Garg: It’s interesting. You mentioned that how your production facility was a few years ago to today. One of the questions we get many times when we are talking to customers and people is: how do you even start on a journey like this? Maybe you can shed some light for someone who’s starting new on this journey of looking to use robots in their manufacturing facility. What are some two or three things you could give them as thoughts or as an input out as advice?

Maximilian Metzner: I think the biggest advice I would give is to start to actually do it. Back in the day, there were also a lot of people in our facility that really doubted if there was any potential at all that could be really exploited with automation. And today, eight years later, we have more than 100 automation systems. If you count our autonomous mobile robots, it’s closer to 140 systems. In production today, I would say 95% of the robot stations we build ourselves. We have in house engineering, set up, and implementation. It really helped us increase our edge in competitiveness quite a lot.

I think technology today is at a point where I would say probably 90% of the people that could be using automation today just don’t start because they have the feeling it’s too much of an effort. It’s too prohibitive to hire people that can automate the process and that can operate a robot and stuff like that. But I don’t think that’s really the case anymore today.

Without going into specific brands, you have a lot of systems that are explicitly made to be easy to operate. To be operatable for somebody who hasn’t worked with a robot before. For example, you have sensors like smart cameras which are made for people who don’t have a background in computer vision and that are also now easy to integrate. So, a lot of automation providers, including Siemens, have an ecosystem of like things that just fit that. You can combine and pick and choose what you might need for your application.

And last of all, for a lot of like standard applications that you might have in your factories like palletizing, depalletizing, welding, gluing—all that kind of stuff, there’s ready to use solutions out there. There are marketplaces where you can basically buy an off the shelf kind of system and it won’t cost you $1,000,000, it’s relatively easy and cheap to integrate if you account for what it can actually save you in manual labor cost.

Rahul Garg: Yeah, that’s a good point. Because that’s another fear people have about starting. The cost of these systems could be very high and that becomes a barrier in their mind, and they think that hey, you know if the effort involved in the cost involved will be so high. So, let me continue doing the way I’ve been doing it for the last many years. And I think what you’re highlighting is that’s not the case.

Maximilian Metzner: Yes, I definitely think it’s not the case. I don’t say every industry is fully automatable. Our facility isn’t fully automatable as well, but that doesn’t mean that there aren’t use cases or there’s not plenty of use cases that it would make sense.

And as I said, the entry barriers today are very low, and I think the pressure to do it is much higher today because continue doing it the way you’ve always done it with manual labor will be not as good an option tomorrow as it was maybe yesterday. Because of demographic change and everything, it will be always harder to find new people to do that work. And let’s face it, the work that we’re talking about for automation is usually the repetitive tasks. Dull tasks the tasks nobody really wants to do. I think it will get harder and harder to even find somebody to do that. Here, automation can also be leveraged to actually reduce the need for people to do those jobs.