Skip to main content

In Between The Tech - Industry 5.0

Published Date

A Conversation with Larry Sweet of Advanced Robotics for Manufacturing Institute

Mark Patrick
Welcome to In Between the Tech. Today we're wrapping up Mouser’s look at Industry 5 as we sit down with Larry Sweet, Director of Engineering at the Advanced Robotics Manufacturing Institute, or A-R-M, to hear his thoughts on the lasting impact behind this next industrial revolution. Larry shares the opportunities for engineers within this space and how Industry 5 is redefining the way in which humans and robots work together.

Host
Larry, thank you for joining us today, please tell us a little about yourself and the Advanced Robotics for Manufacturing Institute.

Larry Sweet 
I'm Larry Sweet. I'm the director of engineering at the Advanced Robotics for Manufacturing Institute, otherwise known as ARM, A-R-M. And it was established about seven years ago. It’s one of the 17, so-called Manufacturing Innovation Institutes, MMIs, that were established maybe like 12 years ago, the first of which was called America Makes, which was focused on additive manufacturing. The other 17 are all focused on different disciplines. So in our case, we're focused obviously on robotic manufacturing, which is very interesting because it really enables us to be involved with it. Things that are very wide diversity from automotive, textile automation, obviously a lot of DOD, defense and aerospace type things.
We basically do two things. On the one hand, we sponsor now well over a hundred projects, technical projects to advance the state-of-the-art of Manufacturing Automation with robots going beyond what's available commercially. So we're breaking new ground either in terms of a new manufacturing process, but in some cases, let's say with hypersonics, it's enabling the manufacturer of totally new materials that don't exist. So the new materials, new products and new manufacturing processes go hand in hand. We are not doing basic research. We take things that have already been proven in a laboratory and our focus is how do you industrialize them and make them available to both large companies and small companies.
The second aspect, which is very important is workforce development. I remember about six years ago; I was involved with a presentation sponsored by the American Society of Engineers to members of Congress. And the main topic back then was, are robots going to take away jobs? And now, with maybe 600,000 unfilled manufacturing jobs in the US right now. No one's really asking that question or at least not quite as much. 
I'd say every company that I talk to, whether it's a small business or a major manufacturer, they're all saying the number one priority right now is workforce. They cannot keep up with the demand. So actually it's a good problem to solve. And so, the workforce development part is something that ARM is taking the leadership role for the country in. 

Host
As the former Head of Engineering at Frito-Lay, can you tell us how the workforce adapted to advancing technology?

Larry Sweet
I was recruited from outside the food industry. So it was interesting to learn how do you make a consistent product like a potato chip when you have 1,800 different varieties of potatoes coming in and how do you make that product consistent? And the new technology that we developed was able to increase the capacity of our factories. We added 800 million pounds of capacity to the system by developing three new technologies that unleashed basically the productivity. A lot of our automation was underutilized and we had to understand why that was happening and then what to do about it. And we developed it, we piloted it, we had a lot of involvement of the workers engaged as we got it closer to going into production, they made some critical contributions.
So, when we had thought we had the technology working, we brought the workers in to operate the equipment as if it was production. And they make really important observations about the user interface, how to set it up in a way that people could understand, but also there's a lot of know-how that is involved in any manufacturing process. It's on the factory floor that it may not be written down. It's sort of handed down generation to generation of the workers, the employees, in the factory. And so they taught us things that we didn't know. And so, by the time we were ready to take the technology into the plant, it was more mature, more ready to go. The other thing is because they had been involved in it, they became our ambassadors, to the factory workers in saying, this is okay, this is not something that's scary. And that acceptance by an experienced workforce was very important. Not only did it get the business productivity savings, we got extra margin growth. And so actually that turned out to be a bonus that we didn't anticipate. So, the workers were very helpful to us. 

Host
How as Industry 4.0 effected a company's operations and manufacturing?

Larry Sweet 
Manufacturing 4.0 has actually been implemented in a very positive way. It's been primarily successful in big companies. So, I'll give you an example that I was at a conference last fall with a gentleman who was head in charge of all machine learning applications at Proctor & Gamble. And he said that there were, I think 90 applications around the world, where they were using machine vision with image recognition and machine learning to detect the defects on products that they were producing and using that as a reliable way with high confidence to identify defects before they got to the customer.
So just as a simple example, you think about the label on the back of a Whisk bottle that it wasn't properly applied, there was a wrinkle in that, that made the barcode unreadable or very small print might make it difficult for a consumer to read the instructions or other information. So that would be something that they were able to train their technology to be able to recognize and then pull that item off. It wouldn't be shipped to the customer. So I was surprised that there were the high number of applications that had been developed and the investment that that company had made in that kind of technology. So that's an example of 4.0. And now if you say, okay, how many companies have actually implemented something like that? It's really relatively few. There are a lot of companies that are advertising capabilities. The primary limitation that they have today is the availability of data that can be used to train the machine learning models to understand. So just typically the way machine learning is developed is you start off with something called a training set. You take a certain amount of data that's for training purposes, and then you have a separate set of data that's called your validation set. The idea here is you're taking the model that you validated during the training and you try it out on this separate set of data to see how well did it actually work. So it's like a blind test. And what happens is that often the first times you try that with a certain particular algorithm, machine learning algorithm, it often doesn't work. And so you have to go back and modify the training algorithm to make sure until you get to the point where the confidence level is sufficiently high. And so that would be a good example. 
So, if you don't have access to enough data, then you really aren't sure whether your algorithm is going to be successful. So if you go to conferences now, they tell you one of the biggest constraints to making more use of machine learning is data. And if you have a company like Proctor & Gamble where you have millions of products, the production is in the millions. So you have access to a huge amount of data. But if you're making something in much smaller quantities, then you may not have enough data. You may try to use the data from everything you're building, but if the number of units that you're making is too small, then you don't have enough data. 
One of the projects at the ARM Institute that we rewarded was something called the AI Data Foundry. And the idea was to be able to collect data from different sources and make that available to some of the smaller companies so it could get the benefit from machine learning, as an example. 

Host
With Industry 4.0 still being implemented, do you think some companies will forego that iteration and skip to 5.0?

Larry Sweet 
I think that would be risky. So let me go back and actually even at the 4.0 level, one of the things that I think most experienced automation robotic and automation practitioners would say, the first thing you should do on your factory is, if you have an area that's a problem area, do what you can using lean principles, six sigma type of approaches, that kind of thing. Clean up your process before you automate. If you go in and you try to automate a process that's a mess, automation is probably going to make it worse. And so basically, even though sometimes they may have an interest in wanting to rush the automation, what they're going to hear from the system integrators and the robot companies or any conference that you go to, you're going to hear that message.
4.0 sets the foundation of how to use the data effectively and how to successfully automate a process when you get to 5.0, the big difference is this is going to involve much more interaction with a human being. I'd say it's a two-way interaction. So let me give you an example, Amazon is very dependent on barcodes. So when I first got to Amazon, they had a system where, let's say if you're picking an order, the human being who's picking the order out of the yellow pods has no information about what the order is. What they're doing is the cloud software is saying you've picked three items, three different items, somewhere out in that building. There might be well in the number of different items of different skews, different distinct products, well it's going to be in the millions.
And the number of items in there is going to be in the tens or hundreds of millions in that building. And so, what the software is doing is it wants to identify where all the choices of the three things that you've ordered and it's going to identify which pod has each of those items. So, you can have three different robots at the same time go find those pods and bring them over to three different picking stations. You'll have three different people pick those items and more or less at the same time. And then they're going to be put in yellow totes and sent down to the packing area where they'll be combined into the single order. So that's basically how the process works. Every step that I described was barcoded. So, if you're an operator at a station, the computer screen comes up and it says, okay, here's the next pallet, next pod is coming in. 
What I described was the process. So, the challenge they have there's a lot of barcoding going on. And so they tried to figure out is there a way to eliminate all the barcoding? And so they developed a system. I can't tell you how it worked as that would be getting into proprietary area, but actually they used basically industry 5.0 approach to try and understand based on observing the operators, how they could increase the confidence, by not eliminating all the barcode, but eliminating some of the barcodes. And so it was less movement for the operator. It took less time, it was easier. It took a long time to develop it, to validate this, but because they were saving time, that meant the operators could do more picks per hour. And so you get a productivity benefit and you get hopefully an ergonomic benefit for the operator. And so, it's not just a machine anymore. The operator is part of a combined interactive system between the two. That's a pretty good example of 5.0. 

Host
What’s beyond Industry 5.0?

Larry Sweet 
Well, I think a robot that would be more intuitive and easier to use, to teach. So the cobot had limitations on the force and the speed that it could operate when a human is around, you could actually put it into a mode where you could grab a hold of the robot and guide it around physically say, I want you to go over here and pick something up. I want you to take it over here and drop it off, or I want you to weld something, so I'm going to say, here's the path that I want you to operate on. And so, the ability to make it more intuitive for people that weren't experienced in robot programming was an advantage of the cobot.
The other thing is that the idea was that you could operate this in an environment where the robot, you don't have to enclose it in a cage. And because that was always a constraint, if the robot ever got in trouble, say picking up an item that dropped it, you would have to go through a procedure to lock out, tag out to allow the operator to go in to make it safe with a conventional industrial robot, the process of going in, fixing the problem, coming back out, telling the robot, okay, the person you're now outside the cage, that might take a minute or two to do that, that's all lost production. So, the idea is if you have a so-called cobot that you could eliminate that as long as it had this force limiting, velocity limiting characteristic. Now the existing robot industry from the major market industrial robot leaders, they countered that by saying, look, I can put LIDAR or other kind of barcode, but other laser sensors to sense the approach of a human being.
And so, when there's no human being around, it can operate at higher speeds. And then when a human being gets closer, it can slow down. And if the human being actually gets literally within the reach of that robot, it'll stop it. The industry is going through an inflection point right now, and the agencies like the Association for Advancing Automation, A3, which is basically the industry group in the US that oversees the robot industry, the motion control, the computer vision industry.
They've teamed up with ISO, the International Standards Organization, and essentially, they want to get away from using the word cobot. What they talk about is they talk about the robot that has the characteristics of being force limiting and velocity limiting, which was kind of the original idea of the universal robot, but that also includes a robot that has those proximity sensors around it. So what they're saying is if you take the robot and now you combine that robot now with the tool on the end, not just the robot, but you might have a welder or gripper or something like that. Now you have a system and then you put that into a factory in a robot cell, that becomes an application. And so what they talk about, it's a collaborative application and it can be implemented in different ways than people anticipated.
And what's most important is that the entities that are responsible for assuring the safety is different for each of those. So you have a robot system, you have a robot, you have a robot system, you have a robot application. So the robot vendor has responsibility for the force and velocity that the robot is going to move. The system integrator is the one that puts a tool on the end of it. And so it might be that it's not the robot that hits you, but it's the tool that hits you and maybe there's a sharp edge on that tool. And so you care about that. So the system integrator's responsible for that, and then the end user, the manufacturer, they're the ones who have to do a risk assessment. This now takes into account the whole work environment, how the space is organized, how the person is with the skill and experience of their workforce, et cetera.
What's next as we get into 5.0. The way people have been using cobots up until now is basically having the cobot doing the job the industrial robot was doing, and the human is basically standing back and maybe feeding a part and the robot can come pick it up. But it was really not a lot of cooperation going on, even though they called it a cooperative robot. It wasn't cooperative. And I think this is getting into, let's say the 6.0 is where human and robots are actually sharing tasks that neither of them can do it on their own. And the human can understand the intentions of the robot. The robot can understand the intentions of the human.
If you're going to interact with a robot and you know exactly what it's going to do, then you can trust that. But now if you're on the other side, the robot is trying to understand what you're trying to do, how much trust do you have that the robot really knows? And so, I think that's going beyond 5.0. 

Mark Patrick
We hope you enjoyed this episode of In Between the Tech. This podcast is part of Mouser’s in-depth look at Industry 5 and how humans and robots can work together more efficiently. Explore the entire Empowering Innovation Together offering on this subject with articles, videos, and more at mouser.com/empowering-innovation.