How Rockwell is bundling AI into robotics, with Nvidia’s help

Rockwell Automation recently announced a partnership with Nvidia for the development of industrial robots with artificial intelligence (AI) processing capacity.

The company will provide autonomous mobile robots and production automation applications to industrial customers seeking greater efficiency in factories through the integration of Nvidia's Isaac robotics platform.

The companies will also integrate Nvidia's Omniverse Cloud application programming interfaces (APIs) with Rockwell's Emulate3D software to create digital twins.

In this interview, Renato Luciano dos Santos, Latin America commercial manager for Rockwell Automation, talks about this new venture in advanced robotics.

BNamericas: How did this agreement with Nvidia come about and what do you hope to achieve with it?

Santos: Since the beginning of the 1980s, Rockwell has been operating in the mechatronics area. The 'marriage' of the electrical front with the mechanical front gave rise to a child, which is robots.

Since then, all of this has evolved, from mechanical to electronics, from analog to digital, until arriving at the industry 4.0 movement in the early 2020s. Digitization, cloud, reduction of dependence on people in processes. I would say we’re at the end of the cycle of this first industry 4.0 movement, but AI has come running over that.

BNamericas: In what sense?

Santos: What we expected to happen in the 2030s is happening now, which is autonomous manufacturing – manufacturing where you no longer need to program, but to configure. Telling a generative AI what it needs to do for us.

This is now coming to the factory floor, with robots and systems with autonomous operation capability.

In October last year, we acquired Canadian firm Clearpath Robotics and its Otto Motors robots.

They are autonomous machines, which have full decision-making capacity, what to do, where to go. We’re in the process of integrating this company, but since then we’ve seen absurd demand in the market for this type of technology.

BNamericas: For what, for example?

Santos: Projects to optimize distribution centers, such as material movement on the factory floor. Also mission-critical process projects. For example, for safety inspections of a boiler, or of dangerous chemicals.

This robot has this ability to decide which items it needs to inspect and, when it detects an anomaly, what the next step is; whether it’s necessary to do a 'drill down', to deepen the analysis, for example.

Most of the cases we work on are under contractual confidentiality. Companies see this as a business differentiator and don’t allow us to disclose it.

But there's a project I'm working on right now in Brazil with a chemical company. It’s to operate a robot for when there are leaks of these products, which are corrosive and alkaline and can cause risk to a human operator carrying out this inspection.

The robot enters the area, knows the route it needs to take. If it encounters an obstruction, it can avoid it and when it arrives at the location it can make a decision on what to do.

BNamericas: And what about the partnership with Nvidia?

Santos: It largely bolsters this. We need to increase the processing capacity of these robots, so that they’re more efficient and more autonomous. This is mission-critical for inspections.

But there’s also use for material movement, for example in a car assembly plant. Today, the possibilities are gigantic.

BNamericas: What does the partnership entail?

Santos: Support, material, consultancy, solutions. At Rockwell, in addition to this line of intelligent devices, which includes Otto's operations, we have another arm which is a tool called Emulate3D. It’s basically a tool for developing digital twins. 

The robot area with the digital twins area is a perfect marriage, because I can simulate in the virtual replica what and how the robot will perform in the real environment.

The partnership with Nvidia tackles these two pillars. With robots, we’re partnering specifically for an Nvidia platform called Isaac, which is focused on neural processing.

Another Nvidia platform we’re using is Nvidia Omniserve Cloud. It’s a platform for cloud data management. We will integrate it with our Emulate3D digital twin tool.

BNamericas: How does Emulate3D currently work?

Santos: Today Emulate runs in local processing, on-premise. The idea is to migrate this to a cloud environment, without any dependence on local processing so that we can work on the platform independently of local hardware.

BNamericas: Is the idea to migrate all customers who use Emulate on-premise to the cloud?

Santos: We like to give options to our customers. The idea is to offer both possibilities. Existing robots can benefit from this type of processing.

For new platforms, the idea is that we can ship more processing capacity in a hybrid way. If there’s network access, they run these processing resources in the cloud. If not, they make the decision locally.

The truth is that some Emulate3D customers don’t want to do this processing in the cloud. Especially when the cloud isn't theirs. Many of these customers use this with either a private cloud or local processing.

BNamericas: And what is the public cloud behind it?

Santos: We don't reveal it, but we have a partnership with a large provider to create our cloud, which is hosted in the public system. 

But we can create a private cloud for the customer, and we provide the infrastructure for that. They can have part of the cloud solutions running locally at their plant and part in our cloud. Or they can have everything on-premise at their plant.

I would say that 80% of our customers use local cloud and 20% use hybrid cloud – on-premise backed by a public cloud.

BNamericas: What sectors and companies are you prioritizing with this partnership? And are there other cases of digital twins in the region?

Santos: We have a case that we’re just handing over. The OEM [original equipment manufacturer, contracted for manufacturing] is building a machine based on our ICT platform, Independent Cart Technology, which replaces mechanical systems with linear motors of very high precision and speed.

This new machine has twice the production capacity of a traditional machine. The end-customer is removing two packaging machines and putting one in their place, which will produce the same as the previous two machines, and with more flexibility, because it can do different types of packaging.

BNamericas: But where does the digital twin come in? In the manufacturing process of this machine?

Santos: That's right. This was a machine that, when the end-customer asked the OEM, and passed on the characteristics, in the traditional world, the OEM would take between three and four years to develop it. Using the digital twin, they take eight months to develop this machine.

They’re building the mechanical part of the machine now, but thanks to the digital twin, the development time has been reduced. All the programming part is already ready. Tests have already been done. 

The end-customer is also being trained on how to use this machine – on the digital twin.

Today, machine manufacturers have benefited greatly from this technology. The automotive industry as well. For several large automakers in Latin America, as a rule, the assembly line has had prototypes in digital twins. We’re currently carrying out a digital simulation project with an automaker in the region for a car model that will be launched in 2026.

In terms of robotics, autonomous robots, we have projects in sectors such as e-commerce, the petrochemical industry, the mining industry. There’s a large company in the sector in Brazil using autonomous robots to inspect dangerous areas. The same as in the petrochemical sector.

BNamericas: In e-commerce, MercadoLibre recently announced a major investment in autonomous robots, but the supplier is a Chinese company.

Santos: Yes. We’re working with the US company that’s their big competitor in the region. It’s possibly our biggest client today, globally speaking.

When we talk about generative AI, specifically, in line with our processing partnership with Nvidia, our main programming tool has CoPilot embedded in it. And we will increasingly see this embedded in Rockwell products, not only for programming, but mainly for the maintenance area. To identify where a problem is and the possible failure.

BNamericas: How is the manufacturing and delivery of these products done in Latin America?

Santos: In the case of Otto, they’re headquartered in Canada. 

What we’re doing is increasing production capacity. We have Otto assembly lines in Europe and we’re creating new lines in Mexico. Our idea is to use our global network of authorized distributors for each region who resell and deliver the products.

In terms of software, the development is centralized in the United States.