In this episode of Manufacturing Matters, Gerard Andrews, product marketing manager for robotics at NVIDIA, discusses how AI will be the key enabling technology for robot autonomy. A critical piece to achieving this goal is the NVIDIA Isaac robotics platform, an end-to-end solution for the development, training, testing, simulation, and deployment of AI-based robots designed to simplify development and accelerate time to market for customers. NVIDIA also uses an AI tool called reinforced learning to significantly reduce the trial and training time of a specific task, such as the dexterous manipulation of a robot hand the company has recently demonstrated. Although reinforced learning is primarily used in research, industrial applications are starting to deploy the technique — expanding the number of tasks and markets that will take advantage of these fine motor skills.
Jimmy Carroll: [00:00:06] Hi, everybody. My name is Jimmy Carroll. I’m the vice president of operations at Tech B2B Marketing. I’m here today at the A3 Business Forum in Orlando, and I have the pleasure of being joined by Gerard Andrews of Nvidia. Now I think most of our listeners or viewers know who Nvidia is, so you don’t really need an introduction. But Gerard, if you could talk about what you do there, that’d be helpful.
Gerard Andrews: [00:00:26] Great. I appreciate that. So again, I am Gerard Andrews from Nvidia. I handle product marketing for our robotics effort. We have a platform called Isaac, which is an end-to-end platform for the development, training, testing, and deployment of AI-based robots. So a lot of fun to work with all these companies that are attending a conference like A3, that are trying to accelerate automation, and we’re excited to bring our platform to help move that mission forward.
Jimmy Carroll: [00:00:55] Yeah, that’s great. So when I think about the intersection of AI and robots, I saw a really good keynote earlier from Anthony Jules at Robust AI. So when you think about not just at Nvidia, of course, but when you think about the intersection of AI and robots, how does AI open up new opportunities in robotics?
Gerard Andrews: [00:01:12] Well, we have a saying at Nvidia that everything that moves will be autonomous, and AI is going to be the key enabling technology for autonomy. And what we mean by autonomy is where robots can have a better awareness of their surroundings and have more intelligence in their decision making as they are deployed to do numerous automation tasks. And so AI is again that key enabler for that autonomy which we are all chasing.
Jimmy Carroll: [00:01:44] It would be a miss for me if I didn’t talk about processing power in the context of Nvidia. So Gerard, I saw you speak at the AI and Smart Automation Conference, the A3 event earlier this year in Columbus, and you talked about something with the simulation, 42 years of simulation in one day. What kind of new applications does that open up? What’s that really mean for somebody?
Gerard Andrews: [00:02:06] I’m glad you brought that one up. In that talk I was referencing some work that we have done around reinforcement learning, which is a technique under the AI umbrella. And we are using reinforcement learning and accelerated computing to accelerate the training processes far more that you can do on conventional CPU-type architectures. And this allows, in that one case, a significant reduction in the trials, in the training time to train that robotic hand to do that specific task. Now reinforcement learning right now is very heavily being used in the research side of the robotics. But we are starting to see some industrial applications take advantage of reinforcement learning for robot locomotion and for more complex tasks like we demonstrated there, which was dexterous manipulation of a hand. And so it’s really exciting to watch the transition of how these things will move from the research into the applications.
Jimmy Carroll: [00:03:14] Yeah, Gerard, I’m glad you mentioned applications. I was wanting to ask you about, with this capability to do that kind of simulation with this processing power, what are some applications in the industrial world that can benefit the end user?
Gerard Andrews: [00:03:27] Okay. Really appreciate that question. We’re already seeing people leveraging reinforcement, learning to train locomotion policies. And what I mean by locomotion is how the robot, like a quadruped, can get around and move in different environments on different surfaces. But I think the future is using reinforcement learning to train fine motor skill tasks like the dexterous hand demonstration that we did. And what this is going to open up is the amount of tasks that robots can do in the home, in retail, in health care applications that are different than today, because a lot of robots today are deployed doing fairly simple gripper applications. But once you can do these finer motor skills, it opens up whole new applications for robotics, and I believe reinforcement learning and the type of tools that we are putting out there will help advance these use cases for robotics.
Jimmy Carroll: [00:04:25] Yeah. So even in terms of some of the well-defined applications in the industrial space, like pick-and-place or bin picking, it sounds like this kind of simulation will also be to benefit those, enhance their existing capabilities. Would you say that’s the case too?
Gerard Andrews: [00:04:40] Absolutely. We believe that simulation has tremendous amounts of value for the robot developer and the robot user. The developer can test and train all of the software that’s going in these more complicated, more intelligent robotic platforms. And you want to test these things out in simulation, and especially when you’re dealing with corner cases. These robots are going to be working more closely and closely with their human partners. And so you want to test out those interactions, validate your safety systems, and continuously, constantly test your software.
Jimmy Carroll: [00:05:16] So staying in the industrial space, what are you most excited about in automation? Not just in terms of the intersection of AI and robotics but the overall space?
Gerard Andrews: [00:05:27] Yeah, I think we’ve heard it in many of the addresses that we’ve had so far, that the macro trends for supply chain disruption and labor shortages are really accelerating, which was already happening, which was the embrace of automation and how reshoring and bringing these jobs and manufacturing back to the home countries where the products will be used. So it has many benefits for that, I think for society. And what’s gratifying for me personally is to be working in a space to enable that because automation is going to be critical to executing on this vision that so many of the companies here at A3 and outside of A3 share for our future.
Jimmy Carroll: [00:06:19] Absolutely. I mean Alan Beaulieu and his global economic outlook said automation will keep the economy afloat during hard times, and that’s really telling of where we’re at in the automation space. So I think what you said really reflects that nicely. Gerard, what else haven’t I asked about? What are you excited about at Nvidia and what are you excited about in the automation space and machine vision and robotics?
Gerard Andrews: [00:06:44] From the Nvidia perspective, like I mentioned earlier, we’re building this end-to-end platform with the idea that AI is going to play an important role in the future of robotics. And what excites me the most is once we put the tools in the hands of our partners, the types of things that they do, the type of solutions that they are working on for their particular markets. And so that’s the most fun part after meeting with someone, giving them the tools, doing some meetings, and they come back and show you how they are using these applications.
Jimmy Carroll: [00:07:21] Yeah, that’s exactly right. When I think about Nvidia, I’ve followed that company for years, and it’s great to get the chance to talk to somebody that works for the company and talk about how they work with their partners and how they’re innovating to open up these new applications. And it’s very interesting. So it’s really been a pleasure to speak to you. I appreciate it. Again, my name is Jimmy Carroll. This is the Manufacturing Matters podcast. Gerard, thank you so much.
Gerard Andrews: [00:07:48] Thanks for having me.