Exploring the future of humanoid robotics with Novanta

Novanta Robotics and Automation has been a leader in motion control solutions since 2022, partnering with top robotics platforms across various industries. Novanta offers drives and coders, motors, and force torque sensors to reduce your risk and costs, helping you get to market faster. Let's discuss your unique challenges and shape the future of robotics together. Visit solaramotion.com to learn more. Well, on the podcast today we're excited to have Nick Damiano. He's the senior business development manager at Novanta Robotics and Automation. Nick, welcome to the show. Hey, thanks Mike. Happy to be here. Thanks for having me. Yeah, great. Well, I'm excited about our conversation today and for folks that are in the audience that may not be familiar, why don't you give us an overview of Novanta Robotics and Automation in some of the markets that you guys are serving? Yeah, thanks. Novanta does a lot of different things in robotics and automation. The easiest way to explain it is if you look at the guts to any given robot, things like motors and coders, drives, that's kind of the precision motion side of our business, and then there's the ATI portion that has end effectors, things that make robots more flexible, like tool changers and force torque sensors that enable more value applications out of robotics. Yeah, and I think that's exactly what I think of Novanta for robotics accessories and in-in-the-varm tooling pieces. What are some of the applications that those solutions are used today? So we go where robots go in general, but I'd say some of the more exciting applications are in surgical robotics, warehouse automation, and humanoid robotics. Obviously, humanoids have been a huge buzzword over the past couple years, and we're seeing a lot of traction, especially with the force torque sensor side of things. You can imagine a humanoid needs a number of different sensors that mimic what a human uses, so obviously cameras to visualize the world around them, but also these force torque sensors to enable them to interact physically with their world. It enables applications like assembly, fitting things together by the feel, or something simple like walking upstairs and grabbing a handrail and understanding the balance with those three points of contact. They're hugely valuable in those sorts of applications, and then also on the motion control side, our miniature drives are hugely valuable when we're looking at fitting a lot of axes into a very small, tight, compact space while keeping high efficiency, so keeping heat generation and power consumption on the lowest end possible. Yeah, I mean, I know you've been in the industry for quite a bit of time now, and it seems like we're talking about humanoid robots every week on the podcast and on the robot report. So lots of interesting things happening there. What excites you most about what's going on in the humanoid space today? That's a huge question. I mean, if you look at all the futurists that are talking about humanoids, I think there's a pretty common perspective that humanoids could be the biggest product of our lifetime. I'm optimistic. I think that we're heading in that direction. If we can get all things right, they could be a transformative technology that we are lucky enough to kind of witness in our lifetimes. They could transform in the way society works, and hopefully for the better, taking people out of positions where the classic dull, dirty, and dangerous tasks that unfortunately required human intervention, that's a perfect fit for something like a humanoid robot to step in and allow somebody to do a higher value, more fulfilling task instead. So it's just the opportunities are endless. I'm optimistic and hopeful that those impacts that humanoids will have will be positive, but I think without a doubt they're going to be one of the biggest, you know, transformations that we'll see in our lifetime here. Yeah, no doubt. Again, you've worked closely with a variety of machine builders over the years and integrators and the rest helping to solve some of their unique problems for assembly and handling and all those different applications. From where you sit, what are some of the unique motion challenges that humanoid robots present, you know, compared to some of these other applications that you've experienced over the last decade or more? Yeah, I think one of the biggest differentiators is that when we have conventional six axis robots like a cuckoo or a fan egg or an ABB, you have a fixed coordinate system, a fixed kind of space that they're working in. So the robot is really good at understanding where it is at any given point in time. So you can program just positionally go from point A to point B to a task, go back to point A and then over and over and over again. With humanoids, the environment is always variable. You're going to have sensors and you're going to have AI and machine learning that take in that data from all the sensors, not only the cameras, but also the force torque sensors and tactile sensors if they may have them and be able to adapt and make decisions at a much higher level, more like behavioral level, than what a conventional fixed robot would. It really kind of changes the way that we perceive how robots act. And sensors, force torque sensors, cameras go from something that is a nice to have to something that is absolutely required to, you know, get the one-to-one ratio of capability from human to humanoid robotics. You know, I hear from friends in the industry all the time, I mean, everybody's asking me what's going on with humanoids and asking what should we be doing in this space? And obviously you guys are taking existing product line, trying to figure out where you can take your IP and stretch it to the needs of the humanoid market. And why do you think this is the right time for a company like Novante to be focused on where humanoids are going? Well, I think it starts with, you know, we're getting to this point, this inflection point in humanoids that we really start to see a potential path to commercial viability. So on top of, you know, the large language models in generative AI, the hardware aside is also getting to a point that is much more capable, we're packing a lot more power into a lot smaller spaces, a lot more, you know, compute power into a very small space, batteries are also more dense. So everything's kind of aligning to this point where in the next, you know, five to ten years, we'll see a rapid progression of technology that gets to a point that is pretty remarkable, I think. And I think Novante is particularly well positioned with forced torque sensors and drives, being the leaders in those spaces kind of leading up to this point that we've only had to tweak our kind of development product roadmaps slightly just to address the specific needs that are involved with humanoids. Things like what I mentioned earlier with power efficiency, I like to use the term strength density for forced torque sensors. So packing the most strength, the most robustness into the smallest lightest package. And that's because typically you want a forced torque sensor as distal as possible, right at the wrist or even further down to get a more accurate reading from the sensor. But having, you know, a big chunky, heavy piece of metal very far down the arm, or by the ankles isn't really ideal because you have, you know, inertia and it means your joints have to be stronger and you can't react quite as quickly. But if we can pack, you know, a strong lightweight high performance force torque sensor at that point, then that's best case scenario for everybody. So that's kind of where where we've been moving with our product line and how we've been attacking this, this market is, you know, looking at the pain points, looking at what the needs of humanoids are. And just tailoring our product offering the things that, you know, I've made our sensors, our drives the best products on the market over the last 20, 30 years. And just, you know, tailoring them to the humanoid market specifically. Yeah, I think that makes sense because as we've talked about where the goals for many manufacturers in terms of price points that they want to get these systems to from a from a street price, it's going to require a bomb cost that's that's, you know, gives them some profit in that at some point. And if you're talking about the supply and components, I expect you're looking for the component designs that you can share across, you know, supply to multiple manufacturers and that helps again, gets the volumes up for the types of things that you're delivering and ultimately, you know, gets the numbers into the the right place to reach this holy grail price point. We've made lots of fun of on the podcast the best because it seems Unachievable to me in the short term Yeah, it's interesting that I think the price discussion is it'll be a long kind of back and forth between the humanoid companies and the suppliers because everybody's barking for this this target price point that is the you know 10 20 50 hundred thousand robot unit, you know Yeah, economy of scale and they're asking for it at the you know 50 robot number. So I think we'll get there It's just it's a chicken in the egg who's gonna provide that 100,000 piece price Early and you know, it's just it'll it'll be a back and forth for a while But I think there's a path and there are people that are trying to make that happen. I think Novance is specifically is investing really heavily in Those cost-down initiatives to optimize our production processes and make sure that everybody's happy As we kind of build through this development timeline. I think you're in a unique seat to see some of the leading indicators Is to win those types of volumes are gonna start moving which will be interesting to watch Now you guys recently launched anybody expects Oh, that's great to hear and you recently launched that you know your safety-related servo drives another example of Bringing everything together into one package especially for you know smaller Axis like fingers. I would imagine in in other smaller joints How's that contributing to the performance and reliability of being a humanoid form factor? That's a great question So the safety conversation it's probably the the hardest technical challenge that humanoid developers have at least in the short term Because you have not only the the component level safety ratings to think about but also the higher level behavioral safety How do you how do you achieve safety on a you know AI behavioral level? But it really does start at the component level it starts at the joint level and the drive level So if you if you have a system that is made up of components that are just off the shelf non-rated And you're trying to make a behavioral like a safety behavioral algorithm and Never gonna really be safe because the components that make up that system aren't safe So our our team has really put a lot of focus on safety rating drives at the smallest form factor and Really focusing on on what makes the most sense for a humanoid specifically a lot of really cool functions that kind of rethink the way that Joint level control happens from the from the drive That really blends itself to enabling a lot of safety functions on the the higher level for humanoids I can't about divulge a whole lot of detail here, but I think what what our team is doing on the the safety front for humanoids is pretty transformative and Is going to be a huge Value driver for for a lot of our good good customers Yeah, no doubt we talk about safety all the time on the show and in the robot report anything related to humanoids We're always you know concerned about the future trajectory not the part of the pun of You know where the industry is going and how we're gonna get there and you talked a few minutes ago About your rich history with force and saying in some of the physical aspects of that, but again We've seen you know force become mainstream for a mechanical assembly and and surgical and other sort types of operations Where the force is important to make sure that you know the ultimate action happens When you think about humanoids and how we want them to interact with the world around them in some cases The same way we use our fingers not quite you know Survelling in our fingers on something until we touch something and then you know holding a mouse picking up a glass of water whatever the case Maybe how do you think that's going to sort of Change and make humanoids more lifelike is as they evolve Yeah, that's a that's an interesting point and I think a lot of the the humanate companies out there have differing perspectives on this You've seen obviously the the ones that are going Like very hands first approach developing a very dexterous very sensitive hand Like the sanctuary Yeah, and then they're the more kind of industrial I don't want to say more practical, but you know more conventional approach where you have this is a robot. It's an industrial tool If we wanted to hold a rivet gun We're going to make a tool that is designed to hold a rivet gun and that robot's going to do riveting all day long I think the common end effector that is always involved in in this case is a force torque sensor or something that Is sensing the the feedback of a forces at the the hand But what the hand looks like I think could be up for debate That that is one of the biggest challenges on the hardware side is is packing all that functionality into a very small space is You know hands are what 11 to 20 degrees of freedom Depending on who you talk to. So I don't know that that is really the most valuable thing at least in the short term I think just to get Humanoid robots up and running A force torque sensor maybe a tool changer and then the standard industrial tools that have made our industrial robots so successful over the last 25 30 years Will probably be the the thing that kind of gets the foot in the door and makes them functional and effective Right off the bat once we get you know into the The applications where humanoid is in the home, which is you know the one of the hardest Technical challenges not only on the Dexterous manipulation side but also on the safety side Um then yes, you're going to probably want a five-finger hand that can do just about anything can handle anything the way that a humanoid or a human would But for the the kickoff applications like Manufacturing warehouse automation They can I think benefit from just a dedicated industrial tool Hmm, so I want to come back to again To close in the loop on on the control behaviors And and what you guys do it in the real world today What about the integrated sensing and motion control? I mean how is that going to help with more human-like or coordinated behaviors Between you know the types of use cases these systems will be deployed into Yeah, so I think force torque sensing is it's a I'm biased. I think it's just about as valuable as vision and when it comes to enabling new applications in robots You think about anything that you do from the day to day where you're really just feeling it You can close your eyes and you can still accomplish that task just based on how how it feels So if you ever want a humanoid to do something like assembly which I've already mentioned Feeling those two things together feeling when something clicks or when threads start to engage and come together But also things like like material removal standing or polishing where you have to apply a constant force along you know a changing contour Those are really important Once you start interacting also physically with with humans if you want to give them a high five or a handshake or The human wants to move that humanoid around and get them out of the way Being able to understand that there is an external force that is changing and I need to react as a robot to that Those are they're all going to be extremely valuable Anytime that there's really a changing environment or there's there's a lot of variables with the parts that are being handled Having a force torque sensor that feeds that data back up into a decision-making algorithm Is always valuable we talk about AI and machine learning all the time There are buzzwords a lot of time, but the one commonality is that they work based off of sensory input It's not just you know you put in a prompt and the robot's going to do it There's a feedback loop and that feedback loop relies on Visual visual feedback force feedback torque feedback All of those kind of base-level sensors to make higher level decisions Yeah, so so I mean this all very exciting are you able to share with us any any of these partners that you're Working with with some of these you know exciting new cutting-edge Designs that you're bringing to market Unfortunately, I can't share specific names But keep an eye out. I think if you have a keen eye you might see an ATI logo on on the rest of a given humanoid Most of the the big winners today. I think have have used our sensors Not only in the risks, but also in the ankles, you know enabling balance and You you see some of the the robots with I would say lesser sensory input on the foreside and they're just kind of stomping around and very kind of archaic and you can feel that they're going to tip over at any given point Those don't necessarily have the the force torque sensing But if you have you know a sensor in the ankle that understands how hard you're you're Contacting the ground or where on the foot is making first contact you can walk a little bit more Gently a little bit more human like So yeah, I think the answer is if you're thinking of a humanoid robot company than we likely are working with them to some extent. But yeah, unfortunately, I can't divulge who specifically. >> Well, it's, I mean, it's a big opportunity at the moment. So, could those two, if you're talking to most of those companies, that's awesome. But if we've got someone in the audience who's listening and is embarking on their own humanoid development project, so they're thinking about, again, what sensors do I need, what drives do I need, all those sort of elements as they pull this together? How do you work with those customers to accelerate their time to market? What is it about Navanta in the products that you have that helps them get where they need to be, faster? >> Yeah, so I'm actually, I'm on a small team of business development folks and advanced solution engineers, so application engineers that understand how our products work in the end use, and our team is really focused on these target markets, warehouse automation and humanoids, and on top of the physical products that we're really leaning into the products that we've developed for this space, so we can get demos very, very quickly, but we also have personnel that are there side by side with you, making sure that we're going into completely unknown waters with a lot of these applications. Making sure that we're side by side and kind of there to support initial integrations and get over that initial hump of, this is a new product, this is a highly technical product, how do I make use of this? We understand how it works, how you're going to get the most value out of it, and we'll be there tomorrow if need be on site to help you get where you're wanting to go. Yeah, I think we're investing a lot in these markets, and our team is really highly dedicated, highly focused, and we're here to help in any way we can. Great, well, I mean, again, coming back to the seat that you sit in, able to talk to some of these companies about their roadmap, about what their needs are and where they're going, what can you share publicly about your view of this next generation of humanoid robots? We've seen some of the first generation now, the last five years, companies like Figure already on version three, iterated multiple times, and in many of the other companies are now getting their second and third generations. How are you preparing to support them, and maybe what can you share about any potential roadmap additions? Yeah, so I think when we look at that look of specific companies and their strategies and who might be a winner in five to 10 years, I generally look at, again, I'm biased, I look at their sensory capability, and if you look at some of the players, particularly coming out of APEC and China, a lot of times they are focused on getting to a viable product and mass production of said product first, and then maybe dialing it back to say, okay, we can we can make a lot of these things, but how do we make it more aware from the sensory perspective, and more functional, more kind of application focused, rather than just like a walking humanoid thing? Then there's the more researchy perspective that are starting with every sensor you can think of, like RGB cameras, a lidar, four-storek sensors, and the wrists, and the ankles, and then over time they'll probably scale back and figure out and optimize what's the last bolt we can remove, and still have a very functional product. We're going to land on a spectrum, and there's going to be a robot up at the more premium end that has all of these sensory inputs and sensory capabilities and can do everything under the sun, and then there's going to be the less sensory aware and maybe less performance and more focused application-wise, and I think it'll work its way out in the long term. Again, I think that the more sensory aware to begin with, I think that's where I would be betting on if I were a betting man, but also the ones that are focusing on safety today. Safety is kind of the longest lead item in the whole whole roadmap, and those that are kind of in the trenches with the ISO groups today developing standards that drive what a safe humanoid is and how it needs to function. That is a huge hugely beneficial play in the overall strategy. A lot of companies are seeing safety and safety ratings from an official perspective as kind of an afterthought, which is a little concerning. I think they acknowledge that there's a challenge and it's a need in the long term, but if you're not building your prototypes with safety in mind, then when do you start bringing that in? Because it is going to take years and years before we get full certifications and start to say, "Okay, this thing is safe that I can have it next to my children in my house folding laundry or doing dishes." So yeah, I think sensors and safety are the biggest things to be thinking about right off the bat. Where do you want to land on that spectrum of price versus performance, where sensors versus price and sensors versus performance? There's a whole, that's a strategy that each kind of robot company is going to have to come up with themselves, and then there's the safety part. You got to be thinking about that right off the bat so you can be certified when the time comes. Yeah, I agree wholeheartedly, and obviously there's quite a few standards that commit projects kicking off as we speak here to get in front of that, but it's still going to be a certain amount of time frame that can't be shortened to produce that. The other thing that's evolving probably faster than anybody ever imagined is AI. I mean, I talk about a half-life of about six months now, four to six months for some of the things that are happening in AI and hardware rarely moves as fast as software. In this case, I don't see it moving as fast, but again, from where you sit in the type of conversations you're having with, especially with the control algorithms, the supervisory decision making, how are you guys reconciling the hardware capabilities with where AI is going at levels above that? So yeah, I think from the novanta perspective, we generally say that we provide hardware rather than software, but when we think about AI and machine learning and the data that it takes to get to something that is value added in a given environment, that AI is only going to be as valuable as the sensors that are feeding into it from a component level. If you're using a bad inaccurate, drifty sensor to begin with, then that algorithm is going to learn based off of that inaccurate and error-prone data, and it's going to be a sloppy performance robot. If you're using an ATI sensor that is the best in the market in terms of resolution, sensitivity, accuracy, drift performance, then you're working with the best on the market to develop your algorithms to be as valuable as possible, as quickly as possible. I think with bad data, enough bad data might make a good valuable algorithm, but you're starting crippled if you're starting with a bad sensor. Start with a good sensor, it'll optimize the learning process right off the bat, and it'll make it so that you can trust the data, you can rely on that data and understand that, yeah, that data is trustworthy in the long run. On that topic of training the models, I'm really curious, are you able to bring in simulated data for for sensing into some of the simulation environments and actually deploy that in ways that the models can learn from simulation and seem to real type of environments? I've had that question myself, and I haven't come up with a good answer yet. I think for the short term, no, I haven't seen anybody with that capability yet, but yeah, you see the digital twins, the physical space, where we can say, okay, the robots here, it's moving around, it's generally balancing, these are the rough physics, but I haven't seen a model that takes in interaction forces so that we can understand how the robot feeling two things together is going to react and work. I think the one thing that we are focused on is making sure that those folks that are developing the digital twin software have the tools from us to be able to easily import an ATI force torque sensor on the 3D model perspective into your system so that at least you have the hockey puck that is modeled into the robot. But yeah, I would love to see that come into fruition, but I haven't seen it yet. I think it'd be really cool. We'll see, leave that up to people smarter than myself. Yeah, I was curious, that's why I had to ask that question. So for companies that are are full tilt into the humanoid development space working on their first-second generation of technologies. What's your advice when it comes to selecting motion and sensing technologies? - Yeah, that's a great question. I think first, come reach out to us. We're happy to help. Our team is dedicated and focused to this market and we're happy to invite our insight in any way we can. I don't think there's a single right answer, whether you're working with novanta products or our forced torque sensors or what have you. But we have a team that is focused and working on these solutions and we're passionate. We're happy to help in any way we can. - Well, Nick, it's been a great conversation. I've enjoyed learning more about the Vanta products and solutions and digging back under the hood as to what it is that you guys do every day. So can you tell all of our listeners or they can go to learn more about your product lines? - Yeah, and thanks for having me, Mike. Listeners can go to www.novanta.com. That's novanta.com. You'll find all of our miniature drive products and forced torque sensors available there. And feel free to reach out to me. I'm on LinkedIn. You'll find me under Nick Damiano at ATI or Novanta. I'm happy to have a chat. - Thanks. - Thanks a lot, Mike. (upbeat music)