How Boston Dynamics Is Building Its Robot Empire

How Boston Dynamics Is Building Its Robot Empire

Show Video

You may have been one of the 140 million people who've seen this video on YouTube. Boston Dynamics has made a name for itself by building robots like Spot, which move in ways that many of us have only seen in sci-fi movies. We look at ourselves as having mastered the ability to create robots with mobility, vision and manipulation, and we think we can combine those and just almost an infinite number of ways at different applications. The public's fascination with Boston Dynamics' robots is likely owned to equal parts awe and fear.

A 2017 episode of the popular dystopian Netflix show Black Mirror depicted a woman running from a pack of killer robot dogs. The show's creator later admitted in an interview with Entertainment Weekly that Boston Dynamics videos were the inspiration. Most people think when they see Spot. That that's cool. There is a small fraction of people who say it's creepy.

We've been telling ourselves fictional stories about dystopian robots for one hundred years. I think it's people's desire to kind of explore that fear that makes them want to react so negatively to Spot. But while Boston Dynamics has sparked the imaginations of the general public with its viral videos, the company is now hoping to get the attention of paying customers. I expect that we will become a serial producer of novel robots with advanced capabilities. I think we'll build every three to five years.

We're going to roll out a new robot targeting a new industry. CNBC got a rare look at Boston Dynamics office in Massachusetts, where the team showed off two of the robots that are working to commercialize Spot and Stretch. The yellow robot that you see walking around is a mobility platform. It gets really interesting when you start adding payloads on it and integrating it into things. We have really nice visual cameras, thermal cameras, microphones, gas sensors and those let you take the robot into interesting places and collect data about what's going on. We see Stretch as ultimately a general purpose box moving machine that can be used anywhere in the warehouse.

Those two markets, in particular, the robotics inspection and the warehouse market, are what we perceive to be the fastest growing market within all the robotics segments. The machines can look at more things faster in greater detail. They never get bored. They don't need breaks. They're just relentless. With nearly 30 years of experience, Boston Dynamics has proven its ability to make robots that move in remarkable ways. Now the question is, can the company turn a profit?

Boston Dynamics was founded in 1992 by Marc Raibert. The company was an offshoot of Raibert's work on dynamic robots at Carnegie Mellon and MIT. Early on, much of Boston Dynamics research and development was funded by the U.S. military and DARPA. In addition to robots, Boston Dynamics also developed 3D computer simulation software. This robot LS3 was part of a project that we worked with DARPA and the Marines.

This was a proof of concept project to show that the robot could actually carry equipment in a useful way for a marine squad. Some of Boston Dynamics, as other early creations included Big Dog, which, like LS3, was meant to traverse rough terrain and help soldiers carry gear, and Cheetah, which at its fastest could reach speeds of 28.3 miles per hour. Pittman was one of the company's first humanoid robots. It was developed for the U.S.

Army to test the special clothing used by soldiers for protection against chemical warfare agents. We're very proud of our history that involved DARPA and government funding for creating some of the fundamental technology. It was really that visionary funding that sustained over long periods of time that was required to solve truly hard problems.

But in 2013, the company's focus shifted after Boston Dynamics was purchased by a well-known technology behemoth. What does a Big Dog, Wildcat and Cheetah have in common? While it may sound like the beginning of a joke, those are the names of robots by Boston Dynamics. Google continues its recent string of acquisitions with engineering company whose repertoire includes designing research robots for the Pentagon. When Google purchased Boston Dynamics for an undisclosed amount, the internet giant was gobbling up robotics companies left and right.

The New York Times first reported the acquisition, noting that it is the eighth robotics company Google has bought in just the last six months. There is speculation that Google wants to build a new class of robots that could perhaps do everything from warehouse work to package delivery. At that time, Google was a bit more ambitious, and they are trying to look for those of the next big thing and highly specialized robotics they so perceive as big emerging area that industry was expecting or anticipating at that time. Google itself did not say much about the acquisition. Other than that, it would honor Boston Dynamics existing military contracts, but was not interested in becoming a military contractor itself.

I think there are some concerns around Google dabbling with military contracts. They might hurt their public reputation. But without government contracts, Boston Dynamics would lose its revenue stream. Even existing contracts that the company had with the military did not always pan out, and R&D was expensive. Owning Boston Dynamics reportedly cost Google $50 million a year. Internal movement at the company also shook things up when Google bought Boston Dynamics.

Its robotics effort was being run by Andy Rubin, who was well known at the company for leading the development of Android. But after Rubin left Google in 2014, Google's robotics division stalled. Then in 2015, Google, which by then was renamed to Alphabet, got a new CFO, Ruth Porat. Over the next few years, Porat went on a cost cutting spree, and Boston Dynamics was one of the casualties.

That relationship didn't really last very long, mainly because the fruit of their relationship is not as big as what Google perceived it to be. Boston Dynamics again found itself with a new owner. Softbank, adding to its artificial intelligence portfolio by buying robot maker Boston Dynamics and Tokyo based Schaft. Once the company moved on to SoftBank, Softbank tried to inject more of a commercial DNA into the startup itself. When SoftBank purchased Boston Dynamics in 2017 for a reported $165 million, the Japanese conglomerate already had a few robotics companies in its portfolio, as well as a consumer robot known as Pepper. The idea was for SoftBank was that it really wanted to become a large robotics supplier to the industry in general.

Under SoftBank, Boston Dynamics did open up sales to its first commercial product Spot in June of 2020. Prior to that, the inspection robot had only been available to lease as part of the company's early adopter program. But like Google before it, SoftBank found that breaking into the robotics market was harder than it seemed. The market still largely concentrated around industrial robots, as well as for warehouse robots. The robots type, the form factors that Boston Dynamics excels in are

not what market wanted at the end of the day. In December of 2020, Boston Dynamics changed hands again. South Korean car company Hyundai announced that it was taking an 80 percent stake in Boston Dynamics to the tune of $880 million.

This year, tens of thousands of robots will be built and put to work. Out in the in the world making money, but in the next 10 years, millions of robots will be built and get put to work in the world. And in most of these markets, no matter how big they are, a small number of companies end up dominating a majority of the market share. And so I think Hyundai is positioning itself to be the leading robot manufacturer.

Hyundai bought us for our mission. They're excited about our mission to launch novel robots that can help industry. And ultimately, I think they see their own mission evolving beyond cars to generalize mobility and see the things that we've been developing as ultimately complementing them. Hyundai and Boston Dynamics are so far from dominating the robotics market, but it's a market with a lot of potential revenue from mobile robots in the warehouse. Manufacturing and infrastructure spaces is expected to increase dramatically from about six point one billion in 2020 to about sixty four point nine billion by 2030. Something like eight hundred million containers are shipped around the world each year.

Many of those are full of boxes. There's probably trillions of boxes that are loaded and unloaded by hand each year in the United States. It's a huge job. It's a mountain of material that has to get moved. Stretch is really power tools to help people move that material. Stretch is made up of a few different parts. The robot uses a mobile base to move around tight spaces and go up loading ramps, and an arm gripper.

Vision cameras and sensors allow the robot to identify and handle a variety of different objects. We're starting with truck unloading, but we think truck loading as well as what we call order building. So when you go through a warehouse and assemble a bunch of different boxes onto a pallet, that order building is a big job and we're actually developing that at present.

Similarly, with Stretch, we're building in intelligence so that the robot can deal with unexpected conditions if suddenly it knocks down a cascade of boxes. The robot can deal with that and pick them up and clean up its own mess. How do you feel about competitors? There's Amazon that has their own robotics for warehouses.

There's a lot of automation companies that are entering the logistics space, and of course, the logistics space has a huge variety of problems to solve. We're uniquely focused on mobile box moving, which almost nobody else is looking at, so we think we have a nice, safe entry into that market that we can own for a significant portion of time. Amazon, I think, is lighting terror in the hearts of all the warehouse owners across the world, so they have a very motivated buyer that wants to buy tech that's comparable to what Amazon is developing. Boston Dynamics says it expects Stretch to go on sale next year, though it would not provide a price point.

Customers can also opt to purchase just the computer vision software that powers stretch, which Boston Dynamics calls Pick. The company says it's working with a few early adopters to test the robot, but would not say who those partners are. Unlike Stretch, Spot has already hit the market. We've sold several hundred Spot's thus far. We launched it really commercially just last year. The entry level Spot explorer is about $75,000.

We have another version that has additional capability and a recharging station that's a little bit more than that. Sales have exceeded expectations, really. Payloads like pan-tilt cameras and lidar are sold separately and can be quite pricey. Even so, Spot has received interest from a number of industries. So far, the robot has been used to do inspections at construction sites, oil rigs, nuclear plants to check the vital signs of COVID-19 patients in hospitals and even remind people to maintain social distance amid the pandemic. So I would take a substation like this and subdivide it into maybe four sections, and I would have the robot do a detailed inspection on each of those four sections.

This is Dean Berlin. He is the lead technology analyst for National Grid and electric and gas utility company that serves customers in Massachusetts, New York and Rhode Island. For the past year and a half, Berlin and his team have been testing to spot robots by programing them to do inspections of National Grid's equipment. Berlin says the team also considered other robots for the job, but chose spot for its mobility capabilities.

Some of the advances that we find for the Boston Dynamics robot is that because it's a four-legged walking terrestrial robot, it can actually transverse our substations fairly easily. The robot also can be equipped with a variety of different payloads, making it mission specific. So we're going to power on and undock the enterprise robot. A typical inspection looks something like this.

First, engineers program Spot so that the robot knows where to go and what gauges to pay attention to. The instructions for each mission are relayed to the robot using a QR code. The robot that National Grid is using is equipped with light bar to help it navigate, as well as visual and thermal cameras to take detailed photos and thermal images of the equipment in the substation. The robot will be focusing on switches and circuit breakers.

It will also take zoomed in visual images of some of the gauges on the equipment to monitor the levels of those gauges. It can use this information to diagnose if there are hot spots on our assets that might indicate areas of concern or areas that we might need to attend to in upcoming maintenance. Prior to using Spot, most inspections at National Grid substations were done by people. In some cases, operation of the substation would have been temporarily shut down since it wouldn't have been safe for humans to do the inspections while the equipment was still on. National Grid has noticed savings in terms of time because the robot is able to perform these missions in parallel to a human performing other task. So in the sense you get efficiency from having a human and the robot operating at the same time.

Spot may even go to space. NASA has been sending teams of the robotic dogs into caves to see if they can one day be used to search for life on other planets. Caves, protection from cosmic rays and stable temperatures, NASA says, makes them the most unlikely of places to find life. My team's focus at NASA is on developing next generation A.I. and autonomy algorithms for robotic systems. Our solution, which we call Nebula, analyzes the terrain, analyzes the environment and risk.

In a certain sense, you can look at nebulizer robot brain that we integrate with robot body that could be a real rover and legged platform, or even a flying drone. In the case of exploring caves, Aiga says his team chose spot for several reasons. You often come across narrow constrictions and narrow passages that you really want to be able to traverse with a system that is not large and bulky, and the Boston Dynamics robot offers an interesting combination of a small size at the same time, high locomotion capabilities and enough power to be able to traverse through those sorts of narrow passages.

The ability for spot to be equipped with different payloads also comes in handy. Sending in one large robot to carry all the scientific equipment needed for a mission would make navigation in the cave difficult. Instead, NASA sends in teams of spot robots, which can each simultaneously accomplish different tasks. NASA says there are a number of things that need to be worked out before Spot can explore new planets, but the potential opportunities have scientists excited.

We have a region on Mars called Tharsis region. There are a lot of interesting caves that if we can explore one day we will learn a tremendous amount of knowledge and we can get a lot of science from it, but it's highly uneven. That's where different and new locomotion ideas need to come in, and legged bathrooms is one of those that can transform how we think about exploring. Although demand for warehouse and inspection robots is expected to increase in the future, Boston Dynamics still faces a number of challenges on its road to commercialization. Many of which the company has already been running into with its Spot robot. Over the past 10 years or so, we've basically solved most of those fundamental research problems.

When things are working well enough, we started to ask how can we make this practical and how can we start making this useful in real applications manufacturing? Having a service department, having a sales team, those have all been new skills for us marketing, you know, all new things for us that have encouraged us to really expand our staff. Playter says his team has grown from 100 employees during the R&D stage to 400 employees now so far. Boston Dynamics says it's been able to keep up with demand, producing around 40 to 50 spot robots a month with the help of contract manufacturers. Final testing, assembly and repairs are done in-house.

But in order to compete in the larger market and have a shot at profitability, experts say Boston Dynamics needs to seriously ramp up production. Boston Dynamics biggest challenge to commercialization today is just cost pure and simple. The hardware itself is the other piece.

The hardware itself for something like a Spot is today probably in the $20,000 ballpark. They've got a great robot hugely technically capable right now, but it's far too expensive to to really sell that huge scale today because people's alternative is a robot that's not nearly as capable, but one fifth of the cost. And as long as it's capable enough, they're always going to go with that alternative. Boston Dynamics hopes that something else Hyundai can help with.

Hyundai is really going to be a great partner for this stage of commercialization because they bring expertise and manufacturing and obviously a worldwide footprint in sales and service. And those things are going to help us get our robots out into a broader audience in the inspection space. Spot will have to contend with wheeled inspection robots and flying inspection robots or drones. A number of companies, including Swiss-based ANYbotics and China-based Unitree robotics, for example, are also making four-legged dog like robots. Though the latter seems to be marketed more as a human companion rather than an industry tool.

We do enjoy a pretty good lead in terms of the robots having reliable behavior in complex environments. Our robots have dramatically better integration between their vision systems and their balance systems, and we also do a lot of the higher level functions that you don't see our competitors doing, like complex mission planning and actually completing a lot of these end use applications that produce value. Spots really the only robot in the world that can deal with the complexity of the terrain that people can. Boston Dynamics faces just as much competition in the warehouse robot market, which includes companies like Six River Systems, Fetch Robotics, GreyOrange and Geek Plus.

And then, of course, there's Amazon. The interesting part with Amazon when it comes to logistics is that they tend to develop solutions just for themselves. Amazon doesn't really supply the technology to external parties, but then again, there definitely can be a competitor because at some point Amazon may feel like they are ready to now serve the wider market, and then they can now go out there with their solution.

In the end, some experts believe the market could be regionally segmented, with Western customers choosing Western robot manufacturers due to security concerns. When you deploy a robotic solution into the field, it's also about collecting data. The current market perception is that if you were to go with the Western suppliers in this case, Boston Dynamics, there is a lot more respect towards how the data is being used. Boston Dynamics says it's off to a good start, so I'm sure developing these products is very well one time consuming and too costly. How are you covering the cost right now? Really through the investment of our owners such as SoftBank and Hyundai, they're really covering the cost.

Now we do have significant revenue at this point, which is also helping. We're generating revenue currently from the sales of our robots, primarily Spot. We have a little bit of time before we're profitable, but we're on a great trajectory so far. As for its R&D roots, Boston Dynamics says it's not leaving those behind. It's critical that we maintain leading edge in R&D work within the company as we commercialize, but we've only covered a fraction of what we need to do with robotics.

There's additional work in vision and manipulation that we want next generation robots to have that we currently can't do. So it's important to keep pushing that forefront. Playter says the company is also looking at other industries, including construction, manufacturing and entertainment. As for Atlas, it will likely remain in the lab. I do think robots will ultimately have relationships with their owners that could be rewarding and interesting. I don't see Atlas as being that product, I hope in the long run. Robots aren't the exception. Instead, they become the rule and that we're all comfortable around them and they delight us and enhance our productivity and safety.

2021-10-17 22:30

Show Video

Other news