MIT Associate Professor of Aeronautics and Astronautics
Returning Two–Year Sabbatical at Google X
I think that we're all going to take for granted that things are flying around looking at things all the time.
By Heidi Legg
If you’re not a drone junkie, you may still have heard that the US dronies” are waiting for the FAA to rule on who can own a drone and where they can fly it, you may have heard that a Hoverboard or Drone are on the top of every tech-head and 10-year old kid’s Santa list, but do you know how pervasive drones will be in your lifetime? When your coffee shop, bike path and grocery line collides with MIT and Harvard, you hear these things. I sat down with drone expert Nick Roy, Associate Professor of Aeronautics and Astronautics who is fresh off a two-year Sabbatical at Google X under his former Carnegie Mellon advisor, Sebastian Thurn who invented the self-driving car inside GoogleX.
What is Project WING?
Project WING is a project at the Google X Lab. Our product at MIT is people and understanding. Part of the reason why I'm back here after two years at Google is because I realized that what I really like doing is teaching undergrads and graduate students about the research enterprise. There isn't a perfect parallel between Project WING at Google and the research enterprise at MIT, because they're meant to do different things.
What were you working on while you were at Google X?
I can't really talk that much about Project WING. I can tell you what's publicly known: it's a drone delivery service and the idea is that we can improve the world by making it easier to move things from place to place at a relatively higher speed. But that’s it.
Can you tell me what a day is like at Google X?
No. I’m sorry.
How long before drones will be delivering goods to homes in America?
I think that we'll have UAVs delivering things in the US in relatively short order, somewhere between one to three years. The regulations are set up so that a company that wants to operate a UAV in a controlled part of the environment may, such as along a pipeline where they own the entire ground space or a closed movie set. I don't think the step between using UAVs to do imaging and inspection to using UAVs to move things around in those environments is that far off.
Is it legal to have UAVs move around movie sets or take photos of bridges or pipelines?
There are some wrinkles to that. To operate on a closed movie set you have to follow the appropriate paperwork and go through steps similar to those of getting your drivers license. You have to show you know how to operate it. Same thing with drones, but the short answer is yes. We know that people use drones for pipeline inspection. We know that they're on movie sets. Imaging of infrastructure such as a bridge strikes me as well within reason.
Is a drone a robot?
You’re work inside the MIT Computer Science and Artificial Intelligence (CSAIL) lab. What happens there?
CSAIL’s mission is to create computational solutions that benefit humanity, and it is the largest lab at MIT with a hundred faculty members. Robotics at MIT is also really large. There are about thirty faculty across campus working on robots of different kinds. We have underwater robots. We have surface robots. We have self-driving cars and trucks. We certainly have space vehicles and we have air vehicles, as well, and I think there are about four groups on campus working on drones of different kinds.
Is a drone a robot?
How many students are at CSAIL?
The number is probably easy to look up but my guess is about 700 undergrad and grad students. It's a very big group. [There are 1028 members on the MIT CSAIL website directory]
Is it hard getting into CSAIL?
If you are an undergraduate and want to work with a faculty inside CSAIL, then I think it's relatively straightforward - but you've already been accepted at MIT. If you're a graduate student and want to work in CSAIL and you've already been admitted to a graduate program, then I think it would be relatively straightforward to find a faculty member who wants to advise your research interest. MIT acceptance into the various graduate programs is somewhere between 2% and 5% and then I have no idea what the statistics are in terms of the faculty applications.
Do engineering students worldwide know about CSAIL?
MIT actually has a matrix structure. There are labs in the departments. The departments are about the educational mission, and they decide teaching responsibilities, which course requirements students have to take, etc. The research enterprise is in the labs. CSAIL is a lab, not a department. CSAIL and the labs decide about funding, about space, about research resources, etc.
I feel very privileged to be a part of the AeroAstro Department and I work with colleagues, in terms of teaching, that are a rich source of interesting research questions. And then I get to collaborate with computer scientists in CSAIL who are often rich sources of solutions. I get the best of both worlds.
AeroAstro is the number one aeronautics department in the US and in the world, and I don't think that labs typically get ranked. It's usually only the departments that get ranked, but the computer science enterprise at MIT is first in the world – probably tied with Stanford and Berkeley. A lab like CSAIL doesn't really have an equal in the world. It's really a cross disciplinary thing. CSAIL is the biggest lab at MIT.
Are you privy to the drone development for war inside this group?
How much do you know about the history of Draper?
Draper Lab actually originated in this department at AeroAstro in the '60s. It was called the Instrumentation Lab at the time and they were charged with developing the avionics and the guidance navigation controls for the moon missions. It was enormously successful and at that time in the '60s – to be part of the Apollo missions, you were working closely with the military. As student groups protested against the amount of military research and classified research that was going on inside campus, MIT decided that basic research was what MIT was about and they would no longer do military and classified research. Draper Labs was spun out.
Do you do any military research today inside your group?
MIT does not take money for military research and can't do anything above basic research. In terms of research that may have military applicability, we obviously do that but that [basic research philosophy] governs everything from the Internet to basic biology, synthetic biology, and basic computer science algorithms.
Application of that research to a military product is done by organizations like Draper and others who can take on classified research.
Inside the drone world, do you all have this sort of good/bad drone bifurcation? Is there a conflict between military vs. consumer use?
I would imagine that there's some but the drone community's large. Here at MIT, again, we have this very strong commitment to basic research. There are some large basic research challenges that are going to have to be solved before we can get our 24/7-drone service doing any one of these things, and therefore we don't differentiate.
The basic technology that we're developing will service a NASA (National Aeronautics and Space Administration) program for search and rescue that will service a DARPA (Defense Advanced Research Projects Agency) program. That’s how you know its basic research in that it's solving real problems. Our research is broadly applicable and when it comes time to figure out how to actually get the drone to do the real thing for Bin Laden or whatever, that is when Draper Labs, or another entity like them, gets involved. They really take it from the basic research into the applied research domain.
You guys are the top of this field.
Well, we hope to be.
Are you ever afraid of the power of what you're unleashing?
I think there are a couple of answers to that question. One is that I generally think of technology as relatively neutral and it's the application of the technology that is something we need to think about. It’s interesting: 3D printing has been hugely enabling. 3D printing is giving us promise of lighter components, more fuel-efficient rapid fabrication, less waste, and fabricating things like medical devices. These are things that we couldn't dream of before. There are obviously huge applications that improve life but at the same time, it’s made possible the 3D printing of a gun that can't be detected. Australia has just banned the use of 3D printing for hand held guns and so that seems like an appropriate response.
The creation of the Internet then would be analogous?
Exactly. But we also have to teach our engineering students to be ethical and you can't blindly develop the technology and not think about the applications. You do have to think about the applications and as part of that ethics MIT is a place where we do basic research and not classified military research. You need to go somewhere else if you want somebody to solve those problems.
George Church is a big voice for transparency at Harvard in genetics. Is there someone here at MIT when it comes to drones and ethics?
That’s a great question. Sheila Widnall is somebody who thinks about that and is a voice for engineering ethics. [First female Secretary of the Air Force and the first woman to lead an entire branch of the U.S. military in the Department of Defense] Somebody else you should probably add to your list. She's amazing.
Where would you predict we will use drones most – space, war, or everyday life?
They are going to be everywhere. What is interesting is how we think about what it means to be a robot. What we expect of autonomous systems has evolved in even the last ten or fifteen years. The idea of a drone being able to fly to a point above the earth and come back with relatively little intervention from a pilot would have been intelligent and fully autonomous twenty or thirty years ago. But now we don't really think of it as intelligent or autonomous because we know that it's driven by GPS and relatively simple algorithms on board, and we also know that any time something goes wrong with it, bad things happen: a drone's going to fall from the sky.
What we're looking for now in terms of robotics and autonomy is much more independent operation in the face of the unexpected and the ability to do more complicated things than simply get from point A to point B. That technology, I think, is going to show up in cars, in drones, in spacecraft, underwater craft, and in offices and homes. I can't predict the specific morphology. I don't know exactly what they're going to look like but I know that we already have these things and they were almost unthinkable twenty years ago.
What are some obvious applications for widespread drone use?
I think the obvious ones are things that we can't see: Precision agriculture is already enormously lauded.
Would you explain?
The US produces a significant fraction of the world's cotton right now. Cotton is incredibly easy for people to grow in Bangladesh and Ecuador and other places. US labor is incredibly expensive. How on earth is it possible for the US labor force to out-produce cotton competitively with Bangladesh and other emerging markets? It’s because US precision agriculture on the ground is so good. The cotton is so high quality that it's actually much higher value cotton than growing in other countries.
The drone introduction will be for inspection. Right now we have to put down an enormous amount of fertilizer in places where we might want to put a little bit of fertilizer.
The drone becomes a foreman?
The drone becomes a foreman and in come cases the worker, too. The drone goes and does inspection and then comes back. Maybe another drone goes out and puts a little bit of nitrogen in just a few targeted places.
What's a lifestyle change we will see with a drone?
We are going to have a lot more information about the world around us. Think about the change we have in terms of the Google Maps. Imagine that you get instantaneous imagery or near-instantaneous imagery of arbitrary locations. It certainly would change traffic patterns. It would change management of infrastructure. I predict we will take for granted the fact that we can know the state of the world around us on a much bigger scale all the time.
Do you care about privacy anymore? You'll be able to spy at any time and place.
Privacy ebbs and flows. Look, 150 or 200 years ago there wasn't much privacy, and everyone in the little village knew exactly what was going on and it was much harder to get away with things. Then we moved out to the suburbs and bigger cities and people know less and all of a sudden privacy became a bigger value, and now it's receding again a little bit. I don't know what will happen next.
What do you say to those concerned?
I don't think it's only drones. Police are wearing body cameras more and more and I was in the UK and the UK has very strong data privacy acts, but at the same time everyone is being imaged all the time. Different cultures will also have different solutions. The Australians have a very strong privacy act but at the same time culturally, as long as you tell everybody what you're doing and you're not hiding anything, they're actually pretty relaxed about imaging. It’s going to vary from country to country and culture to culture.
Where will we see great progress because of drones?
Agriculture is going to be one for sure. City imaging is going to be another and that's going to include infrastructure inspection. In Europe, wind turbines are really difficult to inspect the blades and they have to have people climb up and it's dangerous for people. Today, in places where there are big power lines in this country, they have somebody dangling from a helicopter and they carry them from pylon to pylon.
I think that we're all going to take for granted that things are flying around looking at things all the time. The big change is that there'll be companies and jurisdictions that make it easy to get an instantaneous picture of what's happening in the local neighborhood and then I think the next change we will see will be in transport.
Transport in that we will be carried by drones?
I don't think we'll be in them, at least initially. I don't know if your kids have ever forgotten a lunch at home but boy, would it be nice if instead of having to drive a big piece of steel across town to deliver their lunch to school, you'd be able to put it in a drone? I can imagine that.
I also think remote communities are going to benefit enormously from this. My wife comes from a small town in Northern Ontario and the biggest town nearby is Thunder Bay, and it's three hours away. So people are constantly driving back and forth between Marathon and Thunder Bay, transporting things. Small drones traveling from Thunder Bay to Marathon along the shore of Lake Superior seems like a really, really good idea.
How much weight can drones carry?
How big is your airplane? A drone can be as big as a 747. I think it can be as small as a deck of cards. You pick your drone. Bigger drones bring bigger safety risks. You probably don't want an unmanned 747 flying from Thunder Bay to Marathon and back again.
Will there be unmanned aerial devices before we have unmanned land vehicles?
We have both now.
Which will proliferate more quickly?
I am going to guess the unmanned air vehicles, but we won't see them. They'll be there but we won't realize they're there. They’re going to be flying in places where if they crash, nobody cares.
Jake Ware studied under you and he is now being sponsored here at MIT to study wind turbulence. Can you talk about this?
I'll talk about him. I love this.
What solution is Google chasing by funding robust wind?
Google gave us some money to be able to build better models of urban wind fields. We are using a computational model to look at wind.
Do you know about Weather Underground? It is basically this network of weather stations with members sending in their weather reports and it's relatively high density. You pull it up on the web and it gives you the instantaneous feed, and it's more highly detailed with more data than you might get from CNN Weather. Weather Underground pulls data from the weather station on top of the Green Building, and then there's another weather station in Central Square. You can spend $100 and get a weather station to put in your back yard and all of a sudden you'll be part of the Weather Underground network. It's pretty cool. [From WU: We have 140,000+ members sending real-time data from their own personal weather stations, they provide us with the extensive data that makes our forecasts and products so unique.]
The problem is that it's still relatively sparse and I don't have a weather station out in front of this building. But if I wanted to fly a drone down Mass Ave or through campus, especially for the smaller drones, I might actually like to know the state of the wind field. I would not like to fly into a head wind because that'll be expensive in terms of the power and I'll go slowly. Buildings shed vortices and there's turbulent flow and if I'm going to do a delivery from a drone, I might not like to deliver something in the midst of a vortex stream.
What is Jake mapping? Only Cambridge, or larger swathes of areas?
He's doing a single square because this is really hard to do, but given the instantaneous measurement from the Green Building in Central Square, we'd like to be able to instantaneously predict the wind field at one-meter resolution throughout campus and fold that into the intelligence the drone is using to make decisions about where to fly and how to fly, so that it can be more power efficient and more reliable. That's what we want.
What public opinion would you like to change?
The level of autonomy that we're getting from unmanned systems is increasing and I would like very much to be a part of that revolution because it is really going to revolutionize how we live our lives. We're going to have much smarter cars and much smarter airplanes and much smarter houses and much smarter buildings and things to help us live our lives. Not only would I would be very happy to be part of that revolution, but also to have people understand that that is coming and that it is going to increase productivity and quality of life across the board. I'd like to be part of it and I'd like to do it in a way where the world understands that this is a good thing.
What concerns you about your work?
We're very good, or largely good, at getting these vehicles to go from Point A to Point B, right? The thing that keeps me up at night that I'm excited about is the fact that we want them to do more things. We want them to be smarter, and part of the reason they're not smarter right now is that these robots think very much in terms of geometry. They don't think about the door as a door. They think about it as a flat place. They don't think about this room as my office that has a function. The robots think about this room as a chunk of empty space with some obstacles inside it. If you had a really good robot, it might understand that stuff can move and it better be ready for the stuff to move, but that's all you've got. That's true of the Google self-driving car. It's true of the drones.
We, as humans, think about the world very differently. We think in terms of function and place and appearance, not geometrically. We think about what we want to get done in the world of high level tasks and what that does is it gives us tremendous resilience to when the world changes. It gives us tremendous ability to actually think about more complicated plans.
I'm really excited about this idea of trying to get out of the geometric representation and the geometric model, and thinking more about symbolic representations and more human sensory representations. This will allow us to bring in language.
And social rules?
I wasn't being quite that ambitious but yeah, absolutely.
Great, more robot-people missing EQ!
I am excited about the idea of being able to talk to robots and about being able to team with robots, and when I express what I want out of the world and what I know about the world, the robot has to take that stream of words and turn that into something that the robot can understand, work with and use.
Do you wish the FAA would speed up? They have been under a lot of pressure to get the drone regulations right.
I have a lot of sympathy for the FAA and I think they've actually done a pretty good job. They are the standard bearer for aviation safety without question. The world, to a certain extent, looks to the FAA for the right answer and the FAA has aviation as one of the safest industries. It's one of the safest forms of transport that exists.
They're simultaneously being told to go fast and at the same time go slowly enough to make sure the world is safe and I have to give them credit. They did move a little slowly because the drone revolution happened really fast, but they caught up really quickly and I think they've done a good job.
The real issue is that everybody is showing prototypes and demonstrations of ideas. The technology is not there yet to provide a reliable 24/7 service of any kind. The vehicles that people are talking about building services on – DGI phantom vehicles you order off the web from Hobby King, or whatever, have a mean time to failure of about twenty hours. As a result, drones are literally starting to fall out of the sky. We see stories of drones falling out the sky at football games or sporting events. You often see stories of reporters saying, 'I thought these things were super easy to fly and then I bought one and it flew away and now I'm out $59.'
There are hard technical challenges to get to the level of reliability needed to actually deliver on the commercial promise. The FAA is doing a really good job of holding people to that and they are asking, ‘show me the technology; show me that it's reliable and we will sign off on it.'
When will the new rules go into effect for drone use in the US?
I don't know the answer to that. 2016 seems a reasonable guess.
They've announced that they're going to ask everybody to register their drones, which seems like a pretty good first step and they're going to have that operational in time for Christmas this year, which again is remarkably quick. I say, good for them.
How do you use this opportunity?
I do my best. I try not to break more things than I fix.
Is this drone era analogous to a period in history in the past?
I feel like generally in technology and Artificial Intelligence in particular, it feels an awful lot like the Renaissance – Florence of the 1500s with an independent city-state. There is a book called 1927 by Bill Bryson that talks about the zeitgeist of the Wright Brothers and flying that could be similar, but this feels more like that time in Italy where everybody was working on these different problems and there was this flowering of ideas. That seems pretentious. Ha. I don't know.
As we always ask at the end of our interviews, where is your favorite place to unwind?