Later this year, Zipline will bring its fleet of medical delivery drones (read more about them here) to North Carolina to take part in the Federal Aviation Administration’s UAS Integration Pilot Program (IPP). Zipline will be working with the North Carolina Department of Transportation “to set up a network of medical distribution centers that can use drones to make medical deliveries,” which is what Zipline has been doing for several years now in Africa. The FAA is being very careful and methodical with the IPP, because there are a lot of unknowns about how commercial drones can be safely and effectively integrated into a complex, crowded airspace over a crowded, complex country.
Zipline is in a unique position to show both North Carolina and the FAA what’s possible with drones, since the company is (as far as we know) the only one doing continuous commercial drone delivery on a national scale. They have more experience than anyone else with things like operational-drone safety, which the FAA is understandably very concerned about, as we reported just last week.
Practical delivery drones will need to fly beyond line-of-sight, probably at night, and almost certainly over people, things which the FAA generally doesn’t allow at the moment. These drones will also need to be large enough to carry a kilogram or two of useful payload, which definitely makes them large enough to be potentially dangerous. The FAA’s notice of proposed rulemaking includes some suggestions about how delivery drones might be made safe enough for operating over people, like emergency parachutes or structures that crumple or break up on impact to absorb energy.
It’s probably not a coincidence that Zipline is already using techniques like these in its delivery drones in Rwanda, and ahead of its North Carolina debut, we spoke with Zipline CEO Keller Rinaudo about the present and near future of operational drone safety in the United States.
First, a video from Zipline that shows some of its drone safety systems in operation:
The video is a bit of a Zipline puff piece, of course, but we can’t really fault Zipline for that, and they do, in fact, take safety very seriously—if they didn’t, they wouldn’t survive as a company, simple as that. It’s important to remember, though, that Zipline operates in Rwanda (and very soon in Ghana), which is a very different environment (both in the geographic sense and the regulatory sense) than North Carolina will be. We asked Rinaudo about these differences, as well as about how Zipline’s drone safety systems work.
IEEE Spectrum: Can you describe what Zipline will be doing in North Carolina?
Keller Rinaudo: What we’re going to be doing starting in June is servicing hospitals and health centers covering half, if not more, of the state, putting millions of people within access of this service. Starting to operate at that kind of scale in the U.S. is going to depend on working closely with government regulators to build the most reliable operating environment in the world, and that raises a bunch of questions, like how we can set a high bar for public trust.
The notice of proposed rulemaking that the FAA published last week included some specific figures for drone safety, like using various means to make sure that a drone “will not cause injury to a human being that is equivalent to or greater than the severity of the injury caused by a transfer of 25 foot-pounds of kinetic energy upon impact from a rigid object.” Do you have a sense of whether the current generation of Zipline’s drones would be considered safe by the FAA under that proposed standard?
It’s a little challenging because the FAA is just starting to talk about this, so it’s not like they’re saying, “Here are the numbers, meet them or not.” It’s still a conversation. But what the FAA is doing here is exciting—a lot of people think of the FAA as a slow moving bureaucratic organization, but actually, this is an amazing step forward. They’re saying, “How do we enable this innovation in the U.S.?”
I guess I’m trying to say that the answer of, do we specifically meet what the FAA said in that notice—that wouldn’t necessarily mean that we’d be ready to begin operating tomorrow, there are a lot of processes we have to go through with the FAA, some of which have yet to be influenced by these standards that they’re starting to talk about.
How do you find the right compromise between maximizing range and payload of your drones by making them as light as possible, and adding the redundancy necessary for you to be confident in their safety?
There’s no easy answer—the answer is engineering, and understanding our customers’ needs really well. And also understanding the interdependencies in the system, making sure that if we are iterating on one part of the system that needs to be independently safe, the product can still operate safely even if there’s a problem with that system.
There’s often this feeling in the U.S. of, it just needs to be safe, it needs to be perfect. And of course, how could we build something that isn’t safe? Safety is paramount. But you can make something safer and safer and I think that the U.S. airline industry is a good example of taking that approach really far in an impressive way. But at the same time, all innovation in the general airline industry has ceased. So for us, the challenge with these conversations with the FAA around risk-based safety is to enable us to make a lot of good engineering decisions to make sure that you have the necessary redundancy to achieve a given level of risk, but where you can continue to make really big technological leaps forward.
If all else fails, your drones are able to deploy a parachute to help them land more gently than they would otherwise. Can you give me an example of when this has been necessary and what the cause was?
There are plenty. We can guess 99 percent of the potential problems, but there’s always that 1 percent edge case where we just couldn’t have imagined it. A really nice example is about two years ago, in the first 100 flights that we did in Rwanda—we had shipped our drones on an airplane, I think it was a 737 that flew from a UPS distribution center in Kentucky all the way to Kigali. We set up the distribution center and began operating and we were having propulsion problems where the vehicle would be out trying to make a delivery, and we were finding that the motors were failing at a much higher rate than we’d predicted given all of our testing in the U.S.. That did cause several planes to land by parachute because if both of your motors are failing it’s pretty hard to get home.
When we dug in from an engineering perspective, we found that what was happening was actually the magnets around the stator in the electric motors were detaching because the adhesive was failing. It turned out what had happened was, when the UPS plane had a layover somewhere in West Africa, it was sitting in a really humid location, and moisture had gotten into those motors. And when the plane took off again and went up to like 35,000 feet in 10 minutes, the moisture didn’t have time to evaporate, and instead just froze, and it was that freezing that caused the adhesive in the motor to fail. That’s just one really specific example of something that’s very hard to predict until you start doing it, and it’s the reason why having a safety system like the “paraland” is so important.
How often does the parachute landing system get used during delivery operations?
This is something we’re really proud of. Today in Rwanda, it’s around one in a thousand flights. And about a third of those are life-saving emergency flights [rather than routine scheduled blood delivery flights]. By the end of the year, we expect to be closer to one in two thousand flights.
One other thing to mention is that this is one of the interesting conversations to have with regulators. Okay, one in a thousand. Is that good, or is that bad? Is it good enough? Our approach is, we think that one in a thousand number starts to be a really obvious decision for a regulator, because if for every one time you have a plane using a safety system to bring itself gently to the ground using a parachute, 333 patients are getting emergency life-saving deliveries of medical products, we just feel like that’s the point where it’s becoming obvious that it’s a pretty good trade for society.
When it’s necessary to use the parachute landing system, how often do you still have some communication with or control over the drone such that you can direct it to a preferred place to land?
It’s definitely the majority of the time, it might be the vast majority of the time, that the vehicle has enough time to choose a specific place to use the paraland system. Those tend to be places that we know are easy to access, and are totally absent of any people.
Zipline has a lot of experience flying drones in Rwanda, where it’s very difficult for consumers to purchase or legally fly drones. It’s much different in the United States, where consumer drones are cheap, mostly legal, and the people using them often ignore FAA regulations. How much of a concern is that for Zipline?
That basically gets built into the risk model that Zipline uses, that we’re basically handing to the FAA and allowing them to do the math themselves and conclude that we’re meeting their required risk profile and safety profile. But suffice it to say that the sky is a very big place, so mid-air collisions even for totally blind vehicles are going to be extraordinarily rare. Also, there are a whole bunch of new technologies that Zipline is building and slowly rolling out that improve our ability to detect those kinds of instances and take evasive action.
Given the scale of the industry today, although that is a legitimate concern, the real issues that you’re going to see over the next two are three years will come from irresponsible drone owners losing control or flying too close to airports, not one of our vehicles colliding with a random drone up in the sky.
There’s a lot of interest in drone delivery in the United States, and there are a lot of people working on drone delivery, but many of these companies don’t have the experience that Zipline does with safety. Are you concerned that the enthusiasm to adopt this technology will lead to safety incidents and consequent regulation that will prevent drone delivery from being as effective as it could be?
Certainly it’s a risk, but I think we have a lot of reasons to be optimistic based on the direction that the FAA is taking. It would be a big mistake for the FAA to say, “anything goes!” And then have a bunch of companies come in and operate in really unsafe ways and possibly hurt people, and then have the regulator come in and say, “nothing goes!” But I really don’t think that will happen. The better way to do it is the approach that the FAA is currently taking, which is to look at the state-of-the-art, and look at what safety is reasonably achievable while still having a positive impact on healthcare systems.
With all the experience Zipline has in Africa, what do you think is going to be the biggest challenge of operating in North Carolina?
I think in general, people assume that the difference is much bigger than it actually is—many people in the U.S. don’t have a good sense for what it’s actually like to be in Africa. There’s a sense that logistics systems are so different, infrastructure is so different, and certainly that’s true a little bit, but they’re not true nearly as much as most Americans believe. I think in general, we’re kind of just going to do the same thing.
We’ve had the advantage so far of operating in smaller countries, and smaller governments can make decisions more quickly in a more unified way. I think a lot of people look at African countries and assume it must be difficult to work with those kinds of governments, but the reality is Africa is a very big place, and you can find governments that are a lot more efficient when it comes to decision making than the United States government. As we launch in bigger countries, it’s just a little bit different—getting the momentum to actually have the government make a decision about something like this is going to be harder. But again, the FAA should really be commended for the fact that they’re leaning in right now.
We should be clear that the FAA’s notice of proposed rulemaking is just that— a proposal, and the FAA is soliciting feedback from industry (and anyone else) about whether what they have in mind is realistic and achievable, so Zipline should probably be allowed some wiggle room on their non-answer about whether their drones currently meet the proposed standards. Zipline’s point is that they’ve been working on these problems for years already, which should give the FAA some confidence about their proposal to start delivery flights in North Carolina this June.
[ Zipline ]
Evan Ackerman is a senior editor at IEEE Spectrum. Since 2007, he has written over 6,000 articles on robotics and technology. He has a degree in Martian geology and is excellent at playing bagpipes.