André Prager turns away from me for a moment, rummaging through a pile of stuff on the cart he has pulled into the small conference room. There are lots of cut-up pieces of cardboard, with a few bags of colorful plastic odds and ends mixed in.
“I think the most valuable things in this building are cardboard and tape,” he says. He shows me a rectangle of foam-core with a straw, a broken pen, and a few thumb tacks stuck to it.
It looks like junk, but because we’re at 100 Mayfield Avenue in Mountain View, California, the headquarters of X, Alphabet’s secretive division dedicated to cranking out new Googles, it’s actually anything but. The building is stuffed with cool gear: self-driving car hardware, an 80-foot flatbed scanner that inspects stratospheric balloons, a sprawling tool shop populated by hulking machines with names like Metabeam. But for Prager, this grown-up arts and crafts project represents years of work, and one of many steps on the way to a system that could make drone deliveries actually happen.
Prager is a mechanical engineer for Wing, the drone-delivery effort that, along with internet-beaming balloon scheme Loon, just “graduated” from X and is now a stand-alone company under the Alphabet umbrella. The engineers at Wing, which took shape in 2012, have had to deal with lot of complexities. They designed an H-shaped drone that has a range of 6 miles, can carry a 3-pound package, and tops out at 80 mph. They developed a traffic-management system that could keep the skies safe as they get more crowded with little flyers. And they spent a whole lot of time making a new kind of hook.
The importance of the hook dates to the early days of the project. One of the first decisions the Wing team made was that they didn’t want their delivery drones to land when dropping off packages. Vertical flight is a drain on power, and putting the drone on the ground means putting it within reach of curious (and destructive) children and animals.
The WIRED Guide to Drones
But by the time Prager joined the team in 2016, they still didn’t have a workable idea of how to get the packages onto the ground. They couldn’t just drop them. Parachutes would soften the landing, but what good’s a drone-delivery service if your burrito ends up in your neighbor’s yard or a tree?
So the team decided they'd have the drone hover about 15 feet off the ground and lower its payload, sealed in a cardboard container, to the ground on a string. To start, they wrapped a piece of string around a bobbin and stuck it on the belly of the drone; a programmed motor would start the unspooling when the aircraft reached its destination, the way a yo-yo drops from your hand. But what seemed simple quickly proved a mess. Winding the things was a pain. You had to make one for each individual package. And then, the package would arrive on the customer’s doorstep trailing a long piece of string—hardly an elegant solution. “We’ve got to get rid of this bobbin, was the conclusion,” Prager says.
The post-bobbin era started with a more complex system, a little box that would descend with the package, release it when it reached the ground, then climb back up to the drone. “It was really easy to come up with,” says Trevor Shannon, who is video-conferencing into the meeting from Australia, where Wing runs its trials. And for skilled engineers, it was easy to make. They stuck in actuators and motors to control the movement, sensors to detect touchdown, a radio to tell the drone when it was time to winch the thing back up, and a battery to power it all. It worked well enough, but the team found it overly complex and full of components that might fail. Making drone deliveries work as a business, they felt, hinged on making the novel process as simple as possible for potential consumers.
They scrapped the box setup and looked for a mechanically simpler way to lower the package and release it when it reached the ground. They took inspiration from clicky pens and cabinet doors that swing open when you push them in a bit. Over time, Prager piled his cart with those cardboard odds and ends, which the engineers used to visualize whatever idea popped into their head. But they still felt they could go simpler, more frictionless. “From there,” Shannon says, “it evolved into 'What if we could do it without any moving parts?'”
And so they began designing the perfect hook, something so simple it couldn’t possibly fail. The thing just had to hold the package on the way down, unhook when it reached the ground, then come back up. Once again, what sounded easy proved baffling. Some designs wouldn’t come off the package. Others would release, then swing through the air and reattach themselves. Some would slip through the opening they were supposed to hold onto and get hopelessly tangled.
Each failure led to a new idea, often generated over lunch or in casual conversations with other engineers working on X’s array of projects—what Prager calls “ping-pong with creative people.” And for each creative solution, the team needed creative ways to test it, to root out whatever weakness the latest prototype was hiding. They set up fans to simulate wind and used obstacles like traffic cones to mess with the package as it reached the ground. And they brought in help.
“Our testing is, ‘Hey random person from the hallway, do to this the worst thing you can imagine,’” Prager says. They had people grab the hook and throw it, or hold it as they ran away from the drone. They brought in dogs. After each session, Prager and his colleagues would make an adjustment, hit the 3-D printer, and come back to their lab a few hours later with another iteration.
The final result is the size and shape of a fingerling potato. It’s yellow and made of plastic, but each detail of its design makes Prager and Shannon wince at a painful failure or smile at an unexpected breakthrough. When the package touches the ground, the hook keeps dropping, so it naturally slips off. The motor recognizes that it’s not working so hard anymore and that it’s OK to winch the string back up. The indentation that makes it a hook has an underbite, so that while it’s easy to attach by hand, it can’t grab back on haphazardly. The sides of the hook are shaped differently (one is rounded, the other pointed), so when it fits into the belly of the drone, the aerodynamically shaped package is pointed in the right direction. If by some chance, the hook catches on something as the drone starts to zoom away, the string will unspool and fall to the ground so it doesn’t pull the aircraft down with it. “It’s inherently safe,” Shannon says.
And most of all, it works. It’s the sort of thing you look at and say, “Of course! How else would you make it?” And if you happen to ask Prager or anyone on his team that question, they’ve got a cart piled high with failures to show you.