Antoine Balaresque and Henry Bradlow are worried. They have developed an autonomous flying video camera for use in filming action sports on land and water, capturing scenery while hiking and sightseeing, and covering family events (so everyone gets in the picture). They think they’ve made it simple enough that a parent could just toss it in the air and forget about it while coaxing a child to take her first steps. They have enough seed money ($1 million in investment) to get the prototype they’ve been developing for the past year into production. The technology is coming along nicely; they’ve been able to hire the experts in computer vision, controls, and industrial design that they need, and they’re on track to ship in February 2016.
So what’s the problem? The problem is that their product looks an awful lot like a drone—and they don’t want to be a drone company, they want to be a camera company.
Says Bradlow: “People buy drones because they want to fly something. That’s not what this is, this is a camera. You frame shots with it; you can’t control its flight.”
There are drones, Balaresque says, that carry cameras. These cameras send images to a cell phone, and users can look at the images while separately using the drones’ flight controls to position them for particular shots. Lily is nothing like that, he insists. You don’t fly it, you take pictures with it.
Balaresque got the idea for the project that became Lily Robotics in 2013. On vacation with his family after graduating from college, he browsed through the photos they’d taken—and realized his mother wasn’t in any of them. Most people would probably think, “Oh, I should take some of the pictures.” But Balaresque, who, despite having been a business major, had worked on a number of robotics projects in college, instead thought, “A robot could solve this problem.” When he got back to California, he contacted Bradlow, with whom he’d worked on several of those collegiate robotics projects, and they started developing Lily.
Lily could be the poster child for a classic Silicon Valley startup. Its founders, fresh out of college (in this case, University of California at Berkeley), met at a hackathon. They are working side-by-side, assembling hardware and coding software, with the company’s first few employees in a crowded garage. The garage sits behind an even more crowded hacker hostel owned, I am told, by a long-retired venture capitalist. The house’s porch is jammed with bikes and Segways and other wheeled vehicles; the grassy lawn around the house is overgrown but spacious—plenty big for testing a drone. A table near the pool, surrounded by mismatched chairs, serves as a conference room; the pool itself is used for testing water takeoffs and landings.
Balaresque and Bradlow started working on their technology shortly after that summer 2013 epiphany. Initially, they focused on developing software, particularly computer vision software, in order to make a camera “watch” the user. Their thinking at that point was that they’d mount it on a commercially available drone. But after about six months of development, they realized that existing drones lacked the processing power needed to handle flight controls and the computer vision system simultaneously. They also wanted a waterproof drone small enough to fit in a normal backpack; that just didn’t exist at the time.
The system they came up with uses a video processor to handle the main, high resolution camera. There are three microcontrollers: one controls the motors to stabilize the drone; the other two control the low-resolution cameras for tracking the user, keeping an eye on the ground in order to manage takeoffs and landings, and providing addional help with stabilizing the gadget. Rounding out the list of parts are an accelerometer, a gyroscope, a barometer, a magnetometer, and a GPS device.
A separate puck, designed to be tucked in a pocket or worn on an arm, tells the drone, via WiFi, whom to track. It first identifies the user’s location with a GPS receiver (though once that happens, the vision system, with a faster response time, does most of the tracking. The puck contains some basic controls: to switch the flying camera from following, to, say, circling, or to trigger it to take off or land. (More precise adjustments can be made with a mobile phone app.). The puck also contains an accelerometer to detect sudden changes in motion such as a jump or a fall. These sensors can be used to switch the camera to slow-motion mode at a critical moment. The puck also carries a microphone so the drone can pick up audio as it’s buzzing around recording video.
The Lily flying camera itself is still a work in progress. The prototypes I saw were 3-D printed, and carried low-res cameras; the production models will record HD quality video. But the system indeed does what its creators intended: It keeps the camera trained on the user in several different modes—following from behind, tracking from a set distance in front, circling, or hovering in midair as if held by an invisible selfie-stick (see video, above).
Indeed, as I set up a camera and tripod to photograph the company’s founders and a short demo, I could definitely see the appeal of this gadget—not for extreme sports, but for day to day journalism. Wouldn’t I love camera gear that I didn’t have to juggle, but would instead follow me like an obedient puppy!
Lily’s camera will be priced at US $499 for preorders. The founders indicated that at this price, they wouldn’t be making money, but are looking for these early adopters to act as beta testers. Standard retail pricing will be $999. See Lily’s promotional video, below.