Skydio, a startup based in San Francisco and founded by 3 MIT alumni—Adam Bry SM ’12, Matt Donahoe SM ’11, and Abraham Bacharach PhD ’12—has recently designed a self-governing video-capturing drone also dubbed “selfie drone” that follows and pictures an entity while navigating freely in any setting.
The drone, entitled as R1, entails 13 cameras that record omnidirectional video. The drone can be controlled by an application through which it can set to particular flying and filming environment or be managed manually. The system of R1 incorporates advanced algorithm constituents spanning planning, control, and perception, which offer it distinctive intelligence “that is equivalent to how an individual would steer a surrounding.
The system, on the perception side, utilizes computer vision to find out the objects’ location. Making use of a deep neural network, it accumulates data on every object and recognizes every person by, say, size and clothing. “For every individual it spots, it builds up a distinctive visual identification to set people apart and remains concentrated on the correct individual,” Bry says.
That information is transmitted to a motion-planning system that locates the position of the subject and envisages their next step. It also makes out planning limits in one region to improve filming. “All data is continuously exchanged and balanced … to record a smooth video.” Eventually, the control framework obtains all information to implement the automaton’s arrangement constantly.
A user places the automaton on the land or in their clutch and swipes up on the Skydio app. The R1 fly off, identifies the user, and begins capturing and tracking. From that moment, it functions completely self-rulingly, remaining somewhere in the span of 10–30 ft from a subject, autonomous, or 300 ft away, physically, dependent upon Wi-Fi availability.
Skydio, in the future, expects to discover other applications, such as examining commercial real estate, energy infrastructure and power lines for damage.