Drones of the future will require small but powerful chips in order to become more than just toys.
For many, the word “drone” conjures up images of autonomous, militarized technology. But in the context of small aircraft with multiple rotors that you often see carrying cameras, drones are more accurately associated with hobbyist sport and commercial applications. They’ve only recently begun attracting mainstream attention as drone makers such as Parrot offer a greater and more affordable selection, putting them in the hands of a broader audience.
With increasingly sophisticated capabilities and new technologies, smart drones are also likely to become a topic of discussion next month at the International Consumer Electronics Show where for the first time, show organizers have created a dedicated Unmanned Systems Marketplace, where over a dozen companies will be showing off their new offerings.
According to the CEA, the sales of consumer smart drones are expected to reach 400,000 units and approach $130 million in revenue next year, and exceed $1 billion over the next five years. The FAA acknowledged drones as a popular gift during the holiday season by helping to create a website and video informing the public of acceptable drone use, hoping to mitigate some of the inevitable drone-related accidents.
Their newness isn’t just a matter of increased selection and recent affordability, though. The fundamentals underlying multi-rotor flight are sophisticated far beyond familiar remote-controlled planes and helicopters. Sensors feeding data to a microprocessor keep a number of propellers spinning clockwise and counterclockwise at just the right speed…and that’s just to hover. A pilot’s command to pitch, roll or yaw is received, calculated and translated to motor response in a matter of milliseconds. Aggressive miniaturization means low-power CPUs, RAM, flash, sensors and I/O can be crammed into compact spaces and sold affordably, compelling us to wonder what tomorrow’s smart drones will be capable of, given even greater processing power.
The space is evolving rapidly. Modern drone “brains” add orientation control, GPS-based points of interest, failsafe mechanisms, cruise and cameras to the list of features siphoning compute cycles. Interestingly enough, though, none of those capabilities are considered new or innovative, per se. Large craft designed for the defense industry, such as the RQ-4 Global Hawk, can pack advanced sensor packages operated through a common signal processor marketed as “equivalent to an airborne super computer.” They’re just immensely expensive (think hundreds of millions of dollars). So, the manufacturers of more cost-sensitive smart drones find themselves implementing features based on the potential of light, small and affordable SoCs.
“It’s really not too different from the benefits of advanced computing on a mobile device,” says Brandon Basso, senior research and development engineer at 3D Robotics. “Specific to drones, there are a couple of application domains that make sense for advanced computing. The first is computer vision—being able to sense the environment more effectively. Optimization and routing are also math-intensive problems.”
3D Robotics already sells a sub-$1,000 quad-copter able to track and follow GPS-enabled devices running Android. Its 32-bit Cortex-M4-powered controller can keep the drone’s camera pointed at you. Or, using DroidPlanner 2 software, it’ll follow a flight plan of your creation and train its GoPro on a specific region of interest.
Recently, the company demonstrated a version of its Third-Person View Follow Me technology that employs optical tracking of a subject, rather than following a GPS signal. According to Basso, the task is anywhere from 10 to 100 times more compute-intensive. So 3D Robotics’ engineers augmented their existing controller with Intel’s Edison module, including a dual-core, 500 MHz Atom SoC with 1GB of LPDDR3 RAM, 4GB of flash memory, Wi-Fi and Bluetooth.
“The Edison hardware is better suited for vision tracking because of its high clock rates and more RAM,” says Basso. “Plus, it integrates well with open-source libraries for vision processing.” The fact that Edison fits on a postage stamp-sized module illustrates how next-gen devices will chew through complex algorithms with PC-like performance to enable advanced features, even as they maintain sensitivity to size and cost.
Armed with enough compute horsepower to keep up with lots of incoming data, the next big thing in smart drones technology will likely be affordable radio and laser remote sensing subsystems to address routing—getting from one point to another in the presence of obstacles or adverse environmental conditions. Basso continues, “One of the enabling technologies that got drones to where they are today is cheap MEMS (microelectromechanical systems) sensors—accelerometers and gyroscopes. That was wave one. The next wave of lidars and radars, which used to cost thousands of dollars on their own, are coming down in price. So in five or 10 years, I think you’re going to see a lot more precision in how aircraft navigate, avoiding trees and compensating for wind.”
Expect significant advancements to the cameras strapped to unmanned aerial vehicles too. Right now, much of the emphasis is on higher resolutions and more stable picture quality. But hyperspectral imaging, which involves the collection of information across the electromagnetic spectrum (and not just the three bands that humans see), promises a greater understanding of the environments drones are flown in. 3D Robotics currently does a lot of work surveying farms using point-and-shoot cameras. However, company representatives acknowledge they could generate even more useful assessments of crop health in near-infrared, for instance.
Of course, as processing potential increases and sensors evolve, new business applications will materialize. Delivery is one example. Companies like Amazon didn’t have any interest in drones previously. They do want to serve customers more quickly at lower cost, though, which is evident by its Prime Air drone delivery concept. As the technology improves and becomes less expensive, it appears more likely that multi-rotors will change the way they do business.
We’re forced, then, to consider a world in which unmanned craft buzz around over our heads around the clock, at home and work. Beyond the social implications of that, smart drones in the future could require some form of communication with a central traffic control agent for avoiding collisions, likely enabled by cellular radios. In order to take off, you’d need a link to whatever service keeps craft from running into each other. Such a mandate would have to be passed down from agencies like the FAA, which is already rumored to be considering a pilot’s license as a requirement. Until that happens, today’s low-cost hobbyist flight controllers benefit from return-to-home functionality that monitors the link between drones and their radios to prevent fly-aways. Should a craft go out of range, the autopilot returns to the point of takeoff using GPS navigation.
Not bad for technology typically priced in the hundreds of dollars.
Top image credit: Lima Pix/Flickr