This market research report was originally published at Tractica's website. It is reprinted here with the permission of Tractica.
Self-driving cars are gaining quite a bit of attention lately, with a new headline on the subject just about every day. Meanwhile, as autonomous vehicles get all the publicity, self-piloting drones are quietly making progress and are getting progressively closer to being ready for prime time.
Today, commercial drones are flown manually and they have limited battery capacity that limits their flight time to around 30-40 minutes. Consequently, one drone requires one pilot to control and navigate the craft. The pilot must navigate the drone so it doesn’t bump into an obstacle such as a tree. The pilot also needs to keep an eye on the position of the drone so that it captures the best images from the right angles, in addition to ensuring that the drone returns to its starting position on a timely basis to avoid running out of battery power during its mission.
A self-piloting drone would not only eliminate the need for an operator, but would also provide other advantages such as extended battery life and a reduction in accidents while providing better imaging capabilities. Just as it does for self-driving cars, computer vision is the key enabling technology that makes these improvements possible.
Self-piloting technology in drones, of course, is quite different from cars. For instance, GPS data, one of the key enabling technologies in cars, is not as important when flying drones. Drones may be flown over farmland or forest settings, where precise maps are less relevant. A GPS signal is accurate to within a few meters, which is just enough lack of precision to cause problems for drones in terms of increasing the risk of accidents. Drones must also take into account real-world spatial factors, such as trees and geographic features, and determine a corrective course in real time. And drones flying indoors will not even have the luxury of GPS data – they must rely on what they see.
A self-piloting drone must also be able to process incoming camera signals in real time and quickly determine the appropriate actions based on that data. Another challenge is posed by the fact that there is a delay from the time the propeller is moved to the time its effect is realized and the course is corrected. This means that the system must be able to capture details of obstacles in real time from a safe distance, process the results, and adjust the propellers before catastrophe can strike.
Self-piloting also poses another interesting challenge in terms of power requirements. Propellers consume most of the power used by drones today, so the navigation system must consume as little power as possible to reserve as much power as possible for the propellers. An article published in MIT Technology Review suggests that self-navigation capabilities could reduce flight time by as much as 50%.
Despite the challenges, self-piloting technology is making quite a bit of progress. DJI, one of the top manufacturers of drones, has already introduced its Phantom 4 drone that has self-piloting capabilities. The Phantom 4, introduced earlier this year, can dodge obstacles and find its way around trees. Startups have also risen to the challenge, and a company called Skydio, founded by veterans of MIT and Google, has already created a self-piloting drone. A report on The Verge pointed out how drones outperformed humans in search and rescue missions in a forest while avoiding obstacles.
Regulation will certainly be a major hurdle before such drones can be sold in large quantities, but self-navigation capabilities could easily be used in agricultural or forest environments today. The regulations are less important in these settings and the value proposition can be quite compelling. Use cases such as a drone that can fertilize crops by locating rows would immediately alleviate the need for an operator. A drone that drops supplies to a designated location where a hiker is stranded would also be welcomed irrespective of its cost.
So, while autonomous vehicle guidance systems continue to wrestle with issues such as poorly marked roads, drones are continuing to quietly make progress. We believe that the technology to enable self-navigation in drones is much closer to widespread deployment than autonomous cars, and we expect to see rapid adoption once the right regulatory framework is established.
By Anand Joshi
Principal Analyst, Tractica
