Last week’s panel discussion at the Churchill Club titled “Civilian Drones: The Opportunity Takes Flight” touched varied topics from open source autopilot designs (such as the Pixhawk PX4) that have democratized access to technologies, to auto-navigation and collision avoidance, increasing battery life and payload weight, FAA regulations, and everything in between. The past year has indeed seen frenzied activity in the UAV (unmanned aerial vehicle) or drone space. Several use cases have taken shape, but how many of them are going to be commercially viable and sustainable in the long run? Businesses and large corporations are actively looking to use drones to monitor their assets, infrastructure, and operations. Start-ups in this space are trying to carve a niche by differentiating with unique underlying technologies such as collision avoidance and auto-navigation, which seek to unlock the potential for new use-cases. New services that offer businesses and corporate entities instant access to professional “drone pilots” with their fleet of UAVs for hire, provide imagery and data collection for a fee. Here are a few quick takeaways from the discussions.
Collision avoidance and auto-navigation
Dave Merrill, VP of Enabling Technology, 3DRobotics spoke about the spurt of drones with follow-me technologies—how good are they? Will these drones know their environment? Will it hit a tree or an obstacle? We have also seen several flyaways and crashes despite the skills of expert drone operators. Adam Bry, CEO of SKYDIO spoke about the challenges to autonomously navigate obstacles. What if you have to deal with a GPS-denied flight? (say you go below the tallest structure near you or in between buildings?). How about using smaller drones indoors in a search and rescue operation after a natural disaster such as an earthquake? Onboard intelligence and awareness of the surroundings is the gateway to autonomy, opening up several exciting things that can be done with a drone. Adam thinks that navigation, flight control, and obstacle avoidance all become extremely critical when you look at use cases such as drone delivery systems.
According to SKYDIO, the on-board video has all the data that’s needed to become situationally aware and to make it useful for a drone’s mission. Computer vision and motion planning algorithms running on heavy-duty processors found in today’s ubiquitous smartphones (such as the Qualcomm® Snapdragon™ 805) can give drones the capability to navigate intelligently and with full knowledge of their environment. The underlying technologies promise to deliver a paradigm shift in usability, safety, reliability, and performance for the burgeoning drone market, enabling current applications to scale and opening up many new use-cases.
Battery life and payload capacity
Battery capacity, weight, and longevity are everyone’s concern. Batteries, to a large extent determine the range of a drone’s mission and its payload capacity. How reliably can a commercial drone takeoff and land a thousand times and how long can it stay airborne before bringing it down for a re-charge or replacing the battery? Christian Sanz, CEO of Skycatch avers that companies are not innovating enough on battery usage such as designing more power efficient semiconductor chips. Battery technology hasn’t kept pace with Moore’s law while compute technologies have grown exponentially over the past 30 years. Battery technology companies have no incentive to innovate if there’s no demand for it. There’s great value in extremely power efficient embedded systems and simultaneously increasing battery capacity—what would people want to do then? Would they increase their payload capacity or reduce payload to fly longer? It may depend on the end-application or use-case, but we need to look deeper into utilizing readily available contemporary mobile/smartphone technologies that already operate at ultra-low power envelopes and run on batteries.
Meaningful and high quality data gathering
No matter what the use case or end application may be, the value-add lies in being able to produce meaningful, high quality, and actionable data in real time. Christian Sanz says that autonomous aerial mapping and surveying has really picked up with commercial drones being deployed by large corporations (such as Komatsu and Bechtel) in the construction, mining, and oil & gas industries to improve operational efficiencies, worker safety, reduce environmental impact and costs.
Interesting use-cases in agriculture include crop scouting (to reduce or eliminate economic risk due to pest and weed infestations by spotting them early enough), proper irrigation, and yield improvement measures especially in drought prone areas. These applications generate a voluminous amount of data that needs to be processed in near real-time to obtain deeper and actionable insights. If the drone is not equipped with a higher degree of image processing and compute power, it’d mean uploading the data to the cloud where there’s literally infinite compute resources. Uploading the data to the cloud via a high-bandwidth connection would increase operational costs significantly and in remote areas that would not even be an option, not to mention the ensuing time delay. There are a variety of distributed computing models in order to help optimize the network spend, processing power, and data traffic. By decentralizing CPU intensive algorithms, computing can be placed closer to the edge of the network, thereby improving responsiveness in drone deployments and perhaps reducing the total cost of ownership. This can be accomplished by readily available embedded compute platforms (ultra-light system on modules and single board computers) based on the Qualcomm Snapdragon processors.
More compute power at the edge of the network
For drones with distributed systems running computer-vision/machine learning intensive applications and 4K Ultra HD video/high MP sensor image processing, popular DIY boards that’re based on previous generation processors will just not cut it. These underpowered processors don’t have the teeth to bite into such CPU intensive applications. Besides, given the explosion in the number of onboard sensors collecting volumes of data, decisions (such as collision avoidance) have to be made in real-time as much closer to the drone as possible. Across the board, the panelists agreed that more and more compute resources are bound to be placed right on the drone as we move forward.
Inforce Computing eliminates the mumbo-jumbo of designing such high-end compute hardware out of the equation by providing Snapdragon processor based modular platforms on top of which an end-user can throw their secret sauce and build the entire drone hardware. Inforce’s compute modules are designed to be upward compatible with successive generations of Snapdragon processors, ensuring current designs will not be obsoleted. To learn more on how such system on modules (SOMs) can help streamline your embedded designs and stay competitive, read this downloadable whitepaper.
Using low-power and small form-factor smartphone technologies in drones
Sometime ago, Chris Anderson, the CEO of 3DRobotics, famously said “The personal drone is basically the peace dividend of the smartphone wars.” The insides of a drone today appear more like what’s inside a smartphone than what you’d find in an airplane, which makes auto-piloting and running CPU intensive algorithms an easy proposition. The plug and play Inforce 6501 micro SOM captures the essence of it with very low SWaP (size, weight, and power) dimensions, making it an ideal fit for commercial drones. Designing a custom carrier board for the micro SOM is easy, reducing time to market. Whether it is flight control, video processing, the ability to connect multiple cameras up to 51MP for high-quality streaming or offload compute to the power efficient Hexagon V50 DSP or the Adreno 420 GPU, the Snapdragon 805 based Inforce 6501Micro SOM makes an ideal foil.
If you’d like to find out more about Inforce’s products and discuss how it can be a good fit for your next commercial drone related project, please stop by our exhibit booth #904 at the upcoming InterDrone conference in Las Vegas, NV, September 9-11, 2015.
The Inforce Computing Team
© 2015 Inforce Computing, Inc. All rights reserved.