DALLAS — At AUVSI Xponential here earlier this month, the drone industry’s ongoing shift from military and consumer offerings to commercial applications was clear. Intel Corp. has moved from making chips to sensors and now complete drone offerings. Intel drones demonstrated the company’s commitment to new markets at the show.
“This is the biggest movement in the industry,” said Anil Nanduri, vice president of the New Technology Group and general manager of the new Unmanned Aerial Vehicle Segment at Intel. “The opportunity is easier to understand, as drones for commercial inspections and delivery provide a clear benefit.”
“The systems are more capable for a wide variety of applications,” he told Robotics Business Review. “For instance, UAVs will transform delivery, search and rescue, and agriculture.”
Intel drones benefit from cheaper components
“Drones are able to leverage processors from higher-volume industries, such as smartphones and consumer electronics, lowering costs,” Nanduri explained. “Inertial motion and GPS are much lower-cost, thanks to consumer electronics.”
But software and data are becoming more important to end users than hardware, he added.
“Users are thinking of drones as a vehicle for doing something like capturing high-res imagery,” said Nanduri. “We’re making it easier for a pilot to say, ‘Fly to that tower to inspect it.’ The UAVs are smarter, and the pilot could even be a computer.”
“It’s very much a waypoint-based approach today, but the challenge is how to capture data in a repeatable way,” he said. “The system should help the pilot avoid guywires, and it’s all done through data fusion and compute onboard.”
Greater autonomy in the air
Nanduri echoed the keynote addresses, in which Brian Wynn, president of the Association for Unmanned Vehicle Systems International (AUVSI), observed that unmanned vehicles are giving way to more autonomous systems.
“This is similar to the transformation in the automotive industry,” Nanduri said. “We have the car and the driver, and the car learning to drive itself, thanks to sensors and compute engines.”
“The capabilities are similar, but the pain points and requirements are different,” he acknowledged. “UAVs have constraints like flight time. A 2 kg (4.4 lb.) sensor is too heavy for a drone.”
“We need miniaturization and safety,” said Nanduri. “The process of getting data from drones is still very manual — often, it must land so the operator can swap out an SD card.”
“In the future, the Intel drones will be able to take off, do work, transfer data, and be aware of their environment,” he added. “A UAV can go back and recharge, and it will have safety and redundancy.”
Intel’s open-source Aero platform is designed to help developers customize drones for commercial viability. It’s based on Linux and provides support to help reprogram I/O interfaces.
“I call it a test bench. We want to continue to learn from this.” Nanduri said. “Developers can connect different sensors — including Intel’s RealSense — connect a modem, a USB device, or an SSD card. They can also process data onboard, and they deploy the systems.”
‘Data is the new oil’
“Data is the new oil,” declared Intel CEO Brian Krzanich during his keynote in Texas. He rode onstage on a Loomo, a robot designed by Intel and Segway Robotics Inc. It uses Intel’s RealSense camera and depth-perception system.
“By 2021, we believe that robotics and autonomous systems will be prevalent,” Krzanich said. “The coming flood of data … means the future is in artificial intelligence and machine learning.”
“Automation can simplify workflows, but there is still the challenge of how to manage that data,” Nanduri said. “The first step is classification and analysis. Our data is georeferenced and geotagged automatically in real time.”
“To apply machine-learning techniques and AI, you must first collect the data to train it,” he noted. “There’s a lot more work to do; we need more compute.”
“More data, more compute are part of Intel’s strategy for growth,” said Nanduri. “From a consumer, you might get 1.5 GB a day, but a single commercial flight could generate 50 GB, and a fleet of drones could equal 25 TB a day.”
Falcon 8+ flies at AUVSI
“The Falcon 8+ drone is built for commercial use with redundancy in mind,” Nanduri said. “It has eight motors, triple-redundant IMUs [inertial measurement unit sensors], and multiple buses. With two batteries, there’s no single point of electrical failure.”
“The Falcon 8 also has algorithms to guard against electromagnetic interference, which could affect compass headings during, say, electrical tower inspections,” he added. “The Falcon’s whole payload is swappable with a single screw, and it could synchronize data from, say, both thermal and RGB cameras.”
“This makes maintenance and certification easier,” Nanduri said. “We also have a fixed-wing platform for longer endurance, with high-endurance GPS and RTK [real-time kinematics].”
Krzanich announced a deal with Airbus to use Intel drones for aircraft inspection. A curtain next to the main stage pulled away, revealing a bridge, around which a Falcon 8+ conducted a mock inspection.
Features such as separating the Falcon 8’s sensors from its flight system and ruggedizing the “cockpit” are part of Intel’s play to being a complete drone solution provider.
“These machines and data will drive the needs of the cloud, and we need to build end-to-end capability,” said Nanduri. “We do the reference designs with OEM partners. The unmanned ecosystem is still fragmented, so it’s a really good opportunity for us to provide whole systems.”
Intel unleashes Shooting Star UAVs
One of the highlights of Xponential’s keynotes was a demonstration of the Intel drone known as Shooting Star, which was used for light shows at this year’s Super Bowl and Coachella. At Disney World in Florida, 500 drones set a Guinness World Record by flying twice per night for 45 nights.
“We could wait for someone else to invent it or do it ourselves,” Nanduri said. “It’s easier for us to work on at a solution level.”
“A person who wants a flag or a Christmas tree in the sky doesn’t need to know programming and drone controls. An artist just wants to create, and we make that possible through software automation,” he asserted. “All the artist needs to know is how to transform that creation into physical reality as simply as possible for the audience.”
“For the light shows, we had to build the whole thing from end to end — not just the drone, but the animations, the software,” Nanduri said. “It’s more than just the systems. It’s a leap ahead. We built the concept, and interest has been phenomenal since then.”
“The light show is reusable, and while Shooting Star isn’t a full replacement for fireworks, you can tell a story with them,” he said. “We’re operating worldwide and getting a lot of learning from the events.”
The choreographed swarms are currently being used for entertainment, but these Intel drones also demonstrate the potential for coordinated fleets for precision agriculture, infrastructure inspection, and potentially drone deliveries.
“The FAA has been a very willing partner — it just wants to prove it’s safe,” Nanduri noted. “With the last-mile problem or the last 100 feet, the safety requirements get more stringent as you get closer and denser.”
More on Commercial Drones and Robotics:
- Agriculture Automation Needs Economic Incentives to Grow, Says U.K. Expert
- Autonomous Systems Takeaways From Xponential Day 1
- How Intel Joule Brings Power to Robotics Prototyping
- Cloud Robotics Promises to Simplify Machine Tending
- Police Drone Market Increases With FAA Rules, Test Cases
- Mobile Robots, Cobots Steal the Show at Automate, ProMat 2017
- Autonomous Cars Accelerate Toward the Future
- Smart Machines Increasingly Driven by Connectivity, Political Choice
- Delivery Robots Ready to Satisfy the On-Demand Economy
Intel drones seek partners and markets
“We’ve been operating Shooting Star for the past six months,” said Nanduri. “It’s not yet scalable, we’re helping others operate it.”
“We’re not just building chips for drones,” he added. “We have the ability to test it from end to end, and commercialization will be with partners. We’re working with Movidius on machine-learning chips. It’s more about the systems — customers want turnkey solutions.”
“In some cases, we’re offering drones through an RaaS [robotics as a service] model as we’re building, testing, and gathering data ourselves,” Nanduri said. “As the technology matures, we’ll enable partners to do it.”
“Clearly, our technology can scale into other verticals, which will build with partners,” Nanduri said. “There’s a significant effort in autonomous driving with Mobileye, as well as our work with Ninebot and RealSense. We use cars to collect useful data, the same as with our drones.”
“We look at the market from three angles,” said Nanduri. “Aero is for developers, standardization, and open APIs [application programming interfaces]. The second part is a few verticals, for commercial high-res imagery for oil and gas, commercial inspections, and construction.”
“The third part is the light show, which we’re looking at from a technology viewpoint,” he said. “It’s a good way to introduce awareness to people.”Read More