Design, deliver, and run enterprise blockchain workloads quickly and easily.
All servers and systems
It hasn’t taken drones very long to become ubiquitous. From palm-size toy units to full-scale unmanned aircraft, the drone business has taken off in more ways than one. And while certain industries, such as agriculture and oil and gas exploration, already have taken this technology to heart, other industries are finding uses for drones, especially when combining the data they gather with on-board and ground-based AI and big data analytics. Here are several unique—and cool—commercial uses.
Drones have become so common that children can operate them and create some amazing images from inexpensive devices. But the capability of creating high-quality video and aerial images at low cost has made even inexpensive unmanned aerial vehicles (UAV) good enough to use in commercial video productions.
Operating a drone via remote control offers freedom and flexibility, but it lacks the reliability and control found from using traditional aircraft. The introduction of AI into the market has changed that. “AI in drones often helps us to capture complex shots for film and TV that would normally require a bigger drone, gimbal, or separate camera operator,” explains Christian Tucci, president of Cinematic Aerospace. “AI lets us help lower-budget jobs get unique shots without having to hiring a larger crew to make it happen.”
But video production is only one role AI supports in done-based imaging. Multiple aspects of UAV operation are seeing AI's benefits, including flight control and stabilization, on-board image processing, image post-processing after a flight, and analysis of flight statistics. AI is playing a role in carrying out missions that would otherwise be technologically impossible or too expensive for the desired application, whether the drones are flown in full autonomous mode or remotely controlled by ground operators.
The combination of small flying devices, high-resolution cameras, multiple sensors, and AI is extending across industries because of four important technologies:
When natural disasters damage buildings, the traditional method is to rely on aircraft and ground-level inspection to assess conditions and eventually inform insurance companies. That activity can be slow and inefficient at a time when fast action is needed to save lives and help families and companies get back to normal.
For example, EagleView is applying AI to drones, wherein it post-processes images and delivers detailed analysis. That lets insurance providers assign the right services when their policyholders need them. “During the hurricane season, our AI solutions identified wind and flood damaged properties at multiple resolutions of imagery,” says Shay Strong, EagleView’s director of data science and machine learning.
The company uses satellite data AI models to quickly evaluate neighborhoods where the damage occurred and to what degree. They then collect ultra-high-resolution drone imagery to gather data on the damaged areas and map specific spatial damage for individual properties. “The drone AI categorizes the damage type, where the damage was," says Strong. "It then obscures personally identifiable information like license plates and faces, then ranks the degree of damage for a property." Once the specifics are determined, humans manually evaluate the model's results and use the damage assessment to determine insurance company responses.
Human intervention at the final stages is still critical to call out when the models are wrong. The input is fed back to the system, and the model continues to learn based on this feedback.
Traffic problems are also coming under the eye of AI-supported drones as smart cities try to solve parking problems in congested areas.
It would seem simple enough to fly a hobbyist-type drone over a parking lot to get a picture of the parking congestion, but that's just the start. In one project, says Tucci, AI-plus-drones addressed a hospital’s parking dilemma. Rather than have a large team sit in each lot for a day and count available parking spaces, an engineering company asked Tucci’s team to take photographs of each lot from the air. That let them count the cars at different times during the day.
“Some of the parking lots are big and long, and staying below the 400-foot altitude limit the FAA imposes meant we couldn't take [video of] each lot in one shot,” Tucci explains. The team used AI software to plan a drone flight path to autonomously fly a grid pattern above the campus every hour, with only a single operator watching the controls on-site. Those images would be combined to create a high-resolution composite image. “The AI in the software allowed these calculations to be completed almost instantly and let us play with adjusting parameters in real time.”
AI functions in the drone’s camera system automatically blurred personal details, including faces and license plate numbers, to keep the images and data anonymous. The drone flew the same course every hour of the business day for two days, producing hundreds of individual images. The company used other web-based AI-enabled software to stitch together the images. The engineering firm used the detailed images to analyze traffic flow and parking trends to make recommendations to the hospital. End result: Hospital staff could park closer to the entrances and visitors could be directed to more appropriate locations.
Some drones are more like airplanes than tiny helicopters and larger than hobbyist-style drones. One commercial drone can support loads as heavy as 770 pounds for distances of up to 1,550 miles. The Dronamics UAV can carry more sensors and computing capacity than its smaller siblings, making it smarter and better able to understand the surroundings and flight conditions.
“Our bigger size and our higher speed mean we need to have a farther range of visibility than most sensors nowadays allow, so that the aircraft has ample time to react and maneuver once it detects an object in its path,” says Dronamics co-founder Svilen Rangelov. The company uses AI in the traditional manner in which aircraft autopilot works, and it pilots its drones via remote control. “We're planning for both fully autonomous and semi-autonomous flight because of the uncertainty of regulatory approval.”
Of course, making an aircraft entirely autonomous and leaving it under control of an AI black box presents more issues than self-driving cars, simply because the stakes are higher. As Rangelov points out, “The industry will first have to prove to regulators that semi-autonomous [i.e., human-supervised] flight can achieve or surpass the safety record of manned aviation, and only then will full autonomy get on the agenda.”
For now, AI in aircraft is best used for collision avoidance and deconfliction from unplanned events rather than as a replacement for human guidance, says Rangelov. “It's a very delicate balance to have an object continuously defy gravity in a predictable way.” AI is an enhancer, not an enabler, of unmanned flight. While the algorithms to sustain flight have been known for decades and are used in autopilots like the ones on Boeing and Airbus aircraft, “those systems can be considered to be AI's predecessors," he adds.
As reliability, sensors, communications, and imaging quality improve, industry is finding more uses for drones as lower cost alternatives to traditional aerial tools. AI is increasingly being used for better utilization of these resources, allowing operators to get better data and more accurate results, and further reduce the time and effort required. As flight regulations change in response to industry demands and reliability improves, more industries will adopt the growing set of advantages of AI-enabled drones.
This article/content was written by the individual writer identified and does not necessarily reflect the view of Hewlett Packard Enterprise Company.