Start-up’s AI-enabled system teaches drones to assist fight wildfires
By DRONELIFE Feature Editor Jim Magill
:
When responding to wildfires, time is critical. Response commanders have to have the ability to rapidly access data to reply vital questions – where is the fireplace the most well liked? What direction is the wind blowing? Are there any humans positioned within the danger zone – before making any decisions on fighting the fireplace.
A start-up software and drone company in Washington state has developed a system for answering these questions using artificial intelligence (AI) programs and highly stable and agile coaxial drones.
Launched in 2022 by three aviation industry experts, DataBlanket’s mission is to empower firefighters with latest tactical tools and data.
“To me firefighting felt like an issue that has needed addressing for quite a number of years,” DataBlanket Co-founder and CEO Omer Bar-Yohay said in an interview.
Bar-Yohay has an intensive record of serving within the Israeli military, in addition to being a frontrunner within the civilian defense electric aviation industries. He formed DataBlanket with Yair Katz, a long-time friend and colleague, and Gur Kimchi, who cofounded and led the Amazon Prime Air drone-delivery service.
After serving Within the Israeli defense forces for 12 years, in 2015 he founded Eviation Aircraft, where DataBlanket’s co-founder, Katz served as chief operating officer. Eviation developed, built and flew the world’s largest all-electric commuter aircraft. Bar-Yohay left Eviation in March 2022 to form DataBlanket.
His military background, especially his work in special operations, led Bar-Yohay to imagine that among the technologies and tactics developed for Israel’s defense could possibly be applicable to battling wildfires. “There must be more joint operations per se, more seamless exchanges of data between ground and air operations and higher situational awareness generally,” he said.
The DataBlanket system employs coaxial drones, manufactured by Ascent AeroSystems. Ascent is an American company, whose products are approved to be used by the U.S. military under the Blue sUAS program, a fact appreciated by many public service agencies seeking to avoid the usage of drones produced in China. The fast-flying and agile drones are capable in of operating in inclement weather conditions resembling high winds, typically related to wildfires.
“Large fire events are likely to create their very own weather challenges, with crazy inversions of winds,” Bar-Yohay said. “We wish to have the ability to fly there even when other assets can’t be operated. The drones can operate in winds of over 50 miles an hour.”
To equip them for fighting wildfires, DataBlanket has equipped the Ascent drones with its own sensors and AI-driven technology. “On top of the Ascent platform, we’ve integrated a really significant variety of sensors, obviously to see what’s happening, but additionally for sense-and-avoid and for airspace mitigation,” he said. The corporate also adds a platform developed by Santa Clara, California-based NVIDIA that runs all of the computer-vision pieces, onto the drone.
The system is designed for quick deployment, with the incident commander having the ability to deploy as many as eight drones directly. “Those drones will look towards the realm you’ve asked them to scan, discover major things like smoke, fire, stuff like that, and in case you’re in search-and-rescue mode, they’ll actually map more accurately.”
Under the corporate’s firefighting management system, the AI technology assists within the operation of the drones, but doesn’t have full operational control of them, Bar-Yohay said.
“So, there are very clear boundaries as to what the system is allowed and never allowed to do,” he said. “The management of the assets inside those boundaries may be very much non-deterministic and uses what you might call AI to optimize them.”
Once the drones are deployed, and the system receives input from the computer-vision software, it may make its own decisions based on an algorithm, which tells the system how deeply it should dig into the visual data before determining if that data is important enough to require filing a report back to the incident command.
“Probably the most painful things in automated detection is fake positives. You don’t want the entire system to go crazy for a cloud, if it’s not smoke. You don’t want the entire system to be thrown off since it identifies something that was hot, but was not fire,” Bar-Yohay said.
For instance, if the system detects a figure walking upright throughout the fire zone, it could determine to focus more resources to take a more in-depth look to find out if that figure is a lost person a or a frightened bear. The system could order a second drone to drop beneath the forest cover to fly lower, or use thermal sensors to discover the mysterious figure.
The system uses its assets to reinforce the worldview that it perceives and to continuously create an increasing number of knowledge about that worldview, “helping you, the operator, understand through the app, what’s on the market to see: for real and what’s vital. And we found that these are the 2 areas are barely different, where AI algorithms really do pretty much.”
DataBlanket, which is within the pre-revenue phase of its corporate life, last December closed its seed round with lead investors: Breakthrough Energy Ventures, founded by Bill Gates, and Innovation Endeavors, which funds technological startups and which was co-founded by former Alphabet CEO Eric Schmidt. The corporate has also attracted investment from Latest Vista Funding in addition to from family and friends of the co-founders.
Bar-Yohay said the corporate has secured considered one of the authorizations it must operate from the Federal Aviation Administration and has filed for critically needed authorization to conduct drone flights beyond the visual line of sight. “I believe that it would still take a number of months and a number of more tests for the FAA to present us the total BVLOS capabilities that we’ve asked for but we’re on the correct path there. I might say the FAA has been very, very conscious of what we’re doing and so they’ve asked the correct questions,” said.
He said DataBlanket has already received inquiries into the usage of its system from firefighting agencies, resembling in Orange County California and other jurisdictions across the West Coast. The corporate plans to start conducting pilot programs in the sector next 12 months with a broader deployment of the technology within the late-2024, early- 2025 time-frame.
Given his background, Bar-Yohay also expressed his deep concern over the fate of his countrymen during these fraught times within the Middle East.
“We just hope all the pieces works out alright. Let’s hope it would end soon.” he said.
Read more: