Military operators may soon be able to control overhead unmanned aerial vehicles (UAVs) with just a quick word into a microphone.
Connecticut-based start-up Primordial Labs is developing what it calls a “tactical [artificial intelligence] assistant” the company says will enable troops to coordinate the movement and actions of a UAV using verbal orders.
“We built an AI system that combines conversational voice interface and a platform independent autonomy,” says Mick Adkins, director of product at Primordial. “Those put together allow operators to have less cognitive burden and heads-up, hands-free control of unmanned systems,” he adds.
Known as “Anura,” the voice recognition AI will allow a UAV operator to cue the airborne craft using pre-set vocal cues delivered over a radio. These could be instructions to change the flight pattern or proceed to a designated point.
Primordial aims for Anura to be so-called platform independent, meaning it will be able to function with a variety of UAV types, including teaming those different aircraft to work collaboratively.
“That’s really important to a lot of user communities,” Adkins notes.
The company has been developing its voice command AI assistant in collaboration with Teledyne FLIR, the Oregon-based producer of small UAVs. Specifically, Primordial has been testing Anura on FLIR’s Black Hornet “nano UAV”.
With a total length of just 168mm (6.6in) and weighing only 1.3kg (2.8lb), the Black Hornet is designed to be carried by a single dismounted soldier, who can use the diminutive helicopter to inspect the interior of buildings or provide overhead reconnaissance.
“Perhaps I have multiple Black Hornets and I want to deploy them to recon a certain sector,” Adkins says. “So, I would key the [microphone] and say, ‘team search Point Alpha’.”
The Black Hornet’s standard control tablet would still display live video feed and allow for tactile control of the UAVs.
Jennifer Rochlis, vice-president of products at Teledyne FLIR, says it will collaborate with potential customers interested in the voice command technology to develop the language.
“It makes it really customisable,” she says.
In addition to creating an effective voice recognition system, Primordial also had to overcome the challenge of background noise. The battlefield can be an extremely loud place, with the staccato of machine gun fire and booming artillery easily drowning out voice commands.
“We kind of build some of that dirtiness into the model,” Adkins says of the algorithm that powers Anura. “We’ve tested it in cluttered environments.”
Eventually, the airborne UAV may even be able to talk back to its human operator, relaying critical information about its status or programmed alerts using auditory language, rather than just visual cues.
Rochlis says such a concept would take advantage of the human brain’s efficiency at processing speech and other audio cues. She gives the example of an altitude alarm in an aircraft cockpit.
“Rather than having just an alarm beeping at you, and you have to figure out what that means, it says ‘pull up, pull up’ in that sentence,” Rochlis notes.
“In this way, we can just get a huge advantage in the efficiency of exchanging information,” she adds.
Primordial is moving quickly to expand the number of systems on which Anura is available. Adkins says the company is currently developing the auditory AI assistant on four UAV types, with a recent contract set to expand that number to seven by the end of year.
“We know that operators have platforms they prefer, platforms that rely on, platforms they trust,” Adkins says. “So that’s where we want to live; on those platforms.”
The Black Hornet is in service with more than 50 countries according to Teledyne FLIR, including the USA, UK, France and Australia. The company says it has shipped some 10,000 units of the nano UAV.