If Georgia Tech scientists are right you could eventually be mixing it up with autonomous drones that use a listen and avoid technique in the pattern at untowered airports. The researchers are figuring out how to make the drones make sense of position reports in the pattern and near the airport so they can stay out of the way of crewed aircraft. “This is how humans have operated safely for decades around these airports. So, if we start flying robots here, they should operate in the same way,” Sundhar Vinodh Sangeetha, a robotics Ph.D. student and first author of the study, told the university’s engineering journal. “We humans shouldn’t have to change the way we act because there’s a robot operating around us. That was a big motivation.”
The system uses AI and speech recognition technology to listen in on pilot radio calls to determine where the aircraft are based on the position reports. There was no mention of how it might deal with NORDO aircraft or other drones, but the scientists did say they’ve cut the error in determining aircraft trajectory by 65% to about 500 yards. This is how that sounds in scientific language: “We leverage automatic speech recognition and large language models to transcribe and interpret pilot radio calls, identify aircraft, and extract discrete intent labels. These intent labels are fused with observed trajectories to condition a temporal convolutional network and Gaussian mixture model for probabilistic goal prediction.”


I’m sure it’ll be much better at understanding everyone’s radio calls (particularly the ones getting stepped on by calls from the next field over, or the non-standard ones half of everyone makes, or the non-existent ones by folks without a radio) than the robots who answer the phones at most major businesses these days.
”REPRESENTATIVE!”
”Did you say ‘Shipping and Receiving’?”…..
I just press “O”.
I am sorry, but that number is unavailable. Have you tried our Chat service?
1) Those radio calls aren’t required
2) It barely works when people are actively participating
3) A system that is reliable 95-99.5% of the time fails one time out of every 20 to one time out of every 200.
I applaud the effort. Our aircraft typically are equipped with light beacons, ADS-B, radios, the Mark-1 eyeball, and perhaps a pilot who will make proper call-outs in the pattern, none of which alone is sufficient, but taken together provide a greater measure of risk reduction. Same for drones. Every little bit helps, even with the limitations of any one system. Just so long as none of us relaxes our guard…
“perhaps a pilot who will make proper call-outs in the pattern”
Which means using the last 3 of their tail number, rather than a descriptive color. I always challenge anyone who says that a descriptive callsign is better by asking if they can tell what color an aircraft is even 1 mile away. The only exception I have found is a completely-yellow Cub; all other aircraft effectively look white-and-black from a distance.
Agree!
Whichever village idiot CFI started the trend of using colors in lieu of registration numbers, needs to be tarred and feathered, along with those who insist on doing this at nontower airports.
There is ONE time that’s a valid procedure, during the OSH airshow.
As a semi-colorblind individual, I mostly agree – but would everyone understand just exactly what I meant when I identified myself as a National School Bus Chrome Cub?
[Probably so…”It’s that idiot again.”]
I applaud their effort for realizing that the “robots” should conform to us *human* pilots, rather than human pilots having to accommodate drones. However, I wonder how well it handles mis-position reports, i.e. someone reporting 10 miles east when they mean 10 miles west. And how does it correlate someone reporting as a “blue-and-white cessna” with their actual tail number?
I’m sure this will sound like an extraordinarily stupid question, but why do drones need to use the traffic patterns at non-towered airports? If I were operating a drone near one of these places, why wouldn’t I use the ole helicopter strategy of flying beneath the flow of fixed wing aircraft?
I think the reason is the same as for helicopters — do find out where the fixed wing folks are and avoid them on the way to landing. They don’t intend to fly the pattern; just to figure out, based on the radio calls and standard conventions regarding pattern, where those aircraft will be and avoid flying across their path
A ha! That does make much better sense if that’s what they intend. So they’re thinking about drone hobbyists operating in this case?
I don’t think it would be for drone hobbyists, who really should not be operating near an airplane airport.
Ah yes. Good point.
Do you know if there is any recorded instance where a drone ever got away from the operator? Has it been 100% perfect?
I fly mostly at non towered airports and get extremely frustrated at all the improper terminology that is used, practically no two pilots are versed in what the Pilot Controller Glossary teaches us about traffic calls, and maybe 10% of what I hear understand what AIM Chapter 4 and Advisory Curcular 90-66C teach us about non towered airports. So many new pilots being taught at larger 142 flight schools are clueless about non towered airports which they refer to as uncontrolled airports. So if all this incorrect phraseology is used how can a drone be properly “trained”? Scary.
Will the drones be trained to blurt, ‘Any traffic in the area, please advise’, make straight in approaches or ‘modified base entries’, or to be sure to tell everyone what color their plane is, like idiots do these days?
No? Then put them in charge.