How soon before we see Army helicopters and tiltrotors flying without anybody in the cockpit? The service’s aviation modernization director publicly doubted today the automation would be ready for the first production models set to enter service by 2030. “It’ll be a very tall order to get it into Increment 1,” Brig. Gen. Wally Rugen said. “We’ll have it in another spiral [uppgade] for sure.”
But less than an hour later, the senior engineer for Bell’s V-280 Valor, Ryan Ehinger, told reporters that the tiltrotor had already flight-tested “full auto take-off, full auto landing, and all the points in between.”
Now, the V-280 hasn’t yet flown under computer control for an entire flight. Bell understandably wanted to test each automated process in isolation, with humans in control before and after. But Ehinger said it would be a straightforward next step to string all the automated functions together and have the V-280 do an entire test flight under computer control, albeit with human safety pilots aboard in case of mishaps.
“The capability is there,” said Bell’s V-280 program manager, Paul Wilson. It’s just a matter of slotting it into the flight schedule.
Why is Brig. Gen. Rugen skeptical? It’s inconceivable that Bell didn’t report to him in detail about last month’s test flight long before it briefed the press today. if Bell thinks it can do a fully automated test flight this year, why does Rugen doubt the capability will be ready for the first production aircraft 10 years from now?
To square this circle, we need to delve into the specifics of what the V-280 actually did in December.
/https%3A%2F%2Fsites.breakingmedia.com%2Fuploads%2Fsites%2F3%2F2018%2F12%2FBell-V-280-Valor-in-flight-painted-green-2018-12-18-T042_C026_0110GE_0011-300x169.jpg)
Bell V-280 Valor, the leading candidate for the Future Long-Range Assault Aircraft
What Bell Actually Did – and Didn’t
Over two hour-long flights on Dec. 18th – the two-year anniversary of the V-280’s first flight – the aircraft flew itself for about 20 minutes. The two test pilots aboard took their hands off the controls while the automation took over and took off; then the humans took control again, made sure all systems were a go and handed control back to the computer for the next automated event – rinse and repeat. The back-to-back tests included automated take off, conversion between vertical and horizontal flight, navigation from one waypoint to the next, loitering over a specified area, and landing.
Modifying the V-280 to fly itself was all about the software.
Traditional aircraft and ground vehicles have physical controls that a human operator pushes, pulls and presses; converting them for unmanned operation requires mechanically modifying those controls to work without human muscle power. But Bell’s V-280 – like its civilian 525 helicopter and proposed 360 Invictus combat scout – is a fly-by-wire aircraft. All its controls are electronic, so it’s simple to switch them from human input to computer control.
It’s worth noting that none of this is cutting edge AI/machine learning code, which modifies itself and can produce results that surprise its own creators: The Bell automation is straightforward deterministic IF-THEN code that performs precisely the same way every time.
In fact, much of the V-280’s self-flying software is the same code used in other Bell experiments, the V-247 scout drone for the Marine Corps – and civilian: the Nexus Air Taxi and the Autonomous Pod Transport. While the software has to be customized to handle the very different handling qualities of the four different designs, Wilson told reporters, all four versions have many capabilities in common: navigating from point to point, taking in sensor data, reacting to unexpected hazards….
Hold on, I said. Reacting to unexpected hazards in the civilian world – even when lots of aircraft have to share the airspace and any one of them can make a stupid mistake that causes an accident– is a lot different than in the military world, where there are people deliberately trying to kill you.
Yes, Wilson said, different environments definitely require different algorithms. In some cases you may need the computer to respond to danger much more quickly, potentially up to the physical limits of how hard the aircraft can maneuver.
The V-280’s current software isn’t programmed to respond to, say, anti-aircraft fire, Ehlinger acknowledged. At this stage, Bell just wants to prove the aircraft can fly itself in relatively benign conditions. Flying through hostile fire is a long way away.
Is this the limitation Brig. Gen. Rugen had in mind when he said he doubted the unmanned mode would be ready for the first production aircraft? Since he spoke to reporters before the Bell execs did, we didn’t have the chance to ask. The Army is actually likely to confine automated flight to safe areas before it risks in in a combat zone. But it makes sense the service wants its self-flying aircraft to have some kind of automated survival instinct in any case.