Slashdot Mirror
Silicon Valley Takes a (Careful) Step Toward Autonomous Flying (nytimes.com)
Last week, at a tiny airport in the dusty flatlands east of San Francisco, a red-and-white helicopter lifted gently into the air, hovering a few feet over the tarmac. It looked like any other helicopter, except for the small black cube attached to its nose. From a report: Local officials spent the week testing this aircraft for a new emergency service, due for launch in January, that will respond to 911 calls via the air. But as this helicopter moves police officers and medical workers over the San Joaquin Valley, it will feed a more ambitious project. That black cube is part of a growing effort to build small passenger aircraft that can fly on their own. Today, the helicopter is flown by seasoned pilots. But the new emergency service will be operated by SkyRyse, a Silicon Valley start-up that intends to augment small helicopters and other passenger aircraft with hardware and software that allow for autonomous flight, leaning on many of the same technologies that power driverless cars. These include the 360-degree cameras and radar sensors built into the nose of the aircraft.
"There are many things that must come to fruition before autonomous aircraft start flying people," said Mark Groden, a co-founder and the chief executive of SkyRyse. "But we are developing the technology that can take us there." Sikorsky, a subsidiary of the defense contractor Lockheed Martin, and Xwing, another Silicon Valley start-up, are fashioning similar technology. Others, including Aurora, a company now owned by Boeing, are exploring autonomous flight as they build a new kind of electrical aircraft for "flying taxi services." The initial business plan for Uber's air taxi service, which it hopes to start in five to 10 years, said it would eventually remove pilots from the aircraft.
"There are many things that must come to fruition before autonomous aircraft start flying people," said Mark Groden, a co-founder and the chief executive of SkyRyse. "But we are developing the technology that can take us there." Sikorsky, a subsidiary of the defense contractor Lockheed Martin, and Xwing, another Silicon Valley start-up, are fashioning similar technology. Others, including Aurora, a company now owned by Boeing, are exploring autonomous flight as they build a new kind of electrical aircraft for "flying taxi services." The initial business plan for Uber's air taxi service, which it hopes to start in five to 10 years, said it would eventually remove pilots from the aircraft.
Intel can't make CPUs, all software has bugs, with an endless litany of them undiscovered yet.
Google's cars can't turn left (buggy), no self driving car has yet realistically proven itself in 'all conditions', yet people want FLYING tonnes of steel to be zooming around with hackable, unproven self-flying routines.
These aren't even planes, which STILL have pilots... and multiple computers. Those beasts have heavy maintenance schedules, but when this stuff gets "out" there, is there going to be maintenance checks, visual ones, daily? Weekly? Yearly?
Bah.
Madness.
They'll gamify that.
Google's self-driving cars have problems turning left. No self-driving car is able to drive in all conditions. Let's fix that before we try to make self-flying planes.
If someone wants to fly in a computer, that's their problem. But if they come crashing down on me, that's my problem, so this is only going to fly if these things cruise at high altitude (noise) and if they self-destruct into tiny pieces in case of an unscheduled descent.
They are focusing on the wrong problem. Chartering a helicopter with pilot costs around $400-600 (depends a lot on location and type of helicopter). Yet if you search for helicopter pilot wages, it comes out at about $40-60 per hour. So even if their expensive autonomous box completely replaces the need for a pilot it would only reduce the cost of helicopter travel by around 10%. I doubt that is going to cause a sudden explosion in helicopter travel.
The real cost is in the helicopter, which is an expensive machine to start with that has to be regularly maintained and inspected to ensure it doesn't decide to fall out of the sky.
At least electric helicopter startups offer some ideas on how they might reduce maintenance costs (by having fewer moving parts) but even then, I think these people just do not understand how much extra cost and effort separates the buggy, cobbled together PHP code they used in their last insta cat book website startup, and safety critical software/hardware that cannot be allowed to kill people.
So firetrucks will also be hit from above instead from behind?
Weather and 3d movement of a helicopter.... Plus you can't just tell a helicopter to put on the brakes or turn onto the shoulder and stop altogether in an accident or approaching a situation it can't understand...lol
"Today, the helicopter is flown by seasoned pilots"
I've noticed this about the self-driving cars and other vehicles: they sure have a lot of drivers in them.
On an unrelated note, I had the auto-pilot go out at 37,000 feet the other week. The air data computer had a malfunction and the autopilot could not figure out the airspeed and altitude. The passengers in the back didn't notice because we had two human pilots who took over manually and continued to fly the aircraft safely to the destination. We've been working on automation since Sperry put a gyroscope in a primitive biplane back in 1912. After 106 years of constant refinement, contrary to popular opinion, airplanes do not fly themselves reliably and we still need humans up front for safe operation. While I am sure drones are useful for things such as aerial photography, banner towing, etc. I do not see drones carrying people within my lifetime. If they did, it would be more of an aerial stunt than a viable, safe method of transporting people. I've seen automation fail too many times, and would not trust one enough to put my family on it. You shouldn't either.
well an FAA level code audit can be a big pain with a lot of QA costs.
To me, autonomous flying is a lot easier than driving.
In some ways. However, several things make it harder.
1. Any problem is likely to be a fatal problem. Airplanes don't have "fender benders."
2. Three dimensions are a lot harder than two. (And cars actually only have one and a half directions of mobility-- you can't drive sideways).
In addition, for airplanes (but not 'copters): you can't stop. In a car, if you're not sure what to do, you can just stop and wait for traffic to clear. Airplanes fall if they stop.
http://www.geoffreylandis.com
'Silicon Valley Decides to Experiment with Hypothetical Autonomous Flying That Will Never Materialize Because it Defies the Laws of Physics and Mathematics and is Generally a Stupid Sci-Fi Fantasy'. And I'll bet they dump billions into this vapor instead of cleaning up the poop in their cities.
You can have automated flight controls, stabilization, routing, emergency avoidance, and the rest of it. But there should always be a qualified person on board who is monitoring and can take control to stop, change directions, etc.
It doesn't matter how many Mountain Dews and Cheese Pizza you feed your Dev Team, they cannot anticipate every possible situation.
When Fascism comes to America, it will call itself Anti-Fascism, and tell you to give up your guns.
Go home, David Brock. You've had too much to drink.
Controlling a real helicopter at slow speed is quite difficult. But the quadcopters have a lot of smarts in their controllers, so that their operators just say up down left etc. and the machine does it. Years ago I talked to someone that was putting similar capability into some military helicopters.
So this takes the technical skill out of flying them, so that anyone can do it. Converting musicians into disk jockeys.
I worked for an aerospace flight automaton company a decade ago. They had automated flight and landing, and most large companies were using the automation. I guess holding patterns at airports weren't automated though because the airport operators don't have an interface into the autopilot routing system? This article should have focused on autonomous helicopters which are actually new. The FAA is very unlikely to approve flight without a pilot because they are zealous about safety. I also don't see AI being approved because the FAA requires all flight system code to be deterministic and both machine learning and AI are not deterministic.
Making flying a helicopter easier and safer through AI based warning systems definitely would have it's place though.