It’s not really recommended. GPS signals are correlated and so you can get relative positions much more accurately than absolute ones, so it’s not totally insane to do something like that, but for a small plane I think you’d want less error.
You don’t need that, anyway. Modern control theory is great at sensor fusion. You can figure out the true orientation from the acceleration and magnetic orientation. Not in steady-state flight, you say? That’s fine; add gyros and you can correct for any additional accelerations.
The basic strategy is essentially to use dead reckoning (based on the known performance of the craft and current control inputs) to figure out what the plane “should” be doing, and then using the known sensor inputs to correct the dead reckoning. This avoids the inevitable drift and error that you get from dead reckoning, but also allows you to predict values that you may not have direct access to, such as air velocity (not just airspeed). And do so with accuracy that beats any one sensor in isolation (such as a position figure that beats GPS position).
You can also gracefully degrade from sensor failures, such as from a loss of GPS position. You might want that for navigation but it’s not strictly needed for just flying the plane.
Barometric pressure is also useful as an adjunct to GPS altitude. Even with correction, barometric pressure has poor accuracy. But it has great precision: it can detect centimeter-level changes in height. GPS altitude is the opposite; excellent accuracy but low precision. Combine the two and you get excellent accuracy and precision.
Another sensor I’d want is propeller torque. This tells you (along with RPM) how much power the prop is delivering. That acts as an adjunct for other data such as acceleration.
Mass would also be useful. In fact, load cells on the landing gear would be useful for all kinds of reasons. Lots of planes crash because they’re overloaded (either in absolute terms, or relative to conditions).
These are just completely unrelated problems. Self-driving cars have to navigate around other cars, as well as a 2D surface that’s plastered with obstacles. A wide open 3D space is trivial in comparison. It’s like an autonomous car that only has to “navigate” the Bonneville Salt Flats.
Cameras aren’t going to be used at all, so visibility is a non-issue, let alone interpreting the data. The problem doesn’t require AI in any meaningful sense; they may use machine learning for something or other, but really this is all classical control theory.
Again, it’s not a fully autonomous craft. It’s a craft that abstracts the control inputs from the control surfaces. This isn’t even that alien a thing; the throttle in your car is totally computer controlled and only has partial resemblance to the position of the gas pedal.
It’s up to the pilot to avoid birds, let alone other planes. The autopilot will want to correct for aerodynamic disturbances, like wind shear–but one of the nice things about control theory is that you can often prove that you have enough margin to correct for disturbances under a certain magnitude. And then ensure the plane always stays within the bounds required to have some level of margin.
Not particularly. It’s been around for decades, and its safety record is pretty good. Some vulnerability to icing early on was improved upon.
The most notorious fatal incident wasn’t the plane’s fault — it was the pilots who (in Taipei) shut down the one working engine (the other engine failed a few minutes after takeoff).
From an outsider’s perspective, it seems odd that the response to “we can’t make money trying to be that safe” is “ok, you can be a little more unsafe.”
IMHO, if you have to compromise operational safety to have a viable business model, your business model is by definition not viable. But I’m an outsider, what do I know?
You make the mistake of thinking of “safety” as a binary that is either fully present or fully absent. It is never either of those things.
The only sentence structure that makes sense near the word “safety” is “more safe” and “less safe”.
All safety decisions in every industry are about how much more safety can we buy at an affordable cost. If the cost is unaffordable, that incremental safety goes unbought. A wise system and a wise regulator understand that there is a finite (if very hard to calculate precisely) budget for safety and hence they try to obtain the maximum incremental safety benefit from the initiatives they undertake.
Chasing the latest headline-grabbing but ultimately very rare mishap trying to invent ways to prevent it from ever recurring is just about the polar opposite of wise safety management.
NTSB did not do that last move in this instant case. But these recommendations are sort of their perennial Oldies but Goodies that get trotted out whenever they can. Despite their practical impracticality.
One of the reasons to separate the accident investigation / recommendation agency from the industry regulatory agency is precisely to let the former be the “conscience” of the latter without interference from the chain of command. By law, FAA has to consider the dollar consequences of their regulatory decisions. By law, NTSB is forbidden from considering such things. Naturally that results in differing outcomes. As it should.
If you have some small, remote community that only sees enough passengers per day to warrant a flight on a Cessna 208 with one engine and one pilot, you are in some sense “less safe” than you are on a plane with two engines and two pilots. But when you only have at most nine paying passengers among which to spread out the cost of operating that flight, the expense of having a second pilot may very well put the airline in the red. And then the question becomes, should that community simply not have airline service? Or are we ok with being slightly “less safe” on that flight?
Another factor in the safety equation is mentioned in that report - the nature of charter flying is inherently more hazardous than most airline operations. I can attest to this, having worked in both situations.
There are a number of reasons. The big one is that simply by existing as an “on-demand” operation, charter pilots are often sent to airports they are unfamiliar with, and may have little time to plan. Day, night, good weather, bad weather. Whereas airlines generally go to a set of known destinations on a known schedule. Unless you declare most of the charter side of the industry too dangerous to exist, I don’t know how you get around that.
Another factor is less support. When I was an airline guy the flight planning was done by professional dispatchers and reviewed by the pilots, while in in most charter operations the pilots do everything including often booking their own hotel rooms. The training is mostly similar, but not quite identical. Depending on the size of the charter operation, the quality of the safety management system can be an issue. Whereas at airlines they’re usually operating at such a scale that it’s more likely to be substantial and organized.
Dammit - I tried to edit a typo after the time limit and accidentally deleted the entire post.
Here it is again:
Another factor in the safety equation is mentioned in that report - the nature of charter flying is inherently more hazardous than most airline operations. I can attest to this, having worked in both situations.
There are a number of reasons. The big one is that simply by existing as an “on-demand” operation, charter pilots are often sent to airports they are unfamiliar with, and may have little time to plan. Day, night, good weather, bad weather. Whereas airlines generally go to a set of known destinations on a known schedule. Unless you declare most of the charter side of the industry too dangerous to exist, I don’t know how you get around that.
Another factor is less support. When I was an airline guy the flight planning was done by professional dispatchers and reviewed by the pilots, while in most charter operations the pilots do everything including often booking their own hotel rooms. The training is mostly similar, but not quite identical. Depending on the size of the charter operation, the quality of the safety management can be an issue. Whereas at airlines they’re usually operating at such a scale that it’s more likely to be substantial and organized.
I remember the TWA 800 disaster. A spark ignited vapors in a fuel tank, which brought down a 747. I heard suggestions that planes could be fitted with nitrogen canisters that would release the gas into the tanks. Even if there was vapor and spark, there wouldn’t be oxygen, so no explosion.
Which sounds good, and would have prevented that particular accident; but what happens when that pressurized nitrogen system bursts? That could take out a fuel or hydraulic line and cause an accident. Every change you make carries its own risks and vulnerabilities.
So instead what was mandated is a nitrogen generation system that uses electricity to extract nitrogen from the air and inject that at low pressure into fuel tanks. It’s been 20 years since anyone has ridden on a non-nitrogen equipped airliner.
Was that system analyzed for potential failures, and designed to solve more problems than it could potentially cause? My point wasn’t really about that specific system, but that each change carries its own risks.
People jump to conclusions, and suggest fixes with the assumption that they will always work. They won’t, because nothing always works.
Of course it was. And for cost effectiveness. It’s also considered non-essential equipment and, although highly reliable on the types I’ve flown, isn’t required for dispatch as long as it’s fixed in the next week-ish.
Your larger point though is spot-on and deserves reiterating.
Armchair experts are very prone to assuming their understanding is complete and their proposed solution has no downsides and certainly adds no new risks. Which thinking is of course blinkered bunkum.
The classic example was the rudder control problem on the 737 Classic = 300 / 400 / 500. On 2 occasions it went apeshit, rendered the aircraft uncontrollable, and killed ~150 people each time. Once NTSB plus Boeing figured out what happened, Boeing was tasked to find a fix.
The problem was the existing defective system hadn’t killed anybody on something like 25 million other flights. So they needed to be sure their changes were safer than 2 fuckups per 25M. A tall order.
3 years later they had a plan. It was approved, eventually retrofitted into the then existing fleet, and incorporated in all subsequent new build 737s up through current MAX production today. Nobody has been killed by a rudder apeshit since. Or any other 737 rudder anomalies.
The armchair experts went nucking futz that Boeing wasn’t required to fix the whole fleet in 3 weeks.
See that? That means it was a joke. 