Comment by sxyuan | Hacker News Reader

Comment by sxyuan | Hacker News Reader

There's an increase in incidents because ATCs are overworked and understaffed. The knee-jerk reaction is to try to automate the work away. A few problems with that kind of thinking:

- Automation already exists - e.g., the collision detection warning mentioned in the article. I don't know much about air traffic control but I imagine there's some element of defense in depth. ATCs dozing off or running on fumes won't cause collisions right away (these incidents are mostly close calls) but it sounds like it's only a matter of time before all the other defenses fail and a major accident happens.

- There's always going to be work that's hard to automate. Plenty of software teams have on-call rotations to cover production services. You'd think that if any field would have their operations be fully automated, it's software, but nope. We need to stop thinking that the physical world is easier to deal with - it's often much harder.

- Automation costs money to develop and deploy. Apparently one of the early outcomes of union action back in the 70s was faster automation rollout. [1] For an executive, why spend money on automation if you could just squeeze more out of your workers and compromise on safety instead?

[1] https://en.m.wikipedia.org/wiki/Professional_Air_Traffic_Con...

The lack of automation is causing ATC work to be stressful. People orchestrating complex procedures over a shared vhf radio channel is inherently hard and error prone. Which is why it involves people endlessly confirming and double checking known bits of information.

Automation could simplify that a lot and remove a lot of the uncertainty. Making communications easier might save some lives. A lot of accidents are caused by human pilots and typically that involves them making mistakes in stressful situations. Dealing with ATC is very stressful for pilots. Task saturation is something they all fight throughout their career. IFR pilots have to train and fly a lot to stay current; the ability to keep up with the radio communication and fly procedures safely is quickly lost when they don't. That's how hard it is. A pilot that is not current is not a safe pilot. That's why the rules are so strict for this.

If you look at this space from a distance, it's kind of obvious that more automation is going to be inevitable. Airspaces are busy now and they are going to get a lot more busy in the next decades. At least part of the traffic is going to be autonomous. Those planes will likely be communicating with ATC via networked protocols that have yet to be proposed and standardized. There is no reason why those protocols could not work with other planes. And it would free up ATC staff and pilots to do more urgent things. The way ATC is currently done is not scalable or future proof.

> The lack of automation is causing ATC work to be stressful. … > And it would free up ATC staff and pilots to do more urgent things.

There’s a problem that can happen with this, though, wherein the “easy” routine work is automated out and the number of humans involved is scaled down, but the work those people do then winds up being 100% the urgent/stressful stuff.

I’ve heard that ER doctors have been experiencing this in recent years–not quite with automation so much as just offloading, as less serious cases are picked up by nurse practitioners/urgent care centers, staffing has been adjusted and the actual physicians wind up with a steady stream of stressful high-acuity cases back to back to back.

I agree we might need a minimum staff level of, say, 1 air traffic controller per airport, but unless I've been grievously misinformed about the safety of flying, urgent/stressful emergencies should be very rare.

I suspect the main thing limiting ATC automation is the fact ATC is all done by voice using incredibly shitty radios, and equipment upgrades take like 50 years to roll out.

> I’ve heard that ER doctors have been experiencing this in recent years–not quite with automation so much as just offloading, as less serious cases are picked up by nurse practitioners/urgent care centers, staffing has been adjusted and the actual physicians wind up with a steady stream of stressful high-acuity cases back to back to back.

Yet at the same time, ERs (near me at least) are still inundated with so many patients that average wait times are 4-5hrs.

Maybe you are in an oasis of reason where this doesn't happen. But most ERs I have heard of that are overcrowded is because people go to the emergency room when there is no actual emergency.

I’m sure there’s a number of reasons for this - after about 5 PM, your options are typically the ER or an urgent care. UC’s usually require payment up front, and for people who either don’t have insurance or who have high deductibles, the cost of care at a hospital (which is way more but can be put off) vs the up front cost of urgent care probably pushes people towards ERs.

There’s likely other factors - lack of medical education/first aid, consulting Dr Google and then having anxiety that maybe that slight pain is something horrible, etc.

A public health system could alleviate this stress, if done well. Clinics that are open later than “business hours”, readily accessible telehealth providers, and better public health education could all come about from a public system. Privatized, all of these can only exist with a profit motive, and the insurance and health system are costly.

Isn't that how it should be? Why should a highly paid doctor ever be involved in a routine case that could easily be handled by a nurse?

As an engineer I only get tickets that are too difficult to be handled by support and require code changes. Why should medicine be any different?

Software doesn’t quite have the same fungibility to allow for this, but imagine your job was only being responsible for urgent, high severity tickets throughout your company.

There was no time to mull over a problem, pair code with a peer or junior, or talk to a product owner because everything was urgent and on fire.

Another reason not to scale down the number of humans is that they're needed when automation breaks.

Planes could very well fly themselves 99%+ of the time, but you want to have pilots when there are in flight issues that aren't handled by automation.

They've been talking about automated ATC for at least 20 years now, in part for getting the benefit of more flights arriving and departing per hour than humans can safely manage. https://en.wikipedia.org/wiki/Next_Generation_Air_Transporta...

They have been working on automated ATC for a lot longer than 20 years. I worked at Eurocontrol 1991-1996 where they had been developing a system called ARC2000 since well before I arrived. As the name suggests it was intended to take over ATC in Europe by the year 2000. I worked on a less ambitious system that helped visualise conflicts for manual resolution but that was not a big priority so I eventually left.

My internship in the mid-1990s was with a professor studying VR systems for ATC for low-visibility operation. At the speed this field changes (or doesn't), I doubt I'll see it in my lifetime.

I'm sure it would be done better if implemented over audio, but the image of a frustrated pilot yelling "REPRESENTATIVE! REPRESENTATIVE!!!" just came to mind.

An uncurrent pilot is not inherently an unsafe pilot.

> Dealing with ATC is very stressful for pilots.

This is not my experience at all, and I suspect is also not at all for air carrier pilots who deal with ATC on every flight.

Sure, the guy in the 170 or J3 who mostly flies off his farm might be stressed by flying into the big city airport, but the vast majority of pilots who interact with ATC regularly don't find it particularly stressful.

In fact, given a choice (and as a VFR pilot I have the choice), 99 times out of 100, I will voluntarily call up and receive flight following from ATC. There's pretty much no down side.

> Those planes will likely be communicating with ATC via networked protocols that have yet to be proposed and standardized.

The protocols are not the issue, it’s the high-bandwidth, low-latency wireless network that’s missing. LEO satcom may be an option in the future, but it isn’t yet for flight and safety critical real time applications.

I just finished building a pilot information system for an airline that uses ACARS to communicate. It’s a backup system for a backup system that supports a redundant electronic flight bag tablet.

Automation in ATC exists a little yes, as you said for collision detection, ground proximity warning and runway infringement. That's mostly it.

We have some cool toys to make life easier like CPDLC to communicate via data-link instead of voice, multi layer radar screens with tons of functions, ATIS generation etc but those are not automation, just QOL improvement and if anything goes wrong we go back to old school ATC by voice and maths.

There is a lot of automation in the data part of ATC, so flight plans processing, radar correlation, meteo data etc etc. But this is just normal IT stuff in 2023.

Some cool project are getting there, like virtual towers with augmented reality by using cameras instead of looking out a window. My company did recruit an expert in big data and ML to develop new tools and automate stuff, but always data related, nothing to do with controlling and aircraft on a radar screen.

Instead of more automation, perhaps it would be good to occasionally introduce phantom traffic which is designed to require intervention on their part (e.g. fabricate two flights and put them on a collision path).

The air traffic controllers would be expected to flag them, or intervene, or do whatever they would realistically do in that scenario.

From that, you can determine who's asleep at the wheel so to speak, and send them home. If necessary, planes may need to be grounded until properly alert staff can be available

Not sure how that would work.

"Citation one-three-niner-delta-uniform, Denver Tower, adjust altitude to seven three hundred, bearing one zero five."

"Citation one-three-niner-delta-uniform, Denver Tower, please respond."

Also FYI it's not like anyone can actually be asleep because pilot is required to establish two-way ATC contact to enter the airspace.

RAAS (Runway Awareness and Advisory System), the equivalent system installed on aircraft, also needs to become standard equipment. It is for some airlines, but it is by no means mandatory.

> Plenty of software teams have on-call rotations to cover production services.

There are plenty of valid criticisms of the cloud but this is what you are really paying for. The value proposition of the cloud is infrastructure-oncall-as-a-service.

You usually still need someone oncall who determines whether it’s a cloud problem or your bug.

Reply on news.ycombinator.com