I met someone who says he helped develop TCAS, and that it decides which one to climb or descend based on the registration number.
I'm guessing that it does not issue turn instructions because it only gives a Resolution Advisory when a collision is imminent (otherwise they would be giving RAs all the time). At altitude, where the air is thinner, it is much easier for a plane to immediately climb or descend, planes turn very slowly when the air is thin.
> But what if the other aircraft is "thinking" the same thing? That's the heart of this question. The answer is, whichever TCAS makes a decision first, and then in case of a tie, the lower Mode S address "wins":
This story is a moving example of the self-improvement going on in global air traffic industry; perhaps sometimes slowly, but definitely self-improving. The following book compares the transparency of the air traffic industry with the lack of transparency that currently exists in the medical sector, where errors are denied:
Good answers in sibling comments, but for those who don't want to grovel through a lot of stackexchange answers here is the TLDR: There are two reasons:
1. When TCAS was originally developed, the direction information to the other aircraft was not very accurate, being obtained only via the TCAS antenna. Nowadays GPS information is transmitted via ADS-B and so it's much more accurate, but taking advantage of this would require a major redesign. It may happen eventually, but...
2. Aircraft can change altitude faster than they can change heading. Also, most aircraft can change pitch faster than they can roll, and they are longer and wider than they are high. So effecting enough altitude change to avoid a collision can be done faster than effecting enough heading change to do so.
I don't think the GP was asking about changing heading instead of altitude; I think they were asking about changing heading in addition to altitude. Sure, the altitude change is going to achieve better separation faster, but a heading change in addition to the altitude change will act as a backup in case for some reason something goes wrong with the altitude change.
In this particular accident, it seems like commanding a heading change (to the right for the Tupolev and to the left for the DHL flight) in addition to an altitude change would have avoided the collision.
This is fallacious reasoning. "X would have prevented Y, and Y was a bad outcome, therefore we must do X." You can plug in many things for X which reveal this to be an unsound argument.
Yes, intuitively being able to change headings seems like a good thing except that 1) it adds expense and complexity, and thus produces new potential points of failure and 2) it's not necessary. The problem was not that an altitude change was insufficient, the problem was that two sources of command authority (TCAS and ATC) contradicted each other. The solution is simply to add a new rule: if TCAS and ATC disagree, go with TCAS.
> This is fallacious reasoning. "X would have prevented Y, and Y was a bad outcome, therefore we must do X."
I never said we must do X. It's a suggestion which might or might not be worth considering. From the article it appears that the TCAS system was designed with the assumption that pilots would always immediately do what it said. This accident obviously violated that assumption. The question is how best to respond to that as far as the design of TCAS is concerned; the eventual tradeoffs might lead to the answer "do nothing to TCAS" but that doesn't mean it isn't worth considering possible changes.
> it adds expense and complexity
Yes, this is always a tradeoff with safety systems. But the heading change rule would be even simpler than the altitude change rule (which requires negotiation of possible conflicts between the TCAS systems in the two aircraft): the heading rule is simply "turn away from the other aircraft", i.e., if the other aircraft is to your right, turn left, if it's to your left, turn right. That seems like logic that could be implemented, checked, and verified fairly easily, if it were decided to do so.
> it's not necessary
When it comes to safety systems, "necessary" is always a judgment call. Yes, cases in which adding a heading change to TCAS instructions would be helpful are rare (like this crash was). But all aircraft accidents are rare. No rule gets followed exactly as it's given 100% of the time. Taking that into account when deciding on what safety systems to implement is perfectly justified.
> The solution is simply to add a new rule: if TCAS and ATC disagree, go with TCAS.
Or more precisely, to explicitly make that a rule worldwide, instead of just in the US as it was before this accident.
Yes, this is the response the system settled on, but this, too, involves a tradeoff: this rule has to be implemented by humans, and humans are less reliable than automated systems. The international air safety system evidently decided that humans were reliable enough in this case for this rule to be sufficient; but that doesn't mean they didn't make a tradeoff.
> I think they were asking about changing heading in addition to altitude
Heading and altitude are NOT independent. When an aircraft makes a turn, it banks (it's not like a flat turn you make in your car), and inevitably loses some speed because of extra drag, and some altitude because of loss of lift.
It can be compensated by increasing the engine power, but then you really need to predict that as well.
Yes, I know how planes make turns. It is valid to point out that a TCAS instruction that included a heading change would, at least for the plane whose TCAS instruction was to increase altitude, also have to include an engine power increase, which might not be feasible depending on the altitude (cruise power might already be pretty close to max power). Or else make the turn a very slow one, which would decrease its usefulness in collision avoidance.
Also that bank means you have a wing tilted up and a wing tilted down, increasing the vertical height of the plane at a time you’re trying to increase vertical separation.
Doing both reeks to me of a solution in search of a problem that’s going to end up the target of some later NTSB report when you hit some weird edge case where the dynamics and dimensions of the two planes aren’t fully accounted for.
> bank means you have a wing tilted up and a wing tilted down, increasing the vertical height of the plane at a time you’re trying to increase vertical separation
Yes, but it's also increasing horizontal separation (from a predicted value of zero), which might still be a reasonable tradeoff to consider.
ADS-B is not guaranteed to be available from all aircraft in most of the airspace below 10K feet MSL. (Basically, if a transponder used to be optional before, ADS-B out is also optional.) Mode-C (altitude encoded transponder replies) are also not required, but the equipage rate is very high at this point, which is less the case for ADS-B among GA aircraft.
> ADS-B is not guaranteed to be available from all aircraft in most of the airspace below 10K feet MSL.
Sure, but so what? The lack of universality is not an impediment to implementing a new TCAS system that uses ADS-B data when it's available. (And BTW, the places where ADS-B is not required are generally places where traffic is sparse and so the risk of a collision is pretty low to begin with.)
If the expense and lead time for the redesign now that more reliable and precise position (and first derivative thereof) are available from GPS via ADS-B are to be considered, it's fair to ask "well, what are the downsides as compared to using altitude for separation as today?" and one of them is "a higher percentage of aircraft are transmitting mode-C than will be transmitting ADS-B out".
Given that the clear priority is now to follow a TCAS RA (resolution advisory) over an ATC instruction, I think that the current TCAS approach is good enough to cause aircraft to miss each other and that a GPS-based redesign is unlikely.
I have to imagine that if there was a #3 on that list it would be that introducing a random heading change adds follow-on complexity that could be dangerous itself. Now the planes are not just at different altitudes, but pointed who-knows-where.
You could imagine if, in addition to obeying the altitude change, every plane began to turn rapidly to the right, it could only help the situation? As long as we all agree to turn right in case of a collision warning?
I'm guessing that it does not issue turn instructions because it only gives a Resolution Advisory when a collision is imminent (otherwise they would be giving RAs all the time). At altitude, where the air is thinner, it is much easier for a plane to immediately climb or descend, planes turn very slowly when the air is thin.