Originally Posted by KiloKilo
CNN:
He also mistakenly told the Russian plane to descend -- sending it straight into the cargo jet.
Rather debateble.
Not at all debatable. As you imply. It wasn't a mistake.
Giving an aircraft in a conflict situation a clearance which will resolve that situation is absolutely the right thing to do. The only procedural exception is when the controller *knows* that an aircraft is operating under an ACAS RA, in which case heshe is to issue at most informational advisories only.
(It should be clear that I am not going to follow ATC Watcher's plea not to re-open the discussion of the accident. I beg forgiveness. From the comments so far, it seems to me that there is certain information about TCAS that has not made it out into the wider pilot community and I think it might be opportune to point to it here.)
Nielsen was not advised of the RA until some half-minute into the situation, because DHL's FO was in the loo and the CAP was concentrating on his flight duties.
Nielsen undertook two actions which one can characterise as mistaken. One was allowing the loss of required horizontal separation. The other was in reporting the conflicting traffic to Bakshirian at 2 o'clock when the traffic was actually at 10 o'clock and in sight. (I am not going to address here the causes of these actions. These have been well laid out elsewhere.)
This latter turned out to be crucial. As I speculated in a WWW-available paper a month after the collision, and which was confirmed by the report, this led the Bakshirian crew to think they were in a three-aircraft conflict situation: themselves, traffic at 10 o'clock in sight, and further traffic at 2 o'clock that they didn't see. As I argued in my paper
in the 2004 Safety-Critical Systems and Software conference in Australia, available at
crpit.com/Vol47.html alternatively at
http://http://www.rvs.uni-bielefeld.de/publications/Reports/SCSS04.pdf this presented Bakshirian with a decision situation in which it was rational for them to choose to descend.
There are two crucial observations which follow from this, which have unfortunately been lost in the ensuing chatter.
1. Despite the "TCAS philosophy" which says that
TCAS activates when aircraft have violated separation; andtherefore when ATC has already failed; and therefore ATC is "out of the reckoning",
it is still possible for ATC to give legitimate input which will causally influence the outcome of an avoidance manoeuvre. In other words, the "TCAS philosophy" is based on a mistake.
When I say "legitimate", this does not mean "non-mistaken". People do make mistakes; indeed, the cognitive mistake made by Nielsen in his statement is common ("2 o'clock" is the participant-reciprocal of "10 o'clock). A system design which relies for its efficacy on operator actions, and TCAS is such, must explicitly take into account such common forms of human error.
2. One might also think that, had Nielsen known of the RA sooner, the outcome would have been avoided. Not necessarily so: it would have been within ICAO guidelines for Nielsen to have said "Bakshirian, advise intruder at your 2 o'clock" even if he knew of the RA, thus putting the Russians in exactly the same decision situation as they had in the actual scenario.
It is well known that a two-aircraft resolution algorithm such as employed by TCAS cannot resolve all conflicts. However, it is not even known whether the algorithm can even resolve all *three-aircraft* conflicts. (Is it known through work of Nancy Lynch and colleagues at MIT that it can resolve all two-aircraft conflicts.) Since 2002, I have offered a PhD to anyone who can either show that the TCAS algorithm mathematically resolves all three-aircraft conflicts, or alternatively produce a three-aircraft scenario which it cannot resolve (I have reduced the choice of such a scenario to a limited set of cases). Five years later, no one has successfully taken up that challenge. It stands; I issue it again here.
I cannot think it wise to advise people to *always, without exception* follow an RA, when we don't even know that the algorithm is correct (that is, correctly resolves all cases that concern us, such as three-aircraft conflicts), unless a risk analysis has been performed to show that the exceptions are expected to occur in only a negligible number of cases. To my knowledge, no such analysis has ever been performed.
I think it even less wise to suggest *mandating* the following of an RA, as the BFU issued as a safety recommendation. Forcing people to follow an algorithm which you can't show is correct is not something which sits well with modern ideas of society and choice. (BTW, I don't know that the BFU was aware of the issues I mention here at the time they issued their recommendation. But they could have been: I put the info on the WWW in August 2002.)
PBL