303pilot wrote:

"Eric Greenwell" wrote in message


Ony if you think the problem is 40 gliders instead of 3 or 4, which is
all that was involved in the recent collisions.


I'm not a programmer, but I work with them on a daily basis. It seems to me
that even a 3 or 4 glider problem is highly complex because sailplanes fly
in highly irregular paths.

But still more manageable the 40, right :} ? But to answer the
question, and keeping in mind I don't know any details of the Flarm
system, it may not be possible or necessary to have a TCAS-like system.
Though several gliders flying at random may indeed be complex, the
algorithms chosen can make simplifying assumptions based on the nature
of glider flight. Also, the pilot arriving at a thermal might modify his
arrival to keep the threat level low, compared to how he does it now, so
as not to "alarm" the pilots already in the thermal (and for other
situations, also, not just thermals).

We don't even fly straight point to point--we
weave left and right, we dolphin. A ship might be going (more or less)
straight and a couple hundred feet below me. Not a threat, right? Maybe,
maybe not--what if I'm in a thermal and he's seen me and plans to join me.
He suddenly converts speed to altitude and he's in my blind spot. My GPS
has a 4 second polling cycle. Ooops.

Most GPS receivers we use emit at once a second, though _flight
recorders_ might record at a slower rate (the newer ones will also
record at once per second). Rate isn't a problem.

What if two ships are in a thermal but maintaining separation. Everything's
fine, right? Sure, until we get to the top of the lift band and he suddenly
tightens his turn to go through the core as he heads out on course.

This simple situation is likely easy to handle. It should cause some
alarms in both cockpits, unless it is a diving exit, which would put him
well below the still thermalling glider!

I'm sure anyone contemplating these systems has thought of these
situations and more, and intends to cope with them, and use extensive
testing to validate the equipment. As I mentioned before, the equipment
might cause changes in pilot behavior, perhaps because they wish to
avoid causing alarms, or because they now realize better the dangers
involved.

Have you ever flown in a thermal with even 10 gliders? I have many
times. I can not keep track of even 10 gliders, but I can still thermal
safely when there are that many and more. We are not flying around at
random, but circling in an orderly fashion. Only the nearby gliders are
a threat that must be monitored. In any case, a system that deals with
only a few gliders will cover most of the situations.

Orderly? To you and I yes. To a program? Not really.
Think about what happens at cloudbase in a contest gaggle. The (mostly)
orderly and similar actions (mostly same speed & bank angle) get more
random. Some pilots increase their radius purposefully suboptimizing climb
rates, others deploy a bit of spoiler, others leave. How are these actions
to be predicted?
Even if I have a 1 second polling rate on my GPS & "traffic
analysis/collision avoidance system", how many variables can change in that
one second and how fast can my safe separation be erased?

In one second? Very few. Gliders simply don't react quickly in roll, and
pilots don't pitch rapidly in the cloudbase gaggle. The ones I've been
in, everyone is changing direction smoothly and slowly. Away from a
gaggle, pitch changes can occur rapidly, but one second still seems
short enough to me. It would be better to seek the opinion of someone
actually attempting this, of course!

snip

What must we do? Propose something - we're listening.


Here's my modest proposal, eat them. Sorry, trying again--
Don't ask the system to figure collision potential and don't introduce
another screen. Just have a system call out "target NNW, same altitude,
closing @ x". If I see it, I say something like "clear" or "check" and the
system stops alerting me to the known target. If I don't acknowledge the
target the system continues to provide information at regular intervals to
help me find it.

If I were designing a system, I'd make this capability the first phase.
Perhaps it would be good enough. I'd try a button on the stick before
the confirmation.

I might be re-notified of "cleared" targets if we continue to fly in
proximity to one another.
Frankly, I can't imagine a user interface that would be useful in a large
gaggle. That's probably OK because that's where we're likely to be most
alert to this type of threat.

Brent



--
Change "netto" to "net" to email me directly

Eric Greenwell
Washington State
USA