FLARM.....for good, or evil??

On 10/27/2010 7:20 AM, wrote:
in "awareness mode", of course a led will indicate where the closer
glider is, but still alarms will only be triggered when a deviation
makes the collision a real possibility.

aldo cernezzi

Since the GPS accuracy is probably much poorer than 10 feet, when a
glider comes within 10 feet, I figure the FLARM should consider that a
collision, no?
I don't know what Flarm does in that case, but my guess is the
*relative* accuracy is much better than the *absolute* accuracy. If
that's true, then each glider might have position errors of much more
than 10 feet, but they'll have nearly the same errors, giving a more
accurate separation distance.

Maybe someone more familiar with GPS in this situation can jump in here
and tell us?

--
Eric Greenwell - Washington State, USA (change ".netto" to ".us" to email me)
- "Transponders in Sailplanes - Feb/2010" also ADS-B, PCAS, Flarm http://tinyurl.com/yb3xywl
- "A Guide to Self-launching Sailplane Operation Mar/2004" Much of what you need to know tinyurl.com/yfs7tnz

On 10/27/2010 3:48 PM, Eric Greenwell wrote:
On 10/27/2010 7:20 AM, wrote:
in "awareness mode", of course a led will indicate where the closer
glider is, but still alarms will only be triggered when a deviation
makes the collision a real possibility.

aldo cernezzi
Since the GPS accuracy is probably much poorer than 10 feet, when a
glider comes within 10 feet, I figure the FLARM should consider that a
collision, no?
I don't know what Flarm does in that case, but my guess is the
*relative* accuracy is much better than the *absolute* accuracy. If
that's true, then each glider might have position errors of much more
than 10 feet, but they'll have nearly the same errors, giving a more
accurate separation distance.

Maybe someone more familiar with GPS in this situation can jump in here
and tell us?

Even if the relative GPS position computed by each glider has 0 error,
you still have the problem that at 50 Knots, each aircraft is moving ~
75 ft / second. With FLARM (or ADS-B) only transmitting positions every
second, you can't rely on these technologies to protect you from random
course changes that the systems can't possibly predict, if you are in
close proximity.

--
Mike Schumann

On 10/27/2010 5:33 PM, Mike Schumann wrote:
On 10/27/2010 3:48 PM, Eric Greenwell wrote:
On 10/27/2010 7:20 AM, wrote:
in "awareness mode", of course a led will indicate where the closer
glider is, but still alarms will only be triggered when a deviation
makes the collision a real possibility.

aldo cernezzi
Since the GPS accuracy is probably much poorer than 10 feet, when a
glider comes within 10 feet, I figure the FLARM should consider that a
collision, no?
I don't know what Flarm does in that case, but my guess is the
*relative* accuracy is much better than the *absolute* accuracy. If
that's true, then each glider might have position errors of much more
than 10 feet, but they'll have nearly the same errors, giving a more
accurate separation distance.

Maybe someone more familiar with GPS in this situation can jump in here
and tell us?

Even if the relative GPS position computed by each glider has 0 error,
you still have the problem that at 50 Knots, each aircraft is moving ~
75 ft / second. With FLARM (or ADS-B) only transmitting positions
every second, you can't rely on these technologies to protect you from
random course changes that the systems can't possibly predict, if you
are in close proximity.
FLARM does more than transmit positions: it transmits the projected path
of it's glider. Here's what I understand will happen: when the pilot
makes a course change, a new path is calculated and compared to the
paths Flarm has received from nearby gliders. If this new path puts it
on a collision course with any of them, the pilot is warned
"immediately", meaning it does not have to wait one second. The new path
will be transmitted within one second, so the nearby gliders can update
the other glider's path in their database. It might actually be more
sophisticated than that, such as transmitting a new path sooner if the
amount of change is "large", but I don't know what the algorithms are.

This projected path is a key element to the system working properly.
Without it, each FLARM unit would have to calculate the path of every
nearby glider; with it, each unit only has to calculate one path - it's
own. Potentially, it could be using a much higher position rate than
once a second to calculate it's projected path. In any case, the result
is much better than you might think for a system that transmits once a
second.

Does ADS-B transmit a projected path, or just position?

--
Eric Greenwell - Washington State, USA (change ".netto" to ".us" to email me)
- "Transponders in Sailplanes - Feb/2010" also ADS-B, PCAS, Flarm http://tinyurl.com/yb3xywl
- "A Guide to Self-launching Sailplane Operation Mar/2004" Much of what you need to know tinyurl.com/yfs7tnz

On Oct 27, 8:16*pm, Eric Greenwell wrote:
On 10/27/2010 5:33 PM, Mike Schumann wrote:

On 10/27/2010 3:48 PM, Eric Greenwell wrote:
On 10/27/2010 7:20 AM, wrote:
in "awareness mode", of course a led will indicate where the closer
glider is, but still alarms will only be triggered when a deviation
makes the collision a real possibility.

aldo cernezzi
Since the GPS accuracy is probably much poorer than 10 feet, when a
glider comes within 10 feet, I figure the FLARM should consider that a
collision, no?
I don't know what Flarm does in that case, but my guess is the
*relative* accuracy is much better than the *absolute* accuracy. If
that's true, then each glider might have position errors of much more
than 10 feet, but they'll have nearly the same errors, giving a more
accurate separation distance.

Maybe someone more familiar with GPS in this situation can jump in here
and tell us?

Even if the relative GPS position computed by each glider has 0 error,
you still have the problem that at 50 Knots, each aircraft is moving ~
75 ft / second. *With FLARM (or ADS-B) only transmitting positions
every second, you can't rely on these technologies to protect you from
random course changes that the systems can't possibly predict, if you
are in close proximity.

FLARM does more than transmit positions: it transmits the projected path
of it's glider. Here's what I understand will happen: when the pilot
makes a course change, a new path is calculated and compared to the
paths Flarm has received from nearby gliders. If this new path puts it
on a collision course with any of them, the pilot is warned
"immediately", meaning it does not have to wait one second. The new path
will be transmitted within one second, so the nearby gliders can update
the other glider's path in their database. It might actually be more
sophisticated than that, such as transmitting a new path sooner if the
amount of change is "large", but I don't know what the algorithms are.

This projected path is a key element to the system working properly.
Without it, each FLARM unit would have to calculate the path of every
nearby glider; with it, each unit only has to calculate one path - it's
own. Potentially, it could be using a much higher position rate than
once a second to calculate it's projected path. In any case, the result
is much better than you might think for a system that transmits once a
second.

Does ADS-B transmit a projected path, or just position?

--
Eric Greenwell - Washington State, USA (change ".netto" to ".us" to email me)
- "Transponders in Sailplanes - Feb/2010" also ADS-B, PCAS, Flarmhttp://tinyurl.com/yb3xywl
- "A Guide to Self-launching Sailplane Operation Mar/2004" Much of what you need to know tinyurl.com/yfs7tnz

That's my understanding as well - the Flarm algorithm looks at a range
of potential paths that the glider might maneuver to and calculates
potential collisions on the basis of all potential paths within that
maneuvering envelope. I expect within a couple of seconds the path
assumes a fair amount of potential maneuvering while over a longer
period of time it would limit extreme maneuvering assumptions to be
closer to an extrapolation of the current path/turn rate. Because it
understands your trun rate it works well in thermals.

I have not heard anything about similar capabilities for ADS-B and I
am confident that it doesn't extrapolate with a glider performance
envelope in mind, since it isn't designed explicitly for gliders. I
would think ADS-B would be particularly challenged in predicting
likely collisions in thermals - much more so than Flarm.

9B

On Oct 28, 1:11*am, Andy wrote:
On Oct 27, 8:16*pm, Eric Greenwell wrote:





On 10/27/2010 5:33 PM, Mike Schumann wrote:

On 10/27/2010 3:48 PM, Eric Greenwell wrote:
On 10/27/2010 7:20 AM, wrote:
in "awareness mode", of course a led will indicate where the closer
glider is, but still alarms will only be triggered when a deviation
makes the collision a real possibility.

aldo cernezzi
Since the GPS accuracy is probably much poorer than 10 feet, when a
glider comes within 10 feet, I figure the FLARM should consider that a
collision, no?
I don't know what Flarm does in that case, but my guess is the
*relative* accuracy is much better than the *absolute* accuracy. If
that's true, then each glider might have position errors of much more
than 10 feet, but they'll have nearly the same errors, giving a more
accurate separation distance.

Maybe someone more familiar with GPS in this situation can jump in here
and tell us?

Even if the relative GPS position computed by each glider has 0 error,
you still have the problem that at 50 Knots, each aircraft is moving ~
75 ft / second. *With FLARM (or ADS-B) only transmitting positions
every second, you can't rely on these technologies to protect you from
random course changes that the systems can't possibly predict, if you
are in close proximity.

FLARM does more than transmit positions: it transmits the projected path
of it's glider. Here's what I understand will happen: when the pilot
makes a course change, a new path is calculated and compared to the
paths Flarm has received from nearby gliders. If this new path puts it
on a collision course with any of them, the pilot is warned
"immediately", meaning it does not have to wait one second. The new path
will be transmitted within one second, so the nearby gliders can update
the other glider's path in their database. It might actually be more
sophisticated than that, such as transmitting a new path sooner if the
amount of change is "large", but I don't know what the algorithms are.

This projected path is a key element to the system working properly.
Without it, each FLARM unit would have to calculate the path of every
nearby glider; with it, each unit only has to calculate one path - it's
own. Potentially, it could be using a much higher position rate than
once a second to calculate it's projected path. In any case, the result
is much better than you might think for a system that transmits once a
second.

Does ADS-B transmit a projected path, or just position?

--
Eric Greenwell - Washington State, USA (change ".netto" to ".us" to email me)
- "Transponders in Sailplanes - Feb/2010" also ADS-B, PCAS, Flarmhttp://tinyurl.com/yb3xywl
- "A Guide to Self-launching Sailplane Operation Mar/2004" Much of what you need to know tinyurl.com/yfs7tnz

That's my understanding as well - the Flarm algorithm looks at a range
of potential paths that the glider might maneuver to and calculates
potential collisions on the basis of all potential paths within that
maneuvering envelope. *I expect within a couple of seconds the path
assumes a fair amount of potential maneuvering while over a longer
period of time it would limit extreme maneuvering assumptions to be
closer to an extrapolation of the current path/turn rate. Because it
understands your trun rate it works well in thermals.

I have not heard anything about similar capabilities for ADS-B and I
am confident that it doesn't extrapolate with a glider performance
envelope in mind, since it isn't designed explicitly for gliders. I
would think ADS-B would be particularly challenged in predicting
likely collisions in thermals - much more so than Flarm.

9B- Hide quoted text -

- Show quoted text -

I hadn't really considered how the projected flight path system works
before. But after thinking about it for a bit it has a lot of
potentional. How much the Flarm actually uses I do not know. But it
would be possible for the Flarm to actually learn the possible flight
paths for a given glider and dynamically adjust it's algorithm for
where the glider might be able to go from that. Even without that I
can see that gliders could get very close but have potential flight
paths that would make colliding impossible and as a result would not
create a collision alarm.

Brian

On Thu, 28 Oct 2010 06:53:27 -0700 (PDT), Brian
wrote:


I hadn't really considered how the projected flight path system works
before. But after thinking about it for a bit it has a lot of
potentional. How much the Flarm actually uses I do not know.

Flarm uses ONLY projected flight paths to calculate a collision
probability.

Even without that I
can see that gliders could get very close but have potential flight
paths that would make colliding impossible and as a result would not
create a collision alarm.


This is exactly how Flarm works.
Flarm doesn't care about distances - as long as Flarm doesn't detect a
potential collision cource, you can fly very close to each other
without getting a warning - even if you are circling.



Andreas

On Oct 28, 7:32*am, Andreas Maurer wrote:
On Thu, 28 Oct 2010 06:53:27 -0700 (PDT), Brian
wrote:

I hadn't really considered how the projected flight path system works
before. But after thinking about it for a bit it has a lot of
potentional. How much the Flarm actually uses I do not know.

Flarm uses ONLY projected flight paths to calculate a collision
probability.

Even without that I
can see that gliders could get very close but have potential flight
paths that would make colliding impossible and as a result would not
create a collision alarm.

This is exactly how Flarm works.
Flarm doesn't care about distances - as long as Flarm doesn't detect a
potential collision cource, you can fly very close to each other
without getting a warning - even if you are circling.

Andreas

I think that's the secret for how you make it useful in thermals - if
the system knows you are circling it can do a better job predicting
your curved flight path and potential threats along that path. I
presume that if you assume the full maneuvering envelope of each
glider you'd generate a lot of warnings, so it would make sense to
assume something more limited that strikes a balance between false
positive warnings and missing potential maneuvers that could create a
threat with little advanced notice. Think of a glider pulling up into
a thermal as a good example. I assume that Flarm does all this based
on the following explanation where an expanding projected flight path
envelope is depicted:

http://www.gliderpilot.org/Flarm-WhatDoesItDo

As to ADS-B - without some algorithm for projecting flight paths the
only warning you can realistically generate is a proximity warning.
Even warning only for declining separation distance is a crude form of
relative path prediction, just not a very useful one - particularly
for glider operations with multiple targets and circling flight.

9B

On 10/27/2010 11:16 PM, Eric Greenwell wrote:
On 10/27/2010 5:33 PM, Mike Schumann wrote:
On 10/27/2010 3:48 PM, Eric Greenwell wrote:
On 10/27/2010 7:20 AM, wrote:
in "awareness mode", of course a led will indicate where the closer
glider is, but still alarms will only be triggered when a deviation
makes the collision a real possibility.

aldo cernezzi
Since the GPS accuracy is probably much poorer than 10 feet, when a
glider comes within 10 feet, I figure the FLARM should consider that a
collision, no?
I don't know what Flarm does in that case, but my guess is the
*relative* accuracy is much better than the *absolute* accuracy. If
that's true, then each glider might have position errors of much more
than 10 feet, but they'll have nearly the same errors, giving a more
accurate separation distance.

Maybe someone more familiar with GPS in this situation can jump in here
and tell us?

Even if the relative GPS position computed by each glider has 0 error,
you still have the problem that at 50 Knots, each aircraft is moving ~
75 ft / second. With FLARM (or ADS-B) only transmitting positions
every second, you can't rely on these technologies to protect you from
random course changes that the systems can't possibly predict, if you
are in close proximity.
FLARM does more than transmit positions: it transmits the projected path
of it's glider. Here's what I understand will happen: when the pilot
makes a course change, a new path is calculated and compared to the
paths Flarm has received from nearby gliders. If this new path puts it
on a collision course with any of them, the pilot is warned
"immediately", meaning it does not have to wait one second. The new path
will be transmitted within one second, so the nearby gliders can update
the other glider's path in their database. It might actually be more
sophisticated than that, such as transmitting a new path sooner if the
amount of change is "large", but I don't know what the algorithms are.

This projected path is a key element to the system working properly.
Without it, each FLARM unit would have to calculate the path of every
nearby glider; with it, each unit only has to calculate one path - it's
own. Potentially, it could be using a much higher position rate than
once a second to calculate it's projected path. In any case, the result
is much better than you might think for a system that transmits once a
second.

Does ADS-B transmit a projected path, or just position?

I'm not an expert on either FLARM or ADS-B. I believe that ADS-B
currently only transmits absolute position. Future enhancements might
transmit trajectory, which would be most useful for aircraft with Flight
Management Systems where the trajectory is well defined and could be
used by the ATC system for airspace management.

Regardless of whether or not the trajectory is transmitted, a
sophisticated receiving system (either FLARM or ADS-B based), can
remember each aircraft's position data and compute it's current
trajectory. While a glider might be moving 75 ft / sec, this is
obviously in a relatively forward direction.

Neither the transmitting nor the receiving FLARM or ADS-B system can
predict an abrupt change in course that a pilot flying manually might
command. However, every aircraft has physical limits on roll rates,
etc. that restrain the potential change in direction that can occur
within the one second update interval of these systems. As a result,
the systems can, theoretically, compute a pear shaped threat envelope
for each aircraft and limit collision warnings to those situations where
these envelopes intersect.

It would be interesting to get more detailed information on the exact
algorithms that FLARM uses in it's collision threat analysis and compare
this to the actual unit performance in situations where gliders are
flying at close distances in formation or in gaggles. This could also
help pilots understand the limitations of these systems so they don't
develop a false sense of security in situations where these systems are
not reliable.

--
Mike Schumann

On 10/28/2010 7:29 AM, Mike Schumann wrote:
On 10/27/2010 11:16 PM, Eric Greenwell wrote:


This projected path is a key element to the system working properly.
Without it, each FLARM unit would have to calculate the path of every
nearby glider; with it, each unit only has to calculate one path - it's
own. Potentially, it could be using a much higher position rate than
once a second to calculate it's projected path. In any case, the result
is much better than you might think for a system that transmits once a
second.

Does ADS-B transmit a projected path, or just position?

I'm not an expert on either FLARM or ADS-B. I believe that ADS-B
currently only transmits absolute position. Future enhancements might
transmit trajectory, which would be most useful for aircraft with
Flight Management Systems where the trajectory is well defined and
could be used by the ATC system for airspace management.

Regardless of whether or not the trajectory is transmitted, a
sophisticated receiving system (either FLARM or ADS-B based), can
remember each aircraft's position data and compute it's current
trajectory. While a glider might be moving 75 ft / sec, this is
obviously in a relatively forward direction.
You may be underestimating the value of transmitting the projected path.
When another glider is first detected, your unit has only one position
report and can not determine the flight path from that single point, and
it will take several more precious seconds to determine the flight path
of the potential threat; however, because the projected path is
transmitted every second, your unit immediately knows it.

Neither the transmitting nor the receiving FLARM or ADS-B system can
predict an abrupt change in course that a pilot flying manually might
command. However, every aircraft has physical limits on roll rates,
etc. that restrain the potential change in direction that can occur
within the one second update interval of these systems. As a result,
the systems can, theoretically, compute a pear shaped threat envelope
for each aircraft and limit collision warnings to those situations
where these envelopes intersect.

It would be interesting to get more detailed information on the exact
algorithms that FLARM uses in it's collision threat analysis and
compare this to the actual unit performance in situations where
gliders are flying at close distances in formation or in gaggles.
This could also help pilots understand the limitations of these
systems so they don't develop a false sense of security in situations
where these systems are not reliable.

I'm sure the developers have tested their algorithms with thousands of
simulations using IGC files from gliders in many situations. The Parowan
accident simulation at

http://www.gliderpilot.org/Flarm-Par...dairSimulation

shows what can be done. It would be interesting, informative, and
entertaining if there was a website or application that would let us run
IGC files we select in a simulation like this. I'm curious about how
Flarm would react in a few situations I've encountered. Doing
simulations on a pilot's own files might be more persuasive of the value
of Flarm than even the most well-written explanations, and much more
easily understood than the algorithms themselves.

--
Eric Greenwell - Washington State, USA (change ".netto" to ".us" to email me)
- "Transponders in Sailplanes - Feb/2010" also ADS-B, PCAS, Flarm http://tinyurl.com/yb3xywl
- "A Guide to Self-launching Sailplane Operation Mar/2004" Much of what you need to know tinyurl.com/yfs7tnz

On 10/28/2010 2:14 PM, Eric Greenwell wrote:
On 10/28/2010 7:29 AM, Mike Schumann wrote:
On 10/27/2010 11:16 PM, Eric Greenwell wrote:


This projected path is a key element to the system working properly.
Without it, each FLARM unit would have to calculate the path of every
nearby glider; with it, each unit only has to calculate one path - it's
own. Potentially, it could be using a much higher position rate than
once a second to calculate it's projected path. In any case, the result
is much better than you might think for a system that transmits once a
second.

Does ADS-B transmit a projected path, or just position?

I'm not an expert on either FLARM or ADS-B. I believe that ADS-B
currently only transmits absolute position. Future enhancements might
transmit trajectory, which would be most useful for aircraft with
Flight Management Systems where the trajectory is well defined and
could be used by the ATC system for airspace management.

Regardless of whether or not the trajectory is transmitted, a
sophisticated receiving system (either FLARM or ADS-B based), can
remember each aircraft's position data and compute it's current
trajectory. While a glider might be moving 75 ft / sec, this is
obviously in a relatively forward direction.
You may be underestimating the value of transmitting the projected path.
When another glider is first detected, your unit has only one position
report and can not determine the flight path from that single point, and
it will take several more precious seconds to determine the flight path
of the potential threat; however, because the projected path is
transmitted every second, your unit immediately knows it.

Neither the transmitting nor the receiving FLARM or ADS-B system can
predict an abrupt change in course that a pilot flying manually might
command. However, every aircraft has physical limits on roll rates,
etc. that restrain the potential change in direction that can occur
within the one second update interval of these systems. As a result,
the systems can, theoretically, compute a pear shaped threat envelope
for each aircraft and limit collision warnings to those situations
where these envelopes intersect.

It would be interesting to get more detailed information on the exact
algorithms that FLARM uses in it's collision threat analysis and
compare this to the actual unit performance in situations where
gliders are flying at close distances in formation or in gaggles. This
could also help pilots understand the limitations of these systems so
they don't develop a false sense of security in situations where these
systems are not reliable.

I'm sure the developers have tested their algorithms with thousands of
simulations using IGC files from gliders in many situations. The Parowan
accident simulation at

http://www.gliderpilot.org/Flarm-Par...dairSimulation

shows what can be done. It would be interesting, informative, and
entertaining if there was a website or application that would let us run
IGC files we select in a simulation like this. I'm curious about how
Flarm would react in a few situations I've encountered. Doing
simulations on a pilot's own files might be more persuasive of the value
of Flarm than even the most well-written explanations, and much more
easily understood than the algorithms themselves.


There's no question that FLARM or ADS-B based systems could have easily
prevented the Parowan mid-air. My question is how much you can rely on
this type of equipment to accurately warn you of collisions when you are
flying in gaggles.

Obviously knowing the relative position of other gliders in the gaggle
is helpful. I would be very skeptical, however, of putting my faith in
FLARM or any other system to accurately warn me of a collision with
another glider that was in the same thermal, near my altitude, that was
in my blind spot.

Transmitting the project path of the aircraft is really only beneficial
if the equipment on board the transmitting aircraft has some added
information that is not available to the receiver on the transmitting
aircraft's intent.

With both FLARM and ADS-B systems, the initial visibility of the other
aircraft occurs way before there is any collision threat, so the
receiver should have no difficulty computing the project path of the
other aircraft. ADS-B actually transmits the category of aircraft (i.e.
glider, balloon, etc...) so the receiver can get a pretty good hint on
the type of maneuvers that can be expected.

--
Mike Schumann