IGC File / Flight-track Analysis

noel.wade wrote:
Hi All,

I'm working on a project that I can't talk about much, but I was
hoping to prevail upon some experts here. I need to analyze some IGC
files in my project, and try to use the data to determine the behavior
of the glider.
There are a number of programs that do this, such as SeeYou, simulator
software (Condor?), even commercial programs that reconstruct a flight
from gps data for accident and other investigations. T2T (track to
thermal) picks out thermals from the data. Have you investigated
acquiring one of these programs, or getting them to slice of a piece of
it for your purposes?

--
Eric Greenwell - Washington State, USA
* Change "netto" to "net" to email me directly

* "A Guide to Self-launching Sailplane Operation" at www.motorglider.org

On Jan 21, 8:00*pm, Eric Greenwell wrote:

There are a number of programs that do this, such as SeeYou, simulator
software (Condor?), even commercial programs that reconstruct a flight
from gps data for accident and other investigations. T2T (track to

Eric -

Yes I have. The catch is that this project isn't mine; I'm
volunteering my time for it. And the people I'm working with don't
want the end-result to be tied to any licensing fees or restrictions
on use of the code.
*sigh* *shrug*

Thanks all,

--Noel

noel.wade wrote:
On Jan 21, 8:00 pm, Eric Greenwell wrote:

There are a number of programs that do this, such as SeeYou, simulator
software (Condor?), even commercial programs that reconstruct a flight
from gps data for accident and other investigations. T2T (track to

Eric -

Yes I have. The catch is that this project isn't mine; I'm
volunteering my time for it. And the people I'm working with don't
want the end-result to be tied to any licensing fees or restrictions
on use of the code.
*sigh* *shrug*

Thanks all,

--Noel

XC Soar is Open Source, which might work if the GPL restrictions
wouldn't affect the exploitation of the project (this would require you
to make any modified XC Soar code modules available under the GPS terms,
but not interfacing modules which would stay proprietary).

On Jan 22, 3:05*am, Chris Reed wrote:
noel.wade wrote:
On Jan 21, 8:00 pm, Eric Greenwell wrote:

There are a number of programs that do this, such as SeeYou, simulator
software (Condor?), even commercial programs that reconstruct a flight
from gps data for accident and other investigations. T2T (track to

Eric -

Yes I have. *The catch is that this project isn't mine; I'm
volunteering my time for it. *And the people I'm working with don't
want the end-result to be tied to any licensing fees or restrictions
on use of the code.
*sigh* *shrug*

Thanks all,

--Noel

XC Soar is Open Source, which might work if the GPL restrictions
wouldn't affect the exploitation of the project (this would require you
to make any modified XC Soar code modules available under the GPS terms,
but not *interfacing modules which would stay proprietary).

I'd contact the Aerospace Engineering department at a university near
you. Most graduate programs and some undergraduate ones have analytic
tools they use to do optimal estimation for a state-space
representation of aircraft. To do some of the things you are asking
about I suspect you'd need a full 6 degree of freedom model. Depending
on what the inputs are, you may not be able to fully estimate the
aircraft state - there would be too many unknowns. For instance,
because you don't have attitude information (none of the Euler
angles), you wouldn't be able to cleanly distinguish between gusts
displacing the aircraft and control inputs. You would likely need to
assume coordinated flight and no wind gusts. Some simple Kalman
filters should help, as Dave says.

9B

On Jan 22, 7:10*am, Andy wrote:
On Jan 22, 3:05*am, Chris Reed wrote:



noel.wade wrote:
On Jan 21, 8:00 pm, Eric Greenwell wrote:

There are a number of programs that do this, such as SeeYou, simulator
software (Condor?), even commercial programs that reconstruct a flight
from gps data for accident and other investigations. T2T (track to

Eric -

Yes I have. *The catch is that this project isn't mine; I'm
volunteering my time for it. *And the people I'm working with don't
want the end-result to be tied to any licensing fees or restrictions
on use of the code.
*sigh* *shrug*

Thanks all,

--Noel

XC Soar is Open Source, which might work if the GPL restrictions
wouldn't affect the exploitation of the project (this would require you
to make any modified XC Soar code modules available under the GPS terms,
but not *interfacing modules which would stay proprietary).

I'd contact the Aerospace Engineering department at a university near
you. *Most graduate programs and some undergraduate ones have analytic
tools they use to do optimal estimation for a state-space
representation of aircraft. To do some of the things you are asking
about I suspect you'd need a full 6 degree of freedom model. Depending
on what the inputs are, you may not be able to fully estimate the
aircraft state - there would be too many unknowns. For instance,
because you don't have attitude information (none of the Euler
angles), you wouldn't be able to cleanly distinguish between gusts
displacing the aircraft and control inputs. You would likely need to
assume coordinated flight and no wind gusts. Some simple Kalman
filters should help, as Dave says.

9B

Um, depending on what he's trying to do, the Kalman filters
in the GPS could make life very difficult. Adding another
Kalman filter in back of the filter already in the GPS engine
is not likely to help. And every brand of GPS (and potentially
every software revision of the same brand) may have different
filtering.

Here's an experiment conducted a few times, with a GPS
engine in a car:
Drive at high speed
Disable the antenna
Slow rapidly and turn 90 degrees
Speed up
Enable antenna

What do you think the GPS output shows ?

Hope that helps clarify the problem,
Best Regards, Dave "YO electric"

On Jan 22, 5:49*am, Dave Nadler wrote:
On Jan 22, 7:10*am, Andy wrote:





On Jan 22, 3:05*am, Chris Reed wrote:

noel.wade wrote:
On Jan 21, 8:00 pm, Eric Greenwell wrote:

There are a number of programs that do this, such as SeeYou, simulator
software (Condor?), even commercial programs that reconstruct a flight
from gps data for accident and other investigations. T2T (track to

Eric -

Yes I have. *The catch is that this project isn't mine; I'm
volunteering my time for it. *And the people I'm working with don't
want the end-result to be tied to any licensing fees or restrictions
on use of the code.
*sigh* *shrug*

Thanks all,

--Noel

XC Soar is Open Source, which might work if the GPL restrictions
wouldn't affect the exploitation of the project (this would require you
to make any modified XC Soar code modules available under the GPS terms,
but not *interfacing modules which would stay proprietary).

I'd contact the Aerospace Engineering department at a university near
you. *Most graduate programs and some undergraduate ones have analytic
tools they use to do optimal estimation for a state-space
representation of aircraft. To do some of the things you are asking
about I suspect you'd need a full 6 degree of freedom model. Depending
on what the inputs are, you may not be able to fully estimate the
aircraft state - there would be too many unknowns. For instance,
because you don't have attitude information (none of the Euler
angles), you wouldn't be able to cleanly distinguish between gusts
displacing the aircraft and control inputs. You would likely need to
assume coordinated flight and no wind gusts. Some simple Kalman
filters should help, as Dave says.

9B

Um, depending on what he's trying to do, the Kalman filters
in the GPS could make life very difficult. Adding another
Kalman filter in back of the filter already in the GPS engine
is not likely to help. And every brand of GPS (and potentially
every software revision of the same brand) may have different
filtering.

Here's an experiment conducted a few times, with a GPS
engine in a car:
Drive at high speed
Disable the antenna
Slow rapidly and turn 90 degrees
Speed up
Enable antenna

What do you think the GPS output shows ?

Hope that helps clarify the problem,
Best Regards, Dave "YO electric"

Dave - Depending on the gains on the filter and the breaking action on
the car I expect you would get some sort of position overshoot in the
original direction while the antenna was disconnected, followed by an
arc at the modeled maximum cornering G to try to close the position
and velocity vector errors. But I may not recall my optimal estimation
theory precisely enough.

Even better experiment - Do a loop in your glider and play it back in
SeeYou. You will see that they are using an estimation model that
doesn't allow for inverted flight - or maybe there just isn't enough
information in the igc file to get you to inverted. In any case, the
glider flops around and pivots on its vertical axis as I recall.

Back to Noel's problem - I was referring to how to estimate aircraft
attitudes, velocities and angular rates from the data in the igc file.
Whether that data already has been smoothed through some optimal
estimation technique before generating the igc file shouldn't have too
big an impact on your ability to accurately estimate all the other
states (i.e. Euler angles and rates, linear positions and velocities)
from the position data as long as you are assuming reasonably similar
system dynamics. I'm not exactly clear what Noel is trying to
accomplish from the abbreviated description provided, but I thought it
is to estimate the full aircraft state from a reduced data set. You
should be able to use a calculus of variations approach to generate an
estimate for most state variables from the flight path data only. I'm
not sure what other type of approach would work. But then again I may
have misunderstood Noel's goal.

More on the use of Kalman filters: http://en.wikipedia.org/wiki/Kalman_filter

That went nerdy fast eh?

9B

On 22 Jan, 17:04, Andy wrote:
On Jan 22, 5:49*am, Dave Nadler wrote:





On Jan 22, 7:10*am, Andy wrote:

On Jan 22, 3:05*am, Chris Reed wrote:

noel.wade wrote:
On Jan 21, 8:00 pm, Eric Greenwell wrote:

There are a number of programs that do this, such as SeeYou, simulator
software (Condor?), even commercial programs that reconstruct a flight
from gps data for accident and other investigations. T2T (track to

Eric -

Yes I have. *The catch is that this project isn't mine; I'm
volunteering my time for it. *And the people I'm working with don't
want the end-result to be tied to any licensing fees or restrictions
on use of the code.
*sigh* *shrug*

Thanks all,

--Noel

XC Soar is Open Source, which might work if the GPL restrictions
wouldn't affect the exploitation of the project (this would require you
to make any modified XC Soar code modules available under the GPS terms,
but not *interfacing modules which would stay proprietary).

I'd contact the Aerospace Engineering department at a university near
you. *Most graduate programs and some undergraduate ones have analytic
tools they use to do optimal estimation for a state-space
representation of aircraft. To do some of the things you are asking
about I suspect you'd need a full 6 degree of freedom model. Depending
on what the inputs are, you may not be able to fully estimate the
aircraft state - there would be too many unknowns. For instance,
because you don't have attitude information (none of the Euler
angles), you wouldn't be able to cleanly distinguish between gusts
displacing the aircraft and control inputs. You would likely need to
assume coordinated flight and no wind gusts. Some simple Kalman
filters should help, as Dave says.

9B

Um, depending on what he's trying to do, the Kalman filters
in the GPS could make life very difficult. Adding another
Kalman filter in back of the filter already in the GPS engine
is not likely to help. And every brand of GPS (and potentially
every software revision of the same brand) may have different
filtering.

Here's an experiment conducted a few times, with a GPS
engine in a car:
Drive at high speed
Disable the antenna
Slow rapidly and turn 90 degrees
Speed up
Enable antenna

What do you think the GPS output shows ?

Hope that helps clarify the problem,
Best Regards, Dave "YO electric"

Dave - Depending on the gains on the filter and the breaking action on
the car I expect you would get some sort of position overshoot in the
original direction while the antenna was disconnected, followed by an
arc at the modeled maximum cornering G to try to close the position
and velocity vector errors. But I may not recall my optimal estimation
theory precisely enough.

Even better experiment - Do a loop in your glider and play it back in
SeeYou. You will see that they are using an estimation model that
doesn't allow for inverted flight - or maybe there just isn't enough
information in the igc file to get you to inverted. In any case, the
glider flops around and pivots on its vertical axis as I recall.

Back to Noel's problem - I was referring to how to estimate aircraft
attitudes, velocities and angular rates from the data in the igc file.
Whether that data already has been smoothed through some optimal
estimation technique before generating the igc file shouldn't have too
big an impact on your ability to accurately estimate all the other
states (i.e. Euler angles and rates, linear positions and velocities)
from the position data as long as you are assuming reasonably similar
system dynamics. I'm not exactly clear what Noel is trying to
accomplish from the abbreviated description provided, but I thought it
is to estimate the full aircraft state from a reduced data set. You
should be able to use a calculus of variations approach to generate an
estimate for most state variables from the flight path data only. I'm
not sure what other type of approach would work. But then again I may
have misunderstood Noel's goal.

More on the use of Kalman filters: *http://en.wikipedia.org/wiki/Kalman_filter

That went nerdy fast eh?

9B- Hide quoted text -

- Show quoted text -

This is interesting. Are you saying that if you disconnect the aerial
as you approach the turnpoint and turn early the gps will then give
overshoot positions and the logger log you in the turn even though you
are already on the way to the next one?

I must try that!!

Jim

On Jan 22, 12:04*pm, Andy wrote:
On Jan 22, 5:49*am, Dave Nadler wrote:



On Jan 22, 7:10*am, Andy wrote:

On Jan 22, 3:05*am, Chris Reed wrote:

noel.wade wrote:
On Jan 21, 8:00 pm, Eric Greenwell wrote:

There are a number of programs that do this, such as SeeYou, simulator
software (Condor?), even commercial programs that reconstruct a flight
from gps data for accident and other investigations. T2T (track to

Eric -

Yes I have. *The catch is that this project isn't mine; I'm
volunteering my time for it. *And the people I'm working with don't
want the end-result to be tied to any licensing fees or restrictions
on use of the code.
*sigh* *shrug*

Thanks all,

--Noel

XC Soar is Open Source, which might work if the GPL restrictions
wouldn't affect the exploitation of the project (this would require you
to make any modified XC Soar code modules available under the GPS terms,
but not *interfacing modules which would stay proprietary).

I'd contact the Aerospace Engineering department at a university near
you. *Most graduate programs and some undergraduate ones have analytic
tools they use to do optimal estimation for a state-space
representation of aircraft. To do some of the things you are asking
about I suspect you'd need a full 6 degree of freedom model. Depending
on what the inputs are, you may not be able to fully estimate the
aircraft state - there would be too many unknowns. For instance,
because you don't have attitude information (none of the Euler
angles), you wouldn't be able to cleanly distinguish between gusts
displacing the aircraft and control inputs. You would likely need to
assume coordinated flight and no wind gusts. Some simple Kalman
filters should help, as Dave says.

9B

Um, depending on what he's trying to do, the Kalman filters
in the GPS could make life very difficult. Adding another
Kalman filter in back of the filter already in the GPS engine
is not likely to help. And every brand of GPS (and potentially
every software revision of the same brand) may have different
filtering.

Here's an experiment conducted a few times, with a GPS
engine in a car:
Drive at high speed
Disable the antenna
Slow rapidly and turn 90 degrees
Speed up
Enable antenna

What do you think the GPS output shows ?

Hope that helps clarify the problem,
Best Regards, Dave "YO electric"

Dave - Depending on the gains on the filter and the breaking action on
the car I expect you would get some sort of position overshoot in the
original direction while the antenna was disconnected, followed by an
arc at the modeled maximum cornering G to try to close the position
and velocity vector errors. But I may not recall my optimal estimation
theory precisely enough.

You get plus one point. Depending on the implementation you may
see very interesting discontinuities in the output ;-)

Even better experiment - Do a loop in your glider and play it back in
SeeYou. You will see that they are using an estimation model that
doesn't allow for inverted flight - or maybe there just isn't enough
information in the igc file to get you to inverted. In any case, the
glider flops around and pivots on its vertical axis as I recall.

I've found Reverse Cuban Eights really bother StrePla ;-)

Back to Noel's problem - I was referring to how to estimate aircraft
attitudes, velocities and angular rates from the data in the igc file.
Whether that data already has been smoothed through some optimal
estimation technique before generating the igc file shouldn't have too
big an impact on your ability to accurately estimate all the other
states (i.e. Euler angles and rates, linear positions and velocities)
from the position data as long as you are assuming reasonably similar
system dynamics. I'm not exactly clear what Noel is trying to
accomplish from the abbreviated description provided, but I thought it
is to estimate the full aircraft state from a reduced data set. You
should be able to use a calculus of variations approach to generate an
estimate for most state variables from the flight path data only. I'm
not sure what other type of approach would work. But then again I may
have misunderstood Noel's goal.

Noel has not specified his goal adequately enough to
really answer his question (he gets minus 2 points).

More on the use of Kalman filters: *http://en.wikipedia.org/wiki/Kalman_filter

That went nerdy fast eh?

9B