Tramm Hudson
Jun 18, 2003, 04:01 AM
After a year and a half of effort, the autopilot system finally works!
Yesterday, Dennis, Will and I hovered Dennis' Nova for entire tanks
of fuel, with only the collective/throttle under manual control.
All control loops except altitude were closed and tuned. We should
write a detailed guide to tuning the PID loops, which took the better
part of two full days.
Onboard the Nova was our IMU / Magnetometer / Servo controller,
a Linux Cerfboard with 802.11 CF, and our new u-blox TIM-LP / IO board.
The gains and code are all checked into our CVS tree and should be
a good starting point for folks to try tuning for their helicopters.
As always, full source code is available under the GPL from:
http://autopilot.sourceforge.net/
Will filmed the entire flight and took pictures. So far we only have
one low-quality movie available:
http://autopilot.sourceforge.net/movies/hover1.mpg
Once we convert the higher-quality MiniDV footage, we should have
several movies available. We have a few pictures online, too.
This is one of the current system showing the AHRS + GPS + Linux
flight controller + 802.11 CF card mounted on the high-skids under
carriage, and the ground station software showing the artifical horizon,
moving map and PPM data:
http://autopilot.sourceforge.net/images/nova_proto/bench.jpg
Here is Dennis hovering hands-off with the computer handling the
roll, pitch and yaw axes (while I celebrate!):
http://autopilot.sourceforge.net/images/nova_proto/handsoff.jpg
And one of me "flying" the collective + throttle while the computer
handles the attitude and position. If you have ever watched me flying
for real, you would realize that this is an incredible accomplishment...
http://autopilot.sourceforge.net/images/nova_proto/trammell.jpg
The helicopter stayed in about a 1 m CEP during calm winds and about
a 2 m CEP when the wind gusts would hit it. It flew in slow circles
while under computer control, which post flight analysis showed that
our hard iron calibration for the magnetometer on the bench did not
include the engine.
This resulted in about 20 degree offset between our heading and the GPS
north value, so when we would go "north" based on our magnetometer, we
would actually be flying north west. It is a testament to the stability
of the inner attitude loops that this only caused minor position problems
instead of a catastrophic crashes.
As always, this could not have been done without the combined effort of
everyone involved. Especially big thanks go to Aaron Kahn for all
of his code and assistance flying; to Dennis D'Annunzio for risking
his helicopter and being an incredible test pilot; and to Will Barger
for long hours of helping us build assemblies and spending countless
hours at the field with his camera.
Rotomotion is selling kits and assembled units to help you build your
own autonomous helicopters, too. The GPS boards are coming soon and we
hope to have it tuned to work on a Bergen Observer within a few weeks:
http://www.rotomotion.com/
Trammell
--
-----|----- hudson@osresearch.net W 240-283-1700
*>=====[]L\ hudson@rotomotion.com M 505-463-1896
' -'-`- http://www.swcp.com/~hudson/ KC5RNF
Yesterday, Dennis, Will and I hovered Dennis' Nova for entire tanks
of fuel, with only the collective/throttle under manual control.
All control loops except altitude were closed and tuned. We should
write a detailed guide to tuning the PID loops, which took the better
part of two full days.
Onboard the Nova was our IMU / Magnetometer / Servo controller,
a Linux Cerfboard with 802.11 CF, and our new u-blox TIM-LP / IO board.
The gains and code are all checked into our CVS tree and should be
a good starting point for folks to try tuning for their helicopters.
As always, full source code is available under the GPL from:
http://autopilot.sourceforge.net/
Will filmed the entire flight and took pictures. So far we only have
one low-quality movie available:
http://autopilot.sourceforge.net/movies/hover1.mpg
Once we convert the higher-quality MiniDV footage, we should have
several movies available. We have a few pictures online, too.
This is one of the current system showing the AHRS + GPS + Linux
flight controller + 802.11 CF card mounted on the high-skids under
carriage, and the ground station software showing the artifical horizon,
moving map and PPM data:
http://autopilot.sourceforge.net/images/nova_proto/bench.jpg
Here is Dennis hovering hands-off with the computer handling the
roll, pitch and yaw axes (while I celebrate!):
http://autopilot.sourceforge.net/images/nova_proto/handsoff.jpg
And one of me "flying" the collective + throttle while the computer
handles the attitude and position. If you have ever watched me flying
for real, you would realize that this is an incredible accomplishment...
http://autopilot.sourceforge.net/images/nova_proto/trammell.jpg
The helicopter stayed in about a 1 m CEP during calm winds and about
a 2 m CEP when the wind gusts would hit it. It flew in slow circles
while under computer control, which post flight analysis showed that
our hard iron calibration for the magnetometer on the bench did not
include the engine.
This resulted in about 20 degree offset between our heading and the GPS
north value, so when we would go "north" based on our magnetometer, we
would actually be flying north west. It is a testament to the stability
of the inner attitude loops that this only caused minor position problems
instead of a catastrophic crashes.
As always, this could not have been done without the combined effort of
everyone involved. Especially big thanks go to Aaron Kahn for all
of his code and assistance flying; to Dennis D'Annunzio for risking
his helicopter and being an incredible test pilot; and to Will Barger
for long hours of helping us build assemblies and spending countless
hours at the field with his camera.
Rotomotion is selling kits and assembled units to help you build your
own autonomous helicopters, too. The GPS boards are coming soon and we
hope to have it tuned to work on a Bergen Observer within a few weeks:
http://www.rotomotion.com/
Trammell
--
-----|----- hudson@osresearch.net W 240-283-1700
*>=====[]L\ hudson@rotomotion.com M 505-463-1896
' -'-`- http://www.swcp.com/~hudson/ KC5RNF