Glider Racing in the future

All the details are in the paper from the US Naval Research Lab that you may read he

http://www.dtic.mil/get-tr-doc/pdf?AD=ADA614555 (Downloads a PDF)

Thanks Dave!
Reading this is fascinating. Here is a quote:
"Main (control)loop rates faster than 4 Hz were
tested (up to 10 Hz), but did not appear to have a positive effect on the success of the thermal centering.
Slower rates down to 1 Hz were tested, but appeared to consistently fly through lift and have poor centering ability. The rate of 4 Hz appeared to be a good balance between computational usage and thermalling performance."

Wow, does this apply to human pilots too? If so then you have to make a decision of to turn or not and which way to turn in less than one second or you might be missing lift. Somehow this does not sound right, especially because this small scale glider is flying slower than a sailplane.

On Tuesday, February 16, 2016 at 11:22:10 AM UTC-5, Soartech wrote:
All the details are in the paper from the US Naval Research Lab that you may read he

http://www.dtic.mil/get-tr-doc/pdf?AD=ADA614555 (Downloads a PDF)

Thanks Dave!
Reading this is fascinating. Here is a quote:
"Main (control)loop rates faster than 4 Hz were
tested (up to 10 Hz), but did not appear to have a positive effect on the success of the thermal centering.
Slower rates down to 1 Hz were tested, but appeared to consistently fly through lift and have poor centering ability. The rate of 4 Hz appeared to be a good balance between computational usage and thermalling performance."

Wow, does this apply to human pilots too? If so then you have to make a decision of to turn or not and which way to turn in less than one second or you might be missing lift. Somehow this does not sound right, especially because this small scale glider is flying slower than a sailplane.

Could be that the "organic computer" (human pilot) is hearing & feeling little things that indicate you're getting close to lift. This, coupled with experience:
1-Has you primed to do something
2-Allows you to focus on additional info to "premake" a decision
3-All that's left is to decide "when" to actually carry out the plan

Well...... that's how it goes for some people. ;-)

"The rate of 4 Hz appeared to be a good balance between computational usage and thermalling performance."

Wow, does this apply to human pilots too? If so then you have to make a decision of to turn or not and which way to turn in less than one second or you might be missing lift. Somehow this does not sound right, especially because this small scale glider is flying slower than a sailplane.

This is rather interesting from a controls and handling qualities standpoint, given that the quickest most pilots are capable of responding to an aircraft is between 4 and 5 Hz... who would have thought we were optimized for anything related to flight!

On Tue, 16 Feb 2016 08:48:49 -0800, Giaco wrote:

"The rate of 4 Hz appeared to be a good balance between computational
usage and thermalling performance."

Wow, does this apply to human pilots too? If so then you have to make a
decision of to turn or not and which way to turn in less than one
second or you might be missing lift. Somehow this does not sound right,
especially because this small scale glider is flying slower than a
sailplane.

This is rather interesting from a controls and handling qualities
standpoint, given that the quickest most pilots are capable of
responding to an aircraft is between 4 and 5 Hz... who would have
thought we were optimized for anything related to flight!

Don't forget that ALOFT didn't use a vario: it took all its altitude and
climb rate data from the GPS receiver (they preferred climb rates read
off the GPS to autopilot's output because it was a less noisy signal),
but combined that with airspeed and (I think) pitch and roll rates output
by the autopilot. And then they downlinked all that to a laptop and
uploaded instructions, which they fed to the autopilot to control the
thermal search and centring.

I think they used a bidirectional link to a laptop, which did the
calculations, because there was nothing like the RaspberryPi or
BeagleBoard Black available at the time and also because it let them fine-
tune its calculations in real time while the model was in the air.


--
martin@ | Martin Gregorie
gregorie. | Essex, UK
org |

My guess is that the primary consideration here is "poor centering" and not that the slower controller was missing thermals. The fact that the uav is slower can paradoxically require higher rate control. For a given bank angle, the turn rate is inversely proportional to speed. This means that a given amount of vario (or control) latency corresponds to a greater angular displacement around the circle.

This was and is a nontrivial challenge for autonomous sailplanes, especially when we try to implement an algorithmic version of something like Reichmann's method.

I think Charlie's right on another part of it. Humans (or at least this one anyway) fly with a lot of feedforward in their thermalling, that is we remember/anticipate where we felt the bump or vario jump on the last circle, and anticipate when to start making an adjustment.

I had the opportunity to fly our soaring uav two weeks with Dan's feedback and observations. For one of the weeks he had ALOFT out as well and we could compare side-by-side. ALOFT does indeed do everything on a laptop via radio link, which has advantages and disadvantages.

Compared to our carefully simulated and prepared software builds (our plane was running totally on-board), it was bizarre to see him ctrl-c his system, rewrite some code and boot the whole thing back up, all while the plane was a kilometer overhead. On the other hand, if the radio link got shaky, ALOFT couldn't thermal at all.

shameless plug
Autonomous glider presentation on Friday at the convention, learn all about them!
/shameless plug

--
John Bird
autonomous sailplane takeoff, landing, and spin recovery specialist
AVIA Lab, Penn State

On Tuesday, February 16, 2016 at 4:40:52 PM UTC-5, Martin Gregorie wrote:
On Tue, 16 Feb 2016 08:48:49 -0800, Giaco wrote:

"The rate of 4 Hz appeared to be a good balance between computational
usage and thermalling performance."

Wow, does this apply to human pilots too? If so then you have to make a
decision of to turn or not and which way to turn in less than one
second or you might be missing lift. Somehow this does not sound right,
especially because this small scale glider is flying slower than a
sailplane.

This is rather interesting from a controls and handling qualities
standpoint, given that the quickest most pilots are capable of
responding to an aircraft is between 4 and 5 Hz... who would have
thought we were optimized for anything related to flight!

Don't forget that ALOFT didn't use a vario: it took all its altitude and
climb rate data from the GPS receiver (they preferred climb rates read
off the GPS to autopilot's output because it was a less noisy signal),
but combined that with airspeed and (I think) pitch and roll rates output
by the autopilot. And then they downlinked all that to a laptop and
uploaded instructions, which they fed to the autopilot to control the
thermal search and centring.

I think they used a bidirectional link to a laptop, which did the
calculations, because there was nothing like the RaspberryPi or
BeagleBoard Black available at the time and also because it let them fine-
tune its calculations in real time while the model was in the air.


--
martin@ | Martin Gregorie
gregorie. | Essex, UK
org |

On Tue, 16 Feb 2016 15:32:04 -0800, aibrd wrote:

Compared to our carefully simulated and prepared software builds (our
plane was running totally on-board),

Nice one! Can you say what you're using for the onboard computer?

it was bizarre to see him ctrl-c
his system, rewrite some code and boot the whole thing back up, all
while the plane was a kilometer overhead. On the other hand, if the
radio link got shaky, ALOFT couldn't thermal at all.

Sounded nasty at first, but then I realised that Aloft's autopilot would
simply carry on doing what it was last told to do until the laptop came
back online, so actually its a nice trick.


--
martin@ | Martin Gregorie
gregorie. | Essex, UK
org |

An Odroid XU4, similar guts to the Samsung Galaxy S5. Probably overkill, but it has worked really well for us.

I was taken aback the first time he did it, but that's exactly what happened -- it just kept on flying its search pattern until everything was back up.


On Tuesday, February 16, 2016 at 7:06:09 PM UTC-5, Martin Gregorie wrote:
On Tue, 16 Feb 2016 15:32:04 -0800, aibrd wrote:

Compared to our carefully simulated and prepared software builds (our
plane was running totally on-board),

Nice one! Can you say what you're using for the onboard computer?

it was bizarre to see him ctrl-c
his system, rewrite some code and boot the whole thing back up, all
while the plane was a kilometer overhead. On the other hand, if the
radio link got shaky, ALOFT couldn't thermal at all.

Sounded nasty at first, but then I realised that Aloft's autopilot would
simply carry on doing what it was last told to do until the laptop came
back online, so actually its a nice trick.


--
martin@ | Martin Gregorie
gregorie. | Essex, UK
org |