Microsoft Teaches Autonomous Gliders to Make Decisions on the Fly

[Bob]

Here we go...

https://www.nytimes.com/2017/08/16/technology/microsoft-teaches-autonomous-gliders-to-make-decisions-on-the-fly.html?hp&action=click&pgtype=Homepage&clickSource=story-heading&module=first-column-region&region=top-news&WT.nav=top-news

[Dan Marotta]

Well, then, I just have to decide whether to mount a camera in the
glider so I can enjoy the flight from my living room or simply remove
the stick from my glider and go along for the ride.

Will I need a Facebook account to share the flight with friends, or can
I expect OLC to make the necessary software upgrades?

On 8/17/2017 3:55 AM, Bob wrote:
> Here we go...
>
> https://www.nytimes.com/2017/08/16/technology/microsoft-teaches-autonomous-gliders-to-make-decisions-on-the-fly.html?hp&action=click&pgtype=Homepage&clickSource=story-heading&module=first-column-region&region=top-news&WT.nav=top-news

--
Dan, 5J

[WB]

On Thursday, August 17, 2017 at 4:55:38 AM UTC-5, Bob wrote:
> Here we go...
>
> https://www.nytimes.com/2017/08/16/technology/microsoft-teaches-autonomous-gliders-to-make-decisions-on-the-fly.html?hp&action=click&pgtype=Homepage&clickSource=story-heading&module=first-column-region&region=top-news&WT.nav=top-news

Gonna get me a swarm of these guys to be my wingmen. Let them fan out ahead of me and mark the lift.

Actually, R+D of autonomous soaring has been going on for some time.

[Andrew Ainslie]

Sheesh! A brick run by an 8088 chip can soar in Nevada!

Come to Western New York and see how you do :)

[Martin Gregorie[_5_]]

On Thu, 17 Aug 2017 02:55:35 -0700, Bob wrote:

> Here we go...
>
> https://www.nytimes.com/2017/08/16/technology/microsoft-teaches-
autonomous-gliders-to-make-decisions-on-the-fly.html?
hp&action=click&pgtype=Homepage&clickSource=story-heading&module=first-
column-region&region=top-news&WT.nav=top-news

This is all really quite old news. It is annoying to see supposedly
leading newspapers believing company guff about this sort of stuff when
the pioneering work was done ten years ago by a PHD student, Daniel
Edwards, and which was good enough to gain his doctorate. The project was
called ALOFT, and was unveiled (papers and thesis published) in 2008.
ALOFT is an acronym for Autonomous Locator OF Thermals. Here are
relevant links.

The PHD thesis:
http://zoogz.gregorie.lan/reference/reference/docs/aloft_etd.pdf

Control system description: see NRL/FR/5712--15-10,272

This describes the contest the ALOFT system competed in:
http://www.xcsoaring.com/contests/mccc/2008/report.html

Since all this was published back in 2008, I really find it hard to
believe that anybody can have the sheer cheek to claim to have done
original research in this area without acknowledging to previous work
done by Dan Edwards, yet there's no mention of either him or of ALOFT in
the NYT article.

Of interest to glider pilots is that this project managed to successfully
find and centre in thermals using only the data available from a GPS: the
5m model glider carried a pressure-based altimeter but got better results
by using GPS altitude and ROC calculated from it because that was a less
noisy signal than the pressure altimeter could provide.


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

[Bruce Hoult]

On Tuesday, August 22, 2017 at 2:56:40 AM UTC+3, Martin Gregorie wrote:
> On Thu, 17 Aug 2017 02:55:35 -0700, Bob wrote:
>
> > Here we go...
> >
> > https://www.nytimes.com/2017/08/16/technology/microsoft-teaches-
> autonomous-gliders-to-make-decisions-on-the-fly.html?
> hp&action=click&pgtype=Homepage&clickSource=story-heading&module=first-
> column-region&region=top-news&WT.nav=top-news
>
> This is all really quite old news. It is annoying to see supposedly
> leading newspapers believing company guff about this sort of stuff when
> the pioneering work was done ten years ago by a PHD student, Daniel
> Edwards, and which was good enough to gain his doctorate. The project was
> called ALOFT, and was unveiled (papers and thesis published) in 2008.
> ALOFT is an acronym for Autonomous Locator OF Thermals. Here are
> relevant links.
>
> The PHD thesis:
> http://zoogz.gregorie.lan/reference/reference/docs/aloft_etd.pdf
>
> Control system description: see NRL/FR/5712--15-10,272
>
> This describes the contest the ALOFT system competed in:
> http://www.xcsoaring.com/contests/mccc/2008/report.html
>
> Since all this was published back in 2008, I really find it hard to
> believe that anybody can have the sheer cheek to claim to have done
> original research in this area without acknowledging to previous work
> done by Dan Edwards, yet there's no mention of either him or of ALOFT in
> the NYT article.
>
> Of interest to glider pilots is that this project managed to successfully
> find and centre in thermals using only the data available from a GPS: the
> 5m model glider carried a pressure-based altimeter but got better results
> by using GPS altitude and ROC calculated from it because that was a less
> noisy signal than the pressure altimeter could provide.

That seems very strange and completely against the experience of every glider pilot with both a GPS and a modern MEMS pressure sensor based computerised vario.

[Martin Gregorie[_5_]]

On Mon, 21 Aug 2017 17:35:12 -0700, Bruce Hoult wrote:

>> Of interest to glider pilots is that this project managed to
>> successfully find and centre in thermals using only the data available
>> from a GPS: the 5m model glider carried a pressure-based altimeter but
>> got better results by using GPS altitude and ROC calculated from it
>> because that was a less noisy signal than the pressure altimeter could
>> provide.
>
> That seems very strange and completely against the experience of every
> glider pilot with both a GPS and a modern MEMS pressure sensor based
> computerised vario.
>
Yes, I agree. All I know is what it says in the detailed description
published as NRL/FR/5712--15-10,272. It is marked as an unclassified
document.

This work was done in 2006/7: I know nothing about the quality of
electronic pressure sensors at that time and I don't recall any
description of how the pressure sensor was connected to the rest of the
system or anything about what, if any, filtering or smoothing was applied
to the pressure sensor. Nor, for that matter have I heard anything about
people using GPS altitude information as vario input for our use. The
sampling rate, at least, is similar to the response rate of many
electronic varios so that approach may work for us too.

You can find what looks like a full description here:
https://www.dsiac.org/resources/journals/dsiac/spring-2016-volume-3-
number-2/pursuit-persistent-isr

or search on the document reference: the first search result identifies
the {PDF version that I have a private copy of but the site is either
dead or very slow indeed. If the above reference doesn't give enough
information you can contact me off-list: the copy I have is marked
unclassified.


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

[Bruce Hoult]

On Tuesday, August 22, 2017 at 2:24:38 PM UTC+3, Martin Gregorie wrote:
> On Mon, 21 Aug 2017 17:35:12 -0700, Bruce Hoult wrote:
>
> >> Of interest to glider pilots is that this project managed to
> >> successfully find and centre in thermals using only the data available
> >> from a GPS: the 5m model glider carried a pressure-based altimeter but
> >> got better results by using GPS altitude and ROC calculated from it
> >> because that was a less noisy signal than the pressure altimeter could
> >> provide.
> >
> > That seems very strange and completely against the experience of every
> > glider pilot with both a GPS and a modern MEMS pressure sensor based
> > computerised vario.
> >
> Yes, I agree. All I know is what it says in the detailed description
> published as NRL/FR/5712--15-10,272. It is marked as an unclassified
> document.
>
> This work was done in 2006/7: I know nothing about the quality of
> electronic pressure sensors at that time and I don't recall any
> description of how the pressure sensor was connected to the rest of the
> system or anything about what, if any, filtering or smoothing was applied
> to the pressure sensor. Nor, for that matter have I heard anything about
> people using GPS altitude information as vario input for our use. The
> sampling rate, at least, is similar to the response rate of many
> electronic varios so that approach may work for us too.
>
> You can find what looks like a full description here:
> https://www.dsiac.org/resources/journals/dsiac/spring-2016-volume-3-
> number-2/pursuit-persistent-isr
>
> or search on the document reference: the first search result identifies
> the {PDF version that I have a private copy of but the site is either
> dead or very slow indeed. If the above reference doesn't give enough
> information you can contact me off-list: the copy I have is marked
> unclassified.

When in extreme scratching situations, the pressure altitude from my 1994 Cambridge Aero Instruments "GPSNAV" (one of the first units, hired to a competitor in the 1994 pre-Worlds in Omarama) has saved my bacon a number of times by letting me know whether I'd gained three feet or lost five feet in the last circle.

By 2006/7 something used in a military project should be excellent!

[Martin Gregorie[_5_]]

On Tue, 22 Aug 2017 04:48:16 -0700, Bruce Hoult wrote:

> On Tuesday, August 22, 2017 at 2:24:38 PM UTC+3, Martin Gregorie wrote:
>> On Mon, 21 Aug 2017 17:35:12 -0700, Bruce Hoult wrote:
>>
>> >> Of interest to glider pilots is that this project managed to
>> >> successfully find and centre in thermals using only the data
>> >> available from a GPS: the 5m model glider carried a pressure-based
>> >> altimeter but got better results by using GPS altitude and ROC
>> >> calculated from it because that was a less noisy signal than the
>> >> pressure altimeter could provide.
>> >
>> > That seems very strange and completely against the experience of
>> > every glider pilot with both a GPS and a modern MEMS pressure sensor
>> > based computerised vario.
>> >
>> Yes, I agree. All I know is what it says in the detailed description
>> published as NRL/FR/5712--15-10,272. It is marked as an unclassified
>> document.
>>
>> This work was done in 2006/7: I know nothing about the quality of
>> electronic pressure sensors at that time and I don't recall any
>> description of how the pressure sensor was connected to the rest of the
>> system or anything about what, if any, filtering or smoothing was
>> applied to the pressure sensor. Nor, for that matter have I heard
>> anything about people using GPS altitude information as vario input for
>> our use. The sampling rate, at least, is similar to the response rate
>> of many electronic varios so that approach may work for us too.
>>
>> You can find what looks like a full description here:
>> https://www.dsiac.org/resources/journals/dsiac/spring-2016-volume-3-
>> number-2/pursuit-persistent-isr
>>
>> or search on the document reference: the first search result identifies
>> the {PDF version that I have a private copy of but the site is either
>> dead or very slow indeed. If the above reference doesn't give enough
>> information you can contact me off-list: the copy I have is marked
>> unclassified.
>
> When in extreme scratching situations, the pressure altitude from my
> 1994 Cambridge Aero Instruments "GPSNAV" (one of the first units, hired
> to a competitor in the 1994 pre-Worlds in Omarama) has saved my bacon a
> number of times by letting me know whether I'd gained three feet or lost
> five feet in the last circle.
>
> By 2006/7 something used in a military project should be excellent!

It wasn't military in 2006/7 - just the subject a PHD thesis written by a
student aeronautical engineer. I'd read reports about the project in
model flying publications such as the AMA's 'Model Flying' until its
developer graduated. At this point it seemed to go dark fairly rapidly.
My guess is that this was when the UAS surveillance crowd became aware of
it: most likely when the thesis was published. More info only appeared
several years later.

It seems reasonable to guess that the author was pretty well-known in RC
soaring circles because one of his aims, in which he succeeded, was to
have ALOFT accepted as a competitor in one of the Californian cross-
country soaring contests. These were multi-day events that routinely flew
tasks of up to 100 km with the pilots driven round the task in a
convertible or a beach lounger on the back of a pickup truck. ALOFT was
hand flown for launches and landings and autonomous the rest of the time
with the software running on a laptop with a bidirectional radio link to
the glider, an ARTF 5m soarer kit. The laptop and its owner also followed
the glider round the task by road along with the pilots of the other
models.

So, to me this means that the autopilot hardware was more likely to be
state of the art RC model gear rather than military kit.


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