Weather Routing

[rgleason]

Wow, Gilletarom, this is a serious test of ODraw - WxRte Boundaries! It is a lot slower with some many boundaries. Since we now have a fresh new updated GSHHG High Res Background from Pavel, perhaps we can turn some of these off? What was the reason for them all, I am curious.

Did-g has a routine that will draw a new boundary around shorelines a certain distance away, that could be used with GSHHG. Might even have a user variable distance.

[Gilletarom]

Originally Posted by rgleason

Wow, Gilletarom, this is a serious test of ODraw - WxRte Boundaries! ..... What was the reason for them all, I am curious.

Did-g has a routine that will draw a new boundary around shorelines a certain distance away, that could be used with GSHHG. Might even have a user variable distance.

"What was the reason for all, I am curious ?".

On the one hand, this area is infested with pebbles, and on the other hand, some highlands are infested with fishermen's traps. It was worth taking into account all of this to check the ability of WR to go through very narrow corridors. And I must say that I was impressed after some of the tests. Yes, WR manages to clear a path in this maze.

It remains to solve departures and arrivals of wearther-route.. And here it is not won yet.

[rgleason]

The first two work in about 9 - 10 minutes. Now on Cherbourg-Porte-ouest.

It completed both ways. On the return, the outline for calculation has a zig-zag-zig-zag, which is unusual. I wonder if Sean knows why?

[rgleason]

RG wrote:

Yes Giilletarom, it would be nice to have Installation prompts for:

1. "Do you want to remove all Weather_routing settings in the Opencpn.ini file? Yes, No."
2. "Do you want to remove the User filesfor Weather_routing [For Windows: programdata/opencpn/plugins/weather_routing/] ? Yes, No"

The rest is taken care of by the Uninstall Program.

See: http://www.cruisersforum.com/forums/...ml#post2547475

IN Resonse Jon Gough wrote:

Hi,
Modifying the uninstall process would take quite a bit of time and the generation of more special code. I have come at the problem from another direction. I have added a button on the 'General' tab of the OD preferences dialog. This allows you to delete ALL OD settings (those that are in the '[Settings/ocpn_draw_pi]' when OCPN is closed down. When OCPN is restarted and if OD is enabled this section will be created as if you had just installed OD for the first time. The 'delete' of the section only occurs at OCPN shutdown, which must be clean, so that if you decide you really didn't want to delete the section you can undo your action.

This change is available in patch 1.4.46.

Regards
Jon

I think this is a very good idea, Sean.

[rgleason]

Gille, I got interrupted testing with your files. There is a "polar vortex" storm approaching, a big winter hurricane...

Sean, hope you are staying warm.
Gille's examples are showing that WxRte is not recognizing the new GSHHG High Res Background files. I have deleted the gshhs file directory and placed the new high res gshhg files in my chart directory.

See attached routing please. Have several examples like this where land is not recognized.

[rgleason]

Gille, they all completed. Two of them went over land, and it appears that the Background GSHHG was not recognized (the routing relied on your extensive Boundary Mappings).

See weatherrouting.xml I changed some of the configs.

[rgleason]

Sean, maybe you have some ideas about why this is happening?

http://www.cruisersforum.com/forums/...ml#post2547966

[Gilletarom]

Hello Sean, hello WR team,

Here is another anomaly that occurred in another test.

I wished to go from Cherbourg to Cowes. I added a new starting point, inside the port of Cherbourg, Just behind the breakwater. His name is Cherbourg-inside.

I loaded the Skyron Cherbourg file tonight and did some tests.

I asked for the determination of a weather-route from Cherbourg-inside to Cowes. No problem, WR has successfully completed the weather-route.

But, what's more, I noticed while looking at the oeSENC chart, that there is a breakwater in front of the port of Cowes, (north of the Isle of Wight). I added a boudary around this breakwater.
I asked WR again to create the weather-route between Cherbourg-inside and Cowes, with the same configuration.

In this case, WR fails, very close to the end point and announces Polar failed.

Look at the two screenshots. One ends well. In the other, the last isochron is defective and zig-zag, because of the presence of the boundary around the breakwater.

Here the new grib file : DL.FREE.FR (unzip it)
Here join the new xml configuration file
Here join two screen copy , just the area of Cowes and you see that in one case, all is Ok, in the second case, WR do not complete the meteo-route.

[seandepagnier]

Originally Posted by Gilletarom

Hello Sean

Thanks to have read my posts.
,


1 ° You say this:
"The first case where it's going to go ... but continuing to explore the north sea, I think the end point is within the land? (Gshhs) map to see?"

You may notice that the screenshots show no mapping. We just see the basemap GSHHS. I often do weather-route searches without maps displayed. This was the case in my two tests.
Do you want to say that WR uses, marine charts installed, even if they are not displayed? Do you want me to remove the nautical charts installed in OpenCPN?

2 ° In my two last tests put online, the best weather-route came very close to the end point towards which the virtuald boat was going. There remained less than half a mile to go. This is true for the test around the Cotentin, it's true for the test to Newlyn at the southern end of England.

I am well aware that by placing the point of arrival where I placed them for these tests, I ask WR too much. For research around the Cotentin, it would be enough to shift the point a little to the east to eliminate all the problems. For the weather-route search to Newlyn, I placed the arrival point at the harbor entrance while there are two boundary lines that delineate a fairly tight access channel in front of this harbor entrance.

But, it is by observing how WR traces the isochrones and what space is available near the arrival point, that I had the idea to propose that WR modifies "for a local use", two of the configuration settings .

That said, I give you herewith the files you request.
The links for the grib files provided here are those in the posts that report the tests.
You must also set up the boundary that works with Draw.

And dowload also this files (but, they are big files):
DL.FREE.FR > Grib file Cherbourg, Grid 4km.
DL.FREE.FR >Grib file Channel, Grid 12km.
DL.FREE.FR > Boundary

In each case, put out the ".doc" and unzip the file

I hope I have not forgotten anything.

Note that for the configuration from Cherbourg to Newlyn, I did various tests after the one I put in lines.
You will have to look exactly at the screenshots to find the exact settings of this test. The current settings do not manage to find the right weather-route, but do not pass through the land and continue to search north of the southwestern tip of England. But it takes patience because with this configuration, the search takes a very very long time.

I will try it in the next days. I am waiting for a winter storm to pass anyway.

[seandepagnier]

Originally Posted by Gilletarom

Hello Sean, hello WR team,


But, what's more, I noticed while looking at the oeSENC chart, that there is a breakwater in front of the port of Cowes, (north of the Isle of Wight). I added a boudary around this breakwater.
I asked WR again to create the weather-route between Cherbourg-inside and Cowes, with the same configuration.

In this case, WR fails, very close to the end point and announces Polar failed.

Great work making difficult cases!

I think the simplest way to solve this is to reduce the time step. The routing should be able to complete with smaller steps, but it may take longer to complete.

It is difficult to route around boundaries or land when the boundary is small compared to the distance in a time step. I hope this makes sense.

[rgleason]

Sean and Gille, see, please
http://www.cruisersforum.com/forums/...ml#post2548263

[Gilletarom]

Hello Sean, Hello WR team,

1° Good news: I loaded the latest version of Draw compiled by Hakkan from JongOugh's work.
The defect which systematically displayed the boundaries of the boundary has disappeared. See the Jongough Draw opencpn thread.

2 ° Bad news. This concerns WR:
For example: I run the tested configuration in my last test report (between Cherbourg and Cowes). Configuration that ends with "Polar failed", while the weather-route has arrived very close to the end point.
We can try to export this incomplete road weather because the "Export" button remains active.
BUT
This causes an OpenCPN crash each time.

[rgleason]

Gilleterom, Thank you for stress testing weather_routing. I was quite amazed that WR and OD would work together so nicely, with such an extensive and large group of boundaries defined all up and down the french and english coast. Now you add breakwaters! No wonder it is taking 10 -12 minutes to compute a routing, and no wonder final completion is trez difficile. I usually try to keep it simple using high res background and perhaps a few boundaries. These calcs are faster, indeed calcs can be done concurrently now, I've run 7 routes at once!

I use weather routing for a more general understanding of options rather than such detail.

But thank for the stress tests. I think it is showing that they work together pretty well and that we do need a few improvements such as dynamic interval adjustment in difficult land areas and at completion.

With your fresh experience and stress testing. What do you think is needed on a practical level?

[Gilletarom]

hello,

Originally Posted by rgleason

Gilleterom, Thank you for stress testing weather_routing. ......
With your fresh experience and stress testing. What do you think is needed on a practical level?

I wish we were able to answer a "terrible" question .... The weather-routes that are determined by WR are they really the best?

Indeed, taking the responsibility of offering users such tools, we take heavy task to replace the navigation routes they would implement, without us, by those we bring them. And because of this, we are questioning their safety at sea.

As things stand, the answer to my "terrible" question is not yet guaranteed. But I have hope in the future.

After that, if we could answer "YES" to this question, I hope that WR's usability is further improved. It is also necessary that the interface with the core of OpenCPN is improved.
- The operation of the road and track manager does not suit me right now. I already talked about it in this thread. And I also asked Dave Register about it.
- I'm in doubt as to how OpenCPN and WR handle how to account for hours: The core has 3 choices: UTC, PC Hour and LMT. WR has only two choices. And we must also consider the choices of Grib plugin. Nothing currently allows us to guarantee that there are no bugs due to the management of the time. It's too complicated for alpha users.

What I like, because there is something that I like a lot, is that now, for the development of WR, there is a team with Sean, Stelian and Did-g. That's good.

[Gilletarom]

Hello WR team,

And now, a problem around the starting point.

I resume research between Cherbourg and Cowes. In the first tests, I put my starting point completely outside the harbour. Then, I set a starting point in the harbour front, just south of the breakwater: "Cherbourg-inside".

With starting point at "Cherbourg-inside", I did 3 tests with a timestamp of 30Mn, 20Mn and 10Mn. WR does not complete correctly if TS = 30mn. But he finishes for TS = 20mn and TS = 10Mn. So I have two tracks by exporting the virtual tracks of the virtual boat and an incomplete "virtual track". See the screenshot 30mn in pink, 20mn in green, and 10mn in gray.

1 ° Observe that two tracks pass by the west of the Isle of Wight and a track by the east of the Isle of Wight.

And immediately a question comes to mind: Which one is the best? The one that requires a short calculation time with TS = 30mn or the one that requires a long duration of calculation with TS = 10mn?

2 ° Look how the tracks leave the front port of Cherbourg. Two tracks pass through the west exit, and a track goes through the east exit.

And one more question: Since the conditions are the same (wind and current) with TS = 30 or 20 or 10mn, why does WR make different choices? In these conditions can we trust in WR?

I am even downright skeptical in the choice made by WR in both cases of exit to the west. Indeed, it comes out against the current and with a rather head wind while, towards the east, one is current and wind in the back to arrive at the exit.

Is my analysis bad?