WxRouting- QtVlm or the plugin.

[cagney]

I have compared QtVlm with the OpenCPNs Weather Routing plugin.
I can't say that I'm an expert on either, so settings are probably not optimal. What I do see is that for a given route the plugin always predict a longer passage time even when the routes are very close. Below is a route across the Bay of Biscay. Both agrees on the general route, there are differences however, and the time en route differs with 12 hours! Why?
The same grib, start-time and boat polars are used. I assume that QtVlm is thoroughly tested, while the plugin is fairly new......
Is this a real difference, or does it just reflect my inability to tune the plugin settings?




The used grib and polar is available here: http://www.filedropper.com/comp

Thomas

[LJH]

I haven't compared them yet either. I found that QtVLM is very accurate if I set the polar efficiency at 80%. I will have to try the plugin

[rgleason]

First I am trying to reproduce the example given using the files attached, in Weather-Routing. This seems to be a different route than the Wx-Rte done in Linux and I am not sure why yet.

Any ideas?

How this was run.
Start: 9/1/2013 16:00
Route: from Rio Allones Spain to Truro Eng
Polar File: boat_Mono650.csv
Grib File: 20130901_094651_.grb.bz2
Time Steps: 1 hour
Degree Steps: 5 degrees

Time: 5 days 1 hour

[rgleason]

Here's another trial which is 4 days 28min.
I noticed that getting 9/1/2013 16:00 in the grib had to be an interpolated interval.

Also I noticed that it was difficult to get the "Grib Time" selection in Weather Routing to register at 9/1/2013 16:00 took several trials.

The route is a different path entirely. Are you sure it is the same grib file being used?

I'll run the qtVLM trials later.

[cagney]

The route I tested was from UK to Spain.
Start N 49 59.5 W 5 04.3
Finish N 43 09.1 W 9 16.1

[rgleason]

The route I tested was Spain to England.
Test-FR (mislabeld) N43.27 W9.07
Test-Eng N50.12 W5.007
Therefore these tests above are void.

[rgleason]

Route tested from UK to Spain
Start N 49 59.5 W 5 04.3
Finish N 43 09.1 W 9 16.1
4 days 3:54 hours
465 nm

[rgleason]

QtVLM
Route tested from UK to Spain
Start N 49 59.5 W 5 04.3
Finish N 43 09.1 W 9 16.1
3 days 4:75 hours
458 nm

[rgleason]

qtVLM
Same setup, but after conversion of Routings to Route with simplification and optimization. Saved about 15 min.

[rgleason]

1. The optimal routes are very similar.
2. Therefore the routing algorithm for each program appears to have some correlation.
3. The time is different by about 23 hours. Why?

Why is the time different?
-Surely not because one of the programs found a better route! The routes are too similar.
-Perhaps the boats polars are handled differently with resulting speeds that are lower?
-Perhaps there are some other factors to consider in the way the grib file is read.

In any event, without extensive testing of both programs there is no way to establish which program most accurately predicts the arrival time.

[cagney]

I used the default settings for the plugin. I just loaded the polars and left it at that. The plugin said that the average wind speed was 11 kts.
Here is qtVlms polar for that wind speed.

There is no way that the plugins original ~4.5 kts average boat speed could fit into that.
This made me look closer at the plugins polar and polar settings.
Originally the max speed was about 6 kts, while qtVlm gave about 8 kts.
Something fishy here
In both applications you can adjust the polars. In qtVlm this is a simple % efficiency scale. In the plugin things seems more complicated. Have a look at this messy picture.


The Boat dialog is reached through the settings button next to where you choose your polars. Notice that the Sail Plan tab is open. I have adjusted the "Eta" (whatever that is) to 0.061 to get reaching boat speed up to 8 kts in 11 kts of wind. With this setting the boat of-course performed much better. See the wide purple line on the plot.

So ..... to compare the two wx-routers, the polars must agree. In mus previous post, this was not the case.
One problem here is how the plugin handles wind on a dead run.
Almost 7.5 kts of boat-speed in 11 kts of true wind and 4.5 knot of apparent wind. Not quite possible in the real world. Compare with qtVlm, that have a more realistic polar for runs and broad reaches.

Clearly I have a lot more exploring to do, to understand the plugin, and to be able to do a fair test to compare with qtVlm.

[seandepagnier]

Yes, the calculation for boat polars needs a lot of changing. For comparisons, you should load the polar from a file since the plugin can read the same format as qtvlm and then the polars should be identical. A slight change here can give large differences to the results as grib is different at different times.

I would be interested in differences between the two programs with identical input parameters, and also how long the calculation takes.

[rgleason]

Sean, great to have your participation. I am very excited to be working on this with such a great team. OpenCpn users are going to benefit and we will have an unparalleled tool if we can work this out. Cagney and I are trying to learn more about the Weather Routing plugin and to compare and thus improve Weather Routing. Thomas has been helping, with a new perspective.

The Boat dialog is reached through the settings button next to where you choose your polars. Notice that the Sail Plan tab is open. I have adjusted the "Eta" (whatever that is) to 0.061 to get reaching boat speed up to 8 kts in 11 kts of wind. With this setting the boat of-course performed much better. See the wide purple line on the plot.

So ..... to compare the two wx-routers, the polars must agree. In my previous post, this was not the case.
One problem here is how the plugin handles wind on a dead run.
Almost 7.5 kts of boat-speed in 11 kts of true wind and 4.5 knot of apparent wind. Not quite possible in the real world. Compare with qtVlm, that have a more realistic polar for runs and broad reaches.

Clearly I have a lot more exploring to do, to understand the plugin, and to be able to do a fair test to compare with qtVlm.

I am learning too. As I understand it, the WxRte plugin has a number of pretty lofty & sophisticated goals:
1. Polars: Built-in VPP using boat specs to generate polars & accept qtVlm polar files.
2. Polars & Conditions: Select Sea conditions & Sail sets to change polars. - not sure how this works. Need to learn more.
3. Utilize both Climatology and Grib data - wind, currents and waves.
4. Create routings with constraints on waves (max), speed (max & min-engine), wind (max)
5. Run multiple routings to find best weather windows.

This is a lot to program, and considering the fact that Sean has been using this plugin for his routing, as a very practical test and reality check, that may be why the plugin does create optimistic times. We need to understand some of the parameters better.

I suppose the field "ETA" is Estimated Time of Arrival and perhaps that is similar to "Efficiency" in QtVLM. I hope Sean will explain this a little more for us.

Cagney wrote:

One problem here is how the plugin handles wind on a dead run.
Almost 7.5 kts of boat-speed in 11 kts of true wind and 4.5 knot of apparent wind. Not quite possible in the real world. Compare with qtVlm, that have a more realistic polar for runs and broad reaches

Thomas would you please explain this a little more? Why do you think qtVlm has more realistic polar for runs and broad reaches?

[maitai]

I'm interested too in those tests

It would be a nice idea to post or give a link to polar/grib/starting-ending point/date of departure each time you run such a comparaison. I can run also Adrena-pro on this that can serve as a referee, sort of (unless we are better, of course )

[rgleason]

Dear Maitai,

It is super you are on this thread too! I have to say that by the end of the Virtual Vendee, I was really comfortable with qtVlm and my skills were getting better too, with your help (Although I missed the finish and I think I dropped from about 53 to 125!) -Some friends thought I should have lost the boat completely, since it was found in the morning hard against the French shores!

We will include "post or give a link to polar/grib/starting-ending point/date of departure each time you run such a comparison."

Also of course we'd like to take you up on your offer to run Adrena-pro on the same data. I've just become aware of that program through Predictwind.

I hope we are all ready to learn some interesting stuff.