Post by iconPost by PatH | 2012-01-15 | 01:38:04

Hi Cvetan,

I noticed today that often when I choose a larger timestep (5 day or 7 day plan) when tracing roughly the same path, that the larger timestep seems to indicate it will be more efficient (go farther by a set period of time). I would think that the smaller the timestep the more efficient a course the algorithm could find. Are my eyes just going blurry from looking at the screen too much?

Cheers,

commenticon 6 Comments
Post by iconPost by zezo | 2012-01-15 | 09:12:50
There is more error in the longer paths. When you do a 7-day run the step i 1 hour. It will overestimate the boat speed if you are headed into weaker winds, because it will sail for 1 hour at the higher speed instead of decreasing it gradually.

That will also change the path a bit.

The other problem is that you can't really fit integer numbers of 1-hour steps in 3.5 hours, so your starting time is rounded to the nearest hour (it uses old wind data for up to a hour after the changes).
Post by iconPost by PatH | 2012-01-15 | 14:37:08
OK thanks Cvetan. I should have figured that out. In summary, the smaller timesteps really are more accurate (as expected). The apparent "enhanced efficiency" of the routes found by the larger timesteps is a combination of numerical and model error. Thanks!
Post by iconPost by zezo | 2012-01-15 | 14:48:26
Yes. You just have to use 2-pass approach - check the general strategy with on the 5-7 day forecast, then set the destination 15-20 hours ahead along the track and request 1-day routing to get very accurate track (same 10-minute steps as the game).

A variable step algorithm could do it in only one pass, but I don't think that the overall result will be too different. The error in the weather model 2-3 days ahead is bigger than the routing error almost all the time.
Post by iconPost by PatH | 2012-01-15 | 17:48:19
Thanks. Pretty much exactly what I've been doing (at least in quarters). In the open ocean I'm not sure I have time to manage 10 minute route adjustments.
Post by iconPost by PatH | 2012-01-15 | 17:52:48
*close-quarters
Post by iconPost by zezo | 2012-01-15 | 18:25:25
Sometimes the more precise track is more stable and straighter than the coarse one, so I usually check it if the 7-day track makes small random turns.
border
Topics list
Posts
border
5
border
border
Copyright 2009 by ZEZO.ORG. All Rights Reserved.