routing differences bitSailor / zezo / offline
Post by iconPost by Sunkist | 2023-01-12 | 18:07:32

comparing zezo and bitsailor and a self made preparatory offline router

I found that the choose different ideal routes with similar settings. I can see this happening due to memory conservation (underlying grid size?!) or CPU conservation using polar coordinates in 5° steps and best VMG angles maybe. So if several routes are nearly identical in time, each router will find a different one due to rounding errors. So far so obvious.

What I do not understand is that the wave propagation front (or how you want to call it) varies significantly between zezo and bitsailor (offline version kind of similar to bitsailor). - Even making sure both use the same GFS run. what are the uncertainties in the routing that lead to this difference? is there a way to minimize them?
Is there a way to give a error-number or uncertainty to each location while performing the routing?

Cheers,
Sunke/Sunkist

commenticon 3 Comments
Post by iconPost by Hardtack | 2023-01-12 | 22:47:49
Given the same parameters - start time, start location, polars, boat options, forecast cycle - the wave propagation fronts (which we call isochrones) should really be very similar. Bitsailor uses the same wind interpolation as zezo, except for technical differences that may lead to different rounding errors. And bugs, possibly. Differences should be most noticeable in light wind conditions.

Bitsailor handles obstacles like islands less well than zezo. This can affect the isochrones sometimes. As the routers don't sail straight into the wind, adverse wind patterns can have the same effect.

There are a few other minor differences between zezo and bitsailor but it's hard to say what may cause the diverging results.

Regarding robustness of the route, I don't know of any way to assign error margins or confidence values to individual route points (assuming that's what you mean by locations).
An overall uncertainty of the optimal route could be derived from reverse routing. Neither zezo not bitsailor does that currently.

Post by iconPost by zezo | 2023-01-13 | 11:22:30
There might be difference in the temporal interpolation during the update cycle. I still have not implemented the supposed 3-grib algorithm.
Post by iconPost by Sunkist | 2023-01-16 | 17:03:57
I think the temporal interpolation during the upgrade process is something VR should worry about/improve ... don't need to mention the artifacts of moving pressure systems.

As soon as I find the time in the next week, I'll try to pin down the precise differences that can be observed. unfortunately quite busy right now. But thanks for your inputs.
border
Topics list
Posts
border
1
border
border
5
border
border
Copyright 2009 by ZEZO.ORG. All Rights Reserved.