ShipPathFinder

[Chruker]

Here is a pathfinder for ships.

This version can:
- Find its way across open waters
- Build canals
- Build locks
- Build bridges

Some notes though:
- It still have the debugstatements, but they are commented out.

ShipPathFinder wish-list:
- Be able to do some simple terraforming at sealevel.
- Do some simple terraforming at heights to allow construction of bridges, locks and canals

There is also a new AyStar file, which have the following changes:
- Based of AyStar.6
- Added support for the users of Graph.AyStar to pass an additional variable in the neighbour arrays. The variable format for the Cost, Estimate and CheckDirection callbacks have had that variable added.
- Change: Internally the AyStar class was changed to use the variables in AyStar.Path instancies instead of the functions. This gives more speed.
- Change: FibonacciHeap instead of BinaryHeap which gives it a little bit more speed.
- Fix: Comments about which parameters are sent to the callbacks were outdated.
- Fix: When an AyStar.Path instance was used as source it didnt call the estimate function (not a big deal).

Edit 2009-07-06
- License-wise it is the same as the other AI libs (gpl2+)

[Chruker]

A huge thanks must go to Yexo for providing answers to many many many questions on IRC

[Chruker]

What are the opinions on the following things:
- Should the pathfinder build byous?
- The current version contains a BuildPath() function which builds the path, but I'm not sure if its usefull to have that in the library itself.

[fanioz]

well done Chruker. I beleive once its available, many more AI will use it.
Personally I have no time right now to test it and look thoroughly at the code. But I would like to test it and maybe include them in future version of TransAI.

What are the opinions on the following things:
- Should the pathfinder build byous?
- The current version contains a BuildPath() function which builds the path, but I'm not sure if its usefull to have that in the library itself.

Here my opinion:
1. Bouys only build if the route is too long, it is not pathfinding part but path building part.
2. Therefore Buildpath doesnt needed to be included, as it not part of pathfinding. You may just show how to build using the path returned by this pathfinder. Similiarly with road and rail pathfinder, they have documentation on wiki.

Cheers

[SirkoZ]

- Should the pathfinder build byous?

I don't know about that but AI's certainly should and not far apart - as close as possible yet not too crowded - say - 10 to 12 squares apart like the NoCAB AI. This way it offloads CPU from ship pathfinding.

[Morloth]

- Should the pathfinder build byous?

I don't know about that but AI's certainly should and not far apart - as close as possible yet not too crowded - say - 10 to 12 squares apart like the NoCAB AI. This way it offloads CPU from ship pathfinding.

I actually noticed that if you place the buoys to far apart, ships can get lost. I now put one every 25 tiles, at least, to make sure ships are able to find their way. But I haven't really looked into this pathfinder framework, I don't suppose the road / rail pathfinder build anything but only return a list of tiles? In any case do what they do and you should be fine ^^.

[Yexo]

- Should the pathfinder build buoys?

The should be build ever 10/15 tiles, like SirkoZ says, but not by the pathfinder..

- The current version contains a BuildPath() function which builds the path, but I'm not sure if its useful to have that in the library itself.

I don't think the BuildPath function belongs in the pathfinder. The wiki is an excellent place to copy that function, so everyone can adapt it for his/her AI.

[skippern]

Is there still any development on this?

I will try to implement this, or parts of it into my Shipping AI that I just have started to develop (here)