Adding non binary cost to A*

[ricick]

Would it be possible to add a non binary cost (e.g. 0.5) to matrices, so cells can be counted as walkable but costing more, to allow taking terrain speed into account for example.

[qiao]

This feature could be implemented for A*. However, some other algorithms in this library such as the Breadth-First-Search and the Jump-Point-Search can not use non-uniform cost. Therefore, this behavior may lead to inconsistencies between the APIs.

[MasterScrat]

This fork adds cost support for A*: https://github.com/johnste/PathFinding.js

[dwabyick]

FYI - I was also looking for cost support ... Going to check out the fork. (Thank you for the great library Qiao!)

[rohan-deshpande]

Hey has anyone used this Fork and can share a bit about how it is implemented? I think this should really be part of the A* constructor, it really is necessary for more complex path finding.

[imor]

I was thinking about adding support for this as this is quite a common scenario for A*. But as pointed out by @qiao will have to think about which other finders is it relevant for.

[tlhunter]

Any updates on merging the https://github.com/johnste/PathFinding.js fork into pathfinding?

It's two years behind and probably has a lot of merge conflicts by now...

If someone were to go through the effort of cleaning it and making another PR would it get accepted?

[imor]

johnste's fork only adds support for A*. If anyone is interested in doing this please keep in mind that it needs to be done for other finders as well (wherever it makes sense). Before jumping in though, a discussion about the intended plan would be useful in this thread.

[rohan-deshpande]

I don't really have time but I think this is a necessary feature for an A* pathfinder. I think it would be awesome if it was made part of the official repo and offered as an option for the A* algorithm (if possible).

[tarlepp]

+1 for this

[paulrobello]

I have added a fork with support for non uniform node cost
https://github.com/paulrobello/PathFinding.js
I have also added a pull request #100

[bentwonk]

I like you addition paulrebello.

I note that trace has been removed as being the same as bestfirst, but in my application trace consistently outperforms bestfirst by 5-10%, what is the difference?

[PSeitz]

+1

Shouldn't be a problem to throw an error if the cost parameter is used with an unsupported algorithm?

[rohan-deshpande]

@paulrobello thanks for your fork! Does this add cost to other algorithms other than the A*?

[rohan-deshpande]

Sorry just read the PR! Great stuff!

[pajtai]

@qiao any reason not to merge this? would be nice to get it into the npm