Just throwing in my two pennies - You don't to have to have your map representation (hex) constrain the entities within it's spatial paradigm. There's no reason you can't combine two path finding techniques, use hex as the first quick step and refine as the path is traveled using a higher resolution /alternate representation.
Using common game themes eg. buildings can be placed free form onto the hex map and then bounds checked to determine which hexes are totally obscured / partially obscured. Steering algorithms combined with this step as well as an additional heuristic calculation for partially occupied hexes would decouple your entities from the hex paradigm (if you so wish).
Hexes will partition into triangles obviously, but there's nothing stopping you dividing those triangles again and using these smaller triangles to refine the path finding.
Long story short : You can and often should have multiple spatial representations that you keep in sync as different algorithmic techniques work best with different representations. IE a quad-tree, where you can also push a reference to the individual hex tiles into.
Bookmarks