Quote:Original post by ID Merlin
Keep the value small relative to the cost and it doesn't interfer with finding the optimal path.

I find this quite amusing actually. The addition of a penalty redefines the cost function and hence the optimal path, so in one sense you're not interfering with it (because you're finding the optimum for that new cost function) but in the other, you are interfering with it (because you're finding not finding the original optimum)! ;)

Timkin

Quote:Original post by ID Merlin
But in the real world, the path that a unit would choose would either be something that looks like a straight line, and not a diagonal move until you hit a wall.

Obviously path shapes should be considered during cost function design. I wasn't deriding your suggesting of penalising nonlinearity in the paths... I was merely pointing out your paradoxical comment and that I found it amusing.

Quote:The point of path finding in a game is not to find any mathematically optimal path, but to find a path that makes sense to the human player watching the unit move.

I'd disagree on that point, since the human doesn't have all of the information, or might not be able to reasonably process nonlinear interactions within that observed data of the pathing domain. The path should reflect what the player expects, but should deviate from those expectations where necessary. A path that doesn't meet the players expectations indirectly alerts the player to something that they haven't perceived and so may be a useful design tool.

Cheers,

Timkin

There's nothing new in that article... such methods are functionally equivalent to finding a piecewise linear spline interpolant with gradient constraints (spatial and parametric).

I wasn't disagreeing with the merits of such techniques where they are applicable; only with your statement about the 'point of pathfinding in games'.

Cheers,

Timkin