Path finding in a highly dynamic world

I am working on a simple soccer simulation, I am using potential fields for collision avoidance more specifically following technique,

http://www.ibm.com/developerworks/java/library/j-antigrav/

only obstacles on the field are other players and they are constantly moving. Problem is it works if I assign really big push force to the characters since characters move at speed it takes some time to change direction but this has few drawbacks with such a high gravity I can never position an npc to grab the ball cause there is always some force pushing me around.

I though I could solve this by assigning pulling force to the ball but that actually made it worse. nps would go to the ball, ball starts pulling which开发者_JAVA百科 makes npc push the ball it goes into a loop until npc crashes to a wall.

How I've implemented this is, I have a vector that would steer me towards my target then I add to that all gravitational forces acting on the npc and steer towards that direction.

Basically I am wondering what kind of improvements I can make? my current problem is not hitting other players precisely getting behind a ball without getting affected by other players.

I'm not sure that employing potential fields is the way to go here. If all players are directly between you and the ball, how do you get to it?

I'd be tempted to plot a straight line, then iteratively adjust the route for the position of other players, adjusting for their trajectories and — if you're really clever — anticipating changes in the same.

You could put some kind of limit as to the area of effect that another player's gravity well will have. Limit it to a small radius around that player so that when you're clear of other players, there are no forces acting on your character. You could also diminish the collision avoidance force when you're near the ball, reasoning that players care less about not hitting each other when it's time to kick the ball.