So I'm wanting to implement steering behaviours for the ai in my game. It's a top-down 2D arcadish game with a similar control scheme as Asteroids, where the ai agents track the player through a maze (they're hostile, of course). The ai has the same movement controls as the player: thrust, reverse, turn clockwise, and turn counter-clockwise. I've looked at Craig Reynolds' web site, an excellent resource on steering ai. However, I'm having a little difficulty figuring out how to adapt Reynolds' flocking/steering behaviours to the steering behaviours of my own ai. Reynolds' steering seems to depends on putting forces on an agent and have it accelerate along the resultant vector. My bots on the other hand have a fixed velocity they can thrust or reverse and a fixed rotational velocity for turning. Are there any resources online that deal with this kind of steering?

Perhaps it would be an adea to allow your vehicles to accelerate up to the max velocity rather than just applying this all the time. It would look nicer.

The main problem with boids is the way the steering forces are combined. I'd suggest using a method where you are always steering towards some point which may be calculated via a number of inputs.

Quote:Perhaps it would be an adea to allow your vehicles to accelerate up to the max velocity rather than just applying this all the time. It would look nicer.

I plan on doing this, but not right now. I need to clean up my physics engine, and I'd rather figure out most of the ai logic before having to take acceleration into account.

Quote:I'd suggest using a method where you are always steering toward some point which may be calculated via a number of inputs.

Would you mind elaborating on this a bit? From what I can tell you're proposing that for my ai I calculate a point or list of points that an agent needs to traverse, then have it steer itself toward these points. This might be good since I could incorporate pathfinding algorithms into my ai. However, I'd also like some of the dynamic behavior exhibited by Reynolds' simulations such as seek and evade, obstacle avoidance, or wandering.