I'm trying to create a library for steering AI in games, following the article by Craig Reynolds. In the article, he discusses a "simple vehicle model", wherein every vehicle has some maximum velocity and maximum acceleration. Notably, those are isotropic: the vehicle max speed and acceleration is the same in all directions.

A little bit later in the article there is a mention that this is not a very realistic model. For example, the following might be a diagram for acceleration capabilities of a car, where breaking is easier than accelerating:

Indicatrix of acceleration

I would imagine for different vehicles the diagram would be different. For example, a rocket may not have any breaking capabilities at all. For a human running forward is much easier than backpedaling. It is also probably true that acceleration capabilities depend on current speed of movement. For example, there may not be such thing as "maximum speed" per se, just that at certain speeds forward acceleration becomes close to zero.

My question is: how would I approach modeling alternative (to 1999 Reynolds) formulations of vehicle steering capabilities, and how would this affect the implementation of steering behaviors.



You must log in to answer this question.

Browse other questions tagged .