A* and Steering Behaviours

A* is a graph search algorithm that can be used as part of a global path-planning strategy, and steering behaviors can be used to execute the path proposed by the global path planner while providing local obstacle avoidance.

Like Sneftel said, it is possible to combine the two type of algorithms to obtain interesting results. You could use A* to obtain an optimal path and a simple behavior like following a path could be used in conjunction to create realistic movement. By pitching in a behavior to avoid obstacles, you could even prevent your entities from colliding with dynamic objects which might not be considered by your A* search. The following links might help if you're not totally familiar with the implementations or what they should look like:

Steering behaviors in general
Path finding behavior

I have had problems integrating these to methods together in the past. One problem I have had is that the entity might walk through static geometry when trying to avoid a dynamic object. The original path would have yielded no collision, but the dynamic avoidance does. Satic obj collision detection could be added to the dynamic avoidance, but there are lots of instances where u can still get stuck.

How do path engine like AI.Implant deal with this? I was told they avoid steering behaviors completely (by a sale rep that might not have known what he was talking about, i suppose).

Not having the steering avoid static obstacles too is just lazy, I think. As for getting stuck, it should be simple enough to detect when this has happened, perhaps via attaching expected arrival times to path nodes.

Quote:

Original post by Ender1618 How do path engine like AI.Implant deal with this? I was told they avoid steering behaviors completely (by a sale rep that might not have known what he was talking about, i suppose).

They may have meant that they avoid Reynolds-style steering behaviors. Something like RVO is "steering", but not steering behaviors as they are popularly known.