Bit of a dumb question... Back at Uni I was hearing that Real-Time A* was the best for action game pathfinding and what all the cool games were doing. Since then, I've barely seen anyone using it (or at least not mentioning it). Not in work, AI articles in books and the net, middleware, etc. There seems to be more of a leaning of non-real-time methods being used intelligently. So what's the deal? Are people using RTA*?

I would say that "real time A*" would have to be a bit of a misnomer of sorts. A* needs to know the state of the world at any point in time. That means you have to take a snapshot of the world when you begin. If you update that snapshot at some point, you necessarily have to update your path... or at least to check to see that it is valid.

Which brings me to the next part... repathing/replanning. This is probably what was meant. There are two approaches. 1) continually (i.e. every so often) check to see that the path that you have in memory is valid, or 2) rig up a system where a change to a node ("now blocked") notifies any path owner that was planning on using it. Is that real-time A*? Not really... it's just redoing the process.

Another off the wall possibility is that the reference was to steering behaviors - but that's not A*.

Dunno... *shrug*