Plan for Pat Wilson
by Pat Wilson · 01/13/2002 (9:43 pm) · 1 comments
After a long (college induced) break I'm back coding the player AI. Right now this is what works:
-Framework code for the AI player
-AI/Script interaction and callbacks
Doesn't seem like much but it's a pretty big chunk of what needs to be done...which is:
-Example AI script showing use of callbacks to run a waypoint course
-AI namespacing, so you can make multiple classes of AI for different tasks
-Pathfinding (oog)
Pathfinding, as far as I've seen in most games, is graph based. Meaning the AI has a graph and uses that to calculate movement. This is fine for indoor, but as Mark pointed out to me, it's rather cumbersome for outdoor things. So here is the current plan I've been kicking around for the pathfinding:
AI will get instruction to move from point A to point B. It will draw a straight line between them, and start walking. Along the way, it will poll for nearby objects, if it sees one, it will ask it for it's (precompiled) pathing graph. This will be a small graph that will tell the AI how to path around/through/over the object. The AI can then use this data to navigate indoors as well as outdoors. If anyone has any suggestions, I'd be more then happy to take them.
-Framework code for the AI player
-AI/Script interaction and callbacks
Doesn't seem like much but it's a pretty big chunk of what needs to be done...which is:
-Example AI script showing use of callbacks to run a waypoint course
-AI namespacing, so you can make multiple classes of AI for different tasks
-Pathfinding (oog)
Pathfinding, as far as I've seen in most games, is graph based. Meaning the AI has a graph and uses that to calculate movement. This is fine for indoor, but as Mark pointed out to me, it's rather cumbersome for outdoor things. So here is the current plan I've been kicking around for the pathfinding:
AI will get instruction to move from point A to point B. It will draw a straight line between them, and start walking. Along the way, it will poll for nearby objects, if it sees one, it will ask it for it's (precompiled) pathing graph. This will be a small graph that will tell the AI how to path around/through/over the object. The AI can then use this data to navigate indoors as well as outdoors. If anyone has any suggestions, I'd be more then happy to take them.
About the author
Torque 3D Owner Phil Carlisle
The idea behind them is sort of a fluid dynamic, in that normally, the fluid moves around an obstacle in much the same way as a human would. If you add to these vectors a directional impetus away from anything specifically nasty (enemy turret?).
Obviously the vector grid has to be quite sparse, so you would maybe do LOD for grid space where there ARE obstacles.
But anyway, let me recommend my friend alex's site at ai-depot.com/
He's doing a thesis on pathfinding techniques, is a great guy, and would probably love to discuss the whole pathfinding on landscapes issue with you (and I believe already has an article on this too).
Anyway, great work Pat, keep at it!
Phil.