Dynamic 3D pathfinding (Kynapse AI)

This is a demo of the Kynapse AI engine from Kynogon. A character evolves in a terrain with hundreds of dynamic crates creating very complex conglomerates for pathfinding. Dynamic avoidance alone would never be able to solve such a complex pathfinding.

Category:

Entertainment

Tags:

• artificial
• intelligence
• game
• middleware
• Kynogon
• Kynapse
• dynamic
• pathfinding

[kevnar]

Great job, but an actual life form finding paths would have to look around for a bit before finding the path. This little AI guy changed his path instantly, exactly the way a computer would. A bit of trial and error on the dude's part would have made this brilliant.

Amazing job though.

[x13killer]

you certainly sound like you know what your talking about

[TheFerroh]

@UnIXbLueStar Convex collision detection in 3d is very similar to convex collision detection in 2d. Concave collision detection (which is the case you're mentioning) is more rarely done in games as it is significantly more difficult/slower. However, concave collision detection is not much more complex than concave collision detection in 2d. The common case for games today to test for jumps is simply to trace a (or several) short lines through convex objects in front of the actor. Same for 2D.

[UnIXbLueStar]

@TheFerroh sorry dude but i see a big gap of complexity between testing collision through a randomly shaped 3d cavern hole and any random 2d segments. The complexity is obviously to come up with an efficient way of doing it. Otherwise it just involve more maths.

[TheFerroh]

@UnIXbLueStar Testing whether an actor can jump at a given time is no harder in 3D than it is in 2D. For most games, implementing the feature you mentioned is very easy. I think you are confusing the difficulty of collision detection with the difficulty of pathfinding. Although, in the specific case of testing for the ability to jump over an object a small distance away from an actor during a pathfinding search, the collision detection is actually relatively straightforward in most cases.

[UnIXbLueStar]

@TheFerroh the difficulty in adding 3d is if you can jump.. you would have to calculate if the dude can jump over a given object and if its large enough for him to pass etc... you know everything on the map isnt a simple cube. Its pretty difficult in my book. And that was just jumping, there can be dozens of factor to add to that depending what your game allow.

[Blackday2010]

I did a pathfinding in 3D environment using Panda3D engine...but it still static and can't be recalculated if the goal change its position or when a sudden obstacle comes in the way.....I programed it using OOP technique other than Open-Closed lists method...but Yeah, That's great

[TheFerroh]

@SonnyTheWhiteDwarf Pathfinding in 2D is fundamentally the same as pathfinding in 3D. It's certainly not hard to understand 3D pathfinding if you fully understand 2D pathfinding. Ask any game programmer worth his salt whether a 3D A* search is any harder than a 2D A* search -- the answer is "not much harder".

[loledatyou123]

@SonnyTheWhiteDwarf the women you like aren't gonnas take an interest in you

[Walley666]

@Walley666 Afternote two: Well it's also a processor of information, just like a cpu - but also just like a cpu it has no comprehension of what it's processing, so it's blind.

[Walley666]

@Walley666 Afternote: Whilst that imagined in your mind may *draw* from the contents of your mind (i.e. garden = past experience), that is only the data set being worked with - it's something else doing the 'extrapolation'.