Integrating a navigation solution is just the beginning when building believable, living characters for a game. Shortest path algorithms, developed for robotics, optimize for distance travelled and, if used directly, result in exactly what they were built for: robotic, unnatural behaviour. Supposedly relaxed walkers steer from corner to corner, turning suddenly as if in a hurry. Elite soldiers step over the corpses of fallen comrades into the sights of the waiting player. And an angry swarm of enemies forms a “conga line” towards its target as if politely queuing for a turn to get a swing in. This talk shows how to deal with these unwanted behaviours and how to craft believable paths on top of a navigation solution. From cost modification and its performance pitfalls, through path smoothing and corridors, to steering crowds and handing off to an animation system, we’ll show how to make the paths a believable part of the character’s role.
– A understanding of the often hidden role of the navmesh face shapes and its topology on path find outcomes.
– What A* cost modification/filtering can be used for and where performance pitfalls lie (flood filling).
– How to deal with a large number of characters and how to shape their behaviour when using the same general path and also how select alternative paths.
– What string-pulled shortest path can and cannot provide
– how path smoothing can counteract this and how to style paths to be away from obstacles and provide better turning behaviour.
– How animation systems like motion matching integrate with navigation systems and the needs for path and trajectory prediction.
