This session deals with efficient path finding, flocking, and squad movement behaviors in a complex and dynamically changing world. A graph of connected convex volumes, accurately encapsulating the usable free-space of the game world, is quickly computed from an arbitrary polygon soup of level mesh, using a straightforward 3D rasterization and volume extraction technique. Because the planning graph encapsulates the usable free-space, it supports not only path finding queries, but also repulsion field based flocking and squad behaviors, relying only on rapid checks within the planning graph, avoiding ray-casts against the collision geometry. The planning graph, while detailed enough to support these advanced motion behaviors, is simple enough that it can be updated at runtime as dynamic obstacles modify the environment.