Applications of Vector Fields in AI – The Navigation System of Speedball

Navigation meshes are great for finding paths in a static environment where the obstacles, usually predefined in the level data, cut the mesh and increase the cost of a particular polygon, or make it impassable terrain. Often this nav-mesh is generated and stored at cook time, so that at runtime, you just need to read the saved asset data instead of regenerating the mesh. In turn, that means they are not great for dynamic obstacle avoidance as the mesh regeneration and path recalculation would be expensive. This chapter discusses the use of vector fields to alleviate this issue. It explores how to use a grid to calculate and store the vector influence of objects dynamically and use this influence either to directly alter the navigation path or influence the agent’s velocity when they pass through the field to avoid such dynamic obstacles. There is a minor tie-in to steering techniques here, as influencing the agent’s movement velocity would interfere with any steering logic that is already happening in your game.