AI units in Drunk On Nectar need to perform complex tasks such as flying around the map foraging nectar, fighting rivals and capturing land all while evading not just obstacles along the ground as in most games (for which the default NavMesh in Unreal Engine would sufficie), but obstacles high in the air such as tall, dense trees, landscape formations and so on.
To achieve this, I wrote a custom 3D aerial/flying pathfinding solution for Unreal Engine in C++, check out this video to watch the system in action. One part of this solution that admittedly needs work is better decomposition of the navigation volumes across a map with complex obstacles. I’m working on a different NAV building algorithm for this and will post an update with a hopefully tighter decomposition (i.e. lesser number of volumes and a more visually pleasing spread of volumes across the map). For now, the system works well for my game and I haven’t noticed any adverse performance implications either!
While I don’t go into too much detail of technical implementation in this video I can post more if people are interested (and if I can find the time!)