The AI of The Podlands
When we started work on The Podlands, we quickly realized that Unreal Engine 4's stock AI systems were never going to cut it for production: The built-in Character system is massively overcomplicated (and by extension underperforming), drained our CPU budget with dozens of features we had no need for, and worst of all, didn't remotely live up to our visual standards.
UE4's stock Character movement system doesn't have any type of motion constraints; AI can move in any way it wants. Our animal AI, intended to be realistic(-ish), was supposed to almost exclusively move forwards.
I quickly ended up rolling out this custom AI framework to compensate; based on the AActor class, it's massively more performant than the character class (despite being entirely Blueprint-based), as well as working much better for our use-case. Motion is restricted to the AI's forward-direction, and carefully designed velocity and rotation systems result in the animals moving massively more naturally.
The AI of The Podlands - A Custom AI/Pathfollowing Solution for Unreal Engine 4
The Podland's territorial Carnopod intimidating the player. Animal assets created by co-developer Nathan Viller.
Careful integration of Nathan's animations through a well-designed Animation Blueprint results in fluid and natural movements, well-synchronized with the in-world motions.
Not featured in the first video is the custom-built impulse and sense system, designed to emulate an animal's environmental awareness. Despite the low quality, it - and its consequences - are visible in this video.
The AI of The Podlands - Clingy Gallipods (Older Development Video)