Behind the Scenes 9/12/2024 (Patreon)

Hi guys,

The time since the last post has been a bit chaotic as I've been juggling both the project and tweaking the new equipment to get everything running smoothly.

I did researching different AI solutions for NPC behavior in the game. Here are a few I found interesting:

Goal-Oriented Action Planning (GOAP): Used in games like F.E.A.R., Shadow of Mordor, and Tomb Raider. I love the idea behind it—essentially, it looks at all available actions for a character and builds a plan to achieve a specific goal. The downside is that the planner itself can be heavy on the CPU, as it needs to iterate through all possible combinations of actions, which may require optimization.

Hierarchical Task Network Planning (HTN): Similar to GOAP but more optimized. For example, the HTN plugin on the Unreal Marketplace uses a list of predefined plans by a programmer, which makes it much faster overall.

Utility System (Utility AI): Each action has a weight, often defined by a formula. The action with the highest weight gets selected. This system has been used in The Sims series, making it great for more dynamic decision-making.

Behavior Trees (BTs): A classic solution, but limited. They’re great for basic AI but don’t scale well when things get complex. Unreal Engine has built-in Behavior Trees, so it’s still a viable option for simple behaviors.

State Tree: Seen as an improvement over Behavior Trees, this was added in Unreal Engine 5. It was used in The Matrix Awakens demo, though I haven't experimented with it yet.

I’m still exploring the best approach, but I’m leaning toward GOAP or HTN, or maybe even a mix of both to get the best of both worlds.

c.Alexandraus