Difference between revisions of "Artificial intelligence"

Revision as of 16:12, 28 January 2023

A Character Decision-Making System for FINAL FANTASY XV was developed by combining behavior trees and state machines into a single structure called the AI Graph. This is how artificial intelligence is incorporated into FFXIV.

The AI Graph tool allows level designers to make a multilayered decision-making system with visual node graphs.

Each layer of the AI Graph has its own blackboard system, which can be used to register variables used in the game.

All nodes are reused and have four methods: start the process, update process, finalizing process, and termination condition - allowing them to be executed in both behavior trees and state machines.

The Blackboard within the AI Graph enables sharing of variables between local blackboards or global blackboards across all characters' individual AIs; while Parallel Thinking within the AI graph allows concurrent thinking processes for one character; Interrupting also makes it possible for characters to interrupt their current execution when needed through an interrupting node linked with transition conditions being satisfied; Data & Overrides allow customization of each character's behavior through saved asset files as well as override functions much like C++ language while Debugging is made easier with fast iteration enabled via reloading without compiling after changes were made on an existing graph editor file

The AI Graph Editor enables fast iteration in AI development, allowing users to reload an AI graph without compiling when changes are needed. Two debug windows help the user trace active nodes and display detailed logs generated from a character’s AI Graph and variables.

Attack motion analysis was done through simulation by distributing spheres around a monster to find its attack orbit region which is then approximated by simple solid figures with parameters like attack distance assigned to an attack node of the corresponding AI Graph.

The meta-AI (or ‘AI Director’) monitors game situations, giving characters orders such as saving players or buddies in danger or following escaping players for tension control during battles.

A monster's visual sensors consist of two fan-shaped regions that assign enemies to target lists while selecting targets depends on min/max distances & angles plus priority settings via the corresponding node in their respective graphs.

For some monsters, combining rule-based systems with fixed top-layer graphs allow for simpler data maintenance but less freedom due to one condition per graph template structure requirement.

Lastly, FINAL FANTASY XV's three-layered architecture separates concerns between intelligence & animation while avoiding increasing the size of their respective graphs.