Capcom Details Monster Hunter Stories 3 NPC Battle AI at CEDEC 2026
The presentation shows a practical fusion of machine learning with conventional AI that keeps characters expressive while remaining adjustable by planners, a common pain point in RPG development.
Reporting from 1 source: GameBusiness.jp.
At CEDEC 2026, Capcom programmer Mitsuhiko Inaba presented the battle AI system for Monster Hunter Stories 3. The system uses a two-layer structure where machine learning selects an action category, and traditional methods pick a specific skill. This approach lets planners adjust character behavior intuitively without complex conditions, and a shared base model enables efficient production across characters.
Capcom programmer Mitsuhiko Inaba, who handled AI for field and battle characters in Monster Hunter Stories 3, shared the design at CEDEC 2026. The game's ally NPCs, such as the balanced Simon, the genius Tio, and the mixing-savvy Gaul, act fully autonomously in battle. The team wanted these characters to feel human and smart, avoiding both optimal-only actions and frustrating behavior.
Inaba explained that character traits come from combining skills with "thought," a system that maps situations to skill use via conditions and probabilities. Traditional setups grew hard to tune as conditions multiplied. The solution splits thinking into two layers: machine learning picks a broad action category, such as attack or recovery, while conventional methods choose the specific skill within that category. This keeps skill changes from affecting the model and lets the same model work across different weapons and progress states.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.