Skripsi
OPTIMASI PERILAKU KARAKTER BOSS PADA PERMAINAN MULTIPLAYER MENGGUNAKAN FUZZY STATE MACHINE
Adaptive Artificial Intelligence (AI) plays a significant role in shaping player experience in digital games, particularly in action-based combat scenarios where behavioral dynamics influence perceived challenge, fairness, and engagement. This study investigates the implementation of a Fuzzy State Machine (FuSM) compared to a conventional Finite State Machine (FSM) for boss behavior in a 2D multiplayer action game prototype titled Sentinel Siege using a mixed-method approach that integrates quantitative telemetry-based analysis and qualitative player perception evaluation. Quantitative metrics include Behavioral Variability (Standard Deviation of state distribution), Action Change Rate (ACR), Player Damage Taken (PDT), and Time-to-Kill (TTK), collected through an automated telemetry logging system, while qualitative data were obtained from 33 respondents with balanced distribution (11 per AI model) and analyzed using Reflexive Thematic Analysis. The results indicate that FuSM Set B demonstrates the highest behavioral variability (SD 32.78), adaptivity (ACR 0.56), offensive pressure (PDT 170.3 HP), and battle duration (TTK 78 seconds), and is most frequently perceived as less predictable. However, increased algorithmic complexity does not consistently translate into higher perceived intelligence or fairness, as the conventional FSM model shows more stable perceptions in these aspects despite lower technical dynamism. These findings reveal a non-linear relationship between telemetry-based AI performance and subjective player experience, suggesting that while Fuzzy State Machine tuning enhances behavioral dynamics mathematically, optimal player experience requires balancing technical complexity with perceptual clarity and fairness.
No other version available