Research data: Immersion levels and reinforcement learning metrics for VR AI agents (Project ALBG)

Abstract: This dataset contains research data and technical documentation from the "ALBG" project (Automatic Long-term Behavior Generator). The project aimed to create advanced tools for implementing credible AI in Virtual Reality (VR) environments using machine learning and decision trees. The repository includes: Technical Documentation: Manuals for the Strategy Visualization Module (MWS) and decision tree interpretation in the ALBG tool. Simulation Data: Results from the training phase of the AI agents, including performance metrics from over 10,000,000 simulated matches (as per Stage 2 of the project). Validation Metrics: Data comparing the efficiency of the new MPD (Decision Process Model) versus standard Finite State Machine AI. The research addresses the challenge of low immersion in VR games by implementing adaptive AI capable of long-term strategic planning. The data demonstrates the optimization of the AI's "objective function" (time of game, win rate) and the visualization of decision paths.