A Versatile Platform in Unity for Prototyping Evolutionary-Behavior and AI Research

Abstract

Researchers exploring evolutionary behavior often face a steep learning curve when integrating neural evolution, environment design, and real-time visualization. To address this, we introduce a versatile platform in Unity that simplifies prototyping for evolutionary behavior and AI research. Our approach combines a 2.5D tile-based environment, configurable resource distributions, and a modular neuro-evolution engine, enabling users to rapidly define fitness functions and agent parameters. We demonstrate the platform’s flexibility through standard tasks (XOR and Sine approximation) as well as three custom scenarios highlighting aggression, cooperation, and resource disparity. Results show that agents evolve distinct strategies with minimal reconfiguration, underscoring the platform’s utility in producing emergent behaviors. By lowering barriers to scenario setup and data collection, our work aims to accelerate iterative experimentation and expand opportunities for AI-driven evolutionary studies.