AEONS: A Rust-based 3D Artificial Life Evolution Simulation System

AEONS is a large-scale 3D artificial life evolution simulation system built using Rust and the Bevy engine. It aims to simulate the process of cells evolving from single-celled to multicellular organisms, with their behaviors and morphologies controlled by neural networks. The project integrates modern game engine technology and deep learning frameworks, making it an interdisciplinary technical masterpiece with multiple application values such as evolutionary biology research, AI inspiration, and educational science popularization.

Original Author and Source

• Original author/maintainer: twehler
• Source platform: GitHub
• Original title: AEONS
• Original link: https://github.com/twehler/AEONS
• Release date: 2026-05-24

Project Overview

AEONS simulates life evolution in a 3D virtual world. Unlike simplified 2D simulations, it constructs a complete 3D environment where cells gradually evolve from single-celled to complex multicellular organisms. Integrating game engine technology and deep learning frameworks, it is an interdisciplinary technical achievement.

Core Design Concepts

1. From Cells to Multicellular Organisms: Simulate the evolutionary transition from single-celled to multicellular life. Cells can form connections, develop specialized types, and build complex structures.
2. Neural Network-Controlled Behavior: Each organism's behavior is determined by neural networks, which receive environmental inputs (vision, touch, etc.) and output actions. Weights and structures are heritable, creating evolutionary pressure.
3. Diversity of Evolution Modes: Supports three modes—fully neural control, hybrid control, and rule-based simple control—covering the evolutionary spectrum from simple reflexes to complex intelligence.

Technology Stack Selection

1. Rust: Balances performance and safety. Memory safety avoids data races, concurrency performance is suitable for multi-core processing, and zero-cost abstractions do not sacrifice performance.
2. Bevy Engine: Adopts an ECS architecture—entities represent objects, components are attributes, and systems handle logic—making it modular for easy expansion and maintenance.
3. Burn Framework: A deep learning framework written in Rust, ensuring language consistency, easy deployment, and optimization for Rust's memory model. It is responsible for neural network inference and training.

Simulation Mechanisms

1. Energy and Metabolism: Cells need energy to survive, obtained through photosynthesis, predation, or absorption. They die when energy is exhausted, maintaining ecological balance.
2. Perception and Action: Equipped with sensors like vision, chemical detection, and touch. Neural networks decide actions (movement, feeding, reproduction) based on inputs.
3. Heredity and Variation: Offspring inherit parental information with random mutations (including neural network structure/weights, body structure, metabolic parameters, etc.) to ensure population diversity.

Application Scenarios

1. Evolutionary Biology Research: Adjust environmental parameters to observe population adaptation, verify evolutionary theories, or discover new patterns.
2. AI and Reinforcement Learning: Demonstrates open-ended evolution where intelligence emerges from interactions, inspiring multi-agent strategy evolution.
3. Educational Science Popularization: Visual simulations help understand concepts like natural selection and survival of the fittest, stimulating learning interest.

Technical Challenges and Solutions

1. Large-Scale Entity Management: Uses spatial partitioning (e.g., octree) to optimize collision detection, and parallel computing to accelerate neural network inference.
2. Numerical Stability: Uses fixed-point math and regular normalization of neural network outputs to maintain long-term simulation stability.
3. Separation of Visualization and Simulation: Separates simulation logic from rendering, supporting high-speed background simulation and selective display of key frames.

Summary

AEONS is a cutting-edge exploration in artificial life simulation, demonstrating how complex behaviors emerge from simple rules. It provides learning cases for developers, experimental platforms for researchers, and digital life experiences for users.

Outlook

In the future, more complex ecological interactions (symbiosis, parasitism) can be introduced, user intervention supported, and agents exported for other tasks—demonstrating the potential of computational simulation to understand the essence of life.