Zing 论坛

正文

AEON:浏览器中的神经网络文明模拟器,每个生物都有真实进化的大脑

AEON是一款完全在浏览器中运行的开源神祇游戏与文明模拟器,每个生物都拥有真实的神经网络大脑,能够进化、学习并形成社会结构。项目融合了WorldBox的沙盒创造与文明系列的深度策略,无需安装即可体验人工生命与涌现文明的奇妙世界。

神经网络文明模拟神祇游戏人工生命进化算法开源游戏浏览器游戏JavaScript涌现智能策略游戏
发布时间 2026/06/14 09:15最近活动 2026/06/14 09:18预计阅读 6 分钟
AEON:浏览器中的神经网络文明模拟器,每个生物都有真实进化的大脑
1

章节 01

AEON: Browser-Based Neural Network Civilization Simulator (Main Guide)

AEON is an open-source browser game combining god simulation and civilization building, where every creature has a real neural network brain that evolves, learns, and forms social structures. It merges WorldBox's sandbox creation with Civilization's deep strategy, requiring no installation to experience artificial life and emergent civilizations. Key features include dual game modes, genetic evolution of neural networks, and a complete economic system.

2

章节 02

Project Background & Overview

AEON is described as "WorldBox meets Civilization with a soul"—offering both god-like creation fun (WorldBox) and deep strategy progression (Civilization), while giving each life a "real soul" via neural networks. It is not a scripted simulation but a true artificial life ecosystem.

3

章节 03

Dual Game Modes: Survival & Creation

Survival Mode: Play as a civilization leader—name your people, design flags, choose traits, and guide via policies, research, and diplomacy (no god powers). Focus on governance over micromanagement.

Creation Mode: Act as a creator—shape terrain (mountains, oceans, forests), create life, apply blessings/curses, and observe civilizations evolve in your world.

4

章节 04

Core Technology: Neural Networks & Evolution

Each creature uses a fixed-topology MLP neural network as its brain. Key features:

  1. Genetic Evolution: Genomes mutate/hybridize; natural selection shapes adaptive behaviors (e.g., finding water, gathering resources).
  2. Heritable Traits: Size, speed, diet, aggression—visible and passed to offspring. Click any creature to view its neural network.
  3. Need-Driven Behavior: Hunger/thirst drive actions—civilizations spontaneously settle near water sources.

Economic system: Resource nodes (wood, stone, ore) that deplete and regenerate; tech tree (science) and civics tree (government evolution); cities grow from camps to metropolises, claiming territory.

5

章节 05

Social, Diplomatic & Ecosystem Simulations

Social: Units (civilians, warriors, rangers) with distinct roles; military size depends on policies, with local defense bonuses.

Diplomacy: Fair system (AI uses same GovernanceAPI as players) with alliances, trade, and war.

Culture: Unique traits (martial, collectivist, pious) emerge from lifestyle—similar cultures stay peaceful.

Ecosystem: Wildlife (prey/predators) form a dynamic balance, providing food and challenges for early civilizations.

6

章节 06

Technical Implementation & Performance

Browser-Based: Pure JavaScript (ES modules), no external dependencies—runs entirely in the browser, no server/data sent outside.

Performance: Sprint1 update optimized rendering (viewport culling, LOD) from ~34ms to ~1ms.

Structure: Code split into three parts:

  • src/sim/: Core simulation (world, neural networks, etc.)
  • src/render/: Canvas rendering, camera, effects
  • src/game/: UI, input, governance API
7

章节 07

Development Status & Future Plans

Current State: Early development—core simulation works, but features are being refined. Completed: city building, unit pathfinding, performance boost.

Known Issues: Player nations decline passively; no win/lose screen; units don't rest during long marches; command mode lacks discoverability.

Future: Deepen economic system (action costs), "free person" agents (emotions, community), and 3D WebGL2 renderer.

8

章节 08

Open Source & Community Engagement

License: MIT (free to use/modify/distribute).

How to Run: Use project's static server (node serve.mjs → localhost:8123) or any static server (ES modules need HTTP, not file://). Test: npm test (headless evolution test).

Community: Encouraged to contribute; full docs (DESIGN.md, MECHANICS.md, etc.) available. GitHub Actions run invariant tests; development tracked via issues and sprint plans.