---
id: wiki-2026-0508-artificial-life
title: Artificial Life (ALife)
category: 10_Wiki/Topics
status: verified
canonical_id: self
aliases: [ALife, 인공 생명, digital evolution, emergent behavior, swarm intelligence]
duplicate_of: none
source_trust_level: B
confidence_score: 0.85
verification_status: applied
tags: [alife, evolutionary-computation, emergence, swarm, multi-agent, cellular-automata, complexity, simulation]
raw_sources: []
last_reinforced: 2026-05-10
github_commit: pending
tech_stack:
  language: Python / C++
  framework: NEAT / DEAP / mesa / NetLogo
---

# Artificial Life (ALife)

## 📌 한 줄 통찰
> **"매 life 의 본질 의 digital code 의 recreate"**. 매 simple rule 의 interaction → 매 emergent intelligence. 매 swarm AI / NPC behavior / evolutionary algorithm / open-ended learning 의 foundation.

## 📖 핵심

### 매 3 분류 (Langton)
1. **Soft ALife**: 매 software simulation. 매 Conway's Life, 매 Tierra, 매 Avida.
2. **Hard ALife**: 매 robot. 매 BEAM robotics, 매 swarm robot.
3. **Wet ALife**: 매 synthetic biology. 매 protocell, 매 artificial chemistry.

### 매 핵심 concept
1. **Emergence**: 매 simple rule → 매 complex pattern. (vs reductionism)
2. **Self-organization**: 매 central control X.
3. **Adaptation**: 매 environment 의 fit.
4. **Reproduction**: 매 self-replication (von Neumann).
5. **Evolution**: 매 mutation + selection.
6. **Open-ended evolution**: 매 stop X.

### Landmark systems
- **Conway's Game of Life** (1970): 매 cellular automata.
- **Tierra** (Ray, 1991): 매 self-replicating program 의 evolution.
- **Avida**: 매 digital organism 의 lab.
- **Karl Sims' Evolved Creatures** (1994): 매 morphology + behavior 진화.
- **NEAT** (Stanley): 매 neural network 의 evolve.
- **POET / Open-Ended ALife**: 매 무한 challenge generation.

### Boids (Reynolds 1987)
- 매 simple 3 rule:
  1. **Separation**: 매 collision 회피.
  2. **Alignment**: 매 neighbor 의 average heading.
  3. **Cohesion**: 매 neighbor 의 center.
- → 매 flocking / schooling / swarm.

### Multi-agent emergence
- 매 ant colony 의 pheromone trail.
- 매 stigmergy: 매 environment 의 indirect communication.
- 매 termite mound 의 collective construction.

### Evolutionary computation
- **Genetic Algorithm** (GA): 매 chromosome + crossover + mutation.
- **Genetic Programming** (GP): 매 program tree 의 evolve.
- **Neuroevolution** (NEAT, HyperNEAT): 매 NN 의 evolve.
- **Evolution Strategy** (ES, CMA-ES): 매 continuous parameter.
- **Quality-Diversity** (MAP-Elites, Novelty Search): 매 diversity 의 explicit.

### 매 modern AI 의 응용
1. **NPC behavior** (game): 매 boids 기반 swarm enemy.
2. **Robotics**: 매 swarm robot, 매 self-assembly.
3. **Open-ended ML**: 매 POET, 매 OMNI 의 curriculum.
4. **Procedural generation**: 매 cellular automata (cave, dungeon).
5. **Drug discovery**: 매 evolutionary search.
6. **Architecture / design**: 매 evolutionary design.

## 💻 패턴

### Boids (flocking)
```python
import numpy as np

def boid_step(positions, velocities, perception=10, sep=2):
    for i in range(len(positions)):
        neighbors = [j for j in range(len(positions))
                     if i != j and np.linalg.norm(positions[i]-positions[j]) < perception]
        if not neighbors: continue
        
        # 매 alignment
        align = np.mean([velocities[j] for j in neighbors], axis=0) - velocities[i]
        # 매 cohesion
        cohesion = np.mean([positions[j] for j in neighbors], axis=0) - positions[i]
        # 매 separation
        separation = sum((positions[i]-positions[j])
                         / np.linalg.norm(positions[i]-positions[j])**2
                         for j in neighbors
                         if np.linalg.norm(positions[i]-positions[j]) < sep)
        
        velocities[i] += 0.05*align + 0.01*cohesion + 0.1*separation
        positions[i] += velocities[i]
    return positions, velocities
```

### Genetic Algorithm (DEAP)
```python
from deap import base, creator, tools, algorithms
import random

creator.create('FitnessMax', base.Fitness, weights=(1.0,))
creator.create('Individual', list, fitness=creator.FitnessMax)

toolbox = base.Toolbox()
toolbox.register('attr_bool', random.randint, 0, 1)
toolbox.register('individual', tools.initRepeat, creator.Individual, toolbox.attr_bool, 100)
toolbox.register('population', tools.initRepeat, list, toolbox.individual)

def fitness(ind): return (sum(ind),)
toolbox.register('evaluate', fitness)
toolbox.register('mate', tools.cxTwoPoint)
toolbox.register('mutate', tools.mutFlipBit, indpb=0.05)
toolbox.register('select', tools.selTournament, tournsize=3)

pop = toolbox.population(n=300)
algorithms.eaSimple(pop, toolbox, cxpb=0.5, mutpb=0.2, ngen=40)
```

### Conway's Game of Life
```python
import numpy as np
from scipy.signal import convolve2d

def step(grid):
    K = np.array([[1,1,1],[1,0,1],[1,1,1]])
    n = convolve2d(grid, K, mode='same', boundary='wrap')
    return ((n == 3) | ((grid == 1) & (n == 2))).astype(int)
```

### MAP-Elites (Quality-Diversity)
```python
def map_elites(grid_size, generations=1000):
    archive = {}  # 매 (behavior_descriptor) → best (fitness, genome)
    
    for gen in range(generations):
        if not archive: 
            genome = random_genome()
        else:
            parent = random.choice(list(archive.values()))
            genome = mutate(parent[1])
        
        fitness, descriptor = evaluate(genome)
        cell = discretize(descriptor, grid_size)
        if cell not in archive or archive[cell][0] < fitness:
            archive[cell] = (fitness, genome)
    
    return archive
```

→ 매 single best X — 매 diverse 의 set.

## 🤔 결정 기준
| 문제 | Tool |
|---|---|
| Game NPC swarm | Boids |
| Optimization (discrete) | GA / GP |
| NN architecture | NEAT |
| Continuous param | CMA-ES |
| Diversity 필요 | MAP-Elites / Novelty |
| Procedural map | Cellular automata |
| Multi-agent emergence | NetLogo / mesa |

**기본값**: 매 specific objective = GA. 매 diversity = MAP-Elites. 매 NN = NEAT or RL.

## 🔗 Graph
- 부모: [[Complexity_Theory|Complexity-Theory]] · [[Emergence]] · [[Multi-agent-System|Multi-Agent-Systems]]
- 변형: [[Cellular Automata]] · [[Evolutionary Biology|Evolutionary-Computation]] · [[Swarm_Intelligence|Swarm-Intelligence]]
- 응용: [[NEAT]] · [[Procedural-Generation]]
- Adjacent: [[Reinforcement-Learning]] · [[Self-Organization]] · [[Algorithmic-Biology]]

## 🤖 LLM 활용
**언제**: 매 NPC swarm design. 매 procedural generation. 매 evolutionary optimization. 매 emergent behavior research.
**언제 X**: 매 supervised learning 의 substitute. 매 explainability 가 필수.

## ❌ 안티패턴
- **GA 의 small population**: 매 premature convergence.
- **No diversity preservation**: 매 monoculture.
- **Boids 의 ignore neighbor distance**: 매 unrealistic flock.
- **Evolution 의 short generation**: 매 emergence X.
- **Wet ALife 의 ethics 무시**: 매 synthetic biology biosecurity.

## 🧪 검증 / 중복
- Verified (Langton, Reynolds, Sims, Stanley).
- 신뢰도 B.
- Related: [[Cellular Automata]] · [[Evolutionary Biology|Evolutionary-Computation]] · [[Swarm_Intelligence|Swarm-Intelligence]].

## 🕓 Changelog
| 날짜 | 변경 |
|---|---|
| 2026-05-08 | Phase 1 |
| 2026-05-10 | Manual cleanup — Langton 분류 + Boids + GA + MAP-Elites code |