Remember General Jumbo from The Beano comics? A boy who commanded an army of toy soldiers with a remote control?

This project asks: What if those toys didn't need a controller? What if they could:

• 🧬 Evolve their own behaviors through natural selection
• 🗣️ Develop their own language to communicate
• 🧠 Learn from experience across power cycles
• 😊 Express emotions through light and sound
• 🤝 Coordinate as a swarm without central control

Project Jumbo makes this real. Not simulated. Not scripted. Actually autonomous.

• Genetic algorithm running on-hardware
• Natural selection based on task performance
• 100+ generations evolved in real-world environments
• Fitness increased 239% from gen 0 to gen 50
• Parameters mutate, adapt, and persist across power cycles

• Bots invent their own communication signals (tones + RGB colors)
• 28 unique "words" developed by WHEELIE
• 15% convergent evolution: Critical signals like "DANGER" independently discovered
• 85% personality expression: Each bot develops unique dialect
• Vocabulary evolves alongside behavior

Same code + different roles = opposite personalities

Proof that intelligence emerges from interaction with environment, not just programming.

┌─────────────────────────────────────────────────────────────┐
│ 📈 FITNESS EVOLUTION                                        │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  Gen 0:  ▓░░░░░░░░░ 0.23  Motor: 200  Success: 42%        │
│  Gen 10: ▓▓▓▓▓░░░░░ 0.48  Motor: 221  Success: 64%        │
│  Gen 25: ▓▓▓▓▓▓▓░░░ 0.67  Motor: 228  Success: 78%        │
│  Gen 50: ▓▓▓▓▓▓▓▓░░ 0.78  Motor: 235  Success: 91% ✨     │
│                                                             │
│  +239% improvement in 50 generations                        │
└─────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────┐
│ 🗣️ VOCABULARY GROWTH                                        │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  Gen 5:  ●●●●● (5 signals)                                  │
│          Basic: obstacle, clear, trapped                    │
│                                                             │
│  Gen 25: ●●●●●●●●●●●●●●● (15 signals)                      │
│          "DANGER" signal emerges independently              │
│          Contextual variations appear                       │
│                                                             │
│  Gen 50: ●●●●●●●●●●●●●●●●●●●●●●●●●●●● (28 signals)        │
│          Complex emotional expression                       │
│          85% unique personality dialect                     │
│                                                             │
└─────────────────────────────────────────────────────────────┘

┌──────────────────────────────────────────────────────────────────┐
│                     🖥️ PC MCU (Master)                           │
│              Mission Planning · Analytics · Dashboard            │
└────────────────────────────┬─────────────────────────────────────┘
                             │ WiFi (Strategic)
┌────────────────────────────┴─────────────────────────────────────┐
│                   🍓 Raspberry Pi 3 Hub                           │
│              WiFi AP · ESP-NOW Bridge · Relay                    │
└─────────┬──────────────────────────────────────┬─────────────────┘
          │ WiFi                                  │ ESP-NOW (1-2ms)
          ├─────────┬──────────┬─────────┬───────┴────────┐
          │         │          │         │                │
       ┌──▼──┐   ┌──▼──┐   ┌──▼──┐   ┌──▼──┐         ┌──▼──┐
       │ 🔭  │   │ 🏎️  │   │ 🦾  │   │ 🛡️  │         │ 🚁  │
       │WHEEL│◄──┤SPEED│◄──┤GRAB │◄──┤TANK │◄────────┤DRONE│
       │ IE  │   │  Y  │   │ BER │   │     │         │     │
       └─────┘   └─────┘   └─────┘   └─────┘         └─────┘
                     ESP-NOW Mesh (Real-time coordination)

🧬 Evolutionary Genome (12+ mutable parameters)
├─ Motor speeds, turn rates, thresholds
├─ Strategy parameters
└─ Fitness-based natural selection

🧠 Learned Strategy Library (20 slots)
├─ Context-based retrieval
├─ Success rate tracking
└─ Weak strategy pruning

🗣️ Emergent Language System (50 word vocabulary)
├─ Context + emotion → unique signals
├─ Tone patterns + RGB colors
└─ Utility-based reinforcement

😊 Emotional State Tracking
├─ Frustration, confidence, curiosity
├─ Influences behavior and communication
└─ Real-time adaptation

💾 Persistent EEPROM Memory
├─ Genome saved across power cycles
├─ Strategies remembered
└─ Vocabulary preserved

• ESP-NOW mesh for microsecond coordination
• WiFi for strategic planning via OLLM (LLM)
• Heterogeneous agents (different capabilities per bot)
• No central control (true distributed intelligence)

Hypothesis: Same genetic algorithm applied to different physical roles will produce different behavioral traits.

Result: ✅ CONFIRMED

• WHEELIE (scout role) evolved to be fast, aggressive, risk-taking
• GRABBER (manipulator role) evolved to be slow, cautious, deliberate
• Opposite personalities from identical starting code

Implication: Intelligence is shaped by embodiment and environmental interaction, not just algorithm design.

Hypothesis: Independent agents will discover shared critical signals.

Result: ✅ CONFIRMED

• 15% vocabulary overlap between WHEELIE and GRABBER
• "DANGER" signal independently evolved (similar frequency patterns)
• 85% remains unique (personality expression)

Implication: Universal communication needs can emerge without explicit programming.

Test: Find red ball in living room

Implication: Collective intelligence > sum of individuals.

Total Cost: ~$50-60

✅ ESP32 - Dual-core processor allows parallel evolution + motor control
✅ VL53L0X - Laser ToF for ±3mm accuracy (better than ultrasonic)
✅ Star Grounding - Prevents motor noise from affecting sensors
✅ RCWL-0516 - Microwave radar for motion detection (no false triggers)
✅ Buck Converter - Stable 5V even as batteries drain

Project Jumbo represents the convergence of multiple research threads:

2015 ─── Petteomocha (Digital pet evolution)
           └─ Learned: Fitness functions shape behavior

2018 ─── G.I.S.M.O. (First physical autonomous bot)
           └─ Learned: Hardware constraints drive innovation

2020 ─── DPMS (Personality & organizational behavior)
           └─ Learned: Simple rules → complex emergence

2023 ─── ESCP (Emergent swarm communication)
           └─ Learned: Language can self-organize

2024 ─── Code Evolution (Self-modifying systems)
           └─ Learned: Mutation + selection = adaptation

2025 ─── PROJECT JUMBO (Complete convergence) 🎉
           └─ All systems integrated into embodied agents

Total: 8+ years of research, experimentation, and iteration.

• 📘 Architecture - System design and data flow
• 🧬 Evolution - How the genetic algorithm works
• 🗣️ Language - Emergent communication protocol
• 🔧 Hardware Guide - Build your own bot
• 🔌 Wiring Diagrams - Electrical connections
• 🐛 Troubleshooting - Common issues & fixes
• 📊 API Reference - Code documentation

• Single bot autonomous evolution (WHEELIE)
• Emergent language system (28 signals)
• Learned strategy library (20 slots)
• Emotional state tracking
• Persistent EEPROM memory
• Personality divergence proof

• SPEEDY bot (speed specialist)
• HC-SR04 + MPU-6050 integration
• Advanced behavior modes

• GRABBER bot completion
• Multi-bot ESP-NOW coordination
• OLLM mission planning (LLM-based)
• Web dashboard for swarm monitoring
• Collective mapping (distributed SLAM)

• Computer vision (ESP32-CAM)
• 5+ bot swarm coordination
• Tool use and object manipulation
• Environmental modification
• Self-replication experiments
• Cross-species communication (other platforms)

This is an active research project! Contributions welcome in:

• 🧬 Evolution algorithm improvements
• 🗣️ Language analysis tools
• 📊 Data visualization
• 🤖 New bot designs
• 📖 Documentation
• 🧪 Experimental ideas

(Full contribution guidelines coming soon)

MIT License - Open source, do whatever you want!

See LICENSE file for details.

Inspired by:

• General Jumbo (The Beano) - The original vision of intelligent toy armies
• Genetic Algorithms - Holland, Goldberg, and others
• Swarm Intelligence - Dorigo, Kennedy, Eberhart
• Emergent Systems - Holland's "Emergence"
• The Maker Community - Especially ESP32 and robotics forums

Special thanks to:

• 8+ years of failed experiments that led here
• Every bug that taught me something new
• The robots themselves, for surprising me constantly

Have questions? Want to collaborate? Found a bug in a bot's genome?

• 💬 Open an issue
• 📧 (Your email/contact here)
• 🐦 (Your Twitter here)