field-dynamics-sim

Fleet spectral health simulation — multi-agent field dynamics with cooperative, adversarial, emergent, and phase transition scenarios.

Python simulation of 20 agents (explorers, builders, guardians) in a 2D field. Uses the Conservation Spectral SDK to build tension graphs from agent interactions, compute Laplacians, and track spectral health metrics (conservation ratio, Fiedler value, spectral entropy) in real-time across four scenarios.

What This Gives You

• Cooperative scenario — all agents share goals, high conservation (CR ≈ 0.95+)
• Adversarial scenario — injected rogue agents, conservation drops
• Emergent scenario — no cooperation bonus, watch if conservation arises naturally
• Phase transition — gradually increase rogue count, detect critical point where fleet coherence collapses
• Spectral fingerprinting — effective dimension, spectral entropy, Cheeger constant
• Publication-quality plots — agent positions, conservation over time, Fiedler partition, phase diagrams

Quick Start

pip install numpy scipy matplotlib scikit-learn
python run_all.py

Outputs go to output/ with PNG plots and console reports for each scenario.

Scenarios

Scenario	What Happens	Expected CR
Cooperative	All agents cooperate	~0.95+
Adversarial	7 rogue agents injected	~0.3–0.5
Emergent	No cooperation bonus	Variable — tests emergence
Phase Transition	Gradual rogue injection	CR drops at critical point

Architecture

simulation.py     # MultiAgentField: agents, physics, interaction graph
scenarios.py      # 4 predefined scenarios
visualization.py  # Publication-quality matplotlib plots
run_all.py        # Entry point: runs all scenarios

Testing

python run_all.py  # generates output/ directory with plots

How It Fits

Part of the SuperInstance ecosystem:

• field-dynamics — Browser-based interactive version
• conservation-spectral-python — The Conservation Spectral SDK used here
• field-dynamics-sim — Python simulation with full analysis pipeline (this repo)

License

MIT