NEURO-SWARM v2.2

// ARCHITECTURE ANALYSIS

This simulation utilizes a Genetic Algorithm to optimize a Multi-Layer Perceptron (Neural Network). Unlike simple steering behaviors, these agents must learn the relationship between sensory input and motor output.

The Brain: A Feed-Forward Neural Network with:

• > 4 Inputs: [Angle to Food, Distance to Food, Angle to Poison, Distance to Poison]
• > 8 Hidden Neurons: Allows for non-linear decision making (e.g., assessing risk vs reward).
• > 2 Outputs: [Linear Velocity, Angular Velocity]

Evolution: Agents that survive longer and consume "Packets" (Green) increase their fitness score. Upon extinction of a generation, the highest performing "brains" are cloned, mutated slightly, and repopulated.

// BRAIN VISUALIZATION

The dashboard visualizes the live neural activity of the "Alpha" (Top Ranked) agent in the swarm.

Neural Map Legend (Bottom Right):

• > Nodes: Input Layer (Sensors) -> Hidden Layer (Processing) -> Output Layer (Motor Control).
• > Activity: Brighter nodes indicate stronger activation (-1 to 1).
• > Synapses: Green lines are positive weights (excitatory), Red lines are negative weights (inhibitory). Thickness represents connection strength.