Multi-Agent Systems

by Albert Major

Expanding Agent-Based Models

The lecture focused on taking simple models and extending them to increase real-world plausibility. The goal is to move from abstract simulations to models that reflect complex variables (like weather or physics) by modifying the underlying code and logic in NetLogo.

Core Characteristics of ABMs:

• Simple Rules: Basic instructions for agents (e.g., "move forward," "burn neighbor") can generate complex, unpredictable phenomena.
• Randomness: Models utilize random values (stochasticity), yet they produce consistent patterns of behavior over time.
• Self-Organization: Patterns emerge from agent interactions without being explicitly programmed into the system.
• Emphasis: Different models emphasize different aspects of reality depending on the research question (e.g., focusing on how a fire starts vs. how wind affects it).

Model 1: The Fire Model (Forest Fire)

This model simulates a fire spreading through a forest.

The Base Model:

• Setup: A forest consists of green patches (trees) and empty patches. The fire starts as a "wall" on the left edge of the grid.
• Mechanism: If a green patch has a burning neighbor, it catches fire (100% certainty in the base model).
• Critical Point (Threshold): There is a specific density of trees (approximately 59%) that acts as a tipping point.
  • Below 59%: The fire usually dies out before reaching the right side.
  • Above 59%: The fire almost always burns the majority of the forest.

Extensions (Adding Complexity):

1. Probability of Spread

• Concept: In the real world, fire does not guarantee spread to a neighbor. Factors like humidity, wood age, or tree type affect flammability.