Creating a Multi-Agent Base System with NetLogo

NetLogo is a modeling environment tailored for simulating natural and social phenomena, ideal for studying complex systems evolving over time. It enables modelers to direct numerous independent “agents” simultaneously, allowing exploration of how individual behaviors relate to larger-scale emergent patterns.

But First, what’s an AGENT?

in the context of modeling environments like NetLogo, are entities capable of acting on behalf of individuals or representing elements within a simulation. They can range from simple to complex, and their actions contribute to the overall behavior and outcomes of the simulation. In NetLogo, there are four main types of agents:

1. Turtles: These agents are capable of movement within the simulation world, which can be 2D or 3D and is composed of patches. Each turtle can move freely within the grid of patches.
2. Observer: Unlike turtles, observers do not have a physical location within the world. Instead, they provide instructions and guidance to other agents in the simulation.
3. Patches: Patches represent the grid squares in the simulation world. They cannot move but serve as the environment over which turtles traverse. Patches can also create new turtles.
4. Links: Links connect pairs of turtles within the simulation. They do not have coordinates themselves but rely on the locations of the turtles they connect. If either turtle linked by a link dies, the link also ceases to exist.

Understanding the role and capabilities of these agents is essential for creating and interpreting simulations within NetLogo. They allow modelers to explore complex systems by simulating interactions between individual agents and observing emergent patterns at a larger scale.

Procedures

Procedures in NetLogo allow users to instruct agents using commands and reporters. Commands initiate actions for agents, while reporters provide instructions for computing values.

Commands typically start with verbs like ‘create’, ‘die’, ‘jump’, ‘inspect’, or ‘clear’, while reporters often have noun-like names.

NetLogo provides built-in commands and reporters called primitives, while users can define their own procedures. Each procedure has a name preceded by ‘to’ or ‘to-report’, depending on its type. The keyword ‘end’ marks the conclusion of the procedure.

For instance:

to setup
  clear-all
  create-turtles 10
  reset-ticks
end

to go
  ask turtles [
    fd 1            ;; forward 1 step   
    rt random 10    ;; turn right    
    lt random 10    ;; turn left  
  ]
  tick
end

In this program:

• ‘setup’ and ‘go’ are user-defined commands.
• ‘clear-all’, ‘create-turtles’, ‘reset-ticks’, ‘ask’, ‘lt’ (left turn), ‘rt’ (right turn), and ‘tick’ are primitives.
• ‘random’ and ‘turtles’ are primitive reporters. ‘random’ generates a random integer between 0 and 9, while ‘turtles’ reports the agentset containing all turtles.

Variables in NetLogo serve as containers for storing values within agents. These variables can be categorized into global variables, turtle variables, patch variables, or link variables.

• Global Variables: These have a single value that is accessible by all agents.
• Turtle, Patch, and Link Variables: Each agent type (turtle, patch, or link) has its own set of variables with values specific to that agent.

Some variables are built into NetLogo. For instance, all turtles and links have a ‘color’ variable, while all patches have a ‘pcolor’ variable. These built-in variables provide essential properties for agents in a NetLogo simulation.

Again what is NetLogo?

As a first conclusion, we understand that NetLogo serves as an accessible gateway to Agent-Based Modeling, providing a platform for simulating various agents with distinct behaviors. Widely utilized in professional fields such as social simulations, epidemiology, and biology, NetLogo facilitates exploration and experimentation without the need for extensive programming knowledge, making it an ideal introductory tool.

While NetLogo excels in quick testing of assumptions, for more rigorous simulations, specialized software like AnyLogic and FlexSim, or simulation libraries in languages such as Julia and Python, may be recommended. These tools offer enhanced capabilities for tackling complex problems, particularly in industrial settings requiring detailed digital twin production line simulations.

Our Goal (Creation of a Multi-Agent Base System)

Our system is a multi-agent-based simulation environment built in NetLogo. It simulates the dynamics of an ecosystem consisting of multiple agents, including grass patches, cows, lions, a bull, and a dog. The agents interact with each other and their environment, leading to emergent behaviors and dynamic changes within the simulated ecosystem.

The main objectives of the simulation include:

• Simulating the growth and regrowth of grass patches over time.
• Modeling the grazing behavior of cows, which consume grass resources.
• Implementing herd behavior for cows, where they move in groups led by a bull.
• Introducing lions as predators of cows, which hunt and consume them.
• Incorporating a dog that interacts with lions within its perception radius.

the cows exhibit a specific pattern of behavior by following the lead of a dominant bull. As the bull moves through the environment, the cows instinctively trail behind, forming cohesive groups that roam together. This pattern mirrors real-world herd dynamics, showcasing the role of leadership and collective movement in animal groups.

while the cows adhere to this pattern, the dog in our simulation behaves differently. Unlike the cows, the dog wanders randomly through the environment.

This system aims to provide a platform for studying the interactions and dynamics of different agents within a simulated ecosystem, allowing for experimentation and analysis of various scenarios and parameter settings.

Evolving the Environment:

a) We start by defining a countdown mechanism for patches to simulate grass regrowth. Each patch has a timer (‘temps-croissance’) that decrements at each tick. When the timer reaches zero, the patch turns green again, indicating regrowth.

globals [
  temps-croissance
]

to setup
  clear-all
  set-patch-size 20
  set temps-croissance 10
  ; Other setup procedures...
end
to update-grass
  ask patches with [pcolor = brown] [
    set temps-croissance (temps-croissance - 1)
    if temps-croissance = 0 [
      set pcolor green
      set temps-croissance 10  ; Reset the countdown timer
    ]
  ]
end

b) We introduce the ‘taille-plantes’ attribute for patches to represent grass growth. This attribute increments over time and resets to a random value when reaching zero.

patches-own [
  taille-plantes
]

to update-grass
  ; Increment grass size parameter
  ask patches with [pcolor = green] [
    set taille-plantes (taille-plantes + 1)
  ]
  ; Reset patches to brown when grass size parameter reaches zero
  ask patches with [pcolor = green and taille-plantes <= 0] [
    set pcolor brown
    set taille-plantes (random 3) + 3  ; Random initial grass size parameter
  ]
end

Creating Resource Consumers:

We create cows as consumers of grass. Cows decrease the grass size parameter of the patch they are on (‘taille-plantes’) when grazing.

turtles-own [
  animal-type
  ; Other attributes...
]

to eat-grass
  ask turtles with [animal-type = "cow"] [
    if pcolor = green [
      set pcolor brown
      set taille-plantes (taille-plantes - consommation-vache)
    ]
  ]
end

Herds of Cows:

We implement herd behavior for cows, where they move in groups led by a bull.

to move-cows
  ask turtles with [animal-type = "cow"] [
    ; Herd behavior implementation...
  ]
end

Predators:

Lions are introduced as predators of cows. They move randomly and hunt cows within a certain perception radius.

to hunt-prey
  ask turtles with [animal-type = "lion"] [
    let prey-to-hunt other turtles-here with [animal-type = "cow"]
    ; Predator-prey interaction implementation...
  ]
end

In our simulation, we utilize activity diagrams to visually represent the decision-making process of the dog agent when it detects a nearby predator, specifically a lion. The activity diagram outlines two primary courses of action for the dog: either engaging in an attack or retreating from the confrontation.

Diag Activity

When a lion is detected within the dog’s perception radius, the activity diagram showcases the dog’s decision-making process. If the dog decides to attack, the diagram illustrates the sequence of actions involved in the attack, including approaching the lion and engaging in combat. this diagram delineates the potential outcomes of the dog’s attack on the lion, including both winning and losing scenarios. These outcomes influence the subsequent behavior and dynamics within the simulated ecosystem, contributing to the overall realism and complexity of the simulation.

In addition to the activity diagrams, we’ve incorporated two other types of diagrams into our simulation explanation:

• Use Case Diagram: This diagram provides a simple overview of the interactions between different actors and the system.

Diag Use Case

• Sequence Diagram: Illustrating the chronological sequence of interactions between objects in the system, the sequence diagram offers a detailed depiction of how agents and environmental elements interact within the simulation.

Diag Sequence

Conculsion

NetLogo provides a user-friendly platform for simulating complex systems through Agent-Based Modeling. With its intuitive interface and extensive library of primitives, users can explore diverse phenomena and emergent behaviors. While ideal for introductory simulations, specialized software may be necessary for more advanced modeling needs. Overall, NetLogo facilitates experimentation and understanding of complex systems dynamics.