multi-agent-systems

This is the repo for the Self Organising Multi-Agent Systems Assessed Project

Pre-requitisits

dontnet for Ubuntu 18.04 (example, does not need to be this distro or this version)

  wget -q https://packages.microsoft.com/config/ubuntu/18.04/packages-microsoft-prod.deb -O packages-microsoft-prod.deb
  sudo dpkg -i packages-microsoft-prod.deb 
  sudo apt-get install dotnet-sdk-3.1 -y
  sudo apt install fsharp -y

Dependencies for visualisations

  pip3 install pandas
  pip3 install numpy
  pip3 install matplotlib

Dependency for GUI

  pip3 install PyQt5

Method 1 (non-GUI)

0. Build and compile:

Developed with dotnetcore 3.0 (might be compatible w other versions)

To build (from the GitHub cloned directory do) :
```
dotnet build .\multi-agent-systems.sln
```

To run (from the GitHub cloned directory do) :

cd  .\multi-agent-systems\bin\Debug\netcoreapp3.0\
.\multi-agent-systems.exe  --number-days -1 --number-profiles 7 --number-agents 24

1. Run the program w the default configuration:

In order to run from the command line Baseline Configuration (12 balanced agents, 2 from each other category) use the following command line arguments:

 --number-days -1 --number-profiles 7 --number-agents 24 [--number-runs 1]

where:

--number-days x -> x is the number of days the simulation runs for, if -1 runs until all agents are dead
--number-profiles x -> x is the number of agent profiles there are (number of profie_ files in the dir multi-agent-systems\Agent-Config\agent_dir)
--number-agents x -> x is the number of agents that take part in the simulation initially - this has to be consistent w the number given in the python script
--number-runs x -> optional argument - n is number of times the simulation is repeated (produces x csv files in multi-agent-systems/csv)

2. Run the program w your own configuration:

Modify in multi-agent-systems\Agent-Config\agent_init.py TotalProfiles:

TotalProfiles = [(Balanced, 12), (Egotist, 2), (Idealist, 2), (Susceptible, 2),
                 (NotIdealist, 2), (NotEgotist, 2), (NotSusceptible, 2)]

in order to choose how many agents of which types you'd like to have

Run multi-agent-systems\Agent-Config\agent_init.py in order to generate the profiles
Run the f sharp executable with the command line args specified previously (don't forget to change the number-days, number-profiles, number-agents accordingly)

3. Modify other program variables:

In Config.fs there are other config variables that can be changed for testing & experimenting in dev stages. They are not meant to be modified during the runtime when a user runs the simulation.

4. See the output:

There are 3 ways at the moment:

With printf see the output in terminal
The agent and world states are printed every day in output files in multi-agent-systems/output. They are numbered 0..n-1 depending on how many times the simulation is run (--number-runs). Also, n csv files are produced as well in a similar fashion in the multi-agent-systems/csv folder.
Run cd visualisation\ in top level directory folder, then python3 visualistion.py {number of runs} to create zips of each run in multi-agent-systems\csv\. Each zip contains various visualisations for each run to show what occured.

Method 2 (GUI - only on Linux (and Mac?)) <-- currently on visualisation branch

joshjennings98 / multi-agent-systems Goto Github PK