Haskell Ant Sim

Agent-based ant foraging simulation in Haskell. Possibly the buggiest Haskell program ever written.

Features

• Realistic ant movement.
• 1D ant vision inspired by the 1884 book Flatland.
• Neural network ant brains.
• Genetic algorithm to train the ant brains.

Installation

git clone https://github.com/OlegAlexander/haskell-ant-sim.git
cd haskell-ant-sim
stack run ant-sim

Controls

• Use arrow keys to move the red player ant.
• Press W and drag the mouse to start drawing a wall. Press W again to finish drawing a wall.
• Left click to place a food. Holding down the mouse places more food.
• Right click on walls or food to delete them.
• Press D to toggle debug text.
• Press N to toggle a visualization of the neural network. The input vector is on top. The output vector and movement decision text is on the bottom.
• Press R to toggle the player ant's vision rays.
• Press V toggle a 1D first person view for the player ant.
• Press C to toggle the home compass.
• Press T to start a new training session. To start completely from scratch, set usePretrainedBrains to False in shared/src/Constants.hs. Training can take several days.
• Many parameters can be adjusted in shared/src/Constants.hs.

Why?

• Can I learn Haskell in a fun way as opposed to just reading a book?
• Can I simulate realistic ant foraging behavior?
• Can I use a neural network for the ant brains?
• Can I use a genetic algorithm to train the ant brains?
• Is Haskell suitable for game development?