Nengo Pacman

Installation

This project is compatible with Python 2.7.x and has dependencies on Nengo and Nengo GUI. Ensure you have installed both before running the project.

• Install nengo
• Use pip to install the nengo gui
  • pip install nengo_gui

Next, clone/fork this project from Github and run it from a terminal.

nengo pacman.py

Neural Network Architecture

The game is organized into the pacman world and each individual instance of pacman.

The world owns the game environment, score and sensor data for each pacman. Additionally, the world collects information on distance from wall to pacman, food availability, enemy distance, etc. Using its respective nodes, it sends this information to the pacman's ensembles.

The pacman ensembles, consisting of the food, enemy and obstacles, apply functions on the incoming data and pass it on to the move ensemble. Finally, the move ensemble aggregates the sensory input and determines the speed and direction for the pacman to move in, thereby continuing the game.

The control algorithm for the movement of the pacman lies in the functions that are applied by the food, obstacles and enemy ensembles. Their sensitivity, strength and functionality can be edited by changing the transform, nuerons, and synapse parameters.

Examples

The following examples demonstrate features of the API

Add a pacman to the game

First, create a new instance of pacman in pacman.py:

myPacman = body.Player("pacman", 4, 2, "yellow", 70, 20)

The paramaters passed into the pacman signify (respectively):

• Type of player in the world
• ID of player (n number)
• Size of player
• Color of player
• Linear speed of player
• Rotational speed of player

Now, add the pacman to the game:

pacmen.append(myPacman)

Add a ghost to the game

Create a new instance of a ghost:

myGhost4 = body.Player("ghost", "seeking", 2, "orange", 5, 5)

The paramaters passed into the ghost signify (respectively):

• Type of player in the world
• State of player (either seeking or running away)
• Size of player
• Color of player
• Linear speed of player
• Rotational speed of player

Edit sensory input

In pacman.py, simply edit transform parameters to adjust pacman's sensory input

nengo.Connection(food[0], move[1], transform = 3)

This adjusts the way pacman turns toward the food

nengo.Connection(pacnet.obstacles[[1,2,3]], obstacles, transform = 1, synapse = 0.

This makes pacman more sensitive to enemies

Change maze dimensions

mymap = maze.generateMaze(8,20)

Note: Rows must be in multiples of 2 and columns in 10

Usage

When creating your own neural control network & algorithm to run the game, apply changes on the myModel.py using the framework provided. To run your version of pacman after completing myModel.py, simply run nengo myModel.py to execute the game.

Acknowlegments

@arnav-gudibande, @tcstewar