COMP222 2024 Second CA Assignment Individual Coursework
Design and Implement Artificial Intelligence for a Simplified PacMan game
Assignment 2 (of 2)
Submission on Canvas:
Submit 2 files.
-
A pdf design and test report. There is no strict word limit for this, but it should meet the requirements below, and would typically be around 5 pages.
-
A zip file that contains the source code for your AI implementation. Your tasks are to:
-
Design the AI for a simplified PacMan game, described below (in a pdf report)
-
Implement the AI that you designed in code (frameworks are provided for Python, Java).
-
Test and analyse how your AI performs (also in a pdf report)
The Game
This is a simplified PacMan game, which is played on a 30*30 grid. (or 32*32 with outside walls).
There is 1 ‘Ghost’, 1 ‘PacMan’, and 1 Power pill. The walls can be removed to make it simpler(see below).
The Ghost and PacMan characters each move one square at a time, in any direction (including diagonally!).
The game ends when the Ghost catches the PacMan Character (expect for the case below). Or when 1000 moves have been made.
If the PacMan character collides with the Power Pill; the PacMan has 15 moves to ‘eat’ the Ghost, which will send the Ghost back to its starting position and score points.
The aim of the PacMan is score as much as possible. The scoring is:
100 points for ‘eating’ the Ghost
1 point for each unique square visited.
Instructions
There should be an AI design and implementation for the PacMan character. If the Ghost AI has been amended, this should also include a design.
Your design should use any standard AI techniques that we saw in the lectures; Decision Trees, Finite State Machines, Behaviour Trees, GOB, GOAP, HTN.
You should then Implement the AI design in code. You can use one of the code frameworks provided to test your AI code, then you will only need to modify the ‘move’ function in the PacMan.py or PacMan.java file.
You should test your code by running the game with your AI. Describe the paths that your PacMan and Ghost take and analyse the performance of your AI – you can do this with a diagram of how your PacMan moves during the test run. Describe if the AI implementation matches your design and describe the strengths and weaknesses of the design (and implementation if this is different).
Marking Criteria
AI Design: 40%
Marks will be awarded for the detail and complexity of the AI design, as well as the suitability and correctness of the design tools chosen. In the simplest case, this might be a decision tree that enables the PacMan to avoid the Ghost and explore the area until it is caught. More complex designs might try to hide behind the walls as well as to eat the Power Pill and catch the Ghost.
Implementation: 40 %
Marks are awarded for efficient and correct implementation of the AI – including techniques such as State Machines, Decision Trees, Behaviour Trees, and Search algorithms.
Marks will also be awarded based on how closely the implementation reflects the design.
Testing and Analysis:20 %
This is based on how thorough and accurate the analysis understands and describes both the AI design implementation. This includes analysing the strengths and weaknesses of the AI design and comparing it with the implementation.
Notes.
-
Submit all the code files (Java / Python) as a zip file. If you implement your own game framework, or customised walls, then make sure to include these files.
-
You may customise the game (or write your own), for example, by adding or removing walls, or by changing the Ghost AI. Marks are awarded for the complexity and strength of the resulting AI, not for changes to the game (but it may be difficult to demonstrate more complex AI without a more complex game).
3. Look at the PacMan.py or PacMan.java file, as well as the relevant Ghost file to see how to check the locations in the game. The checkPosition function can be used to find the type of object in any given position, and the position of the PacMan and Ghost objects can be read directly.