Pacman Agent

A template for coding a pacman agent.

Setting up the environment

1. Copy or clone the code from this framework to create your Pacman Agent, e.g., git clone git@github.com:aig-upf/pacman-agent.git
2. Go into pacman-agent folder, cd pacman-agent/
3. Run git submodule update --init --remote to pull the last pacman-contest
4. Create a virtual environment, e.g., python3.8 -m venv venv
5. Activate the virtual environment with source venv/bin/activate
6. Go to the pacman-contest folder and install the requirements:
   • cd pacman-contest/
   • pip install -r requirements.txt
   • pip install -e .

Coding a new agent

In the root folder do the following:

1. Create in my_team.py a class with the name of your agent that inherits from CaptureAgent, e.g. class ReflexCaptureAgent(CaptureAgent):
2. In the new class, override the def choose_action(self, game_state): function to return the best next action (check the given source code example).
3. (Optional) Add any other functions to the class for reasoning / learning and improving your agents decision which could also use other code sources in the same folder.

To debug the agent you can run capture.py between the baseline_team and your current agent:

1. cd pacman-contest/src/contest/
2. python capture.py -r baseline_team -b ../../../my_team.py