Animation

In this section, we’ll visualize the game process. It loads pre-trained model to see how it acts after training so if you want to use it, you have to download our folder of pre-trained model. Please follow these steps:

• Download and extract our pre-trained models folder in current working directory. You can download directly here or you can use gdown library to download and unzip by unzip command.

  >>> import gdown
  >>> file_url = 'https://drive.google.com/uc?id=1SmrYhUC9PNyBR_yUJQJxJsBrNjSARZQD'
  >>> gdown.download(file_url, 'checkpoints.zip', quiet=False)
  

  unzip checkpoints.zip
  

  Warning

  Make sure you have checkpoints folder in current working directory, otherwise program can’t find it for loading.

• To run animation, run in a normal python file or interactive mode

  >>> from rbgame.menu import RoboticBoardGameMenu
  >>> from rbgame.utils import astar_construtor, dqn_constructor
  >>> RoboticBoardGameMenu(astar=astar_construtor, dqn=dqn_constructor)
  

• It will show you below screen:

  

  Introdution screen.

  You can right away run with default configuration by clicking Play or click Setting to edit it:

  

  Setting screen.

  All players is human by default. You can chose astar if you want to use agent using A* path finding algorithm or dqn to use DQN agent.

• Click OK to back to main menu and now click Play to animate with new configuration:

  

  Game process screen. Bar progress in top-right corner shows how many mails player has collected. Grid in bottom-right corner shows batteries of the robots.

If you play by yourself, table following will show you how to:

Button details.

Button	Acton ID	Action
space	0	Stand still. Charge if possible.
↑	1	Make move foward. Pick up or drop off if possible.
↓	2	Make move backward. Pick up or drop off if possible.
←	3	Make move to left. Pick up or drop off if possible.
→	4	Make move to right. Pick up or drop off if possible.