Pick'n Del

A HackableGame based on tiled-land.

Install

In your directory we advise to clone side by side hackagames and tiled-land.

git clone https://github.com/iktorz-net/hackagames
git clone https://github.com/imt-mobisyst/tiled-land

Make sure you have the last version, and install them with python3 pip:

git -C hackagames pull
pip install ./hackagames
git -C tiled-land pull
pip install ./tiled-land

get-started:

Everything starts with a launch file. It will define a world and test a bot. As a first try, we recommend playing directly with the human interface (the player.ShellPlayer).

import tiledland.game.pickndel as pnd

world= pnd.World().initializeGrid([
        [00, 00, 00, 00],  # 1  2  3  4
        [-1, 00, -1, -1],  #    5      
        [00, 00, 00, 00],  # 6  7  8  9
        [00, -1, -1, 00]   #10       13
    ])

master= pnd.GameMaster( world, tic=12 )
bot= pnd.player.ShellPlayer()

master.testPlayer( bot, 1 )

The goal is to realize a maximum of pick and delivery missions (cf. market place area).

then you can replace the ShellPlayer with your own Bot:

import tiledland.game.pickndel as pnd

class FirstBot():

    # Constructor:
    def __init__(self):
        super().__init__()
        self._model= pnd.World()
        self._id= 0
        self._tic= 0

    # Player interface :
    def wakeUp(self, playerId, numberOfPlayers, gamePod):
        self._id= playerId
        self._model.fromPod(gamePod)

    def perceive(self, podState):
        self._tic= self._model.setOnPodState(podState)

    def decide(self):
        return "pass"

    def sleep(self, result):
        print( f"End on: {result}" )

We recommend using a _model attributes to model the Pick'n Del world (the map, the robot and the missions).

Somme useful methods:

• carrierTile() returning the current tile position of the carrier.
• carrierMission() returning the current mission identifier.
• carrierGoal() returning the next tile to reach.

More Complexity:

The first increase in complexity consists of growing the size of the world. To this effect, it is possible to download a complete launch file that will start a game with different game configuration, each recorded in a json format.

• Archives to download: .zip

The pickndel.py launch script will start a game accordingly to the configuration given as command argument:

For instance

python3 pickndel.py conf-pnd-medium-02.json

As a first exercise, the goal is to learn, using reinforcement learning how to move in a world knowing that some of the tiles are encumbered. There is a probability to be stuck inside a tile when the carrier try a move from that tile. The encumbered tile and the associated probabilities are not known, and the smallest path is not necessarily the faster.