bramgrooten / hanabi Goto Github PK

A simple Hanabi learning environment for multiple agents.

• Basic run example is located in main.py.
• Basic unittests

• Hands: ndarray(num player, hands size, 4))

  a list for every player with the card information in his hands.

  • Color index:

    • 0 = ?
    • 1 = R(ed)
    • 2 = G(green)
    • 3 = B(lue)
    • 4 = W(hite)
    • 5 = Y(ellow)
    • 6 = M(agenta)

  • Card rank:

    • 0 = UNKNOWN
    • 1 = ONE
    • 2 = TWO
    • 3 = THREE
    • 4 = FOUR
    • 5 = FIVE

  • Informed Color index:

    • See color index

  • Informed Color rank:

    • See rank

  Example:

  [[[4 3 0 0]
   [5 2 0 0]
   [5 1 0 0]
   [3 3 0 0]
   [3 3 0 0]]
  
  [[6 4 0 0]
   [5 3 0 0]
   [3 1 0 0]
   [6 1 0 0]
   [2 1 0 0]]]  
   
  

• Played: ndarray(num of colors,) The current value of the firework stack. Based on the colors order (see below)

  Example: [0 0 0 1 0 0] , the White 1 has been placed (4th value in colors).

• Discard: ndarray(num colors, num discards for that color) The discard list for every color. Based on the colors order (see below)

  Example:

  [[2]     # For Red a 2 has been discarded
   [1, 4]  # For Green a 1 and 4 have been discarded
   [0]     # For Blue no cards have been discarded (0 is a place holder)
   [2]     # For White a 2 has been discarded
   [0]     # For Yellow no cards have been discarded (0 is a place holder)
   [3]]    # For Magenta (Multicolor) the 3 has been discarded (only when using 6 cards)
  

• Info: dict Contains extra information about the hints, fuses, current player and game stats

  Example:

  { 
    'hints': 0, 
    'fuses': 3, 
    'turns_left': 1, 
    'turns_played': 15, 
    'players': 2, 
    'current_player': 0
  }
  

• Moves: tuple((player id, action performed)) Actions are described in constants where

  • id

    • 0 = Play, card at {index}
    • 1 = Inform color {index}, to player {player}
    • 2 = Inform rank {index}, to player {player}
    • 3 = Discard card at {index}

  Example: Player 1, plays the card located at the 5th position in his hand.

  [(1, <class Actions: (id=0, index=4, player=None)>)]

• Colors: ndarray(num colors,)

  Define all the colors that are currently in the game and define the color order for played cards and discarded cards.

  Example:

  [<Colors.RED: 'R'>, <Colors.GREEN: 'G'>, <Colors.LIGHTBLUE_EX: 'B'>, <Colors.BLACK: 'W'>, <Colors.LIGHTYELLOW_EX: 'Y'>]

• log: dict(turn, list of performed actions in that turn.)

  example:

  Key: turns log (unpacked dict)
  
      - Turn 0: Played card: Y1 Drew card: B1
      - Turn 1: Played card: R1 Drew card: Y3
      - Turn 2: Informed player: 0 of rank: 4
      - Turn 3: Informed player: 0 of rank: 3
      - Turn 4: Informed player: 0 of rank: 3
      - Turn 5: Informed player: 1 of rank: 2
      - Turn 6: Tried to play card: B3, but was not possible Drew card: B4
      - Turn 7: Discard card: G1 Drew card: W3
      - Turn 8: Informed player: 0 of rank: 3
      - Turn 9: Tried to play card: W3, but was not possible Drew card: R3
      - Turn 10: Informed player: 0 of rank: 4
      - Turn 11: Informed player: 0 of rank: 3
      - Turn 12: Tried to play card: R3, but was not possible Drew card: B1
      - Turn 13: Informed player: 1 of color: W
      - Turn 14: Discard card: B4 Drew card: Y3
      - Turn 15: Discard card: W4 Drew card: R2
  

Example of a game render.

• First all of the general information is provided, this incldues information over hint tokens and fuses.
• Then all the player hands are shown, except the players hand whose turn it is.
• And finally all the actions performed by that player are displayed.