A5 Option 1: Baroque Chess Agents |
CSE 415: Introduction to Artificial Intelligence The University of Washington, Seattle, Spring 2019 |
OPTION 1: Baroque Chess Agent.
For option 1, you will work with a partner to create a Python module (typically just a single .py file) that will be able to play a game of Baroque Chess and compete in a tournament. Although the tournament is not part of the assignment itself, your program must be technically able to participate in it. (The tournament will take place after the assignment is due; we may offer some extra credit for programs that place highly in the rankings.) Your player's file name should be of the form [free_name]_BC_Player.py, where you get to choose an original name, but that name is followed by "_BC_Player.py". An example might be Magnifico_BC_Player.py. If you need to split your agent to use any additional modules, please name them using the convention: [free_name]_BC_module_[module_name].py. Here an example could be Magnifico_BC_module_static_eval.py. Game rules: A good overview of the rules of Baroque Chess is given in the Wikipedia article on Baroque Chess. However, since there are a number of variations of Baroque Chess, we have to specify some version, and we will use an operational definition of the rules. ''When in doubt, try it out, at the validation page.''
Image courtesy of Chess Guide at multionline.net.
It will be essential that your program use the specified representation of states, so that it is compatible with all the other Baroque Chess programs developed in the class. Some code for representing states of Baroque Chess is shown near end of this page. (This is included as BC_state_etc.py in the Starter code.) Your program should be designed to anticipate time limits on moves. There are two aspects to this: (1) use iterative deepening search, and (2) poll a clock frequently in order to return a move before time runs out. Starter code is available. This code provides a game-master to handle the turn-taking between a pair of playing agents, and it provides a class definition for states of the game. Some simple agent stub files are included to help get the project going. We are playing Baroque Chess with rules that: do not permit (a) any choice of center-counter symmetry (see the Wikipedia description of Baroque Chess), or (b) long-leaper double jumping (it's considered detrimental to having balanced and dynamic games), or (c) "suicide" moves. We will assume that the game ends when either player loses a king. Your program should implement the following functions, so that they can be called by the game administration software: prepare(player2Nickname . This function will be called by the
game administration software before a game starts, and usually before
any of the other functions below are called. If your agent needs to
initialize any data structures, this is a good place for that to be done,
as the time taken to execute this will not be "on the clock" that is
running during game moves.
parameterized_minimax(currentState, alphaBeta=False, ply=3, useBasicStaticEval=True) .
This function is for testing the basic capabilities of your agent. It should be able to work for any combination
of values of arguments, within reason,
except that it is not required to integrate Zobrist Hashing into your agent.
For example, it should be OK with any legal state for currentState ;
either using or not using alpha-beta pruning; a specified number of ply (maximum
depth) from 0 (statically evaluate the current state only) to perhaps 8 (which might not be too bad if there are
only a few pieces left on the board).
When The use of Zobrist hashing in your agent is optional.
However, if you do use it, it should be disabled while running the
In order for Although your
The return value of
introduce() . This function will return a multiline string
that introduces your player, giving its full name (you get to make
that up), the names and UWNetIDs of its creators (you), and
(optionally) some words to describe its character.
nickname() . This function should return a short version of the playing agent's name (16
characters or fewer). This name will be used to identify the player's moves in game
transcripts.
makeMove(currentState, currentRemark, timeLimit=10) .
This is probably your most important function. It should return a
list of the form [[move, newState], newRemark] . The move is a data
item describing the chosen move, of the form
((r, c),(r1, c1)). Here (r, c) gives the starting row and column
coordinates of the piece that is being moved, and (r1, c1) give the coordinates of the
square where it ends up.
(fr, new_fr) .
Here fr is a string such as "e2", where the first character gives the "file" (column) and
and the second character gives the "rank" (row) on the board of the piece
being moved, and new_fr gives the coordinates of the
square where that piece ends up.
The
The
The
The
Your staticEval(state) .
This function will perform a static evaluation of the given state.
The value returned should be high if the state is good for WHITE and
low if the state is good for BLACK.
This function is not the same as your basic static evaluation function mentioned
above in association with parameterized_minimax , but should be
custom-designed by your team to help your agent be a strong player.
The staff plans to test your staticEval function separately at some
point during the grading, so it should be possible to use it by executing
code such
as the following. (This example assumes your agent is named "The_Roman_BC_Player".)
import The_Roman_BC_Player as player staticResult = player.staticEval(some_state) The quality of your agent's playing will probably depend heavily on how well your staticEval function works. You may incorporate Zobrist hashing into your agent. It is optional. The staff may provide some assistance in terms of a module with methods for Zobrist hashing, but using it will not be required whether you choose to use Zobrist hashing or not. In the code example above, the starting board is shown using ASCII text, and the encoding is as follows: (lower case for black, upper case for WHITE): p: pincer l: leaper i: imitator w: withdrawer k: king c: coordinator f: freezer -: empty square on the boardHere are the contents of the starter code file BC_state_etc.py .
It's presented here for convenient references, but it is part of the .tar archive
linked above.
BLACK = 0 WHITE = 1 INIT_TO_CODE = {'p':2, 'P':3, 'c':4, 'C':5, 'l':6, 'L':7, 'i':8, 'I':9, 'w':10, 'W':11, 'k':12, 'K':13, 'f':14, 'F':15, '-':0} CODE_TO_INIT = {0:'-',2:'p',3:'P',4:'c',5:'C',6:'l',7:'L',8:'i',9:'I', 10:'w',11:'W',12:'k',13:'K',14:'f',15:'F'} def who(piece): return piece % 2 def parse(bs): # bs is board string '''Translate a board string into the list of lists representation.''' b = [[0,0,0,0,0,0,0,0] for r in range(8)] rs9 = bs.split("\n") rs8 = rs9[1:] # eliminate the empty first item. for iy in range(8): rss = rs8[iy].split(' '); for jx in range(8): b[iy][jx] = INIT_TO_CODE[rss[jx]] return b INITIAL = parse(''' c l i w k i l f p p p p p p p p - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - P P P P P P P P F L I W K I L C ''') class BC_state: def __init__(self, old_board=INITIAL, whose_move=WHITE): new_board = [r[:] for r in old_board] self.board = new_board self.whose_move = whose_move; def __repr__(self): s = '' for r in range(8): for c in range(8): s += CODE_TO_INIT[self.board[r][c]] + " " s += "\n" if self.whose_move==WHITE: s += "WHITE's move" else: s += "BLACK's move" s += "\n" return s def __eq__(self, other): if not (type(other)==type(self)): return False if self.whose_move != other.whose_move: return False try: b1 = self.board b2 = other.board for i in range(8): for j in range(8): if b1[i][j] != b2[i][j]: return False return True except Exception as e: return False def test_starting_board(): init_state = BC_state(INITIAL, WHITE) print(init_state) test_starting_board() |
What to Turn In:
Turn in your files via Canvas. Do not zip up the files.
|
Updates and Corrections: Due date pushed back to May 15. The spec for the returned move has been changed to match the starter code (May 14, 10:30 AM). The "prepare" function's parameter was clarified in Piazza on May 13, and then updated here. Minor corrections to the BaroqueGameMaster.py program in the starter code were incorporated, and the new version is here. (It's also included in an updated version of the starter code, but nothing else is changed in the starter code.) Any additional changes will be described here. |