File size: 6,452 Bytes
c755ba9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the BSD-style license found in the
# LICENSE file in the root directory of this source tree.

"""
Chess Environment Implementation.

An RL agent plays White against a random bot (Black).
Reward is shaped: per-move material delta + terminal win/loss/draw bonus.
"""

import random
from typing import List
from uuid import uuid4

import chess

from openenv_core.env_server.interfaces import Environment
from openenv_core.env_server.types import State

from models import ChessAction, ChessObservation

# Piece values: P=1, N=3, B=3, R=5, Q=9
PIECE_VALUES = {
    chess.PAWN: 1,
    chess.KNIGHT: 3,
    chess.BISHOP: 3,
    chess.ROOK: 5,
    chess.QUEEN: 9,
}


class ChessEnvironment(Environment):
    """
    Chess environment where an RL agent (White) plays against a random bot (Black).

    Reward shaping:
      - material_delta: change in (White material - Black material) each step
      - terminal bonus: +1 for win, -1 for loss, 0 for draw
    """

    def __init__(self):
        self._board = chess.Board()
        self._state = State(episode_id=str(uuid4()), step_count=0)
        self._captured_pieces: List[str] = []

    def reset(self) -> ChessObservation:
        self._board = chess.Board()
        self._state = State(episode_id=str(uuid4()), step_count=0)
        self._captured_pieces = []

        return ChessObservation(
            board_fen=self._board.fen(),
            legal_moves=[m.uci() for m in self._board.legal_moves],
            white_move="",
            black_move=None,
            material_balance=0.0,
            game_status="ongoing",
            captured_pieces=[],
            done=False,
            reward=0.0,
        )

    def step(self, action: ChessAction) -> ChessObservation:  # type: ignore[override]
        # --- validate ---
        try:
            move = chess.Move.from_uci(action.move)
        except (chess.InvalidMoveError, ValueError) as exc:
            raise ValueError(f"Invalid UCI string: {action.move!r}") from exc

        if move not in self._board.legal_moves:
            raise ValueError(
                f"Illegal move: {action.move!r}. "
                f"Legal moves: {[m.uci() for m in self._board.legal_moves]}"
            )

        self._state.step_count += 1
        balance_before = self._material_balance()

        # --- White's move ---
        self._track_capture(move)
        self._board.push(move)
        white_uci = action.move

        # check if game ended after White's move
        status = self._get_game_status()
        if self._is_terminal_status(status):
            balance_after = self._material_balance()
            material_delta = balance_after - balance_before
            reward = material_delta + self._terminal_reward(status)
            return ChessObservation(
                board_fen=self._board.fen(),
                legal_moves=[],
                white_move=white_uci,
                black_move=None,
                material_balance=balance_after,
                game_status=status,
                captured_pieces=list(self._captured_pieces),
                done=True,
                reward=reward,
            )

        # --- Black's move (random) ---
        black_moves = list(self._board.legal_moves)
        black_move = random.choice(black_moves)
        self._track_capture(black_move)
        self._board.push(black_move)
        black_uci = black_move.uci()

        # --- post-move evaluation ---
        status = self._get_game_status()
        balance_after = self._material_balance()
        material_delta = balance_after - balance_before
        terminal = self._is_terminal_status(status)
        reward = material_delta + (self._terminal_reward(status) if terminal else 0.0)

        return ChessObservation(
            board_fen=self._board.fen(),
            legal_moves=[m.uci() for m in self._board.legal_moves] if not terminal else [],
            white_move=white_uci,
            black_move=black_uci,
            material_balance=balance_after,
            game_status=status,
            captured_pieces=list(self._captured_pieces),
            done=terminal,
            reward=reward,
        )

    # ------------------------------------------------------------------
    # Helpers
    # ------------------------------------------------------------------

    def _track_capture(self, move: chess.Move) -> None:
        """Record a captured piece symbol (handles en passant)."""
        board = self._board
        if board.is_en_passant(move):
            self._captured_pieces.append(chess.piece_symbol(chess.PAWN))
        elif board.piece_at(move.to_square) is not None:
            self._captured_pieces.append(board.piece_at(move.to_square).symbol())

    def _material_balance(self) -> float:
        """Return White material minus Black material."""
        white = 0.0
        black = 0.0
        for sq in chess.SQUARES:
            piece = self._board.piece_at(sq)
            if piece is None:
                continue
            val = PIECE_VALUES.get(piece.piece_type, 0)
            if piece.color == chess.WHITE:
                white += val
            else:
                black += val
        return white - black

    def _get_game_status(self) -> str:
        b = self._board
        if b.is_checkmate():
            return "checkmate"
        if b.is_stalemate():
            return "stalemate"
        if b.is_insufficient_material():
            return "draw_insufficient"
        if b.is_fifty_moves():
            return "draw_fifty"
        if b.is_repetition():
            return "draw_repetition"
        if b.is_check():
            return "check"
        return "ongoing"

    @staticmethod
    def _is_terminal_status(status: str) -> bool:
        return status in ("checkmate", "stalemate", "draw_insufficient",
                          "draw_fifty", "draw_repetition")

    def _terminal_reward(self, status: str) -> float:
        if status == "checkmate":
            # whoever is to move is in checkmate → they lost
            if self._board.turn == chess.BLACK:
                return 1.0   # White delivered checkmate
            else:
                return -1.0  # Black delivered checkmate
        # all other terminal states are draws
        return 0.0

    @property
    def state(self) -> State:
        return self._state