Android开源 - Android黑白棋游戏实现
xshguo
8年前
<p>黑白棋，又叫苹果棋，最早流行于西方国家。游戏通过相互翻转对方的棋子，最后以棋盘上谁的棋子多来判断胜负。黑白棋非常易于上手，但精通则需要考虑许多因素，比如角边这样的特殊位置、稳定度、行动力等。本游戏取名为黑白棋大师，提供了8种难度等级的选择，从菜鸟、新手、入门、棋手到棋士、大师、宗师、棋圣，助你不断提升棋力。</p> <p>本文将着重介绍黑白棋实现过程中用到的算法。</p> <h2>黑白棋游戏规则</h2> <p>游戏规则见黑白棋大师中的截图。</p> <p><img src="https://simg.open-open.com/show/5ebec0331f15af735e59cf68375884fb.png"></p> <h2>黑白棋大师游戏截图</h2> <p>游戏启动界面。</p> <p><img src="https://simg.open-open.com/show/4a2f35f925d9c821dca0e3d59e04bada.png"></p> <p>游戏过程中的一个截图。</p> <p><img src="https://simg.open-open.com/show/36cae1b43ee7617dc89ee689ff7febf1.png"></p> <p>开新局时的选项，选择先后手以及AI的水平。</p> <p><img src="https://simg.open-open.com/show/a0e751e7add1a71611733b91e3b2d126.png"></p> <h2>几个关键的类</h2> <h2>Rule</h2> <p>Rule类实现游戏规则相关的方法，包括</p> <ol> <li>判断某一步是否合法</li> <li>获取所有的合法走步</li> <li>走一步并翻转敌方棋子</li> <li>统计两方棋子个数</li> </ol> <h2>Algorithm</h2> <p>Algorithm类实现极小极大算法，包括</p> <ol> <li>局面评估函数，对当前局面打分，越高对max越有利，越低对min越有利</li> <li>min()方法</li> <li>max()方法</li> <li>获得一个好的走步</li> </ol> <h2>ReversiView</h2> <p>ReversiView继承自SurfaceView，实现棋盘的界面，在该类定义棋盘界面的绘制、更新等操作。</p> <h2>RenderThread</h2> <p>RenderThread继承自Thread，是控制ReversiView以一定fps更新、重绘界面的线程。</p> <h2>具体实现</h2> <h2>棋盘表示</h2> <p>byte[][]二维数组存储棋盘，-1表示有黑子，1表示有白子，0表示棋格为空</p> <h2>游戏规则类Rule的实现</h2> <p>提供几个关于游戏规则的静态方法。</p> <h3>判断某一个位置是否位于棋盘内</h3> <pre> <code class="language-java">public static boolean isLegal(int row, int col) { return row >= 0 && row < 8 && col >= 0 && col < 8; } </code></pre> <h3>判断某一方在某个位置落子是否合法</h3> <p>即判断该子是否能与己方棋子在某个方向上夹住敌方棋子。</p> <pre> <code class="language-java">public static boolean isLegalMove(byte[][] chessBoard, Move move, byte chessColor) { int i, j, dirx, diry, row = move.row, col = move.col; if (!isLegal(row, col) || chessBoard[row][col] != Constant.NULL) return false; for (dirx = -1; dirx < 2; dirx++) { for (diry = -1; diry < 2; diry++) { if (dirx == 0 && diry == 0) continue; int x = col + dirx, y = row + diry; if (isLegal(y, x) && chessBoard[y][x] == (-chessColor)) { for (i = row + diry * 2, j = col + dirx * 2; isLegal(i, j); i += diry, j += dirx) { if (chessBoard[i][j] == (-chessColor)) { continue; } else if (chessBoard[i][j] == chessColor) { return true; } else { break; } } } } } return false; } </code></pre> <h3>某一方走一步子</h3> <p>将各个方向上被翻转的棋子的颜色改变，并返回这些棋子在棋盘的位置，方便显示翻转动画。</p> <pre> <code class="language-java">public static List<Move> move(byte[][] chessBoard, Move move, byte chessColor) { int row = move.row; int col = move.col; int i, j, temp, m, n, dirx, diry; List<Move> moves = new ArrayList<Move>(); for (dirx = -1; dirx < 2; dirx++) { for (diry = -1; diry < 2; diry++) { if (dirx == 0 && diry == 0) continue; temp = 0; int x = col + dirx, y = row + diry; if (isLegal(y, x) && chessBoard[y][x] == (-chessColor)) { temp++; for (i = row + diry * 2, j = col + dirx * 2; isLegal(i, j); i += diry, j += dirx) { if (chessBoard[i][j] == (-chessColor)) { temp++; continue; } else if (chessBoard[i][j] == chessColor) { for (m = row + diry, n = col + dirx; m <= row + temp && m >= row - temp && n <= col + temp && n >= col - temp; m += diry, n += dirx) { chessBoard[m][n] = chessColor; moves.add(new Move(m, n)); } break; } else break; } } } } chessBoard[row][col] = chessColor; return moves; } </code></pre> <h3>获取某一方当前全部合法的落子位置</h3> <pre> <code class="language-java">public static List<Move> getLegalMoves(byte[][] chessBoard, byte chessColor) { List<Move> moves = new ArrayList<Move>(); Move move = null; for (int row = 0; row < 8; row++) { for (int col = 0; col < 8; col++) { move = new Move(row, col); if (Rule.isLegalMove(chessBoard, move, chessColor)) { moves.add(move); } } } return moves; } </code></pre> <h3>统计玩家和AI的棋子个数</h3> <pre> <code class="language-java">public static Statistic analyse(byte[][] chessBoard, byte playerColor) { int PLAYER = 0; int AI = 0; for (int i = 0; i < 8; i++) { for (int j = 0; j < 8; j++) { if (chessBoard[i][j] == playerColor) PLAYER += 1; else if (chessBoard[i][j] == (byte)-playerColor) AI += 1; } } return new Statistic(PLAYER, AI); } </code></pre> <h2>游戏算法类Algorithm的实现</h2> <h3>极大过程和极小过程</h3> <p>这两个过程的函数形式为：</p> <pre> <code class="language-java">private static MinimaxResult max(byte[][] chessBoard, int depth, int alpha, int beta, byte chessColor, int difficulty); private static MinimaxResult min(byte[][] chessBoard, int depth, int alpha, int beta, byte chessColor, int difficulty); </code></pre> <p>chessBoard为棋盘；depth为博弈树搜索深度；alpha和beta用于alpha-beta剪枝，在max方法中alpha不断更新为局面评分的较大值，在min方法中beta不断更新为局面评分的较小值，当alpha >= beta时就进行剪枝；chessColor表示棋子颜色；difficulty表示游戏难度，对应于不同的AI水平。</p> <p>由于黑子先行，黑子总是调用max()方法，白子调用min()方法。</p> <p>下面以极大过程为例。</p> <p>如果深度为0，只要返回当前局面评分即可。如果双方均没有步可走，表示已经达到最终局面，返回该局面评分。如果仅单方无处可走，调用min递归即可。</p> <p>正常情况下有步可走，遍历每个合法的走步，如果alpha大于等于beta，剪枝直接break，否则走步并递归。</p> <p>best是当前max节点维护的一个最佳值，调用的min方法的alpha是取得alpha和best的较大值。</p> <pre> <code class="language-java">private static MinimaxResult max(byte[][] chessBoard, int depth, int alpha, int beta, byte chessColor, int difficulty) { if (depth == 0) { return new MinimaxResult(evaluate(chessBoard, difficulty), null); } List<Move> legalMovesMe = Rule.getLegalMoves(chessBoard, chessColor); if (legalMovesMe.size() == 0) { if (Rule.getLegalMoves(chessBoard, (byte)-chessColor).size() == 0) { return new MinimaxResult(evaluate(chessBoard, difficulty), null); } return min(chessBoard, depth, alpha, beta, (byte)-chessColor, difficulty); } byte[][] tmp = new byte[8][8]; Util.copyBinaryArray(chessBoard, tmp); int best = Integer.MIN_VALUE; Move move = null; for (int i = 0; i < legalMovesMe.size(); i++) { alpha = Math.max(best, alpha); if(alpha >= beta){ break; } Rule.move(chessBoard, legalMovesMe.get(i), chessColor); int value = min(chessBoard, depth - 1, Math.max(best, alpha), beta, (byte)-chessColor, difficulty).mark; if (value > best) { best = value; move = legalMovesMe.get(i); } Util.copyBinaryArray(tmp, chessBoard); } return new MinimaxResult(best, move); } private static MinimaxResult min(byte[][] chessBoard, int depth, int alpha, int beta, byte chessColor, int difficulty) { if (depth == 0) { return new MinimaxResult(evaluate(chessBoard, difficulty), null); } List<Move> legalMovesMe = Rule.getLegalMoves(chessBoard, chessColor); if (legalMovesMe.size() == 0) { if (Rule.getLegalMoves(chessBoard, (byte)-chessColor).size() == 0) { return new MinimaxResult(evaluate(chessBoard, difficulty), null); } return max(chessBoard, depth, alpha, beta, (byte)-chessColor, difficulty); } byte[][] tmp = new byte[8][8]; Util.copyBinaryArray(chessBoard, tmp); int best = Integer.MAX_VALUE; Move move = null; for (int i = 0; i < legalMovesMe.size(); i++) { beta = Math.min(best, beta); if(alpha >= beta){ break; } Rule.move(chessBoard, legalMovesMe.get(i), chessColor); int value = max(chessBoard, depth - 1, alpha, Math.min(best, beta), (byte)-chessColor, difficulty).mark; if (value < best) { best = value; move = legalMovesMe.get(i); } Util.copyBinaryArray(tmp, chessBoard); } return new MinimaxResult(best, move); } </code></pre> <h3>alpha-beta剪枝原理</h3> <p>先解释下alpha和beta的物理含义，alpha表示max节点迄今为止的最佳局面评分，beta表示min节点迄今为止的最佳局面评分。</p> <p>举个例子见下图（数值为虚构），假设深度是两层，每个结点有两行数字，上方的两个数分别是alpha和beta，表示作为参数传到该层的alpha和beta。下方的数表示了该节点best的更新过程。</p> <p><img src="https://simg.open-open.com/show/86cfbad1a7cf584c947caeac01d80930.png"></p> <p>看图中第一个红色的叉号，该位置处会更新beta为正无穷和2的较小值，即2，导致alpha大于等于beta成立，发生剪枝，对应于min方法中相应位置处的break操作。</p> <h3>获得AI计算出的最佳走步</h3> <p>该方法用于AI走步以及提示功能。</p> <pre> <code class="language-java">public static Move getGoodMove(byte[][] chessBoard, int depth, byte chessColor, int difficulty) { if (chessColor == Constant.BLACK) return max(chessBoard, depth, Integer.MIN_VALUE, Integer.MAX_VALUE, chessColor, difficulty).move; else return min(chessBoard, depth, Integer.MIN_VALUE, Integer.MAX_VALUE, chessColor, difficulty).move; } </code></pre> <h3>局面评估函数</h3> <p>局面评估函数决定了AI水平的高低。对应于不同的AI等级，设计了不同的评估函数。</p> <p>菜鸟级别只关注棋子个数，新手、入门、棋手3个级别不仅关注棋子的个数，而且关注特殊位置的棋子（边、角），棋士和大师级别在棋子个数、边角之外还考虑了行动力，即对方下轮可选的下子位置的个数，宗师和棋圣考虑稳定度和行动力。稳定度将在下一小节介绍。</p> <pre> <code class="language-java">private static int evaluate(byte[][] chessBoard, int difficulty) { int whiteEvaluate = 0; int blackEvaluate = 0; switch (difficulty) { case 1: for (int i = 0; i < 8; i++) { for (int j = 0; j < 8; j++) { if (chessBoard[i][j] == WHITE) { whiteEvaluate += 1; } else if (chessBoard[i][j] == BLACK) { blackEvaluate += 1; } } } break; case 2: case 3: case 4: for (int i = 0; i < 8; i++) { for (int j = 0; j < 8; j++) { if ((i == 0 || i == 7) && (j == 0 || j == 7)) { if (chessBoard[i][j] == WHITE) { whiteEvaluate += 5; } else if (chessBoard[i][j] == BLACK) { blackEvaluate += 5; } } else if (i == 0 || i == 7 || j == 0 || j == 7) { if (chessBoard[i][j] == WHITE) { whiteEvaluate += 2; } else if (chessBoard[i][j] == BLACK) { blackEvaluate += 2; } } else { if (chessBoard[i][j] == WHITE) { whiteEvaluate += 1; } else if (chessBoard[i][j] == BLACK) { blackEvaluate += 1; } } } } break; case 5: case 6: for (int i = 0; i < 8; i++) { for (int j = 0; j < 8; j++) { if ((i == 0 || i == 7) && (j == 0 || j == 7)) { if (chessBoard[i][j] == WHITE) { whiteEvaluate += 5; } else if (chessBoard[i][j] == BLACK) { blackEvaluate += 5; } } else if (i == 0 || i == 7 || j == 0 || j == 7) { if (chessBoard[i][j] == WHITE) { whiteEvaluate += 2; } else if (chessBoard[i][j] == BLACK) { blackEvaluate += 2; } } else { if (chessBoard[i][j] == WHITE) { whiteEvaluate += 1; } else if (chessBoard[i][j] == BLACK) { blackEvaluate += 1; } } } } blackEvaluate = blackEvaluate * 2 + Rule.getLegalMoves(chessBoard, BLACK).size(); whiteEvaluate = whiteEvaluate * 2 + Rule.getLegalMoves(chessBoard, WHITE).size(); break; case 7: case 8: /** * 稳定度 */ for (int i = 0; i < 9; i++) { for (int j = 0; j < 9; j++) { int weight[] = new int[] { 2, 4, 6, 10, 15 }; if (chessBoard[i][j] == WHITE) { whiteEvaluate += weight[getStabilizationDegree(chessBoard, new Move(i, j))]; } else if (chessBoard[i][j] == BLACK) { blackEvaluate += weight[getStabilizationDegree(chessBoard, new Move(i, j))]; } } } /** * 行动力 */ blackEvaluate += Rule.getLegalMoves(chessBoard, BLACK).size(); whiteEvaluate += Rule.getLegalMoves(chessBoard, WHITE).size(); break; } return blackEvaluate - whiteEvaluate; } </code></pre> <h3>稳定度计算</h3> <p>我们知道，在黑白棋中，棋盘四角的位置一旦占据是不可能再被翻转的，因此这几个位置上的子必然是稳定子，而边上的子只有可能沿边的方向被翻转，稳定的程度高于中间的位置上的子。</p> <p>因此，试图给每个子定义一个稳定度，描述该子不被翻转的稳定程度。</p> <p>一共有四个方向，即左-右，上-下，左上-右下，右上-左下。举个例子，下面代码中的 (drow[0][0], dcol[0][0]) 表示向左移动一个单位的向量， (drow[0][1], dcol[0][1]) 表示向右移动一个单位的向量。</p> <p>对于棋盘中某个子的位置，向左找到第一个不是该颜色的位置（可以是出界），再向右找到第一个不是该颜色的位置（可以是出界），如果这两个位置至少有一个出界，或者两个均为敌方棋子，稳定度加1。</p> <p>对于另外三个方向作同样操作。可以看到，角上的棋子的稳定度必然为4，其他位置则根据具体情况并不恒定不变。</p> <pre> <code class="language-java">private static int getStabilizationDegree(byte[][] chessBoard, Move move) { int chessColor = chessBoard[move.row][move.col]; int drow[][], dcol[][]; int row[] = new int[2], col[] = new int[2]; int degree = 0; drow = new int[][] { { 0, 0 }, { -1, 1 }, { -1, 1 }, { 1, -1 } }; dcol = new int[][] { { -1, 1 }, { 0, 0 }, { -1, 1 }, { -1, 1 } }; for (int k = 0; k < 4; k++) { row[0] = row[1] = move.row; col[0] = col[1] = move.col; for (int i = 0; i < 2; i++) { while (Rule.isLegal(row[i] + drow[k][i], col[i] + dcol[k][i]) && chessBoard[row[i] + drow[k][i]][col[i] + dcol[k][i]] == chessColor) { row[i] += drow[k][i]; col[i] += dcol[k][i]; } } if (!Rule.isLegal(row[0] + drow[k][0], col[0] + dcol[k][0]) || !Rule.isLegal(row[1] + drow[k][1], col[1] + dcol[k][1])) { degree += 1; } else if (chessBoard[row[0] + drow[k][0]][col[0] + dcol[k][0]] == (-chessColor) && chessBoard[row[1] + drow[k][1]][col[1] + dcol[k][1]] == (-chessColor)) { degree += 1; } } return degree; } </code></pre> <p> </p> <p> </p>