-
Notifications
You must be signed in to change notification settings - Fork 1
/
ai.h
325 lines (281 loc) · 9.23 KB
/
ai.h
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
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
#pragma once
#include <stdio.h>
#include <stdlib.h>
#include <time.h>
#include "global_const.h"
#include "bitboard.h"
#include "sennichite.h"
const int piece_value[10] = { // 盤面上の駒の評価値(暫定値)
10, // 歩
40, // 銀
80, // 角
100, // 飛
50, // 金
100000, // 王
40, // と金
40, // 成銀
100, // 馬
120 // 龍
};
const int dist_value[10] = { // 相手の王との距離が 1 のときの価値
10, // 歩
40, // 銀
0, // 角
0, // 飛
50, // 金
10, // 王
40, // と金
40, // 成銀
0, // 馬
0 // 龍
};
const int handpiece_value[6] = { // 持ち駒の評価値(暫定値)
11, // 歩
41, // 銀
81, // 角
110, // 飛
51, // 金
100000 // 王
};
// turn のプレイヤーから見た盤面の評価値(暫定値)
// 勝敗の判定は必要ない
int eval(int turn)
{
int score = 0;
// 盤上の駒単独の価値
for (int p = 0; p < 10; p++) {
score += popcount(g_board.piecebb[playeridx(turn)][p]) * piece_value[p];
}
// 持ち駒単独の価値
for (int p = 0; p < 6; p++) {
score += g_board.hand[playeridx(turn)][p] * handpiece_value[p];
}
BitBoard opp_OU = g_board.piecebb[playeridx(-turn)][pieceidx(OU)];
int sq = square(opp_OU);
int row = sq / 5, col = sq % 5;
for (int i = 0; i < 5; i++) {
for (int j = 0; j < 5; j++) {
int piece = g_board.state[i][j];
if (piece == EMPTY || piece * turn < 0) continue;
int dist = abs(row - i) + abs(col - j); // 相手の王とのマンハッタン距離
if (dist == 0) continue;
score += (int) dist_value[pieceidx(piece)] / dist;
}
}
//しょぼい駒で王を追い詰めてる方がコスパ良いかも?と思ったので、valueを固定値にしたバージョン
int value = 80; //暫定値
BitBoard movable = 0;
for (int i = 0; i < 5; i++) {
for (int j = 0; j < 5; j++) {
int piece = g_board.state[i][j];
if (piece == EMPTY) continue;
else if (piece * turn <= 0){
movable = movable | get_movable(abs(piece), bitboard(5*i+j), playeridx(-turn));
continue;
}
int manhattan_dist = abs(row - i) + abs(col - j); // 相手の王とのマンハッタン距離
int chebyshev_dist = max(abs(row - i), abs(col - j)); // 相手の王とのチェビシェフ距離
if (manhattan_dist == 0) continue;
if (piece == FU) score += (int) value / manhattan_dist;
else if (piece == KK || piece == promote(KK)){
if (abs(abs(row - i) - abs(col - j)) <= 1) score += value / 2;
//評価値が大きくなりすぎないように2で割っとく。
}
else if (piece == HI || piece == promote(HI)){
if (min(abs(row - i), abs(col - j)) <= 1) score += value / 2;
//評価値が大きくなりすぎないように2で割っとく
}
else if (piece == OU) continue; //王で王を取りに行くのは考えない。
else score += value / chebyshev_dist;
}
}
for (int i=0; i<5; i++){
for (int j=0; j<5; j++){
int piece = g_board.state[i][j];
if (piece == EMPTY || piece * turn <= 0) continue;
else{
//自分のこまが見つかったら、
if (popcount(movable & bitboard(5*i+j)) > 0){
//相手の動ける範囲にいたら、
score -= piece_value[pieceidx(piece)];
}
}
}
}
return score;
}
int alphabeta(int depth, int maxdepth, int alpha, int beta)
{
int turn = g_board.turn;
if (turn == AI && judge_checking(AI)) // AIが必ず勝つ
return INF - depth;
if (turn == USER && judge_checking(USER)) // USERが必ず勝つ
return -INF + depth;
if (depth >= maxdepth)
return eval(AI) - eval(USER);
int lower, upper;
TTEntry* entry = tt_search();
if (!entry) { // 未探索
entry = tt_insert();
lower = -INF;
upper = INF;
}
else { // 探索済み
lower = entry->lower;
upper = entry->upper;
if (entry->searched_depth == maxdepth - depth) { // 同じ局面を同じ深さで探索済み
if (lower == upper) return lower;
else if (lower >= beta) return lower;
else if (upper <= alpha) return upper;
else {
alpha = max(alpha, lower);
beta = min(beta, upper);
}
}
}
int bestscore = (turn == AI) ? -INF - 100 : INF + 100;
move_t movelist[100];
int n = get_movelist(movelist, 100);
if (n == 0) return bestscore;
move_t move;
int score;
for (int i = -1; i < n; i++) { // for each move
move = (i == -1) ? entry->bestmove : movelist[i]; // 最初に最善手で探索して探索範囲を狭める
do_move(move, 1);
if (turn == AI)
score = alphabeta(depth + 1, maxdepth, max(alpha, bestscore), beta);
else
score = alphabeta(depth + 1, maxdepth, alpha, min(beta, bestscore));
undo_move(move, 1);
if (turn == AI) {
if (bestscore < score) {
bestscore = score;
entry->bestmove = move;
}
if (bestscore >= beta) {
break; // beta cut
}
}
else {
if (bestscore > score) {
bestscore = score;
entry->bestmove = move;
}
if (bestscore <= alpha) {
break; // alpha cut
}
}
}
if (bestscore <= alpha) {
entry->lower = lower;
entry->upper = bestscore;
}
else if (bestscore >= beta) {
entry->lower = bestscore;
entry->upper = upper;
}
else {
entry->lower = entry->upper = bestscore;
}
entry->searched_depth = maxdepth - depth;
return bestscore;
}
void print_move(move_t move);
int alphabeta_init(move_t* ret, int maxdepth)
{
if (judge_checking(AI)) // AIが必ず勝つ
return INF;
if (maxdepth <= 0)
return eval(AI) - eval(USER);
int lower, upper;
TTEntry* entry = tt_search();
if (!entry) { // 未探索
entry = tt_insert();
}
else { // 探索済み
lower = entry->lower;
upper = entry->upper;
if (entry->searched_depth == maxdepth) { // 同じ局面を同じ深さで探索済み
if (lower == upper)
return lower;
}
}
int bestscore = -INF - 100;
move_t movelist[100];
int n = get_movelist(movelist, 100);
if (n == 0) return bestscore;
move_t move;
int score;
for (int i = -1; i < n; i++) { // for each move
move = (i == -1) ? entry->bestmove : movelist[i]; // 最初に最善手で探索して探索範囲を狭める
do_move(move, 1);
score = alphabeta(1, maxdepth, -INF, INF);
undo_move(move, 1);
// printf("move: ");
// print_move(move);
// printf(", score: %d\n", score);
if (bestscore < score) {
bestscore = score;
*ret = entry->bestmove = move;
}
}
entry->lower = entry->upper = bestscore;
entry->searched_depth = maxdepth;
return bestscore;
}
// 千日手かどうかの判定(深さ1の手のみ)
void judge_nextmove_sennichite()
{
TTEntry* entry = tt_search();
if (!entry) { // 未探索
entry = tt_insert();
}
move_t movelist[200];
int n = get_movelist(movelist, 200);
int loser;
for (int i = 0; i < n; i++) {
do_move(movelist[i], 1);
loser = judge_sennichite();
if (loser) {
entry = tt_search();
if (!entry)
entry = tt_insert();
entry->lower = entry->upper = (loser == USER) ? INF - 1 : -INF + 1;
}
undo_move(movelist[i], 1);
}
}
void print_move(move_t move);
// 反復深化 alpha-beta 探索(深さ depthlimit まで)による最善手を move に代入
// 返り値 1:成功、-1:指し手無し
int choose_move(move_t* move, int depthlimit)
{
judge_nextmove_sennichite();
int bestscore = 0;
for (int maxdepth = 3; maxdepth <= depthlimit; maxdepth++) {
// printf("\nmaxdepth = %d\n", maxdepth);
bestscore = alphabeta_init(move, maxdepth);
// printf("bestscore: %d\n", bestscore);
if (bestscore > INF - 1000) // AIの勝利
break;
}
if (bestscore == -INF - 100) return -1; // 指せる手がない
return 1;
}
// 次の手を求めて表示する
// 返り値 1:成功、-1:指し手無し
int compute_output(move_t* move)
{
int depthlimit = 8;
clock_t start, end;
start = clock();
int res = choose_move(move, depthlimit);
if (res == 1) {
print_move(*move);
printf("\n");
}
end = clock();
if (DEBUG) printf("time: %f\n", (double)(end - start) / CLOCKS_PER_SEC);
if (DEBUG && res == -1) printf("no move\n");
return res;
}