-
Notifications
You must be signed in to change notification settings - Fork 0
/
MinimaxAgent.cpp
135 lines (130 loc) · 4.06 KB
/
MinimaxAgent.cpp
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
/* =============================================================================
# FileName: MinimaxAgent.cpp
# Desc: MinimaxAgent
# Author: YanlongLi
# Email: [email protected]
# HomePage: http://www.yanlongli.me
# Created: 2017-10-14 01:02:34
# Version: 0.0.1
# LastChange: 2017-10-14 06:43:11
# History:
# 0.0.1 | YanlongLi | init
============================================================================= */
#include "MinimaxAgent.h"
#include "Evaluator.h"
#include "utils.h"
#include <cfloat>
#include <algorithm>
MinimaxAgent::MinimaxAgent(long long maxdepth, bool _alphaBetaPrune, Evaluator* _evaluator)
: MAX_DEPTH(maxdepth), alphaBetaPrune(_alphaBetaPrune), evaluator(_evaluator) {
depth = 0;
}
/*
* the max value we can get under current state
* current board is result of maxer's action
* so we simulate miner's actions and get the max
*/
double MinimaxAgent::maxValue(const Board& board) {
int wl = this->checkWinOrLoss(board);
if (wl != 0) return (*evaluator).maxerEvaluation(board); // terminal state, so maxer win
//
depth ++;
if (depth >= MAX_DEPTH) { // reatch max depth
return (*evaluator).maxerEvaluation(board);
}
//
vector<Action> actions = validActionByMiner(board);
double v = -DBL_MAX;
for (auto action : actions) {
v = max(v, minValue(action.result()));
if (alphaBetaPrune) {
if (v >= beta) break;
alpha = max(alpha, v);
}
}
return v;
}
/*
* the min value we can get under current state
* current board is result of miner's action
* so we simulate maxer's actions and get the min
*/
double MinimaxAgent::minValue(const Board& board) {
int wl = this->checkWinOrLoss(board);
if (wl != 0) return (*evaluator).minerEvaluation(board); // terminal state, so miner win
//
depth ++;
if (depth >= MAX_DEPTH) { // reatch max depth
return (*evaluator).minerEvaluation(board);
}
//
double v = DBL_MAX;
vector<Action> actions = validActionByMaxer(board);
for (auto action : actions) {
v = min(v, maxValue(action.result()));
if (alphaBetaPrune) {
if (v <= alpha) break;
beta = min(beta, v);
}
}
return v;
}
Board MinimaxAgent::nextByMaxer(const Board& board) {
depth = 0;
// assume the input is not a terminal state
vector<Action> actions = validActionByMaxer(board);
Action* decision = NULL;
double v = -DBL_MAX;
for (auto action : actions) {
double t = maxValue(action.result());
if (t > v) {
v = t;
if (decision == NULL) {
decision = new Action(board);
}
*decision = action;
}
}
if (decision == NULL) {
fprintf(stderr, "no action returned\ncurrent board for maxer:\n");
printBoard(board);
fprintf(stderr, "valid action: %ld\n", actions.size());
fprintf(stderr, "%f\n", v);
for (Action action : actions) {
double t = maxValue(action.result());
fprintf(stderr, "%f\n", t);
}
exit(-1);
}
return decision->result();
}
Board MinimaxAgent::nextByMiner(const Board& board) {
depth = 0;
alpha = -DBL_MAX;
beta = DBL_MAX;
// assume the input is not a terminal state
vector<Action> actions = validActionByMiner(board);
Action* decision = NULL;
double v = DBL_MAX;
for (auto action : actions) {
double t = minValue(action.result());
if (t < v) {
v = t;
if (decision == NULL) {
decision = new Action(board);
}
*decision = action;
}
}
if (decision == NULL) {
fprintf(stderr, "no action returned\ncurrent board for miner:\n");
printBoard(board);
fprintf(stderr, "valid action: %ld\n", actions.size());
for (Action action : actions) {
double t = minValue(action.result());
fprintf(stderr, "%f\n", t);
}
exit(-1);
}
return decision->result();
}