qwixx/qwixx_agent_mcts.py at master · stobra/qwixx · GitHub

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
from gym_qwixx.envs.qwixx_env import QwixxEnv
from qwixx_agent_random import QwixxAgentRandom
from qwixx_game_performer import QwixxGamePerformer

import random

# Monte carlo tree search
class QwixxAgentMCTS():


	def __init__(self, playerIndex):
		#print("Init Qwixx agent random, playerIndex=" + str(playerIndex))
		self.playerIndex = playerIndex


	def get_next_action(self, env, diceThrow, availableActionsParams):

		# For each action:
		# 	perform that action, then perform the rest of the game randomly
		actionValues = [-100] * (len(availableActionsParams)) # -100 because must be below -20

		for actionIndex in range(len(availableActionsParams)):
			# Clone env
			envClone = env.clone()


			availableActionsPerPlayer = []
			chosenActionPerPlayer = []

			# Assume other players collect actions this round randomly
			for playerIndex in range(envClone.numberPlayers):
				availableActions = envClone.get_available_actions(diceThrow, playerIndex)
				availableActionsPerPlayer.append(availableActions)

				if playerIndex != self.playerIndex:
					agent = QwixxAgentRandom(playerIndex)
					chosenAction = agent.get_next_action(envClone, diceThrow, availableActions)
					chosenActionPerPlayer.append(chosenAction)
				else:
					chosenActionPerPlayer.append(availableActionsParams[actionIndex])

			# Apply actions
			gamePerformer = QwixxGamePerformer(envClone.numberPlayers, envClone, [QwixxAgentRandom(playerIndex), QwixxAgentRandom(playerIndex)])
			gamePerformer.apply_action_tuple_per_player(diceThrow, chosenActionPerPlayer, availableActionsPerPlayer)

			if envClone.is_game_over() == False:
				envClone.next_round()

			# From then on perform rest of the game randomly
			result = gamePerformer.perform_complete_game()
			actionValues[actionIndex] = result[self.playerIndex]


		chosenActionIndex = self.argmax(actionValues)

		#print("actionValues:")
		#display(actionValues)
		#print("chosenActionIndex: " + str(chosenActionIndex))

		return availableActionsParams[chosenActionIndex]

	def argmax(self, actionValues):
		currentMax = -100
		currentIndex = -1
		for actionIndex in range(len(actionValues)):
			if actionValues[actionIndex] > currentMax:
				currentMax = actionValues[actionIndex]
				currentIndex = actionIndex

		return currentIndex