CODEWITHSUNDEEP
algorithmrecursionartificial-intelligenceminimaxpacman
raghu
raghu

Reputation: 131

issue with implementation of minimax algorithm for multi-agent pacman with multple min layers

The following is the code snippet of minimax algorithm for multi-agent pacman where there are multiple ghosts(min players).

def min_max(self, gamestate, current_depth, min_count):
    if current_depth == 1:
         return (self.evaluationFunction(gamestate),None)
# if max node
    if min_count == 0:
        min_count = gamestate.getNumAgents() - 1
        legal_actions = gamestate.getLegalActions(0)
        max_list = []
        for action in legal_actions:
            max_list.append((self.min_max(gamestate.generateSuccessor(0, action), current_depth - 1, min_count), action))
        return max(max_list, key = itemgetter(0))
# if min nodes...
    else:
        legal_actions = gamestate.getLegalActions(min_count)
        min_list = []
        for action in legal_actions:
            min_list.append((self.min_max(gamestate.generateSuccessor(min_count, action),current_depth - 1, min_count - 1), action))
            print(current_depth,min_count)
        return min(min_list, key = itemgetter(0))

pacman_move = self.min_max(gameState, self.depth * (no_of_ghosts + 1) + 1, 0)[1]

But I'm getting the following error: ValueError: min() arg is an empty sequence.

Any help is appreciated

Upvotes: 0

Views: 3090

Answers (1)

Antti Haapala -- Слава Україні
Antti Haapala -- Слава Україні

Reputation: 133909

You do not have any legal moves at min-step - and that's why min_list is empty.

>>> min([])
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: min() arg is an empty sequence

Thus handle the case of:

if not legal_actions:
    # pass? win? lose?

before doing min()

Upvotes: 1

Related Questions