def __init__(self, problem, steps):
self.original_problem = deepcopy(problem)
start_state, special_things = checker.problem_to_state(problem)
self.steps = steps
self.current_state_of_board, self.current_special_things = checker.problem_to_state(
problem)
self.eval = checker.Evaluator(0, problem, steps)
self.act_list = ACT_LIST
all_states, trans_dict, rewards = self.compute_states
print(all_states)
print(rewards)
mdp.MDP.__init__(self,
init=start_state,
actlist=["U", "D", "R", "L"],
terminals=[],
transitions=trans_dict,
states=all_states,
gamma=0.01)
self.reward = rewards #mpd rewards dictionary
self.U = mdp.value_iteration(self)
self.pi = mdp.best_policy(self, self.U)
# print(mdp.best_policy(self, self.U))
print("end of initialization\n\n\n\n")
return
def __init__(self,
board_state,
last_pacman_action="reset",
e1=0,
e2=0,
e3=0):
# board
self.state, self.special_things = checker.problem_to_state(board_state)
self.size_of_board = len(self.state)
self.last_pacman_action = last_pacman_action
self.last_type_eaten = 0
self.score = 0
self.h_val = 0
self.future_h_val = 0
self.numb_of_ghosts = self.get_num_ghosts()
# Expected Values for each type of dot
self.e1 = e1
self.e2 = e2
self.e3 = e3
self.numb_of_dots_within_three = self.get_dot_neighbors()
self.numb_of_dots_div_board_size = self.numb_of_dots_within_three / self.size_of_board
# initialize location of all the dots
self.list_of_dots = {}
for number_of_row, row in enumerate(board_state):
for number_of_column, cell in enumerate(row):
if cell % 10 == 1 or cell % 10 == 2 or cell % 10 == 3:
self.list_of_dots[(number_of_row, number_of_column)] = cell
def __init__(self, initial, goal=None):
"""The constructor specifies the initial state, and possibly a goal
state, if there is a unique goal. Your subclass's constructor can add
other arguments."""
self.original_problem = deepcopy(initial)
self.initial, self.special_things = checker.problem_to_state(initial)
self.accumulated_reward = 0
#self.initial = initial
self.goal = goal
def __init__(self,board_state,last_pacman_action="reset",e1=0,e2=0,e3=0):
# board
self.state, self.special_things = checker.problem_to_state(board_state)
self.size_of_board = len(self.state)
self.last_pacman_action=last_pacman_action
self.last_type_eaten = 0
self.score = 0
self.h_val = 0
self.future_h_val = 0
self.numb_of_ghosts = self.get_num_ghosts()
# Expected Values for each type of dot
self.e1=e1
self.e2=e2
self.e3=e3
self.numb_of_dots_within_three = self.get_dot_neighbors()
self.numb_of_dots_div_board_size = self.numb_of_dots_within_three / self.size_of_board
def choose_next_action(self, state):
state_of_board, special_things = checker.problem_to_state(state)
eval_state = checker.Evaluator(0, state, 1)
if not "pacman" in special_things:
# check if PACMAN is still in the game
return "reset"
# if pacman is still in the game, then, choose best next step.
s = self.eval_state_to_ab_state_plus_md(eval_state)
if s in self.pi:
new_min_md = 0
# check if we need to update R based on Ghost location:
min_md = self.find_min_md_from_ghosts(eval_state)
# we check if there any ghosts on the board, and if they are very close.
if min_md != -100 and min_md <= 2:
print("performing update to R")
# start scanning for a better position
for action in ["U", "L", "R", "D"]:
child_eval = deepcopy(eval_state)
checker.Evaluator.change_state_after_action(
child_eval, action)
temp_new_md = self.find_min_md_from_ghosts(child_eval)
if temp_new_md != -100 and temp_new_md > new_min_md:
new_min_md = temp_new_md
next_state_md = self.eval_state_to_ab_state_plus_md(
child_eval)
self.rewards[next_state_md] = self.rewards[
next_state_md] + 10 * new_min_md
# TODO: we might be yeilding a state that didnt exist before
self.U = mdp.value_iteration(self)
self.pi = mdp.best_policy(self, self.U)
return self.pi[s]
else:
a = ["U", "D", "L", "R"]
print("random chosen")
# maybe here we should go into a simple dfs to find rest of the route to finish the board? @meir
index = random.randint(0, 3)
return a[index]