def Copy(env):
n_env = helicopter.EnvMakerForestFire(
init_pos_row=env.pos_row,
init_pos_col=env.pos_col,
n_row=env.n_row,
n_col=env.n_col,
p_tree=env.p_tree,
p_fire=env.p_fire,
moves_before_updating=env.moves_before_updating,
reward_type=env.reward_type,
reward_tree=env.reward_tree,
reward_fire=env.reward_fire,
reward_empty=env.reward_empty,
reward_hit=env.reward_hit,
sub_tree=env.sub_tree,
sub_empty=env.sub_empty,
sub_fire=env.sub_fire,
sub_rock=env.sub_rock,
sub_lake=env.sub_lake,
ip_tree=env.ip_tree,
ip_empty=env.ip_empty,
ip_fire=env.ip_fire,
ip_rock=env.ip_rock,
ip_lake=env.ip_lake)
n_env.grid = copy.deepcopy(env.grid)
n_env.total_reward = copy.deepcopy(env.total_reward)
n_env.total_hits = copy.deepcopy(env.total_hits)
n_env.remaining_moves = copy.deepcopy(env.remaining_moves)
return n_env
class ForestFireEnv0(gym.Env):
metadata = {'render.modes': ['human']}
# Parameters
maker_params = {'env_mode': 'stochastic',
'n_row': 16, 'n_col': 16, 'p_tree': 0.100, 'p_fire': 0.005,
'init_pos_row': 8, 'init_pos_col': 8,
'termination_type': 'continuing', 'reward_type': 'cells', 'observation_mode': 'plain'}
cell_symbols = {'tree': 0.77, 'empty': 0.66, 'fire': -1.0, 'rock': 0.88, 'lake': 0.99}
substitution_effects = {'sub_tree': None, 'sub_empty': None, 'sub_fire': 'empty', 'sub_rock': None, 'sub_lake': None}
reward_values = {'reward_tree': 0.0, 'reward_fire': -1.0, 'reward_empty': 0.0, 'reward_hit': None,}
init_cell_probs = {'ip_tree': 0.75, 'ip_empty': 0.25, 'ip_fire': None, 'ip_rock': 0.00, 'ip_lake': None}
# Instantiated Environment
env = helicopter.EnvMakerForestFire(**maker_params,
**cell_symbols,
**substitution_effects,
**reward_values,
**init_cell_probs)
def reset(self):
return self.env.reset()
def step(self, action):
return self.env.step(action)
def render(self):
return self.env.render()
def close(self):
return self.env.close()
N_TRAIN = 15
N_STEPS = 20
N_SAMPLES = 30
K_Rollout = 3 * FREEZE
LOOKAHEAD = 4
if __name__ == '__main__':
env = helicopter.EnvMakerForestFire(n_row=N_ROW,
n_col=N_COL,
p_tree=P_TREE,
p_fire=P_FIRE,
init_pos_row=Init_Row,
init_pos_col=Init_Col,
moves_before_updating=FREEZE,
tree=TREE,
empty=EMPTY,
fire=FIRE,
reward_type=C_TYPE,
reward_tree=C_TREE,
reward_fire=C_FIRE,
reward_empty=C_EMPTY,
reward_hit=C_HIT,
reward_step=C_STEP,
reward_move=C_MOVE)
H = heuristic.Heuristic_m1_v3
exp = rollout.Experiment(env,
H,
H_mode=13,
N_TRAIN=N_TRAIN,
N_STEPS=N_STEPS,
N_SAMPLES=N_SAMPLES,
Init_Col = 7 #Helicopter initial col
P_FIRE = 0.03 #Probability to turn a green cell into ared cell
P_TREE = 0.1 #Probability to turn an empty cell into a green cell
# Symbols for cells
TREE = 0
FIRE = 2
EMPTY = 1
FREEZE = 8 #Movements of Helicopter after update Automata
if __name__ == '__main__':
#Create a new enviroment with initial parameters
env = helicopter.EnvMakerForestFire(n_row=N_ROW,
n_col=N_COL,
p_tree=P_TREE,
p_fire=P_FIRE,
init_pos_row=Init_Row,
init_pos_col=Init_Col,
moves_before_updating=FREEZE,
tree=TREE,
empty=EMPTY,
fire=FIRE)
# First observation
observation = env.reset()
#Create a copy of enviroment with initial observation
env_1 = Copy(env)
observation_1 = observation
# Making checkpoints
checkpoint_env = env.make_checkpoint()
checkpoint_env_1 = env_1.make_checkpoint()