def reset(self):
x = random.randint(0, self.maze.max_x - 1)
y = random.randint(0, self.maze.max_y - 1)
self.maze = Maze.build(bounds=(self.maze.max_x, self.maze.max_y))
if self.viewer is not None:
self.viewer.set_maze(self.maze)
return self.get_state()
def __init__(self, log_name='MazeEnv', maze=None):
if maze is None:
self.maze = Maze.build(bounds=(50, 50))
else:
self.maze = Maze(bounds=(maze.max_x, maze.max_y),
target=maze.target)
# if MazeEnv.logger is None:
# MazeEnv.logger = Logger("MazeEnv")
# self.logger = Logger(log_name, show_in_console=False)
self.viewer = None
def __init__(self, log_name='MazeEnv', maze=None):
if maze is None:
self.maze = Maze.build(bounds=(10, 10), block_cnt=20)
else:
self.maze = Maze(
start=(maze.x, maze.y),
bounds=(maze.max_x, maze.max_y),
door=maze.door,
blocks=maze.blocks)
# if MazeEnv.logger is None:
# MazeEnv.logger = Logger("MazeEnv")
# self.logger = Logger(log_name, show_in_console=False)
self.viewer = None
total_step)
except Exception as e:
print(e)
if __name__ == "__main__":
GLOBAL_NET_SCOPE = 'Global_Net'
N_S = MazeEnv.state_space_dim
N_A = MazeEnv.action_dim
SESS = tf.Session()
with tf.device("/cpu:0"):
# OPT_A = tf.train.RMSPropOptimizer(LR_A, name='RMSPropA')
# OPT_C = tf.train.RMSPropOptimizer(LR_C, name='RMSPropC')
global_maze = Maze.build(bounds=(80, 80))
# sess, name, N_S, N_A, globalAC, maze=None
GLOBAL_AC = A3CNet(SESS, GLOBAL_NET_SCOPE, N_S,
N_A) # we only need its params
workers = []
# Create worker
for i in range(N_WORKERS):
i_name = 'W_%i' % i # worker name
workers.append(
Worker(SESS, i_name, N_S, N_A, GLOBAL_AC, global_maze))
COORD = tf.train.Coordinator()
SESS.run(tf.global_variables_initializer())
worker_threads = []
total_step)
except Exception as e:
print(e)
if __name__ == "__main__":
GLOBAL_NET_SCOPE = 'Global_Net'
N_S = MazeEnv.state_space_dim
N_A = MazeEnv.action_dim
SESS = tf.Session()
with tf.device("/cpu:0"):
# OPT_A = tf.train.RMSPropOptimizer(LR_A, name='RMSPropA')
# OPT_C = tf.train.RMSPropOptimizer(LR_C, name='RMSPropC')
global_maze = Maze.build(bounds=(30, 30), block_cnt=200)
# sess, name, N_S, N_A, globalAC, maze=None
GLOBAL_AC = A3CNet(SESS, GLOBAL_NET_SCOPE, N_S,
N_A) # we only need its params
workers = []
# Create worker
for i in range(N_WORKERS):
i_name = 'W_%i' % i # worker name
workers.append(
Worker(SESS, i_name, N_S, N_A, GLOBAL_AC, global_maze))
COORD = tf.train.Coordinator()
SESS.run(tf.global_variables_initializer())
worker_threads = []
def _calc_block_size(self):
w, h = self.get_size()
area_w = w * self._get_occupy()
area_h = h * self._get_occupy()
block_w = area_w / self.maze.max_x
block_h = area_h / self.maze.max_y
return (block_w, block_h)
def viewer_run():
viewer = MazeViewer(Maze())
while True:
viewer.render()
if __name__ == '__main__':
import time
import random
import logging
maze = Maze.build((10, 10))
viewer = MazeViewer(maze)
while True:
dx, dy = random.randint(-1, 1), random.randint(-1, 1)
print(dx, dy)
maze.move(dx, dy)
viewer.render()
time.sleep(0.1)
