def obstacles_room(plotV=True, num_sample=100, computeV=False):
height = 10
width = 10
reward_location = 18
initial_state = None # np.array([25])
obstacles_location = [
12, 13, 22, 23, 35, 36, 45, 46, 62, 63, 72, 73, 67, 77
]
walls_location = [
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, 9,
19, 29, 39, 49, 59, 69, 79, 89, 99, 90, 91, 92, 93, 94, 95, 96, 97, 98,
99
]
obstacles_transition_probability = .2
domain = domains.GridMazeDomain(height, width, reward_location,
walls_location, obstacles_location,
initial_state,
obstacles_transition_probability)
maze = LearningMazeDomain(domain=domain, num_sample=num_sample)
V = None
if computeV:
V = value_iteration(maze.domain.graph, reward_location,
obstacles_location, walls_location,
obstacles_transition_probability)
if plotV:
fig, ax = plt.subplots(1, 1)
maze.domain.graph.plot_signal(np.array(V), vertex_size=60, ax=ax)
plt.savefig('graphs/obstacleRoom_trueV.pdf')
plt.close()
return maze, V
def oneroom(plotV=True, num_sample=100, computeV=False):
height = 10
width = 10
reward_location = 9
initial_state = None # np.array([25])
obstacles_location = [] # range(height*width)
walls_location = []
obstacles_transition_probability = .2
domain = domains.GridMazeDomain(height, width, reward_location,
walls_location, obstacles_location,
initial_state,
obstacles_transition_probability)
maze = LearningMazeDomain(domain=domain, num_sample=num_sample)
V = None
if computeV:
V = value_iteration(maze.domain.graph, reward_location,
obstacles_location, walls_location,
obstacles_transition_probability)
if plotV:
fig, ax = plt.subplots(1, 1)
maze.domain.graph.plot_signal(np.array(V), vertex_size=60, ax=ax)
plt.savefig('plots/one_room_trueV.pdf')
plt.close()
return maze, V
def update_domain(self, new_reward_location, obstacles_location, num_samples, length_sample):
self.reward_location = new_reward_location
self.obstacles_location = obstacles_location
self.num_samples = num_samples
self.length_sample = length_sample
self.domain = domains.GridMazeDomain(self.height, self.width, new_reward_location,
self.walls_location, self.obstacles_location, self.initial_state,
self.obstacles_transition_probability)
self.compute_samples()
def example_grid_maze():
height = 10
width = 10
reward_location = 9
initial_state = None # np.array([25])
obstacles_location = [14, 13, 24, 23, 29, 28, 39, 38] # range(height*width)
walls_location = [50, 51, 52, 53, 54, 55, 56, 74, 75, 76, 77, 78, 79]
obstacles_transition_probability = .2
domain = domains.GridMazeDomain(height, width, reward_location,
walls_location, obstacles_location, initial_state,
obstacles_transition_probability)
maze = LearningMazeDomain(domain=domain, num_sample=2000)
return maze
def oneroom(plotV=True, num_sample=100, length_sample=100, computeV=False):
height = 10
width = 10
reward_location = 9
initial_state = None # np.array([25])
obstacles_location = [] # range(height*width)
walls_location = []
obstacles_transition_probability = .2
domain = domains.GridMazeDomain(height, width, reward_location,
walls_location, obstacles_location, initial_state,
obstacles_transition_probability)
maze = LearningMazeDomain(domain=domain, num_sample=num_sample, length_sample=length_sample)
return maze
def tworooms(num_sample=100, length_sample=100, discount=0.9):
height = 10
width = 10
reward_location = 18
initial_state = None # np.array([25])
obstacles_location = [] # range(height*width)
walls_location = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9,
10, 20, 30, 40, 50, 60, 70, 80, 90,
9, 19, 29, 39, 49, 59, 69, 79, 89, 99,
90, 91, 92, 93, 94, 95, 96, 97, 98, 99,
41, 42, 43, 44, 46, 47, 48, 49]
obstacles_transition_probability = .2
domain = domains.GridMazeDomain(height, width, reward_location,
walls_location, obstacles_location, initial_state,
obstacles_transition_probability)
maze = LearningMazeDomain(domain=domain, num_sample=num_sample, length_sample=length_sample, discount=discount)
return maze
def obstacles_room(plotV=True, num_sample=100, length_sample=100, computeV=False):
height = 10
width = 10
reward_location = 18
initial_state = None # np.array([25])
obstacles_location = [12, 13, 22, 23,
35, 36, 45, 46,
62, 63, 72, 73,
67, 77]
walls_location = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9,
10, 20, 30, 40, 50, 60, 70, 80, 90,
9, 19, 29, 39, 49, 59, 69, 79, 89, 99,
90, 91, 92, 93, 94, 95, 96, 97, 98, 99]
obstacles_transition_probability = .2
domain = domains.GridMazeDomain(height, width, reward_location,
walls_location, obstacles_location, initial_state,
obstacles_transition_probability)
maze = LearningMazeDomain(domain=domain, num_sample=num_sample, length_sample=length_sample)
return maze
def threerooms(num_sample=5000, length_sample=100):
height = 50
width = 100
reward_location = 198
initial_state = None # np.array([25])
obstacles_location = [] # range(height*width)
walls_location = []
walls_location.extend(range(100))
walls_location.extend(range(4900, 5000))
walls_location.extend(range(0, 5000, 100))
walls_location.extend(range(99, 5000, 100))
walls_location.extend(range(1600, 1670))
walls_location.extend(range(1680, 1700))
walls_location.extend(range(3200, 3220))
walls_location.extend(range(3230, 3300))
obstacles_transition_probability = .2
domain = domains.GridMazeDomain(height, width, reward_location,
walls_location, obstacles_location, initial_state,
obstacles_transition_probability)
maze = LearningMazeDomain(domain=domain, num_sample=num_sample, length_sample=length_sample)
return maze
def threerooms(plotV=True, num_sample=5000, computeV=False):
height = 50
width = 100
reward_location = 198
initial_state = None # np.array([25])
obstacles_location = [] # range(height*width)
walls_location = []
walls_location.extend(range(100))
walls_location.extend(range(4900, 5000))
walls_location.extend(range(0, 5000, 100))
walls_location.extend(range(99, 5000, 100))
walls_location.extend(range(1600, 1670))
walls_location.extend(range(1680, 1700))
walls_location.extend(range(3200, 3220))
walls_location.extend(range(3230, 3300))
obstacles_transition_probability = .2
domain = domains.GridMazeDomain(height, width, reward_location,
walls_location, obstacles_location,
initial_state,
obstacles_transition_probability)
maze = LearningMazeDomain(domain=domain, num_sample=num_sample)
V = None
if computeV:
V = value_iteration(maze.domain.graph, reward_location,
obstacles_location, walls_location,
obstacles_transition_probability)
if plotV:
fig, ax = plt.subplots(1, 1)
maze.domain.graph.plot_signal(np.array(V), vertex_size=60, ax=ax)
plt.savefig('graphs/threeRooms_trueV.pdf')
plt.close()
return maze, V
def example_grid_maze(plotV=True):
height = 10
width = 10
reward_location = 9
initial_state = None # np.array([25])
obstacles_location = [14, 13, 24, 23, 29, 28, 39,
38] # range(height*width)
walls_location = [50, 51, 52, 53, 54, 55, 56, 74, 75, 76, 77, 78, 79]
obstacles_transition_probability = .2
domain = domains.GridMazeDomain(height, width, reward_location,
walls_location, obstacles_location,
initial_state,
obstacles_transition_probability)
maze = LearningMazeDomain(domain=domain, num_sample=2000)
V = value_iteration(maze.domain.graph, reward_location, obstacles_location,
walls_location, obstacles_transition_probability)
if plotV:
fig, ax = plt.subplots(1, 1)
maze.domain.graph.plot_signal(np.array(V), vertex_size=60, ax=ax)
plt.savefig('graphs/simpleMaze_trueV.pdf')
plt.close()
return maze, V
