def encodeMazeProbabilistic(filename, p=1):
maze = None
with open(filename, 'r') as file:
maze = file.readlines()
maze = [line.split() for line in maze]
maze = [[int(num) for num in line] for line in maze]
maze = np.array(maze)
currState = 0
startState = -1
endState = -1
states = np.zeros(maze.shape)
for i in range(len(states)):
for j in range(len(states[i])):
if maze[i][j] != 1:
states[i][j] = currState
currState += 1
else:
states[i][j] = -1
continue
if maze[i][j] == 2:
startState = currState - 1
if maze[i][j] == 3:
endState = currState - 1
mdp = MDP()
mdp.numStates = currState
mdp.numActions = 4
mdp.startState = startState
mdp.endStates = [endState]
mdp.allStates = [None] * mdp.numStates
for idx in range(len(mdp.allStates)):
mdp.allStates[idx] = State(idx, mdp.numActions)
for i in range(len(maze)):
for j in range(len(maze[i])):
if states[i][j] == -1:
continue
validStates = float(
int(states[i - 1][j] != -1) + int(states[i + 1][j] != -1) +
int(states[i][j - 1] != -1) + int(states[i][j + 1] != -1))
valid = [None] * 4
valid[0] = (states[i - 1][j] != -1)
valid[1] = (states[i][j + 1] != -1)
valid[2] = (states[i + 1][j] != -1)
valid[3] = (states[i][j - 1] != -1)
goal = [None] * 4
goal[0] = (maze[i - 1][j] == 3)
goal[1] = (maze[i][j + 1] == 3)
goal[2] = (maze[i + 1][j] == 3)
goal[3] = (maze[i][j - 1] == 3)
if valid[0]:
if goal[0]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 0,
int(states[i - 1][j]), -1,
p + ((1.0 - p) / validStates)
])
if valid[1]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 1,
int(states[i - 1][j]), -1, (1.0 - p) / validStates
])
if valid[2]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 2,
int(states[i - 1][j]), -1, (1.0 - p) / validStates
])
if valid[3]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 3,
int(states[i - 1][j]), -1, (1.0 - p) / validStates
])
else:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 0,
int(states[i - 1][j]), -1, p + (1.0 - p) / validStates
])
if valid[1]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 1,
int(states[i - 1][j]), -1, (1.0 - p) / validStates
])
if valid[2]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 2,
int(states[i - 1][j]), -1, (1.0 - p) / validStates
])
if valid[3]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 3,
int(states[i - 1][j]), -1, (1.0 - p) / validStates
])
if valid[1]:
if goal[1]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 1,
int(states[i][j + 1]), -1,
p + ((1.0 - p) / validStates)
])
if valid[0]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 0,
int(states[i][j + 1]), -1, (1.0 - p) / validStates
])
if valid[2]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 2,
int(states[i][j + 1]), -1, (1.0 - p) / validStates
])
if valid[3]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 3,
int(states[i][j + 1]), -1, (1.0 - p) / validStates
])
else:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 1,
int(states[i][j + 1]), -1, p + (1.0 - p) / validStates
])
if valid[0]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 0,
int(states[i][j + 1]), -1, (1.0 - p) / validStates
])
if valid[2]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 2,
int(states[i][j + 1]), -1, (1.0 - p) / validStates
])
if valid[3]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 3,
int(states[i][j + 1]), -1, (1.0 - p) / validStates
])
if valid[2]:
if goal[2]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 2,
int(states[i + 1][j]), -1,
p + ((1.0 - p) / validStates)
])
if valid[0]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 0,
int(states[i + 1][j]), -1, (1.0 - p) / validStates
])
if valid[1]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 1,
int(states[i + 1][j]), -1, (1.0 - p) / validStates
])
if valid[3]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 3,
int(states[i + 1][j]), -1, (1.0 - p) / validStates
])
else:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 2,
int(states[i + 1][j]), -1,
p + ((1.0 - p) / validStates)
])
if valid[0]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 0,
int(states[i + 1][j]), -1, (1.0 - p) / validStates
])
if valid[1]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 1,
int(states[i + 1][j]), -1, (1.0 - p) / validStates
])
if valid[3]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 3,
int(states[i + 1][j]), -1, (1.0 - p) / validStates
])
if valid[3]:
if goal[3]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 3,
int(states[i][j - 1]), -1,
p + ((1.0 - p) / validStates)
])
if valid[0]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 0,
int(states[i][j - 1]), -1, (1.0 - p) / validStates
])
if valid[1]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 1,
int(states[i][j - 1]), -1, (1.0 - p) / validStates
])
if valid[2]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 2,
int(states[i][j - 1]), -1, (1.0 - p) / validStates
])
else:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 3,
int(states[i][j - 1]), -1,
p + ((1.0 - p) / validStates)
])
if valid[0]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 0,
int(states[i][j - 1]), -1, (1.0 - p) / validStates
])
if valid[1]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 1,
int(states[i][j - 1]), -1, (1.0 - p) / validStates
])
if valid[2]:
mdp.allStates[int(states[i][j])].addTransition([
int(states[i][j]), 2,
int(states[i][j - 1]), -1, (1.0 - p) / validStates
])
mdp.gamma = 1
return mdp
def encodeMazeDeterministic(filename):
maze = None
with open(filename, 'r') as file:
maze = file.readlines()
maze = [line.split() for line in maze]
maze = [[int(num) for num in line] for line in maze]
maze = np.array(maze)
currState = 0
startState = -1
endState = -1
states = np.zeros(maze.shape)
for i in range(len(states)):
for j in range(len(states[i])):
if maze[i][j] != 1:
states[i][j] = currState
currState += 1
else:
states[i][j] = -1
continue
if maze[i][j] == 2:
startState = currState - 1
if maze[i][j] == 3:
endState = currState - 1
mdp = MDP()
mdp.numStates = currState
mdp.numActions = 4
mdp.startState = startState
mdp.endStates = [endState]
mdp.allStates = [None] * mdp.numStates
for idx in range(len(mdp.allStates)):
mdp.allStates[idx] = State(idx, mdp.numActions)
for i in range(len(maze)):
for j in range(len(maze[i])):
if states[i][j] == -1:
continue
if states[i - 1][j] != -1:
if maze[i - 1][j] == 3:
mdp.allStates[int(states[i][j])].addTransition(
[int(states[i][j]), 0,
int(states[i - 1][j]), -1, 1])
else:
mdp.allStates[int(states[i][j])].addTransition(
[int(states[i][j]), 0,
int(states[i - 1][j]), -1, 1])
if states[i][j + 1] != -1:
if maze[i][j + 1] == 3:
mdp.allStates[int(states[i][j])].addTransition(
[int(states[i][j]), 1,
int(states[i][j + 1]), -1, 1])
else:
mdp.allStates[int(states[i][j])].addTransition(
[int(states[i][j]), 1,
int(states[i][j + 1]), -1, 1])
if states[i + 1][j] != -1:
if maze[i + 1][j] == 3:
mdp.allStates[int(states[i][j])].addTransition(
[int(states[i][j]), 2,
int(states[i + 1][j]), -1, 1])
else:
mdp.allStates[int(states[i][j])].addTransition(
[int(states[i][j]), 2,
int(states[i + 1][j]), -1, 1])
if states[i][j - 1] != -1:
if maze[i][j - 1] == 3:
mdp.allStates[int(states[i][j])].addTransition(
[int(states[i][j]), 3,
int(states[i][j - 1]), -1, 1])
else:
mdp.allStates[int(states[i][j])].addTransition(
[int(states[i][j]), 3,
int(states[i][j - 1]), -1, 1])
mdp.gamma = 0.9
return mdp