from UI import UI
from Service import Service
from Repository import Repository
from Test import Test
if __name__ == "__main__":
repo = Repository("repoPuncte.txt", "repoTriunghiuri.txt")
serv = Service(repo)
ui = UI(serv)
teste = Test()
running = True
while running:
ui.display()
option = ui.getOption()
if option == "0":
running = False
elif option == "-1":
teste.runTests()
else:
ui.doOption(option)
class MDP(object):
def __init__(self, r, ui_width, ui_height, number=1):
self.r = r
self.start()
self.ui_width = ui_width
self.ui_height = ui_height
self.number = number
self.ui = UI(ui_width, ui_height, plus=ui_width * number)
def start(self):
self.gridInit()
self.policyInit()
def policyInit(self):
self.policy = []
for i in range(3):
self.policy.append([])
for j in range(4):
self.policy[i].append("*")
def gridInit(self):
self.grid = []
self.generateMatriz()
self.grid[0][3].u = 1
self.grid[0][3].terminal = True
self.grid[1][3].u = -1
self.grid[1][3].terminal = True
self.grid[2][3].u = 0.2
self.grid[1][1].u = -0.5
# self.grid[1][1].terminal = True
def generateMatriz(self):
for i in range(3):
self.grid.append([])
for j in range(4):
cell = Cell(i, j)
self.grid[i].append(cell)
def update(self):
for j in range(4):
for i in reversed(range(3)):
self.maxUtility(i, j)
def maxUtility(self, i, j):
cell = self.grid[i][j]
if not cell.terminal:
maxUtility, action = self.max(i, j)
cell.u = self.r + maxUtility
self.policy[i][j] = action
def max(self, i, j):
cell = self.grid[i][j]
upN = cell.getUpNeighbor()
rightN = cell.getRightNeighbor()
downN = cell.getDownNeighbor()
leftN = cell.getLeftNeighbor()
action = ["UP", "RIGHT", "DOWN", "LEFT"]
aux = [
self.utility(upN),
self.utility(rightN),
self.utility(downN),
self.utility(leftN)
]
index = 0
i = 0
while i < 3:
if aux[index] < aux[i + 1]:
index = i + 1
i = i + 1
return aux[index], action[index]
def utility(self, neighbors):
total = 0
for neighbor in neighbors:
cell = self.grid[neighbor[0]][neighbor[1]]
total = (cell.u * neighbor[2]) + total
return total
def display(self):
self.ui.display(self.grid, self.policy, self.r)
class QLearning(object):
def __init__(self,
r,
d,
ui_width,
ui_height,
dI,
dJ,
rows=3,
cols=2,
a=0.5,
g=0.8):
self.r = r
self.d = d
self.a = a
self.g = g
self.rows = rows
self.cols = cols
self.start()
self.dI = dI
self.dJ = dJ
self.i = self.dI
self.j = self.dJ
self.ui_width = ui_width
self.ui_height = ui_height
self.ui = UI(ui_width, ui_height)
def start(self):
self.gridInit()
self.policyInit()
def policyInit(self):
self.policy = []
for i in range(self.rows):
self.policy.append([])
for j in range(self.cols):
self.policy[i].append("*")
def gridInit(self):
self.grid = []
self.generateMatriz()
self.grid[0][1].setTerminal(10)
def generateMatriz(self):
for i in range(self.rows):
self.grid.append([])
for j in range(self.cols):
cell = Cell(i, j)
self.grid[i].append(cell)
# Q(s,a) <- Q(s,a) + alfa[r + gama*max(Q(s+1,a)) - Q(s,a)]
def up(self):
cell = self.grid[self.i][self.j]
if cell.terminal:
self.i = self.dI
self.j = self.dJ
else:
if (self.i - 1 < 0):
maxUtility, action = self.max(self.i, self.j)
u = self.d + (self.g * maxUtility)
cell.vU = cell.vU + self.a * (u - cell.vU)
cell.setU(cell.vU)
self.policy[self.i][self.j] = cell.policy()
else:
maxUtility, action = self.max(self.i - 1, self.j)
u = self.r + (self.g * maxUtility)
cell.vU = cell.vU + self.a * (u - cell.vU)
cell.setU(cell.vU)
self.policy[self.i][self.j] = cell.policy()
self.i = self.i - 1
def right(self):
cell = self.grid[self.i][self.j]
if cell.terminal:
self.i = self.dI
self.j = self.dJ
else:
if (self.j + 1 > self.cols - 1):
cell = self.grid[self.i][self.j]
maxUtility, action = self.max(self.i, self.j)
u = self.d + (self.g * maxUtility)
cell.vR = cell.vR + self.a * (u - cell.vR)
cell.setU(cell.vR)
self.policy[self.i][self.j] = cell.policy()
else:
maxUtility, action = self.max(self.i, self.j + 1)
u = self.r + (self.g * maxUtility)
cell.vR = cell.vR + self.a * (u - cell.vR)
cell.setU(cell.vR)
self.policy[self.i][self.j] = cell.policy()
self.j = self.j + 1
def down(self):
cell = self.grid[self.i][self.j]
if cell.terminal:
self.i = self.dI
self.j = self.dJ
else:
if (self.i + 1 > self.rows - 1):
maxUtility, action = self.max(self.i, self.j)
u = self.d + (self.g * maxUtility)
cell.vD = cell.vD + self.a * (u - cell.vD)
cell.setU(cell.vD)
self.policy[self.i][self.j] = cell.policy()
else:
maxUtility, action = self.max(self.i + 1, self.j)
u = self.r + (self.g * maxUtility)
cell.vD = cell.vD + self.a * (u - cell.vD)
cell.setU(cell.vD)
self.policy[self.i][self.j] = cell.policy()
self.i = self.i + 1
def left(self):
cell = self.grid[self.i][self.j]
if cell.terminal:
self.i = self.dI
self.j = self.dJ
else:
if (self.j - 1 < 0):
cell = self.grid[self.i][self.j]
maxUtility, action = self.max(self.i, self.j)
u = self.d + (self.g * maxUtility)
cell.vL = cell.vL + self.a * (u - cell.vL)
cell.setU(cell.vL)
self.policy[self.i][self.j] = cell.policy()
else:
maxUtility, action = self.max(self.i, self.j - 1)
u = self.r + (self.g * maxUtility)
cell.vL = cell.vL + self.a * (u - cell.vL)
cell.setU(cell.vL)
self.policy[self.i][self.j] = cell.policy()
self.j = self.j - 1
def auto(self):
if self.grid[self.i][self.j].terminal:
self.i = self.dI
self.j = self.dJ
while not self.grid[self.i][self.j].terminal:
if self.grid[self.i][self.j].policy() == "UP":
self.up()
elif self.grid[self.i][self.j].policy() == "RIGHT":
self.right()
elif self.grid[self.i][self.j].policy() == "DOWN":
self.down()
elif self.grid[self.i][self.j].policy() == "LEFT":
self.left()
def maxUtility(self, i, j):
cell = self.grid[i][j]
if not cell.terminal:
maxUtility, action = self.max(i, j)
cell.u = self.r + maxUtility
self.policy[i][j] = action
def max(self, i, j):
cell = self.grid[i][j]
action = ["UP", "RIGHT", "DOWN", "LEFT"]
aux = [cell.vU, cell.vR, cell.vD, cell.vL]
index = 0
i = 0
while i < 3:
if aux[index] < aux[i + 1]:
index = i + 1
i = i + 1
return aux[index], action[index]
def display(self):
self.ui.display(self.grid, self.policy, self.r, self.a, self.g, self.i,
self.j)
