Я пытался реализовать алгоритм итерации значений. У меня есть сетка
grid = [[0, 0, 0, +1],
[0, "W", 0, -1],
[0, 0, 0, 0]]
Список действий
actlist = {UP:1, DOWN:2, LEFT:3, RIGHT:4}
И функция вознаграждения
reward = [[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 0]]
Я написал функцию T, которая возвращает кортеж из трех кортежей.
def T(i,j,actions):
if(i == 0 and j == 0):
if(actions == UP):
return (i,i,0.8),(i,i,0.1),(i,j+1,0.1)
elif(actions == DOWN):
return (i+1,j,0.8),(i,j,0.1),(i,j+1,0.1)
elif(actions == LEFT):
return (i,j,0.8),(i,j,0.1),(i+1,j,0.1)
elif(actions == RIGHT):
return (i,j+1,0.8),(i,i,0.1),(i+1,j,0.1)
elif (i == 0 and j == 1):
if(actions == UP):
return (i,i,0.8),(i,j-1,0.1),(i,j+1,0.1)
elif(actions == DOWN):
return (i,j,0.8),(i,j-1,0.1),(i,j+1,0.1)
elif(actions == LEFT):
return (i,j-1,0.8),(i,j,0.1),(i,j,0.1)
elif(actions == RIGHT):
return (i,j+1,0.8),(i,j,0.1),(i,j,0.1)
elif(i == 0 and j == 2):
if(actions == UP):
return (i,j,0.8),(i,j-1,0.1),(i,j+1,0.1)
elif(actions == DOWN):
return(i+1,j,0.8),(i,j-1,0.1),(i,j+1,0.1)
elif(actions == LEFT):
return (i,j-1,0.8),(i,j,0.1),(i+1,j,0.1)
elif(actions == RIGHT):
return (i,j+1,0.8),(i,j,0.1),(i+1,j,0.1)
elif(i == 0 and j == 3):
if(actions == UP):
return (-1,-1,0.8),(-1,-1,0.1),(-1,-1,0.1)
elif(actions == DOWN):
return (-1,-1,0.8),(-1,-1,0.1),(-1,-1,0.1)
elif(actions == LEFT):
return (-1,-1,0.8),(-1,-1,0.1),(-1,-1,0.1)
elif(actions == RIGHT):
return (-1,-1,0.8),(-1,-1,0.1),(-1,-1,0.1)
# 2nd row
elif (i == 1 and j == 0):
if(actions == UP):
return (i-1,j,0.8),(i,j,0.1),(i,j,0.1)
elif(actions == DOWN):
return (i+1,j,0.8),(i,j,0.1),(i,j,0.1)
elif(actions == LEFT):
return (i,j,0.8),(i-1,j,0.1),(i+1,j,0.1)
elif(actions == RIGHT):
return (i,j,0.8),(i-1,j,0.1),(i+1,j,0.1)
elif(i == 1 and j ==1):
if(actions == UP):
return (i,j,0.8),(i,j,0.1),(i,j,0.1)
elif(actions == DOWN):
return (i,j,0.8),(i,j,0.1),(i,j,0.1)
elif(actions == LEFT):
return (i,j,0.8),(i,j,0.1),(i,j,0.1)
elif(actions == RIGHT):
return (i,j,0.8),(i,j,0.1),(i,j,0.1)
elif (i == 1 and j == 2):
if(actions == UP):
return (i-1,j,0.8),(i,j,0.1),(i,j+1,0.1)
elif(actions == DOWN):
return (i+1,j,0.8),(i,j,0.1),(i,j+1,0.1)
elif(actions == LEFT):
return (i,j,0.8),(i-1,j,0.1),(i+1,j,0.1)
elif(actions == RIGHT):
return (i,j+1,0.8),(i-1,j,0.1),(i+1,j,0.1)
elif(i == 1 and j == 3):
if(actions == UP):
return (-2,-2,0.8),(-2,-2,0.1),(-2,-2,0.1)
elif(actions == DOWN):
return (-2,-2,0.8),(-2,-2,0.1),(-2,-2,0.1)
elif(actions == LEFT):
return (-2,-2,0.8),(-2,-2,0.1),(-2,-2,0.1)
elif(actions == RIGHT):
return (-2,-2,0.8),(-2,-2,0.1),(-2,-2,0.1)
# 3rd row
elif(i == 2 and j == 0):
if(actions == UP):
return (i-1,j,0.8),(i,j,0.1),(i,j+1,0.1)
elif(actions == DOWN):
return (i,j,0.8),(i,j,0.1),(i,j+1,1,0.1)
elif(actions == LEFT):
return (i,j,0.8),(i-1,j,0.1),(i,j,0.1)
elif(actions == RIGHT):
return (i,j+1,0.8),(i-1,j,0.1),(i,j,0.1)
elif (i == 2 and j == 1):
if(actions == UP):
return (i,j,0.8),(i,j-1,0.1),(i,j+1,0.1)
elif(actions == DOWN):
return (i,j,0.8),(i,j-1,0.1),(i,j+1,0.1)
elif(actions == LEFT):
return (i,j-1,0.8),(i,j,0.1),(i,j,0.1)
elif(actions == RIGHT):
return (i,j+1,0.8),(i,j,0.1),(i,j,0.1)
elif(i == 2 and j == 2):
if(actions == UP):
return (i-1,j,0.8),(i,j-1,0.1),(i,j+1,0.1)
elif(actions == DOWN):
return (i,j,0.8),(i,j-1,0.1),(i,j+1,0.1)
elif(actions == LEFT):
return (i,j-1,0.8),(i-1,j,0.1),(i,j,1)
elif(actions == RIGHT):
return (i,j+1,0.8),(i-1,j,0.1),(i,j,0.1)
elif(i == 2 and j == 3):
if(actions == UP):
return (i-1,j,0.8),(i,j-1,0.1),(i,j,0.1)
elif(actions == DOWN):
return (i,j,0.8),(i,j-1,0.1),(i,j,0.1)
elif(actions == LEFT):
return (i,j-1,0.8),(i-1,j,0.1),(i,j,0.1)
elif(actions == RIGHT):
return (i,j,0.8),(i-1,j,0.1),(i,j,0.1)
Эта функция вызывается в функции итерации значений:
def value_iteration():
U1 = [[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 0]]
while True:
U=U1.copy()
delta = 0
for i in range(len(grid)):
for j in range(len(grid[i])):
U1[i][j] = max(sum(p*(R(k,l)+gamma*U[k][l]) for (k,l,p) in T(i,j,a)) for a in actlist)
print(i,j,U1[i][j])
delta = max(delta, abs(U1[i][j] - U[i][j]))
if delta <= epsilon*(1 - gamma)/gamma:
return U
Проблема в том, что первые две итерации цикла for прошли хорошо с выходом
0 0
0 1
0 2
0 3
1 0
1 1
1 2
1 3
Но потом код остановился с ошибкой
ValueError: too many values to unpack (expected 3)
U1[i][j] = max(sum(p*(R(k,l)+gamma*U[k][l]) for (k,l,p) in T(i,j,a)) for a in actlist)
Ошибка возникает в этой строке. - person Shifat E Arman   schedule 06.08.2018(i, j, action)
, и чьи значения были результатами, которые вы должны вернуть из функции (хотя они должны быть жестко запрограммированы. Если вы не знакомы с словари, я бы посоветовал вам обратиться к официальной документации - person Edward Minnix   schedule 06.08.2018