Simplify the creation of QLearning
Run the following to install:
pip install Easy-QLearningimport EQL
####Initialise QTable####
#Mission: go to the house without run over the human
#Game grid
Grid = [#1 = House, -1 = Human
[0,0,1],
[0,-1,0],
[0,0,0]
]
#First car position coordinates
x = 0
y = 2
#Current state
state = 7
#Action list and travel coordinate
actions = [
[-1, 0], # Up
[1, 0], #Down
[0, -1], # Left
[0, 1] # Right
]
#Make QTable
Qtable = EQL.QLearning(nbAction=4,nbState=10)
####Train the QTable####
#100 games
for _ in range(100):
#Reset the car position
x = 0
y = 2
#Default state
state = 7
#While the car are not in the house
while x != 2 or y != 0:
#Display the grid
print("---------------------")
yTemp = 0
for line in Grid:
xTemp = 0
for pt in line:
print("%s\t" % (pt if yTemp != y or xTemp != x else "X"), end="")
xTemp += 1
yTemp += 1
print("")
#Choose an action
action = Qtable.takeAction(state,epsilon=0.4)
#Move the car
y = max(0, min(y + actions[action][0],2))
x = max(0, min(x + actions[action][1],2))
#Calcul the position in the grid (state)
newState = (y*3+x+1)
#Get the reward of the position
reward = Grid[y][x]
print("state : ", newState)
print("reward : ", reward)
#Update Q function
Qtable.updateQFunction(newState,state,reward)
#Next state
state = newState
#Display the QTable
for s in range(1, 10):
print(s, Qtable.QTable[s])
#Save my QTable in myTable.npz
Qtable.saveQTable("myTable")
#Load my QTable from myTable.npz
Qtable.loadQTable("myTable")$ pip install -e .[dev]