Q-Learning code

import numpy as np

r = np.matrix(
		[[-1, -1, -1, -1, 0, -1],
        [-1, -1, -1, 0, -1, 100],
		[-1, -1, -1, 0, -1, -1],
		[-1,  0,  0, -1, 0, -1],
		[0, -1, -1, 0, -1, 100],
		[-1, 0, -1, -1, 0, 100]])
q = np.zeros((6, 6))
gmma = 0.8
epsilion = 0.3
a = 0.2
for episode in range(1000):
	# give a random state of the agent , 0-5
	state = np.random.randint(0, 6)
	if state == 5:
		print(state, " reach directly")
	else:
		print(state, end="")
	# while the state is not reach the last goal 5
	while state != 5:
		# choose the possible actions, but we can not choose the 
		# action whose R[state, action] = -1
		#record the all possible actions, and the possible Q values
		possibleActions = []
		possibleQ = []
		# a for loop to choose the possible action
		for action in range(6):
			if r[state, action] >= 0:
				possibleActions.append(action)
				#record the possible q values, which will be used in 
				# update the Q matrix
				possibleQ.append(q[state, action])
		# choose the next action, epsilion= 0.4, 
		# means we have the possibility of 40 percent to choose random
		# and the 60 percent to choose the max
		action = -1
		if np.random.random() < epsilion:
			action = possibleActions[np.random.randint(0, len(possibleActions))]
		else:
			# epsilon--greedy, choose the maxQ action
			action = possibleActions[np.argmax(possibleQ)]
		# update the q value
		q[state, action] = 0.2 * (r[state, action] + gmma * q[action].max()) + 0.8 * q[state, action]
		# update the state
		state = action
		print("-->" + str(state), end="")
		if state == 5:
			print()
	if episode % 10 == 0 :
		print()
		print("Training episode: %d" % episode)
		print(q)