def __init__(self, env):
super(LearningAgent, self).__init__(env) # sets self.env = env, state = None, next_waypoint = None, and a default color
self.color = 'red' # override color
self.planner = RoutePlanner(self.env, self) # simple route planner to get next_waypoint
self.q_table = QTable(alpha=0.1, gamma=0.1)
self.q_table_updater = QTableUpdater(self.q_table)
self.total_actions = 0.0
self.total_rewards = 0.0
class LearningAgent(Agent):
"""An agent that learns to drive in the smartcab world."""
def __init__(self, env):
super(LearningAgent, self).__init__(env) # sets self.env = env, state = None, next_waypoint = None, and a default color
self.color = 'red' # override color
self.planner = RoutePlanner(self.env, self) # simple route planner to get next_waypoint
self.q_table = QTable(alpha=0.1, gamma=0.1)
self.q_table_updater = QTableUpdater(self.q_table)
self.total_actions = 0.0
self.total_rewards = 0.0
# self.last_occurence_of_punishment = 0.0
def set_q_table(self, alpha=0.0, gamma=0.0):
self.q_table = QTable(alpha=alpha, gamma=gamma)
self.q_table_updater = QTableUpdater(self.q_table)
def reset(self, destination=None):
self.planner.route_to(destination)
# TODO: Prepare for a new trip; reset any variables here, if required
def update(self, t):
# Gather inputs
self.next_waypoint = self.planner.next_waypoint() # from route planner, also displayed by simulator
inputs = self.env.sense(self)
deadline = self.env.get_deadline(self)
# Update state
self.state = 'light: {}, left: {}, oncoming: {}, next_waypoint: {}'.format(inputs['light'],
inputs['left'],
inputs['oncoming'],
self.next_waypoint)
# Select action according to your policy
action = self.q_table.best_action(light=inputs['light'],
next_waypoint=self.next_waypoint,
left=inputs['left'],
oncoming=inputs['oncoming'])
# Execute action and get reward
reward = self.env.act(self, action)
# Learn policy based on state, action, reward
self.q_table_updater.update(light=inputs['light'],
next_waypoint=self.next_waypoint,
left=inputs['left'],
oncoming=inputs['oncoming'],
action=action,
reward=reward)
self.total_rewards += reward
self.total_actions += 1.0
print "LearningAgent.update(): deadline = {}, inputs = {}, action = {}, reward = {}, next_waypoint = {}".format(deadline, inputs, action, reward, self.next_waypoint) # [debug]
def __init_q_table(self):
self.q_table = {}
def __positions(self):
positions_list = []
for i in range(6):
for j in range(8):
positions_list.append((i+1,j+1))
return positions_list
def set_q_table(self, alpha=0.0, gamma=0.0):
self.q_table = QTable(alpha=alpha, gamma=gamma)
self.q_table_updater = QTableUpdater(self.q_table)
