def run_simulation(MDP, policy):
print "Starting simulation for given MDP"
while MDP.get_parked() == False:
action = policy.choose_action(MDP.get_time())
print "[TIME", MDP.get_time() ,"]:", policy.get_name(), "chose action", action
MDP.take_action(action)
print "[TIME", MDP.get_time() ,"]: Moved to state", MDP.get_state(), "Current reward %.3f." % MDP.get_reward()
print "Exited in (spot, handicapped, available):", MDP.get_spot(), MDP.get_handicapped(), MDP.get_available()
def run_training(MDP, policy, horizon):
t = 0
trajectory = []
while t < horizon:
if MDP.get_parked():
# we need to make it do one more update.
action = policy.choose_training_action()
state = MDP.get_state()
trajectory.append((state, action, MDP.get_state()))
#reset our simulator
MDP.reset()
policy.q_updates(trajectory)
trajectory = []
else:
#record trajectory
action = policy.choose_training_action()
state = MDP.get_state()
policy.take_action(action)
trajectory.append((state, action, MDP.get_state()))
t += 1
def run_simulation(MDP, policy):
#print "Starting simulation for given MDP"
while not MDP.get_parked():
action = policy.choose_action(MDP.get_time())
#print "[TIME", MDP.get_time() ,"]:", policy.get_name(), "chose action", action
policy.take_action(action)