def test_warehouse_04():
env = make_test_warehouse_env_01()
expected_value = None
policy = MCTSPolicy()
we.execute(env, policy)
print('**' * 30)
print('[Result]')
print('Finish time clock value=', env.finish_time_clock,
':uncompleted orders=', len(env.available_orders))
def test_warehouse_greedy(order_count):
start = time.time()
env = make_test_warehouse_env(order_count) #order 60개기준
expected_value = None
policy = GreedyPolicy()
we.execute(env, policy)
print('**' * 30)
print('[Result]GreedyPolicy')
print('Finish time clock value=', env.finish_time_clock,
':uncompleted orders=', len(env.available_orders))
end = time.time()
print('time', (end - start))
def test_warehouse_random(order_count, max_iteration):
start = time.time()
#반복수행
best = 99999999999
for i in range(max_iteration):
env = make_test_warehouse_env(order_count) #order 60개기준
policy = RandomPolicy()
we.execute(env, policy)
if best > env.finish_time_clock:
best = env.finish_time_clock
#print('Finish time clock value=', env.finish_time_clock,':uncompleted orders=',len(env.available_orders))
print('[Result] RandomPolicy')
print("Random Best=", best)
end = time.time()
print('time', (end - start))
def test_warehouse_mcts(order_count,
max_iteration=200,
exploration_constant=1.41):
start = time.time()
env = make_test_warehouse_env(order_count) #order 60개기준
expected_value = None
policy = MCTSPolicy(max_iteration=200, exploration_constant=1.41)
we.execute(env, policy)
print('**' * 30)
print('[Result] MCTS')
print('Finish time clock value=', env.finish_time_clock,
':uncompleted orders=', len(env.available_orders))
print('Order=', policy.result_order)
print('Folklift=', policy.result_folklift)
end = time.time()
print('time', (end - start))
def rollout(self, env, node):
clone_env = env.copy() #초기 상태를 복사해서 실행
#Order의 순서는 현재 상태까지 상태 사용, 나머지는 random 정렬 (fast roll out)
pre_order_sequence = []
random_order_sequence = clone_env.orders.copy()
for order_no in node.order_no_sequence:
order = clone_env.get_order_by_no(order_no)
random_order_sequence.remove(order)
pre_order_sequence.append(order)
np.random.shuffle(random_order_sequence)
order_sequence = pre_order_sequence + random_order_sequence
sequentail_policy = SequentialPolicy(order_sequence)
we.execute(clone_env, sequentail_policy)
reward = 0 - clone_env.finish_time_clock
return reward
