import numpy as np
max_iter_num = 10
max_iter_ue = 50
room_size = 20
c_ns = 0.299792458
np.random.seed(47)
ue_num = int(input())
bs_pos_1 = np.asarray(input().split(" "), dtype=float)
bs_pos_2 = np.asarray(input().split(" "), dtype=float)
data_input = np.zeros([ue_num, 4], float)
for i in range(ue_num):
data_input[i] = np.asarray(input().split(" "), dtype=float)
toa_all = data_input
bs_pos = np.zeros((4, 2))
bs_pos[0, :2] = bs_pos_1
bs_pos[1, :2] = bs_pos_2
bs_pos[2, :2] = np.array([-1, 1]) * 5
bs_pos[3, :2] = np.array([1, 1]) * 5
bs_num = bs_pos.shape[0]
ue_pos_est = np.zeros((ue_num, 2), dtype=float)
bs_pos_est = bs_pos.copy()
min_loss = np.inf
tdoa_measured = toa_all[:, 1:] - toa_all[:, 0:1]
ue_pos_best = ue_pos_est.copy()
bs_pos_best = bs_pos_est.copy()
for iter in range(max_iter_num):
bs_pos_est[2, 0] = (np.random.rand() - 1) * room_size / 2
bs_pos_est[3, 0] = (np.random.rand()) * room_size / 2
bs_pos_est[2, 1] = (np.random.rand()) * room_size / 2
bs_pos_est[3, 1] = (np.random.rand()) * room_size / 2
loss_tdoa = 0
for ue_i in range(ue_num):
min_loss_ue = np.inf
for iter_ue in range(max_iter_ue):
ue_pos_est_i = (np.random.rand(2) - 0.5) * room_size
toa_est_i = (
np.sqrt(
(bs_pos_est[:, 0] - ue_pos_est_i[0]) ** 2
+ (bs_pos_est[:, 1] - ue_pos_est_i[1]) ** 2
)
/ c_ns
)
tdoa_est_i = toa_est_i[1:] - toa_est_i[0:1]
loss_tdoa_i = np.sum((tdoa_est_i - tdoa_measured[ue_i, :]) ** 2)
if loss_tdoa_i < min_loss_ue:
ue_pos_est[ue_i, :] = ue_pos_est_i.copy()
min_loss_ue = loss_tdoa_i
loss_tdoa += min_loss_ue
if loss_tdoa < min_loss:
ue_pos_best = ue_pos_est.copy()
bs_pos_best = bs_pos_est.copy()
min_loss = loss_tdoa
ue_pos_est = ue_pos_best
bs_pos_est = bs_pos_best
output_data_line_1 = np.array([[bs_pos_est[2, 0], bs_pos_est[2, 1]]])
output_data_line_2 = np.array([[bs_pos_est[3, 0], bs_pos_est[3, 1]]])
output_data = ue_pos_est
print(bs_pos_est[2, 0], bs_pos_est[2, 1])
print(bs_pos_est[3, 0], bs_pos_est[3, 1])
for row in output_data:
print(*row)
