aps_vector_x = np.ones((track.duration, 1)) * track.aps[:, 0].transpose() aps_vector_y = np.ones((track.duration, 1)) * track.aps[:, 1].transpose() aps_vector_dir = np.ones((track.duration, 1)) * track.aps[:, 2].transpose() track_x = np.ones((track.duration, track.aps.shape[0])) * track.track[:, 0].reshape((track.duration, 1)) track_y = np.ones((track.duration, track.aps.shape[0])) * track.track[:, 1].reshape((track.duration, 1)) # Calculate distance between APs to MS aps_dist = ((aps_vector_x - track_x) ** 2 + (aps_vector_y - track_y) ** 2) # remove AP if it is saturated aps_sat = aps_dist <= cfg.r_sat_sqrd # Calculate simulated global angles from APs to the MS global_angle_sim = fn.find_global_angle(aps_vector_x, aps_vector_y, track_x, track_y) # Calculate simulated local angles from APs to the MS local_angle_sim = global_angle_sim - aps_vector_dir # Add random error local_angle_sim = fn.add_error(local_angle_sim) """ Get position, remember to remove sat, when positioning """ # Converting to predicted global angle global_angle = local_angle_sim + aps_vector_dir # Converting predicted angles into slopes slopes = 1 / np.tan(np.radians(global_angle))
pos_error_y = np.ones((grid.shape[0], cfg_exp.N / cfg_exp.res + 1))
print 'training covariance'
for k in range(cfg_exp.N):
"""
Simulate
"""
# Calculate distance between APs to MS
aps_dist = ((aps_vector_x - grid_x) ** 2 + (aps_vector_y - grid_y) ** 2)
# remove AP if it is saturated
aps_sat = aps_dist <= cfg_exp.r_sat_sqrd
# Calculate global angles from APs to the MS
global_angle_sim = fn.find_global_angle(aps_vector_x, aps_vector_y, grid_x, grid_y)
# Calculate local angles from APs to the MS
local_angle_sim = global_angle_sim - aps_vector_dir
# Add random error
local_angle_sim = fn.add_error(local_angle_sim)
"""
Get grid, remember to remove sat, when positioning
"""
# Converting to predicted global angle
global_angle = local_angle_sim + aps_vector_dir
# Converting predicted angles into slopes
slopes = 1 / np.tan(np.radians(global_angle))
