forked from mawei/dp
1
0
Fork 0
dp/laika/dgps.py

161 lines
6.5 KiB
Python

import os
import numpy as np
from datetime import datetime
from .gps_time import GPSTime
from .constants import SECS_IN_YEAR
from . import raw_gnss as raw
from .rinex_file import RINEXFile
from .downloader import download_cors_coords
from .helpers import get_constellation
def mean_filter(delay):
d2 = delay.copy()
max_step = 10
for i in range(max_step, len(delay) - max_step):
finite_idxs = np.where(np.isfinite(delay[i - max_step:i + max_step]))
if max_step in finite_idxs[0]:
step = min([max_step, finite_idxs[0][-1] - max_step, max_step - finite_idxs[0][0]])
d2[i] = np.nanmean(delay[i - step:i + step + 1])
return d2
def download_and_parse_station_postions(cors_station_positions_path, cache_dir):
if not os.path.isfile(cors_station_positions_path):
cors_stations = {}
coord_file_paths = download_cors_coords(cache_dir=cache_dir)
for coord_file_path in coord_file_paths:
try:
station_id = coord_file_path.split('/')[-1][:4]
with open(coord_file_path, 'r+') as coord_file:
contents = coord_file.readlines()
phase_center = False
for line_number in range(len(contents)):
if 'L1 Phase Center' in contents[line_number]:
phase_center = True
if not phase_center and 'ITRF2014 POSITION' in contents[line_number]:
velocity = [float(contents[line_number+8].split()[3]),
float(contents[line_number+9].split()[3]),
float(contents[line_number+10].split()[3])]
if phase_center and 'ITRF2014 POSITION' in contents[line_number]:
epoch = GPSTime.from_datetime(datetime(2005,1,1))
position = [float(contents[line_number+2].split()[3]),
float(contents[line_number+3].split()[3]),
float(contents[line_number+4].split()[3])]
cors_stations[station_id] = [epoch, position, velocity]
break
except ValueError:
pass
cors_station_positions_file = open(cors_station_positions_path, 'wb')
np.save(cors_station_positions_file, cors_stations)
cors_station_positions_file.close()
def get_closest_station_names(pos, k=5, max_distance=100000, cache_dir='/tmp/gnss/'):
from scipy.spatial import cKDTree
cors_station_positions_dict = load_cors_station_positions(cache_dir)
station_ids = list(cors_station_positions_dict.keys())
station_positions = []
for station_id in station_ids:
station_positions.append(cors_station_positions_dict[station_id][1])
tree = cKDTree(station_positions)
distances, idxs = tree.query(pos, k=k, distance_upper_bound=max_distance)
return np.array(station_ids)[idxs]
def load_cors_station_positions(cache_dir):
cors_station_positions_path = cache_dir + 'cors_coord/cors_station_positions'
download_and_parse_station_postions(cors_station_positions_path, cache_dir)
with open(cors_station_positions_path, 'rb') as f:
return np.load(f, allow_pickle=True).item() # pylint: disable=unexpected-keyword-arg
def get_station_position(station_id, cache_dir='/tmp/gnss/', time=GPSTime.from_datetime(datetime.utcnow())):
cors_station_positions_dict = load_cors_station_positions(cache_dir)
epoch, pos, vel = cors_station_positions_dict[station_id]
return ((time - epoch)/SECS_IN_YEAR)*np.array(vel) + np.array(pos)
def parse_dgps(station_id, station_obs_file_path, dog, max_distance=100000, required_constellations=['GPS']):
station_pos = get_station_position(station_id, cache_dir=dog.cache_dir)
obsdata = RINEXFile(station_obs_file_path)
measurements = raw.read_rinex_obs(obsdata)
# if not all constellations in first 100 epochs bail
detected_constellations = set()
for m in sum(measurements[:100],[]):
detected_constellations.add(get_constellation(m.prn))
for constellation in required_constellations:
if constellation not in detected_constellations:
return None
proc_measurements = []
for measurement in measurements:
proc_measurements.append(raw.process_measurements(measurement, dog=dog))
# sample at 30s
if len(proc_measurements) > 2880:
proc_measurements = proc_measurements[::int(len(proc_measurements)/2880)]
if len(proc_measurements) != 2880:
return None
station_delays = {}
n = len(proc_measurements)
for signal in ['C1C', 'C2P']:
times = []
station_delays[signal] = {}
for i, proc_measurement in enumerate(proc_measurements):
times.append(proc_measurement[0].recv_time)
Fx_pos = raw.pr_residual(proc_measurement, signal=signal)
residual = -np.array(Fx_pos(list(station_pos) + [0, 0]))
for j, m in enumerate(proc_measurement):
prn = m.prn
if prn not in station_delays[signal]:
station_delays[signal][prn] = np.nan*np.ones(n)
station_delays[signal][prn][i] = residual[j]
assert len(times) == n
# TODO crude way to get dgps station's clock errors,
# could this be biased? Only use GPS for convenience.
model_delays = {}
for prn in station_delays['C1C']:
if get_constellation(prn) == 'GPS':
model_delays[prn] = np.nan*np.zeros(n)
for i in range(n):
model_delays[prn][i] = dog.get_delay(prn, times[i], station_pos, no_dgps=True)
station_clock_errs = np.zeros(n)
for i in range(n):
station_clock_errs[i] = np.nanmean([(station_delays['C1C'][prn][i] - model_delays[prn][i]) for prn in model_delays])
# remove clock errors and smooth out signal
for prn in station_delays['C1C']:
station_delays['C1C'][prn] = mean_filter(station_delays['C1C'][prn] - station_clock_errs)
for prn in station_delays['C2P']:
station_delays['C2P'][prn] = station_delays['C2P'][prn] - station_clock_errs
return DGPSDelay(station_id, station_pos, station_delays,
times, max_distance)
class DGPSDelay:
def __init__(self, station_id, station_pos,
station_delays, station_delays_t, max_distance):
self.id = station_id
self.pos = station_pos
self.delays = station_delays
self.delays_t = station_delays_t
self.max_distance = max_distance
def get_delay(self, prn, time, signal='C1C'):
time_index = int((time - self.delays_t[0])/30)
assert abs(self.delays_t[time_index] - time) < 30
if prn in self.delays[signal] and np.isfinite(self.delays[signal][prn][time_index]):
return self.delays[signal][prn][time_index]
return None
def valid(self, time, recv_pos):
return (np.linalg.norm(recv_pos - self.pos) <= self.max_distance and
time - self.delays_t[0] > -30 and
self.delays_t[-1] - time > -30)