Skip to content

TASGNSS API Documents

Backend

Source code in tasgnss/core.py 
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
class Backend:
    def __init__(self, use_torch=False):
        self.use_torch = use_torch
        if use_torch:
            try:
                import torch
                self.torch = torch
                self.np = np
            except ImportError:
                print("Please install pytorch to enable the torch-based WLS")
                self.use_torch = False
                self.torch = None
                self.np = np
        else:
            self.torch = None
            self.np = np

    def array(self, data, dtype=None):
        if self.use_torch and self.torch is not None:
            return self.torch.tensor(data, dtype=dtype)
        else:
            return self.np.array(data, dtype=dtype)

    def zeros(self, shape, dtype=None):
        if self.use_torch and self.torch is not None:
            return self.torch.zeros(shape, dtype=dtype)
        else:
            return self.np.zeros(shape, dtype=dtype)

    def ones(self, shape, dtype=None):
        if self.use_torch and self.torch is not None:
            return self.torch.ones(shape, dtype=dtype)
        else:
            return self.np.ones(shape, dtype=dtype)

    def eye(self, n, dtype=None):
        if self.use_torch and self.torch is not None:
            return self.torch.eye(n, dtype=dtype)
        else:
            return self.np.eye(n, dtype=dtype)

    def diag(self, v):
        if self.use_torch and self.torch is not None:
            return self.torch.diag(v)
        else:
            return self.np.diag(v)

    def linalg_norm(self, x, axis=None):
        if self.use_torch and self.torch is not None:
            return self.torch.linalg.norm(x, dim=axis)
        else:
            return self.np.linalg.norm(x, axis=axis)

    def linalg_lstsq(self, A, B, rcond=None):
        if self.use_torch and self.torch is not None:
            return [self.torch.linalg.pinv(A) @ B]
        else:
            return self.np.linalg.lstsq(A, B, rcond=rcond)

    def reshape(self, x, shape):
        if self.use_torch and self.torch is not None:
            return x.reshape(shape)
        else:
            return self.np.reshape(x, shape)

    def squeeze(self, x, axis=None):
        if self.use_torch and self.torch is not None:
            if axis is None:
                return self.torch.squeeze(x)  # 不传 dim 参数，让 PyTorch 自动处理
            else:
                return self.torch.squeeze(x, dim=axis)
        else:
            return self.np.squeeze(x, axis=axis)

    def any(self, x, axis=None):
        if self.use_torch and self.torch is not None:
            return self.torch.any(x, dim=axis)
        else:
            return self.np.any(x, axis=axis)

    def unique(self, x, return_counts=False):
        if self.use_torch and self.torch is not None:
            if return_counts:
                return self.torch.unique(x, return_counts=True)
            else:
                return self.torch.unique(x)
        else:
            return self.np.unique(x, return_counts=return_counts)

    def sqrt(self, x):
        if self.use_torch and self.torch is not None:
            return self.torch.sqrt(x)
        else:
            return self.np.sqrt(x)

    def sin(self, x):
        if self.use_torch and self.torch is not None:
            return self.torch.sin(x)
        else:
            return self.np.sin(x)

    def cos(self, x):
        if self.use_torch and self.torch is not None:
            return self.torch.cos(x)
        else:
            return self.np.cos(x)

    def arctan2(self, y, x):
        if self.use_torch and self.torch is not None:
            return self.torch.atan2(y, x)
        else:
            return self.np.arctan2(y, x)

    def degrees(self, x):
        if self.use_torch and self.torch is not None:
            return self.torch.rad2deg(x)
        else:
            return self.np.degrees(x)

    def radians(self, x):
        if self.use_torch and self.torch is not None:
            return self.torch.deg2rad(x)
        else:
            return self.np.radians(x)

    def where(self, condition, x, y):
        if self.use_torch and self.torch is not None:
            return self.torch.where(condition, x, y)
        else:
            return self.np.where(condition, x, y)

    def sum(self, x, axis=None):
        if self.use_torch and self.torch is not None:
            return self.torch.sum(x, dim=axis)
        else:
            return self.np.sum(x, axis=axis)

    def dot(self, a, b):
        if self.use_torch and self.torch is not None:
            return self.torch.matmul(a, b)
        else:
            return self.np.dot(a, b)

    def abs(self, x):
        if self.use_torch and self.torch is not None:
            return self.torch.abs(x)
        else:
            return self.np.abs(x)

    def ones_like(self, x, dtype=None):
        if self.use_torch and self.torch is not None:
            return self.torch.ones_like(x, dtype=dtype)
        else:
            return self.np.ones_like(x, dtype=dtype)

    def transpose(self, x):
        if self.use_torch and self.torch is not None:
            return x.T
        else:
            return self.np.transpose(x)

    def stack(self, arrays, axis=0):
        if self.use_torch and self.torch is not None:
            return self.torch.stack(arrays, dim=axis)
        else:
            return self.np.stack(arrays, axis=axis)

    def vstack(self, arrays):
        if self.use_torch and self.torch is not None:
            return self.torch.vstack(arrays)
        else:
            return self.np.vstack(arrays)

    def hstack(self, arrays):
        if self.use_torch and self.torch is not None:
            return self.torch.hstack(arrays)
        else:
            return self.np.hstack(arrays)

    def concatenate(self, arrays, axis=0):
        if self.use_torch and self.torch is not None:
            return self.torch.cat(arrays, dim=axis)
        else:
            return self.np.concatenate(arrays, axis=axis)

    def copy(self, x):
        if self.use_torch and self.torch is not None:
            return x.clone()
        else:
            return self.np.copy(x)

    def asarray(self, x, dtype=None):
        if self.use_torch and self.torch is not None:
            return self.torch.as_tensor(x, dtype=dtype)
        else:
            return self.np.asarray(x, dtype=dtype)

    def is_tensor(self, x):
        if self.use_torch and self.torch is not None:
            return isinstance(x, self.torch.Tensor)
        else:
            return isinstance(x, self.np.ndarray)

    def to_numpy(self, x):
        if self.use_torch and self.torch is not None and isinstance(x, self.torch.Tensor):
            return x.detach().cpu().numpy()
        else:
            return x

    def from_numpy(self, x):
        if self.use_torch and self.torch is not None:
            return self.torch.from_numpy(x)
        else:
            return x

    def requires_grad_(self, x, requires_grad=True):
        if self.use_torch and self.torch is not None and isinstance(x, self.torch.Tensor):
            return x.requires_grad_(requires_grad)
        else:
            return x

    def grad(self, x):
        if self.use_torch and self.torch is not None and isinstance(x, self.torch.Tensor):
            return x.grad
        else:
            return None

    def no_grad(self):
        if self.use_torch and self.torch is not None:
            return self.torch.no_grad()
        else:
            class NoGrad:
                def __enter__(self):
                    pass
                def __exit__(self, exc_type, exc_val, exc_tb):
                    pass
            return NoGrad()

    def zeros_like(self, x, dtype=None):
        if self.use_torch and self.torch is not None:
            return self.torch.zeros_like(x, dtype=dtype)
        else:
            return self.np.zeros_like(x, dtype=dtype)

    def block_diag(self, *arrays):
        """
        Create a block diagonal matrix from provided arrays.
        """
        if self.use_torch and self.torch is not None:
            return self.torch.block_diag(*arrays)
        else:
            # Manual implementation for NumPy
            if len(arrays) == 0:
                return self.np.array([])
            # Calculate total shape
            total_rows = sum(arr.shape[0] for arr in arrays)
            total_cols = sum(arr.shape[1] for arr in arrays)
            result = self.np.zeros((total_rows, total_cols))
            row_offset = 0
            col_offset = 0
            for arr in arrays:
                rows, cols = arr.shape
                result[row_offset:row_offset + rows, col_offset:col_offset + cols] = arr
                row_offset += rows
                col_offset += cols
            return result

    def astype(self, tensor_or_array, dtype):
        """
        Convert the data type of the input array/tensor.
        Automatically dispatches to torch.to(dtype) or np.astype(dtype)
        """
        if self.use_torch and self.torch is not None:
            return tensor_or_array.to(dtype)
        else:
            return tensor_or_array.astype(dtype)

    def to(self, tensor_or_array, device):
        """
        Move tensor to device (if using torch), or do nothing (if using numpy).
        """
        if self.use_torch and self.torch is not None:
            return tensor_or_array.to(device)
        else:
            return tensor_or_array

    # ========== 常用数据类型属性 ==========

    @property
    def float16(self):
        return self.torch.float16 if self.use_torch else self.np.float16

    @property
    def float32(self):
        return self.torch.float32 if self.use_torch else self.np.float32

    @property
    def float64(self):
        return self.torch.float64 if self.use_torch else self.np.float64

    @property
    def int8(self):
        return self.torch.int8 if self.use_torch else self.np.int8

    @property
    def int16(self):
        return self.torch.int16 if self.use_torch else self.np.int16

    @property
    def int32(self):
        return self.torch.int32 if self.use_torch else self.np.int32

    @property
    def int64(self):
        return self.torch.int64 if self.use_torch else self.np.int64

    @property
    def uint8(self):
        return self.torch.uint8 if self.use_torch else self.np.uint8

    @property
    def bool(self):
        return self.torch.bool if self.use_torch else self.np.bool_

    @property
    def complex64(self):
        return self.torch.complex64 if self.use_torch else self.np.complex64

    @property
    def complex128(self):
        return self.torch.complex128 if self.use_torch else self.np.complex128

astype(tensor_or_array, dtype)

Convert the data type of the input array/tensor. Automatically dispatches to torch.to(dtype) or np.astype(dtype)

Source code in tasgnss/core.py 
283
284
285
286
287
288
289
290
291
def astype(self, tensor_or_array, dtype):
    """
    Convert the data type of the input array/tensor.
    Automatically dispatches to torch.to(dtype) or np.astype(dtype)
    """
    if self.use_torch and self.torch is not None:
        return tensor_or_array.to(dtype)
    else:
        return tensor_or_array.astype(dtype)

block_diag(*arrays)

Create a block diagonal matrix from provided arrays.

Source code in tasgnss/core.py 
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
def block_diag(self, *arrays):
    """
    Create a block diagonal matrix from provided arrays.
    """
    if self.use_torch and self.torch is not None:
        return self.torch.block_diag(*arrays)
    else:
        # Manual implementation for NumPy
        if len(arrays) == 0:
            return self.np.array([])
        # Calculate total shape
        total_rows = sum(arr.shape[0] for arr in arrays)
        total_cols = sum(arr.shape[1] for arr in arrays)
        result = self.np.zeros((total_rows, total_cols))
        row_offset = 0
        col_offset = 0
        for arr in arrays:
            rows, cols = arr.shape
            result[row_offset:row_offset + rows, col_offset:col_offset + cols] = arr
            row_offset += rows
            col_offset += cols
        return result

to(tensor_or_array, device)

Move tensor to device (if using torch), or do nothing (if using numpy).

Source code in tasgnss/core.py 
293
294
295
296
297
298
299
300
def to(self, tensor_or_array, device):
    """
    Move tensor to device (if using torch), or do nothing (if using numpy).
    """
    if self.use_torch and self.torch is not None:
        return tensor_or_array.to(device)
    else:
        return tensor_or_array

covecef(pos, Q)

Convert covariance from ENU to ECEF Parameters: pos: tuple or array of (lat, lon) in degrees Q: 3x3 covariance matrix in ENU frame Returns: Q_ecef: 3x3 covariance matrix in ECEF frame

Source code in tasgnss/core.py 
551
552
553
554
555
556
557
558
559
560
561
def covecef(pos, Q):
    """
    Convert covariance from ENU to ECEF
    Parameters:
        pos: tuple or array of (lat, lon) in degrees
        Q: 3x3 covariance matrix in ENU frame
    Returns:
        Q_ecef: 3x3 covariance matrix in ECEF frame
    """
    E = xyz2enu(pos)
    return E.T @ Q @ E

doppler_observe_func(vel, dT, pos, satpos, satvel, sdT, sys, enable_torch=False, device='cpu')

Computes Doppler (range-rate) observation residuals and their Jacobian matrix for GNSS velocity estimation. This function models the geometric relative velocity, Earth rotation (Sagnac) correction, and receiver-satellite clock drift difference. The state vector assumed is: [vx, vy, vz, dT] — velocity + public clock drift (no position or clock bias). Clock bias is handled externally (e.g., in pseudorange module), enabling modular design via block_diag fusion.

Parameters: vel : receiver velocity vector, shape (3,) dT : public pseudorange rate for all systems, shape (1,), unit: m/s pos : approximate receiver position (ECEF), shape (3,) — precision ~100m sufficient satpos : satellite positions (ECEF), shape (n_obs, 3) satvel : satellite velocities (ECEF), shape (n_obs, 3) sdT : satellite clock drifts (from broadcast/SP3), shape (n_obs,), unit: s/s sys : list of GNSS system identifiers for each observation, e.g., ['G', 'E', 'C'] enable_torch : if True, use PyTorch backend for computations device : if enable_torch is True, specifies the device ('cpu' or 'cuda')

Returns: v : predicted Doppler velocity residuals (m/s), shape (n_obs, 1) H : Jacobian matrix w.r.t state [vx, vy, vz, dT], shape (n_obs, 4)

Source code in tasgnss/core.py 
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
def doppler_observe_func(vel, dT, pos, satpos, satvel, sdT, sys, enable_torch=False, device='cpu'):
    """
    Computes Doppler (range-rate) observation residuals and their Jacobian matrix for GNSS velocity estimation.
    This function models the geometric relative velocity, Earth rotation (Sagnac) correction, and receiver-satellite clock drift difference.
    The state vector assumed is: [vx, vy, vz, dT] — velocity + public clock drift (no position or clock bias).
    Clock bias is handled externally (e.g., in pseudorange module), enabling modular design via block_diag fusion.

    Parameters:
    vel : receiver velocity vector, shape (3,)
    dT : public pseudorange rate for all systems, shape (1,), unit: m/s
    pos : approximate receiver position (ECEF), shape (3,) — precision ~100m sufficient
    satpos : satellite positions (ECEF), shape (n_obs, 3)
    satvel : satellite velocities (ECEF), shape (n_obs, 3)
    sdT : satellite clock drifts (from broadcast/SP3), shape (n_obs,), unit: s/s
    sys : list of GNSS system identifiers for each observation, e.g., ['G', 'E', 'C']
    enable_torch : if True, use PyTorch backend for computations
    device : if enable_torch is True, specifies the device ('cpu' or 'cuda')

    Returns:
    v : predicted Doppler velocity residuals (m/s), shape (n_obs, 1)
    H : Jacobian matrix w.r.t state [vx, vy, vz, dT], shape (n_obs, 4)
    """
    backend = Backend(enable_torch)

    # 转换所有输入为 float64 并移动到 device
    dT = backend.asarray(dT, dtype=backend.float64)
    pos = backend.asarray(pos, dtype=backend.float64)
    satpos = backend.asarray(satpos, dtype=backend.float64)
    satvel = backend.asarray(satvel, dtype=backend.float64)
    sdT = backend.asarray(sdT, dtype=backend.float64)

    vel = backend.to(vel, device)
    dT = backend.to(dT, device)
    pos = backend.to(pos, device)
    satpos = backend.to(satpos, device)
    satvel = backend.to(satvel, device)
    sdT = backend.to(sdT, device)

    # 计算单位向量 e
    diff = satpos - pos  # shape: (n_obs, 3)
    norm = backend.linalg_norm(diff, axis=1).reshape(-1, 1)  # shape: (n_obs, 1)
    e = -diff / norm  # shape: (n_obs, 3)

    # 几何相对速度项
    rel_vel = satvel - vel  # shape: (n_obs, 3)
    v1 = -backend.sum(e * rel_vel, axis=1)  # shape: (n_obs,)

    # Sagnac 修正项
    v2 = prl.OMGE / prl.CLIGHT * (
        satvel[:, 1] * pos[0] - satvel[:, 0] * pos[1] +
        satpos[:, 1] * vel[0] - satpos[:, 0] * vel[1]
    )  # shape: (n_obs,)

    # 钟差项
    v_clock = dT - prl.CLIGHT * sdT  # shape: (n_obs,)

    # 总 Doppler 预测值
    v = v1 + v2 + v_clock  # shape: (n_obs,)
    v = v.reshape(-1, 1)  # shape: (n_obs, 1)


    # 构造 Jacobian H: [∂v/∂vx, ∂v/∂vy, ∂v/∂vz, ∂v/∂dT]
    H_t = backend.ones(len(e), dtype=backend.float64).reshape(-1, 1)  # shape: (n_obs, 1)
    H_t = backend.to(H_t, device)
    H = backend.hstack((e, H_t))  # shape: (n_obs, 4)

    return v, H

ecef_to_enu_direct(satpos, recv_pos)

Convert satellite ECEF coordinates to receiver ENU coordinate system. Parameters: satpos : ndarray Satellite ECEF coordinates (n, 3) recv_pos : ndarray Receiver ECEF coordinates (1, 3) Returns: enu : ndarray Satellite ENU coordinates (n, 3)

Source code in tasgnss/core.py 
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
def ecef_to_enu_direct(satpos, recv_pos):
    """
    Convert satellite ECEF coordinates to receiver ENU coordinate system.
    Parameters:
        satpos : ndarray
            Satellite ECEF coordinates (n, 3)
        recv_pos : ndarray
            Receiver ECEF coordinates (1, 3)
    Returns:
        enu : ndarray
            Satellite ENU coordinates (n, 3)
    """
    # Step 1: Calculate the latitude and longitude of the receiver
    lat, lon, _ = p3d.ecef2geodetic(recv_pos[0], recv_pos[1], recv_pos[2])
    # Step 2: Calculate the relative vector from satellite to receiver (u, v, w)
    rel_pos = satpos - recv_pos  # (n, 3)
    # Step 3: Construct the rotation matrix from ECEF to ENU
    sin_lat, cos_lat = np.sin(np.radians(lat)), np.cos(np.radians(lat))
    sin_lon, cos_lon = np.sin(np.radians(lon)), np.cos(np.radians(lon))
    R = np.array([
        [-sin_lon, cos_lon, 0],
        [-sin_lat * cos_lon, -sin_lat * sin_lon, cos_lat],
        [cos_lat * cos_lon, cos_lat * sin_lon, sin_lat]
    ])
    # Step 4: Batch convert to ENU coordinates
    enu = rel_pos @ R.T  # Matrix multiplication
    return enu

enu_to_azel(enu, degree=False)

Convert ENU coordinates to azimuth and elevation angles. Parameters: enu : ndarray ENU coordinates (n, 3), each row represents [E, N, U] Returns: azimuth : ndarray Azimuth angle array (n,) elevation : ndarray Elevation angle array (n,)

Source code in tasgnss/core.py 
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
def enu_to_azel(enu, degree=False):
    """
    Convert ENU coordinates to azimuth and elevation angles.
    Parameters:
        enu : ndarray
            ENU coordinates (n, 3), each row represents [E, N, U]
    Returns:
        azimuth : ndarray
            Azimuth angle array (n,)
        elevation : ndarray
            Elevation angle array (n,)
    """
    E, N, U = enu[:, 0], enu[:, 1], enu[:, 2]

    # Calculate azimuth
    azimuth = np.arctan2(E, N)  # Return value range [-pi, pi]
    #azimuth = np.degrees(azimuth)  # Convert to degrees
    azimuth = (azimuth + 2*np.pi) % (2*np.pi)  # Ensure range in [0, 360]

    # Calculate elevation
    horizontal_distance = np.sqrt(E**2 + N**2)  # Horizontal distance
    elevation = np.arctan2(U, horizontal_distance)  # Return value range [-pi/2, pi/2]
    #elevation = np.degrees(elevation)  # Convert to degrees
    if not degree:
        return azimuth, elevation
    else:
        return np.degrees(azimuth), np.degrees(elevation)

get_atmosphere_error(gtime, satpos, satprns, nav, p)

Computes modeled atmospheric delay errors (ionospheric, tropospheric) and associated observation variances for a set of satellites at a given epoch. Designed for use in GNSS positioning and quality control — especially in multipath-prone environments like urban canyons.

This function uses RTKLIB’s built-in models:

Ionosphere: IONOOPT_BRDC (broadcast Klobuchar model) Troposphere: TROPOPT_SAAS (Saastamoinen model)

gtime (gtime_t): Epoch time in RTKLIB’s internal time format.

satpos (List[ArrayLike] or ndarray of shape (n, 6)): Satellite positions and velocities for each satellite, packed as [x, y, z, vx, vy, vz] (ECEF, meters and m/s).

satprns (List[int]): List of satellite PRN numbers (e.g., [1, 5, 12, 19]).

nav (nav_t): Navigation data structure containing ionospheric/tropospheric model parameters (e.g., broadcast iono coeffs).

p (ArrayLike, length=3): Receiver approximate position in ECEF coordinates [X, Y, Z] (meters). Used to compute elevation/azimuth and atmospheric delays.

iono_error (ndarray, shape=(n,)): The modeled ionospheric delay per satellite: dion (meters).

trop_error (ndarray, shape=(n,)): The modeled tropospheric delay per satellite: dtrp (meters).

var_el (ndarray, shape=(n,)): Elevation-dependent variance (empirical model, e.g., for multipath suppression in urban canyons). Computed via RTKLIBvar(azel[1], sys).

var_iono (ndarray, shape=(n,)): Broadcast ionospheric model variance (squared standard deviation, m²), output from ionocorr.

var_tropo (ndarray, shape=(n,)): Saastamoinen tropospheric model variance (m²), output from tropcorr.

Processing Steps

Converts receiver ECEF position p to geodetic coordinates (pos) using ecef2pos. For each satellite: Computes line-of-sight vector and satellite elevation/azimuth using geodist and satazel. Calculates ionospheric delay (dion) and its variance (vion) via ionocorr(..., IONOOPT_BRDC, ...). Calculates tropospheric delay (dtrp) and its variance (vtrp) via tropcorr(..., TROPOPT_SAAS, ...). Computes elevation-based variance (vel) using RTKLIBvar(elevation, system) — useful for downweighting low-elevation satellites. Returns arrays of total atmospheric error and component variances.

Source code in tasgnss/core.py 
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
def get_atmosphere_error(gtime, satpos, satprns, nav, p):
    """
    Computes modeled atmospheric delay errors (ionospheric, tropospheric) and associated observation variances for a set of satellites at a given epoch.
    Designed for use in GNSS positioning and quality control — especially in multipath-prone environments like urban canyons.

    Note: This function uses RTKLIB’s built-in models:
        Ionosphere: IONOOPT_BRDC (broadcast Klobuchar model)
        Troposphere: TROPOPT_SAAS (Saastamoinen model)

    Parameters:
    gtime (gtime_t):
        Epoch time in RTKLIB’s internal time format.

    satpos (List[ArrayLike] or ndarray of shape (n, 6)):
        Satellite positions and velocities for each satellite, packed as [x, y, z, vx, vy, vz] (ECEF, meters and m/s).

    satprns (List[int]):
        List of satellite PRN numbers (e.g., [1, 5, 12, 19]).

    nav (nav_t):
        Navigation data structure containing ionospheric/tropospheric model parameters (e.g., broadcast iono coeffs).

    p (ArrayLike, length=3):
        Receiver approximate position in ECEF coordinates [X, Y, Z] (meters). Used to compute elevation/azimuth and atmospheric delays.

    Returns:
    iono_error (ndarray, shape=(n,)):
        The modeled ionospheric delay per satellite: dion (meters).

    trop_error (ndarray, shape=(n,)):
        The modeled tropospheric delay per satellite: dtrp (meters).

    var_el (ndarray, shape=(n,)):
        Elevation-dependent variance (empirical model, e.g., for multipath suppression in urban canyons). Computed via RTKLIBvar(azel[1], sys).

    var_iono (ndarray, shape=(n,)):
        Broadcast ionospheric model variance (squared standard deviation, m²), output from ionocorr.

    var_tropo (ndarray, shape=(n,)):
        Saastamoinen tropospheric model variance (m²), output from tropcorr.

    Processing Steps:
        Converts receiver ECEF position p to geodetic coordinates (pos) using ecef2pos.
        For each satellite:
            Computes line-of-sight vector and satellite elevation/azimuth using geodist and satazel.
            Calculates ionospheric delay (dion) and its variance (vion) via ionocorr(..., IONOOPT_BRDC, ...).
            Calculates tropospheric delay (dtrp) and its variance (vtrp) via tropcorr(..., TROPOPT_SAAS, ...).
            Computes elevation-based variance (vel) using RTKLIBvar(elevation, system) — useful for downweighting low-elevation satellites.
        Returns arrays of total atmospheric error and component variances.
    """
    n = len(satprns)
    satsys = get_list_sat_name(satprns,True)
    e = prl.Arr1Ddouble(3)
    rr = prl.Arr1Ddouble(3)
    rr[0] = p[0]
    rr[1] = p[1]
    rr[2] = p[2]
    azel = prl.Arr1Ddouble(2)
    pos = prl.Arr1Ddouble(3)
    prl.ecef2pos(rr,pos)
    dion = prl.Arr1Ddouble(1)
    vion = prl.Arr1Ddouble(1)
    dtrp = prl.Arr1Ddouble(1)
    vtrp = prl.Arr1Ddouble(1)

    vels = []
    vions = []
    vtrps = []
    ion_err = []
    trop_err = []
    sysname = prl.Arr1Dchar(4)

    for i in range(n):
        sp = make1Darray(satpos[i],prl.Arr1Ddouble)
        prl.geodist(sp,rr,e)
        prl.satazel(pos,e,azel)
        prl.ionocorr(gtime,nav,satprns[i],pos,azel,prl.IONOOPT_BRDC,dion,vion)
        prl.tropcorr(gtime,nav,pos,azel,prl.TROPOPT_SAAS,dtrp,vtrp)
        vel = RTKLIBvar(azel[1],satsys[i])
        vels.append(vel)
        vions.append(vion.ptr)
        vtrps.append(vtrp.ptr)
        ion_err.append(dion.ptr)
        trop_err.append(dtrp.ptr)
    return np.array(ion_err),np.array(trop_err), np.array(vels), np.array(vions),np.array(vtrps)

get_obs_pnt(obs, nav, prcopt=None)

Performs Single Point Positioning (SPP) by directly calling RTKLIB’s pntpos() function. Returns the computed position/velocity solution, success/failure status, and diagnostic message.

Ideal for quick, standalone positioning without filters or ambiguity resolution.

obs (obs_t): GNSS observation data structure for one epoch (must contain data[0..n-1] of type obsd_t).

nav (nav_t): Navigation data structure with ephemerides, ionospheric/tropospheric models, and satellite biases.

prcopt (prcopt_t, optional): Processing options (e.g., iono/tropo model, elevation mask, positioning mode). If None, defaults to prl.prcopt_default.

sol (sol_t): Solution structure containing: sol.rr[0:3]: ECEF position [x, y, z] (meters) sol.rr[3:6]: ECEF velocity [vx, vy, vz] (m/s) — if computed sol.time: Epoch time Other metadata (refer to RTKLIB documentation for full details).

status (bool): True → Positioning succeeded (solution is valid) False → Positioning failed (check msg for reason)

msg (str): Diagnostic message from RTKLIB. Common failure reasons: "insufficient satellites" "gdop error" (GDOP too high) "chi-square error" (residual validation failed) "no navigation data"

Processing Flow

Initializes solution and satellite status buffers. Sets solution time to match first observation. Calls RTKLIB’s pntpos() — computes SPP using pseudoranges, applies models (iono/tropo/dcbs), solves least-squares. Returns raw result — no post-filtering or smoothing.

Source code in tasgnss/core.py 
 956
 957
 958
 959
 960
 961
 962
 963
 964
 965
 966
 967
 968
 969
 970
 971
 972
 973
 974
 975
 976
 977
 978
 979
 980
 981
 982
 983
 984
 985
 986
 987
 988
 989
 990
 991
 992
 993
 994
 995
 996
 997
 998
 999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
def get_obs_pnt(obs, nav, prcopt=None):
    """
    Performs Single Point Positioning (SPP) by directly calling RTKLIB’s pntpos() function.
    Returns the computed position/velocity solution, success/failure status, and diagnostic message.

    Ideal for quick, standalone positioning without filters or ambiguity resolution.

    Parameters:
    obs (obs_t):
        GNSS observation data structure for one epoch (must contain data[0..n-1] of type obsd_t).

    nav (nav_t):
        Navigation data structure with ephemerides, ionospheric/tropospheric models, and satellite biases.

    prcopt (prcopt_t, optional):
        Processing options (e.g., iono/tropo model, elevation mask, positioning mode).
        If None, defaults to prl.prcopt_default.

    Returns:
    sol (sol_t):
        Solution structure containing:
            sol.rr[0:3]: ECEF position [x, y, z] (meters)
            sol.rr[3:6]: ECEF velocity [vx, vy, vz] (m/s) — if computed
            sol.time: Epoch time
            Other metadata (refer to RTKLIB documentation for full details).

    status (bool):
        True → Positioning succeeded (solution is valid)
        False → Positioning failed (check msg for reason)

    msg (str):
        Diagnostic message from RTKLIB. Common failure reasons:
            "insufficient satellites"
            "gdop error" (GDOP too high)
            "chi-square error" (residual validation failed)
            "no navigation data"

    Processing Flow:
        Initializes solution and satellite status buffers.
        Sets solution time to match first observation.
        Calls RTKLIB’s pntpos() — computes SPP using pseudoranges, applies models (iono/tropo/dcbs), solves least-squares.
        Returns raw result — no post-filtering or smoothing.
    """
    if prcopt is None:
        prcopt = prl.prcopt_default
    m = obs.n
    sol = prl.sol_t()
    sat = prl.Arr1Dssat_t(prl.MAXSAT)
    sol.time = obs.data[0].time
    msg = prl.Arr1Dchar(100)
    azel = prl.Arr1Ddouble(m*2)
    prl.pntpos(obs.data.ptr,obs.n,nav,prcopt,sol,azel,sat.ptr,msg)
    if msg.ptr and "chi-square error" not in msg.ptr and "gdop error" not in msg.ptr:
        return sol,False,msg.ptr
    else:
        return sol,True,msg.ptr

get_sagnac_corr(satpos, p)

Computes the Sagnac correction (relativistic range correction due to Earth’s rotation) for GNSS positioning. This effect arises because the Earth rotates during signal propagation, causing a relative motion between satellite and receiver in the ECEF frame. The magnitude is typically ~3 meters and must be corrected for precise positioning.

Note: The receiver position p does not need to be highly accurate — even an error of ~100 meters introduces negligible change in the Sagnac correction.

satpos (ndarray, shape=(n, 3) or (n, 6)): Satellite positions in ECEF coordinates [X, Y, Z] (meters). If 6-element vectors are passed (including velocity), only the first three are used.

p (ArrayLike, length=3): Approximate receiver position in ECEF coordinates [X, Y, Z] (meters). Accuracy requirement: ~100 m is sufficient.

sagnac_corr (ndarray, shape=(n,)): Sagnac correction in meters, one value per satellite.

Why It Matters

The Sagnac effect is a relativistic correction that accounts for the fact that the ECEF frame is rotating. As the signal travels from satellite to receiver (~0.07s), the Earth rotates slightly, causing a geometric discrepancy if positions are treated as static in ECEF. This correction ensures consistency with the inertial frame assumption in GNSS signal models.

Source code in tasgnss/core.py 
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
def get_sagnac_corr(satpos, p):
    """
    Computes the Sagnac correction (relativistic range correction due to Earth’s rotation) for GNSS positioning.
    This effect arises because the Earth rotates during signal propagation, causing a relative motion between satellite and receiver in the ECEF frame.
    The magnitude is typically ~3 meters and must be corrected for precise positioning.

    Note: The receiver position p does not need to be highly accurate — even an error of ~100 meters introduces negligible change in the Sagnac correction.

    Parameters:
    satpos (ndarray, shape=(n, 3) or (n, 6)):
        Satellite positions in ECEF coordinates [X, Y, Z] (meters). If 6-element vectors are passed (including velocity), only the first three are used.

    p (ArrayLike, length=3):
        Approximate receiver position in ECEF coordinates [X, Y, Z] (meters). Accuracy requirement: ~100 m is sufficient.

    Returns:
    sagnac_corr (ndarray, shape=(n,)):
        Sagnac correction in meters, one value per satellite.

    Why It Matters:
        The Sagnac effect is a relativistic correction that accounts for the fact that the ECEF frame is rotating.
        As the signal travels from satellite to receiver (~0.07s), the Earth rotates slightly, causing a geometric discrepancy if positions are treated as static in ECEF.
        This correction ensures consistency with the inertial frame assumption in GNSS signal models.
    """
    sagnac_corr = prl.OMGE*(satpos[:,0]*p[1]-satpos[:,1]*p[0])/prl.CLIGHT
    return sagnac_corr

get_sat_pos(obsd, n, nav)

Computes satellite positions, velocities, clock biases, and associated ephemeris variances for a given epoch of observations. Filters out satellites with invalid or missing ephemeris data and returns only valid entries.

Note: The input is obsd_t (i.e., obs.data), not obs_t. Typical usage: get_sat_pos(obs.data, obs.n, nav)

obsd (Arr1Dobsd_t or equivalent): Array of observation data for one epoch (e.g., obs.data). Must be contiguous and correspond to n satellites.

n (int): Number of satellites (i.e., number of elements in obsd) for this epoch.

nav (nav_t): Ephemeris and clock data structure, typically populated by read_obs() or equivalent.

rr (Arr1Ddouble, length = 6 * len(mask)): Satellite positions and velocities, packed as [x, y, z, vx, vy, vz] for each valid satellite. Units: meters and meters/second.

dts (Arr1Ddouble, length = 2 * len(mask)): Satellite clock bias and drift, packed as [bias, drift] for each valid satellite. Units: seconds and seconds/second.

var (Arr1Ddouble, length = len(mask)): Ephemeris variance (squared standard deviation) for each valid satellite. Unit: m².

mask (List[int]): List of indices (from original 0..n-1) corresponding to satellites with valid ephemeris data. Used to map back to original observation order.

Processing Details: Internally calls RTKLIB’s satposs() to compute satellite states. Identifies satellites with invalid position data (where |x| < 1e-10) as “no ephemeris”. Constructs a mask of valid satellite indices. Uses helper function arr_select to filter rs, dts, and var arrays according to the mask. Returns filtered arrays and mask.

Source code in tasgnss/core.py 
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
def get_sat_pos(obsd, n, nav):
    """
    Computes satellite positions, velocities, clock biases, and associated ephemeris variances for a given epoch of observations.
    Filters out satellites with invalid or missing ephemeris data and returns only valid entries.

    Note: The input is obsd_t (i.e., obs.data), not obs_t.
    Typical usage: get_sat_pos(obs.data, obs.n, nav)

    Parameters:
    obsd (Arr1Dobsd_t or equivalent):
        Array of observation data for one epoch (e.g., obs.data). Must be contiguous and correspond to n satellites.

    n (int):
        Number of satellites (i.e., number of elements in obsd) for this epoch.

    nav (nav_t):
        Ephemeris and clock data structure, typically populated by read_obs() or equivalent.

    Returns:
    rr (Arr1Ddouble, length = 6 * len(mask)):
        Satellite positions and velocities, packed as [x, y, z, vx, vy, vz] for each valid satellite. Units: meters and meters/second.

    dts (Arr1Ddouble, length = 2 * len(mask)):
        Satellite clock bias and drift, packed as [bias, drift] for each valid satellite. Units: seconds and seconds/second.

    var (Arr1Ddouble, length = len(mask)):
        Ephemeris variance (squared standard deviation) for each valid satellite. Unit: m².

    mask (List[int]):
        List of indices (from original 0..n-1) corresponding to satellites with valid ephemeris data. Used to map back to original observation order.

    Processing Details:
    Internally calls RTKLIB’s satposs() to compute satellite states.
    Identifies satellites with invalid position data (where |x| < 1e-10) as “no ephemeris”.
    Constructs a mask of valid satellite indices.
    Uses helper function arr_select to filter rs, dts, and var arrays according to the mask.
    Returns filtered arrays and mask.
    """
    svh = prl.Arr1Dint(prl.MAXOBS)
    rs = prl.Arr1Ddouble(6*n)
    dts = prl.Arr1Ddouble(2*n)
    var = prl.Arr1Ddouble(1*n)
    prl.satposs(obsd[0].time,obsd.ptr,n,nav,0,rs,dts,var,svh)
    noeph = []
    for i in range(n):
        if abs(rs[6*i]) < 1e-10:
            noeph.append(i)
    mask = list(set(range(n))-set(noeph))
    nrs = arr_select(rs,mask,6)
    var = arr_select(var,mask)
    ndts = arr_select(dts,mask,2)
    nrs = np.array(nrs).reshape(-1,6)
    ndts = np.array(ndts).reshape(-1,2)
    var = np.array(var)
    return nrs,ndts,var,mask

huber_kernel_factory(k=1.345, scale=1.0, enable_torch=False)

Factory function to create a Huber robust kernel function with configurable parameters.

The Huber kernel downweights large residuals to provide robustness against outliers.

k (float): Tuning constant that determines the transition point between quadratic and linear behavior. Typical values range from 1.0 to 2.0. Default is 1.345 (commonly used value). scale (float): Scale factor for the residuals. Can be used to normalize the residuals. Default is 1.0.

Returns: callable: A Huber kernel function that takes residuals as input and returns weights.

Source code in tasgnss/core.py 
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
def huber_kernel_factory(k=1.345, scale=1.0, enable_torch=False):
    """
    Factory function to create a Huber robust kernel function with configurable parameters.

    The Huber kernel downweights large residuals to provide robustness against outliers.

    Parameters:
    k (float): Tuning constant that determines the transition point between quadratic and linear behavior.
               Typical values range from 1.0 to 2.0. Default is 1.345 (commonly used value).
    scale (float): Scale factor for the residuals. Can be used to normalize the residuals.
                   Default is 1.0.

    Returns:
    callable: A Huber kernel function that takes residuals as input and returns weights.
    """
    def huber_kernel(residuals, enable_torch=enable_torch):
        """
        Huber robust kernel function.

        Parameters:
        residuals (array-like): Input residuals

        Returns:
        array-like: Weights for each residual
        """
        backend = Backend(enable_torch)
        residuals = backend.asarray(residuals)
        scaled_residuals = backend.abs(residuals) / scale

        # Huber weighting function: w(r) = 1 if |r| <= k, else k/|r|
        weights = backend.where(scaled_residuals <= k, 
                               backend.ones_like(scaled_residuals),
                               k / (scaled_residuals+1e-12))

        return weights.squeeze()

    return huber_kernel

irls_pnt_pos(o, nav, use_cache=True, return_residual=False, enable_torch=False, robust_kernel=None, b=None, device='cpu')

Performs Iterative Reweighted Least Squares (IRLS) positioning using GNSS observations with robust estimation.

This function implements an iterative IRLS algorithm with robust kernel function to solve for receiver position, velocity, and clock parameters. It supports caching for performance optimization and PyTorch backend for gradient-based optimization.

Key Features: - Uses robust kernel function for outlier rejection - Uses caching to accelerate repeated calls with the same observation data - Supports PyTorch backend for gradient propagation, enabling use in neural network optimization - Handles multiple GNSS constellations with separate clock bias parameters - Includes atmospheric and relativistic corrections

o (obs_t): GNSS observation data structure for one epoch. nav (nav_t): Navigation data structure with ephemerides and satellite biases. use_cache (bool, optional): Whether to use cached preprocessing results. When True, if the same observation object is processed multiple times, the preprocessing results (satellite positions, atmospheric corrections, etc.) are cached and reused, significantly speeding up repeated calls. Default is True. return_residual (bool, optional): Whether to return residuals, Jacobian matrix, and weight matrix. When True, the returned dictionary includes a "residual_info" key containing: - "residual": Observation residuals vector - "H": Design matrix (Jacobian) - "W": Weight matrix These can be used to compute Dilution of Precision (DOP) metrics. Default is False. enable_torch (bool, optional): Whether to use PyTorch backend for computations. When True, the function uses PyTorch tensors and operations, allowing gradients to flow through the computation graph. This enables the use of this function in neural network training, where weights (w) and bias (b) can be optimized using gradient descent. Default is False. robust_kernel (callable, optional): Robust kernel function that takes residuals as input and returns weights. Function signature: weights = kernel(residuals) If None, uses standard least squares without robust weighting. Default is None. b (array-like, optional): Bias vector to be subtracted from pseudorange observations. If enable_torch=True, gradients will propagate through b to the position solution, allowing b to be optimized in neural networks. Default is None (zero vector). device (str, optional): Device to run PyTorch computations on ('cpu' or 'cuda'). Default is 'cpu'.

dict: A dictionary containing positioning results, status, and additional information with keys: - "status" (bool): True if positioning succeeded, False otherwise - "pos" (array): Receiver position [x, y, z] in ECEF coordinates - "cb" (array): Receiver clock bias for each GNSS system - "cd" (array): Receiver clock drift - "msg" (str): Status message - "data" (array): Processed observation data - "solve_data" (dict): Preprocessed data used in solving - "raw_data" (dict): Raw observation data - "residual_info" (dict, optional): Residuals and Jacobian matrix if return_residual=True

Source code in tasgnss/core.py 
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
def irls_pnt_pos(o, nav, use_cache=True, return_residual=False, enable_torch=False, 
                 robust_kernel=None, b=None, device='cpu'):
    """
    Performs Iterative Reweighted Least Squares (IRLS) positioning using GNSS observations with robust estimation.

    This function implements an iterative IRLS algorithm with robust kernel function to solve for receiver position, 
    velocity, and clock parameters. It supports caching for performance optimization and PyTorch backend
    for gradient-based optimization.

    Key Features:
    - Uses robust kernel function for outlier rejection
    - Uses caching to accelerate repeated calls with the same observation data
    - Supports PyTorch backend for gradient propagation, enabling use in neural network optimization
    - Handles multiple GNSS constellations with separate clock bias parameters
    - Includes atmospheric and relativistic corrections

    Parameters:
    o (obs_t): 
        GNSS observation data structure for one epoch.
    nav (nav_t): 
        Navigation data structure with ephemerides and satellite biases.
    use_cache (bool, optional): 
        Whether to use cached preprocessing results. When True, if the same
        observation object is processed multiple times, the preprocessing results (satellite positions,
        atmospheric corrections, etc.) are cached and reused, significantly speeding up repeated calls.
        Default is True.
    return_residual (bool, optional): 
        Whether to return residuals, Jacobian matrix, and weight matrix.
        When True, the returned dictionary includes a "residual_info" key containing:
        - "residual": Observation residuals vector
        - "H": Design matrix (Jacobian)
        - "W": Weight matrix
        These can be used to compute Dilution of Precision (DOP) metrics. Default is False.
    enable_torch (bool, optional): 
        Whether to use PyTorch backend for computations. When True,
        the function uses PyTorch tensors and operations, allowing gradients to flow through the
        computation graph. This enables the use of this function in neural network training, where
        weights (w) and bias (b) can be optimized using gradient descent. Default is False.
    robust_kernel (callable, optional): 
        Robust kernel function that takes residuals as input and returns weights.
        Function signature: weights = kernel(residuals)
        If None, uses standard least squares without robust weighting. Default is None.
    b (array-like, optional): Bias vector to be subtracted from pseudorange observations. If
        enable_torch=True, gradients will propagate through b to the position solution, allowing
        b to be optimized in neural networks. Default is None (zero vector).
    device (str, optional): Device to run PyTorch computations on ('cpu' or 'cuda'). Default is 'cpu'.

    Returns:
    dict: A dictionary containing positioning results, status, and additional information with keys:
        - "status" (bool): True if positioning succeeded, False otherwise
        - "pos" (array): Receiver position [x, y, z] in ECEF coordinates
        - "cb" (array): Receiver clock bias for each GNSS system
        - "cd" (array): Receiver clock drift
        - "msg" (str): Status message
        - "data" (array): Processed observation data
        - "solve_data" (dict): Preprocessed data used in solving
        - "raw_data" (dict): Raw observation data
        - "residual_info" (dict, optional): Residuals and Jacobian matrix if return_residual=True
    """
    maxiter = 100
    o_id = id(o)
    if use_cache and o_id in cache_data:
        p, p_t, v, v_t, data, cdata, raw_data = cache_data[o_id]
    else:
        p, p_t, v, v_t, data, cdata, raw_data = preprocess_obs(o, nav, use_cache)

    # test cdata['sys'], check the number of systems and the number of unknowns
    if np.unique(cdata['sys']).shape[0] + 3 > cdata['pr'].shape[0]:
        return {
            "status": False,
            "pos": np.zeros(4),
            "msg": "insufficient satellites for the number of systems",
            "data": {},
            "solve_data": cdata,
            "raw_data": raw_data
        }


    iter = 0

    backend = Backend(enable_torch)

    def ensure_array(x, shape, dtype=backend.float64):
        if x is None:
            return backend.zeros(shape, dtype=dtype)
        return backend.asarray(x, dtype=dtype).reshape(shape)

    # initalize variables
    p = ensure_array(p, (3,))
    p_t = ensure_array(p_t, (len(SYS_NAME),))


    # ensure all variables are on the correct device
    p = backend.to(p, device)
    p_t = backend.to(p_t, device)

    dp = backend.asarray(np.array(100.0), dtype=backend.float64)
    dp = backend.to(dp, device)

    # construct p_t_mask
    idx = backend.array([list(SYS_NAME).index(s) for s in data[:,2]], dtype=backend.int64)  # 索引用 int64
    eye_matrix = backend.eye(len(SYS_NAME), dtype=backend.float64)
    p_t_mask = eye_matrix[idx] 
    p_t_mask = backend.to(p_t_mask, device)

    # process b
    if b is None:
        b = backend.zeros_like(backend.asarray(cdata['pr']), dtype=backend.float64)
    else:
        b = backend.asarray(b, dtype=backend.float64).reshape(-1,1)
    b = backend.to(b, device)

    # Initialize base weight matrix using observation variance
    var = backend.asarray(cdata['var'], dtype=backend.float64)
    W_base = backend.diag(1.0 / var)

    # ensure cdata pr and dop are tensors on the correct device
    pr_tensor = backend.asarray(cdata['pr'], dtype=backend.float64)
    pr_tensor = backend.to(pr_tensor, device)

    try:
        while iter < maxiter and backend.linalg_norm(dp) > 0.001:
            psr, H = pseudorange_observe_func(
                p, backend.dot(p_t_mask, p_t), cdata['satpos'][:,:3], cdata["sdt"][:,0],
                cdata['I'], cdata['T'], cdata['sagnac'], cdata['sys'], enable_torch=enable_torch,
                device=device
            )

            p_residual = pr_tensor - psr - b
            residual = p_residual

            # Apply robust kernel function if provided
            if robust_kernel is not None:
                # Compute robust weights from residuals
                robust_weights = robust_kernel(p_residual)
                # Combine with base weights
                W = backend.diag(robust_weights)
            else:
                W = W_base

            # solve least squares
            W_H = backend.dot(W, H)
            W_r = backend.dot(W, residual)
            result = backend.linalg_lstsq(W_H, W_r, rcond=None)
            dp = result.solution if hasattr(result, 'solution') else result[0]

            # update
            p = p + backend.squeeze(dp[:3])
            p_t = p_t + backend.squeeze(dp[3:])
            iter += 1

    except Exception as e:
        zero_pos = backend.zeros(4, dtype=backend.float64)
        zero_pos = backend.to(zero_pos, device)
        return {
            "status": False,
            "pos": zero_pos,
            "msg": str(e),
            "data": {},
            "solve_data": cdata,
            "raw_data": raw_data
        }

    if iter > maxiter or backend.linalg_norm(dp) > 1e-2:
        zero_pos = backend.zeros(4, dtype=backend.float64)
        zero_pos = backend.to(zero_pos, device)
        return {
            "status": False,
            "pos": zero_pos,
            "msg": "not converge",
            "data": {},
            "solve_data": cdata,
            "raw_data": raw_data
        }

    # store to cache
    cache_data[o_id][0] = backend.to_numpy(p)
    cache_data[o_id][1] = backend.to_numpy(p_t)


    # optional: return residuals and H
    if return_residual:
        psr, H = pseudorange_observe_func(
            p, backend.dot(p_t_mask, p_t), cdata['satpos'][:,:3], cdata["sdt"][:,0],
            cdata['I'], cdata['T'], cdata['sagnac'], cdata['sys'], enable_torch=enable_torch
        )

        pr_tensor = backend.asarray(cdata['pr'], dtype=backend.float64)
        pr_tensor = backend.to(pr_tensor, device)

        residual = pr_tensor - psr - b
        residual_info = {
            "residual": residual,
            "H": H,
            "W": W
        }
    else:
        residual_info = {}

    return {
        "status": True,
        "pos": p,
        "cb": p_t,
        "msg": "ok",
        "data": data,
        "solve_data": cdata,
        "raw_data": raw_data,
        "residual_info": residual_info
    }

prange(obs, nav, opt, var)

Computes the corrected pseudorange for a single satellite observation, applying differential code bias (DCB) and ionospheric delay corrections based on configuration. Supports both single-frequency and ionosphere-free dual-frequency combinations.

obs (obsd_t): Single satellite observation record from an epoch (e.g., obs.data[i]). Must contain pseudorange measurements (P[0], P[1]) and signal codes (code[0], code[1]).

nav (nav_t): Ephemeris and satellite bias structure containing DCB (differential code bias) and TGD (time group delay) corrections.

opt (prcopt_t or equivalent): Processing options structure. Key field: opt.ionoopt: Specifies ionospheric correction mode (e.g., IONOOPT_IFLC for ionosphere-free linear combination).

var (Arr1Ddouble, length ≥ 1): Output array — var[0] is set to the default code variance (0.3² m²) for single-frequency cases. Not modified in dual-frequency mode.

p (float): Corrected pseudorange in meters. Returns 0.0 if: P1 is missing, or Dual-frequency mode is enabled but P2 is missing.

Processing Logic: DCB Correction (C1→P1, C2→P2): Applied for GPS and GLONASS if code type indicates C/A code (CODE_L1C or CODE_L2C). Uses satellite-specific biases from nav.cbias.

Ionosphere-Free Combination (if opt.ionoopt == IONOOPT_IFLC): Uses dual-frequency pseudoranges (P1, P2) to form ionosphere-free linear combination: p = (P2 - γ·P1) / (1 - γ) where γ = (f1/f2)² (frequency ratio squared).

For BeiDou and Galileo, additional TGD/BDG corrections are applied before combination.
System-specific frequency constants and bias models are used (GPS, GLO, GAL, CMP, IRN).

Single-Frequency Mode (default): Applies only TGD or BGD correction to P1. Sets var[0] = 0.3² (default code variance). System-specific TGD models applied (e.g., GPS TGD, GLO –dtaun, GAL BGD, etc.).

Supported GNSS

GPS / QZSS: L1-L2 (IFLC) or L1-only GLONASS: G1-G2 (IFLC) or G1-only Galileo: E1-E5b (IFLC) or E1-only (BGD applied) BeiDou: B1-B2 (IFLC) or B1I/B1Cp/B1Cd (TGD/ISC applied) NavIC (IRNSS): L5-S (IFLC) or L5-only

Source code in tasgnss/core.py 
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
def prange(obs, nav, opt, var):
    """
    Computes the corrected pseudorange for a single satellite observation, applying differential code bias (DCB) and ionospheric delay corrections based on configuration.
    Supports both single-frequency and ionosphere-free dual-frequency combinations.

    Parameters:
    obs (obsd_t):
        Single satellite observation record from an epoch (e.g., obs.data[i]). Must contain pseudorange measurements (P[0], P[1]) and signal codes (code[0], code[1]).

    nav (nav_t):
        Ephemeris and satellite bias structure containing DCB (differential code bias) and TGD (time group delay) corrections.

    opt (prcopt_t or equivalent):
        Processing options structure. Key field:
        opt.ionoopt: Specifies ionospheric correction mode (e.g., IONOOPT_IFLC for ionosphere-free linear combination).

    var (Arr1Ddouble, length ≥ 1):
        Output array — var[0] is set to the default code variance (0.3² m²) for single-frequency cases. Not modified in dual-frequency mode.

    Returns:
    p (float):
        Corrected pseudorange in meters. Returns 0.0 if:
        P1 is missing, or
        Dual-frequency mode is enabled but P2 is missing.

    Processing Logic:
    DCB Correction (C1→P1, C2→P2):
        Applied for GPS and GLONASS if code type indicates C/A code (CODE_L1C or CODE_L2C). Uses satellite-specific biases from nav.cbias.

    Ionosphere-Free Combination (if opt.ionoopt == IONOOPT_IFLC):
        Uses dual-frequency pseudoranges (P1, P2) to form ionosphere-free linear combination:
        p = (P2 - γ·P1) / (1 - γ)
        where γ = (f1/f2)² (frequency ratio squared).

        For BeiDou and Galileo, additional TGD/BDG corrections are applied before combination.
        System-specific frequency constants and bias models are used (GPS, GLO, GAL, CMP, IRN).

    Single-Frequency Mode (default):
        Applies only TGD or BGD correction to P1. Sets var[0] = 0.3² (default code variance).
        System-specific TGD models applied (e.g., GPS TGD, GLO –dtaun, GAL BGD, etc.).

    Supported GNSS:
        GPS / QZSS: L1-L2 (IFLC) or L1-only
        GLONASS: G1-G2 (IFLC) or G1-only
        Galileo: E1-E5b (IFLC) or E1-only (BGD applied)
        BeiDou: B1-B2 (IFLC) or B1I/B1Cp/B1Cd (TGD/ISC applied)
        NavIC (IRNSS): L5-S (IFLC) or L5-only
    """
    P1, P2, gamma, b1, b2 = 0.0, 0.0, 0.0, 0.0, 0.0
    var[0] = 0.0

    sat = obs.sat

    sys_name = prl.Arr1Dchar(4)
    prl.satno2id(sat,sys_name)
    sys = SYS[sys_name.ptr[0]]
    P1 = obs.P[0]
    P2 = obs.P[1]

    if P1 == 0.0 or (opt.ionoopt == prl.IONOOPT_IFLC and P2 == 0.0):
        return 0.0

    # P1-C1, P2-C2 DCB correction
    if sys == prl.SYS_GPS or sys == prl.SYS_GLO:
        if obs.code[0] == prl.CODE_L1C:
            P1 += nav.cbias[sat - 1,1]  # C1->P1
        if obs.code[1] == prl.CODE_L2C:
            P2 += nav.cbias[sat - 1,2]  # C2->P2

    if opt.ionoopt == prl.IONOOPT_IFLC:  # dual-frequency
        if sys == prl.SYS_GPS or sys == prl.SYS_QZS:  # L1-L2, G1-G2
            gamma = (prl.FREQ1 / prl.FREQ2) ** 2
            return (P2 - gamma * P1) / (1.0 - gamma)
        elif sys == prl.SYS_GLO:  # G1-G2
            gamma = (prl.FREQ1_GLO / prl.FREQ2_GLO) ** 2
            return (P2 - gamma * P1) / (1.0 - gamma)
        elif sys == prl.SYS_GAL:  # E1-E5b
            gamma = (prl.FREQ1 / prl.FREQ7) ** 2
            if prl.getseleph(prl.SYS_GAL):  # F/NAV
                P2 -= gettgd(sat, nav, 0) - gettgd(sat, nav, 1)  # BGD_E5aE5b
            return (P2 - gamma * P1) / (1.0 - gamma)
        elif sys == prl.SYS_CMP:  # B1-B2
            gamma = (((prl.FREQ1_CMP if obs.code[0] == prl.CODE_L2I else prl.FREQ1) / prl.FREQ2_CMP) ** 2)
            b1 = gettgd(sat, nav, 0) if obs.code[0] == prl.CODE_L2I else gettgd(sat, nav, 2) if obs.code[0] == prl.CODE_L1P else gettgd(sat, nav, 2) + gettgd(sat, nav, 4)  # TGD_B1I / TGD_B1Cp / TGD_B1Cp+ISC_B1Cd
            b2 = gettgd(sat, nav, 1)  # TGD_B2I/B2bI (m)
            return ((P2 - gamma * P1) - (b2 - gamma * b1)) / (1.0 - gamma)
        elif sys == prl.SYS_IRN:  # L5-S
            gamma = (prl.FREQ5 / prl.FREQ9) ** 2
            return (P2 - gamma * P1) / (1.0 - gamma)
    else:  # single-freq (L1/E1/B1)
        var[0] = 0.3 ** 2

        if sys == prl.SYS_GPS or sys == prl.SYS_QZS:  # L1
            b1 = gettgd(sat, nav, 0)  # TGD (m)
            return P1 - b1
        elif sys == prl.SYS_GLO:  # G1
            gamma = (prl.FREQ1_GLO / prl.FREQ2_GLO) ** 2
            b1 = gettgd(sat, nav, 0)  # -dtaun (m)
            return P1 - b1 / (gamma - 1.0)
        elif sys == prl.SYS_GAL:  # E1
            b1 = gettgd(sat, nav, 0) if prl.getseleph(prl.SYS_GAL) else gettgd(sat, nav, 1)  # BGD_E1E5a / BGD_E1E5b
            return P1 - b1
        elif sys == prl.SYS_CMP:  # B1I/B1Cp/B1Cd
            b1 = gettgd(sat, nav, 0) if obs.code[0] == prl.CODE_L2I else gettgd(sat, nav, 2) if obs.code[0] == prl.CODE_L1P else gettgd(sat, nav, 2) + gettgd(sat, nav, 4)  # TGD_B1I / TGD_B1Cp / TGD_B1Cp+ISC_B1Cd
            return P1 - b1
        elif sys == prl.SYS_IRN:  # L5
            gamma = (prl.FREQ9 / prl.FREQ5) ** 2
            b1 = gettgd(sat, nav, 0)  # TGD (m)
            return P1 - gamma * b1
    return P1

preprocess_obs(o, nav, use_cache=True)

Preprocesses GNSS observation data for positioning.

Parameters: o (obs_t): GNSS observation data structure for one epoch. nav (nav_t): Navigation data structure with ephemerides and satellite biases. use_cache (bool, optional): Whether to use cached results if available.

Returns: dict: A dictionary containing preprocessing results and status information.

Source code in tasgnss/core.py 
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
def preprocess_obs(o, nav, use_cache=True):
    """
    Preprocesses GNSS observation data for positioning.

    Parameters:
    o (obs_t): GNSS observation data structure for one epoch.
    nav (nav_t): Navigation data structure with ephemerides and satellite biases.
    use_cache (bool, optional): Whether to use cached results if available.

    Returns:
    dict: A dictionary containing preprocessing results and status information.
    """
    o_id = id(o)
    if o.n < 4:
            return [[None] * 7][0]
    # intialize with pntpos can help fix some position-related parameters, such as sagnac, ionosphere, troposphere, or it need a iteration to solve it.
    sol, status, msg = get_obs_pnt(o,nav)
    # if not status:
    #     return [[None] * 7][0]
    p = np.array([sol.rr[0],sol.rr[1],sol.rr[2]])
    time = o.data[0].time
    satpos,sdt,var,mask = get_sat_pos(o.data,o.n,nav)
    opt = prl.prcopt_default
    vmeas = prl.Arr1Ddouble(1)
    o.data = arr_select(o.data,mask)
    o.n = len(mask)
    o.nmax = o.n
    data = []
    raw_datas = []
    sagnac = get_sagnac_corr(satpos[:,:3], p)
    iono_error, trop_error, var_els, var_ions, var_tropos = get_atmosphere_error(time,satpos[:,:3], [o.data[i].sat for i in range(o.n)], nav, p)
    satpos_enu = ecef_to_enu_direct(satpos[:,:3],p)
    az,el = enu_to_azel(satpos_enu)
    for i in range(o.n):
        obsd = o.data[i]
        # store raw data
        raw_data = {}
        raw_data['P'] = np.array(obsd.P[0:3])
        raw_data['L'] = np.array(obsd.L[0:3])
        raw_data['D'] = np.array(obsd.D[0:3])
        raw_data['SNR'] = np.array(obsd.SNR[0:3])/1000
        raw_data['LLI'] = np.array(obsd.LLI[0:3])
        raw_data['code'] = np.array(obsd.code[0:3])

        sname = get_sat_name(obsd.sat)
        s_sys = sname[0]

        corrected_p = prange(obsd,nav,opt,vmeas)
        dop = obsd.D[0]
        if corrected_p == 0.0:
            continue

        freq = prl.sat2freq(obsd.sat,obsd.code[0],nav)

        data.append(
            (
                obsd.sat,
                sname,
                s_sys,
                satpos[i],
                sdt[i],
                corrected_p,
                sagnac[i],
                iono_error[i],
                trop_error[i],
                obsd.SNR[0]/1000,
                var[i]+vmeas.ptr+var_els[i]+var_ions[i]+var_tropos[i],
                az[i],
                el[i],
                -prl.CLIGHT/freq*dop,
                freq
            )
        )
        raw_datas.append(raw_data)
    data = np.array(data,dtype=object)
    cdata = {
        'satpos': np.array(data[:,3].tolist()).astype(np.float64),
        'pr': data[:,5].astype(np.float64).reshape(-1,1),
        'dop': data[:,13].astype(np.float64).reshape(-1,1),
        'sdt': np.vstack(data[:,4]).astype(np.float64),
        'sagnac': data[:,6].astype(np.float64),
        'I': data[:,7].astype(np.float64),
        'T': data[:,8].astype(np.float64),
        'sys': data[:,2].tolist(),
        'var': data[:,10].astype(np.float64),
    }
    p = None
    p_t = None
    ret_data = [None,None,None,None,data,cdata,raw_datas]
    # if it's initialization, do not store the position and receiver clock bias
    if use_cache:
        cache_data[o_id] = ret_data
    return ret_data

pseudorange_observe_func(pos, dt, satpos, sdt, I, T, sagnac, sys, keep_states=True, enable_torch=False, device='cpu')

Computes pseudorange observation residuals and their Jacobian matrix for GNSS positioning. Models geometric range, Sagnac effect, receiver-satellite clock bias difference, and optional iono/tropo delays. The state vector assumed is: [x, y, z, dt_sys1, dt_sys2, ...] — position + per-system clock bias (no velocity or clock drift). Clock drift is handled externally (e.g., in Doppler module), enabling modular design via block_diag fusion.

Parameters: pos : receiver position (ECEF), shape (3,) dt : receiver clock bias per GNSS system, shape (n_sys,), unit: m → e.g., dt = [dt_GPS, dt_GAL, dt_BDS, ...] satpos : satellite positions (ECEF), shape (n_obs, 3) sdt : satellite clock biases (from broadcast/SP3), shape (n_obs,), unit: s I : ionospheric delay (optional, can be zero), shape (n_obs, 1), unit: m T : tropospheric delay (optional, can be zero), shape (n_obs, 1), unit: m sagnac : precomputed Sagnac correction term (from get_sagnac_corr), shape (n_obs, 1), unit: m sys : list of GNSS system identifiers for each observation, e.g., ['G', 'E', 'C'] keep_states : if False, removes Jacobian columns corresponding to systems with no observation (for WLS compatibility) enable_torch : if True, use PyTorch backend for computations device : if enable_torch is True, specifies the device ('cpu' or 'cuda')

Returns: psr : predicted pseudorange residuals (m), shape (n_obs, 1) H : Jacobian matrix w.r.t state [x, y, z, dt_sys1, dt_sys2, ...], shape (n_obs, 3 + n_active_sys)

Source code in tasgnss/core.py 
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
def pseudorange_observe_func(pos, dt, satpos, sdt, I, T, sagnac, sys, keep_states=True, enable_torch=False, device='cpu'):
    """
    Computes pseudorange observation residuals and their Jacobian matrix for GNSS positioning.
    Models geometric range, Sagnac effect, receiver-satellite clock bias difference, and optional iono/tropo delays.
    The state vector assumed is: [x, y, z, dt_sys1, dt_sys2, ...] — position + per-system clock bias (no velocity or clock drift).
    Clock drift is handled externally (e.g., in Doppler module), enabling modular design via block_diag fusion.

    Parameters:
    pos : receiver position (ECEF), shape (3,)
    dt : receiver clock bias per GNSS system, shape (n_sys,), unit: m
        → e.g., dt = [dt_GPS, dt_GAL, dt_BDS, ...]
    satpos : satellite positions (ECEF), shape (n_obs, 3)
    sdt : satellite clock biases (from broadcast/SP3), shape (n_obs,), unit: s
    I : ionospheric delay (optional, can be zero), shape (n_obs, 1), unit: m
    T : tropospheric delay (optional, can be zero), shape (n_obs, 1), unit: m
    sagnac : precomputed Sagnac correction term (from get_sagnac_corr), shape (n_obs, 1), unit: m
    sys : list of GNSS system identifiers for each observation, e.g., ['G', 'E', 'C']
    keep_states : if False, removes Jacobian columns corresponding to systems with no observation (for WLS compatibility)
    enable_torch : if True, use PyTorch backend for computations
    device : if enable_torch is True, specifies the device ('cpu' or 'cuda')

    Returns:
    psr : predicted pseudorange residuals (m), shape (n_obs, 1)
    H : Jacobian matrix w.r.t state [x, y, z, dt_sys1, dt_sys2, ...], shape (n_obs, 3 + n_active_sys)
    """
    backend = Backend(enable_torch)

    satpos = backend.asarray(satpos, dtype=backend.float64)
    sdt = backend.asarray(sdt, dtype=backend.float64)
    I = backend.asarray(I, dtype=backend.float64)
    T = backend.asarray(T, dtype=backend.float64)
    sagnac = backend.asarray(sagnac, dtype=backend.float64)

    pos = backend.to(pos, device)
    dt = backend.to(dt, device)
    satpos = backend.to(satpos, device)
    sdt = backend.to(sdt, device)
    I = backend.to(I, device)
    T = backend.to(T, device)
    sagnac = backend.to(sagnac, device)

    #dt = dt/prl.CLIGHT # convert m to s

    # 几何距离 + 单位向量
    diff = satpos - pos  # shape: (n_obs, 3)
    norm = backend.linalg_norm(diff, axis=1).reshape(-1, 1)  # shape: (n_obs, 1)
    e = -diff / norm  # shape: (n_obs, 3)

    psr1 = norm + sagnac.reshape(-1, 1)  # shape: (n_obs, 1)
    psr2 = dt - prl.CLIGHT * sdt       # shape: (n_obs,)
    psr3 = I + T                         # shape: (n_obs, 1)

    psr = psr1 + psr2.reshape(-1, 1) + psr3.reshape(-1, 1)  # shape: (n_obs, 1)

    # 构造钟差映射矩阵 H_t: (n_obs, n_sys)
    idx = backend.array([list(SYS_NAME).index(s) for s in sys], dtype=backend.int64)
    eye_matrix = backend.eye(len(SYS_NAME), dtype=backend.float64)
    H_t = eye_matrix[idx]  # shape: (n_obs, n_sys)

    # 可选：移除无观测系统的列
    if not keep_states:
        non_zero_cols = backend.any(H_t != 0, axis=0)  # shape: (n_sys,)
        H_t = H_t[:, non_zero_cols]  # shape: (n_obs, n_active_sys)

    H_t = backend.to(H_t, device)

    # 构造完整 Jacobian: [∂psr/∂x, ∂psr/∂y, ∂psr/∂z, ∂psr/∂dt1, ∂psr/∂dt2, ...]
    H_pos = e  # shape: (n_obs, 3)
    H_clock = H_t  # shape: (n_obs, n_sys) or (n_obs, n_active_sys)
    H = backend.hstack((H_pos, H_clock))  # shape: (n_obs, 3 + n_sys)

    return psr, H

read_obs(rcv, eph, opt='', ref=None)

Reads GNSS observation and ephemeris data from RINEX files using RTKLIB’s internal structures.

rcv (str or list of str): Path(s) to RINEX observation file(s) containing receiver measurements. These files are read as observation data (type 1 in RTKLIB).

eph (str or list of str): Path(s) to RINEX navigation/ephemeris file(s). These files are read as ephemeris data (type 2 in RTKLIB).

opt (str, optional): Options string passed to RTKLIB’s readrnx function for customized reading behavior. Default is "-SYS=GCREJI" to include GPS, GLONASS, Galileo, BeiDou, QZSS, and IRNSS systems.

ref (str or list of str, optional): Path(s) to RINEX observation file(s) from a reference station, used for Real-Time Kinematic (RTK) processing. If provided, these are also read as type 2 (ephemeris-type) data for reference station handling.

Returns: obs_t: RTKLIB structure containing GNSS observation data (pseudorange, carrier phase, etc.). nav_t: RTKLIB structure containing satellite ephemeris, clock, and ionospheric model data. sta_t: RTKLIB structure containing station information (e.g., antenna position, receiver info), primarily populated when reading reference station files.

Source code in tasgnss/core.py 
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
def read_obs(rcv, eph, opt = "", ref=None):
    """
    Reads GNSS observation and ephemeris data from RINEX files using RTKLIB’s internal structures.

    Parameters:
    rcv (str or list of str):
        Path(s) to RINEX observation file(s) containing receiver measurements.
        These files are read as observation data (type 1 in RTKLIB).

    eph (str or list of str):
        Path(s) to RINEX navigation/ephemeris file(s).
        These files are read as ephemeris data (type 2 in RTKLIB).

    opt (str, optional):
        Options string passed to RTKLIB’s readrnx function for customized reading behavior.
        Default is "-SYS=GCREJI" to include GPS, GLONASS, Galileo, BeiDou, QZSS, and IRNSS systems.

    ref (str or list of str, optional):
        Path(s) to RINEX observation file(s) from a reference station, used for Real-Time Kinematic (RTK) processing.
        If provided, these are also read as type 2 (ephemeris-type) data for reference station handling.

    Returns:
    obs_t: RTKLIB structure containing GNSS observation data (pseudorange, carrier phase, etc.).
    nav_t: RTKLIB structure containing satellite ephemeris, clock, and ionospheric model data.
    sta_t: RTKLIB structure containing station information (e.g., antenna position, receiver info), primarily populated when reading reference station files.
    """
    obs = prl.obs_t()
    nav = prl.nav_t()
    sta = prl.sta_t()
    if type(rcv) is list:
        for r in rcv:
            prl.readrnx(r,1,opt,obs,nav,sta)
    else:
        prl.readrnx(rcv,1,opt,obs,nav,sta)
    if type(eph) is list:
        for f in eph:
            prl.readrnx(f,2,opt,obs,nav,sta)
    else:
        prl.readrnx(eph,2,opt,obs,nav,sta)
    if ref:
        if type(ref) is list:
            for r in ref:
                prl.readrnx(r,2,opt,obs,nav,sta)
        else:
            prl.readrnx(ref,2,opt,obs,nav,sta)
    return obs,nav,sta

split_obs(obs, ref_obs=False)

Splits a monolithic obs_t structure into a list of obs_t objects, each containing observations from a single epoch. This facilitates per-epoch processing in applications such as RTK or time-series analysis.

obs (obs_t): The input observation structure, typically generated by read_obs(). All observations across all epochs are stored in obs.data.

ref_obs (bool, optional, default=True): If True, includes reference station observations (receiver ID = 2) in the split epoch data. If False, only observations from the primary receiver (receiver ID = 1) are retained per epoch.

List[obs_t]: A list of obs_t structures, each representing one epoch’s worth of observation data. Each element contains: .data: Array of observations for that epoch. .n: Actual number of observations in the epoch. .nmax: Maximum allocated size (equal to total observations detected for the epoch).

Behavior: First calls pyrtklib.sortobs(obs) to ensure observations are sorted chronologically by time and receiver. Iterates through epochs using nextobsf(obs, i) to locate epoch boundaries. For each epoch: Allocates a new obs_t structure. Copies observations from receiver 1 (primary). If ref_obs=True, also copies observations from receiver 2 (reference station), appending them after receiver 1’s data. Skips epochs with no primary receiver data. Returns the list of per-epoch obs_t structures.

Source code in tasgnss/core.py 
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
def split_obs(obs, ref_obs=False):
    """
    Splits a monolithic obs_t structure into a list of obs_t objects, each containing observations from a single epoch.
    This facilitates per-epoch processing in applications such as RTK or time-series analysis.

    Parameters:
    obs (obs_t):
        The input observation structure, typically generated by read_obs(). All observations across all epochs are stored in obs.data.

    ref_obs (bool, optional, default=True):
        If True, includes reference station observations (receiver ID = 2) in the split epoch data.
        If False, only observations from the primary receiver (receiver ID = 1) are retained per epoch.

    Returns:
    List[obs_t]:
        A list of obs_t structures, each representing one epoch’s worth of observation data. Each element contains:
        .data: Array of observations for that epoch.
        .n: Actual number of observations in the epoch.
        .nmax: Maximum allocated size (equal to total observations detected for the epoch).

    Behavior:
    First calls pyrtklib.sortobs(obs) to ensure observations are sorted chronologically by time and receiver.
    Iterates through epochs using nextobsf(obs, i) to locate epoch boundaries.
    For each epoch:
        Allocates a new obs_t structure.
        Copies observations from receiver 1 (primary).
        If ref_obs=True, also copies observations from receiver 2 (reference station), appending them after receiver 1’s data.
        Skips epochs with no primary receiver data.
    Returns the list of per-epoch obs_t structures.
    """
    prl.sortobs(obs)
    i = 0
    m = nextobsf(obs,i)
    obss = []
    while m!=0:
        tmp_obs = prl.obs_t()
        tmp_obs.data = prl.Arr1Dobsd_t(m)
        rcv1 = 0
        rcv2 = 0
        for j in range(m):
            if obs.data[i+j].rcv == 1:
                tmp_obs.data[rcv1] = obs.data[i+j]
                rcv1+=1
        if rcv1 == 0:
            i+=m
            m = nextobsf(obs,i)
            continue
        if rcv1 != m and ref_obs:
            for j in range(m-rcv1):
                if obs.data[i+j+rcv1].rcv == 2:
                    tmp_obs.data[rcv1+rcv2] = obs.data[i+rcv1+j]
                    rcv2+=1
        tmp_obs.n = rcv1+rcv2
        tmp_obs.nmax = m
        i+=m
        obss.append(tmp_obs)
        m = nextobsf(obs,i)
    return obss

wls_pnt_pos(o, nav, use_cache=True, return_residual=False, enable_torch=False, w=None, b=None, device='cpu')

Performs Weighted Least Squares (WLS) positioning using GNSS observations.

This function implements an iterative WLS algorithm to solve for receiver position, velocity, and clock parameters. It supports caching for performance optimization and PyTorch backend for gradient-based optimization.

Key Features: - Uses caching to accelerate repeated calls with the same observation data - Supports PyTorch backend for gradient propagation, enabling use in neural network optimization - Handles multiple GNSS constellations with separate clock bias parameters - Includes atmospheric and relativistic corrections

o (obs_t): GNSS observation data structure for one epoch. nav (nav_t): Navigation data structure with ephemerides and satellite biases. use_cache (bool, optional): Whether to use cached preprocessing results. When True, if the same observation object is processed multiple times, the preprocessing results (satellite positions, atmospheric corrections, etc.) are cached and reused, significantly speeding up repeated calls. Default is True. return_residual (bool, optional): Whether to return residuals, Jacobian matrix, and weight matrix. When True, the returned dictionary includes a "residual_info" key containing: - "residual": Observation residuals vector - "H": Design matrix (Jacobian) - "W": Weight matrix These can be used to compute Dilution of Precision (DOP) metrics. Default is False. enable_torch (bool, optional): Whether to use PyTorch backend for computations. When True, the function uses PyTorch tensors and operations, allowing gradients to flow through the computation graph. This enables the use of this function in neural network training, where weights (w) and bias (b) can be optimized using gradient descent. Default is False. w (array-like, optional): Weight matrix for pseudorange observations. If enable_torch=True, gradients will propagate through w to the position solution, allowing wp to be optimized in neural networks. Default is None (uses inverse of observation variance). b (array-like, optional): Bias vector to be subtracted from pseudorange observations. If enable_torch=True, gradients will propagate through b to the position solution, allowing b to be optimized in neural networks. Default is None (zero vector). device (str, optional): Device to run PyTorch computations on ('cpu' or 'cuda'). Default is 'cpu'.

dict: A dictionary containing positioning results, status, and additional information with keys: - "status" (bool): True if positioning succeeded, False otherwise - "pos" (array): Receiver position [x, y, z] in ECEF coordinates - "cb" (array): Receiver clock bias for each GNSS system - "cd" (array): Receiver clock drift - "msg" (str): Status message - "data" (array): Processed observation data - "solve_data" (dict): Preprocessed data used in solving - "raw_data" (dict): Raw observation data - "residual_info" (dict, optional): Residuals and Jacobian matrix if return_residual=True

Source code in tasgnss/core.py 
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
def wls_pnt_pos(o, nav, use_cache=True, return_residual=False, enable_torch=False, w=None, b=None, device='cpu'):
    """
    Performs Weighted Least Squares (WLS) positioning using GNSS observations.

    This function implements an iterative WLS algorithm to solve for receiver position, velocity,
    and clock parameters. It supports caching for performance optimization and PyTorch backend
    for gradient-based optimization.

    Key Features:
    - Uses caching to accelerate repeated calls with the same observation data
    - Supports PyTorch backend for gradient propagation, enabling use in neural network optimization
    - Handles multiple GNSS constellations with separate clock bias parameters
    - Includes atmospheric and relativistic corrections

    Parameters:
    o (obs_t): 
        GNSS observation data structure for one epoch.
    nav (nav_t): 
        Navigation data structure with ephemerides and satellite biases.
    use_cache (bool, optional): 
        Whether to use cached preprocessing results. When True, if the same
        observation object is processed multiple times, the preprocessing results (satellite positions,
        atmospheric corrections, etc.) are cached and reused, significantly speeding up repeated calls.
        Default is True.
    return_residual (bool, optional): 
        Whether to return residuals, Jacobian matrix, and weight matrix.
        When True, the returned dictionary includes a "residual_info" key containing:
        - "residual": Observation residuals vector
        - "H": Design matrix (Jacobian)
        - "W": Weight matrix
        These can be used to compute Dilution of Precision (DOP) metrics. Default is False.
    enable_torch (bool, optional): 
        Whether to use PyTorch backend for computations. When True,
        the function uses PyTorch tensors and operations, allowing gradients to flow through the
        computation graph. This enables the use of this function in neural network training, where
        weights (w) and bias (b) can be optimized using gradient descent. Default is False.
    w (array-like, optional): Weight matrix for pseudorange observations. If enable_torch=True,
        gradients will propagate through w to the position solution, allowing wp to be optimized
        in neural networks. Default is None (uses inverse of observation variance).
    b (array-like, optional): Bias vector to be subtracted from pseudorange observations. If
        enable_torch=True, gradients will propagate through b to the position solution, allowing
        b to be optimized in neural networks. Default is None (zero vector).
    device (str, optional): Device to run PyTorch computations on ('cpu' or 'cuda'). Default is 'cpu'.

    Returns:
    dict: A dictionary containing positioning results, status, and additional information with keys:
        - "status" (bool): True if positioning succeeded, False otherwise
        - "pos" (array): Receiver position [x, y, z] in ECEF coordinates
        - "cb" (array): Receiver clock bias for each GNSS system
        - "cd" (array): Receiver clock drift
        - "msg" (str): Status message
        - "data" (array): Processed observation data
        - "solve_data" (dict): Preprocessed data used in solving
        - "raw_data" (dict): Raw observation data
        - "residual_info" (dict, optional): Residuals and Jacobian matrix if return_residual=True
    """
    maxiter = 20
    o_id = id(o)
    if use_cache and o_id in cache_data:
        p, p_t, v, v_t, data, cdata, raw_data = cache_data[o_id]
    else:
        p, p_t, v, v_t, data, cdata, raw_data = preprocess_obs(o, nav, use_cache)

    # test cdata['sys'], check the number of systems and the number of unknowns
    if np.unique(cdata['sys']).shape[0] + 3 > cdata['pr'].shape[0]:
        return {
            "status": False,
            "pos": np.zeros(4),
            "msg": "insufficient satellites for the number of systems",
            "data": {},
            "solve_data": cdata,
            "raw_data": raw_data
        }


    iter = 0

    backend = Backend(enable_torch)

    def ensure_array(x, shape, dtype=backend.float64):
        if x is None:
            return backend.zeros(shape, dtype=dtype)
        return backend.asarray(x, dtype=dtype).reshape(shape)

    # initalize variables
    p = ensure_array(p, (3,))
    p_t = ensure_array(p_t, (len(SYS_NAME),))


    # ensure all variables are on the correct device
    p = backend.to(p, device)
    p_t = backend.to(p_t, device)

    dp = backend.asarray(np.array(100.0), dtype=backend.float64)
    dp = backend.to(dp, device)

    # construct p_t_mask
    idx = backend.array([list(SYS_NAME).index(s) for s in data[:,2]], dtype=backend.int64)  # 索引用 int64
    eye_matrix = backend.eye(len(SYS_NAME), dtype=backend.float64)
    p_t_mask = eye_matrix[idx] 
    p_t_mask = backend.to(p_t_mask, device)

    # process b
    if b is None:
        b = backend.zeros_like(backend.asarray(cdata['pr']), dtype=backend.float64)
    else:
        b = backend.asarray(b, dtype=backend.float64).reshape(-1,1)
    b = backend.to(b, device)

    # process Wpr
    if w is not None:
        if isinstance(w, (int, float)) and w == 1:
            W_pr = backend.eye(len(cdata['pr']), dtype=backend.float64)
        else:
            w_tensor = backend.asarray(w, dtype=backend.float64)
            W_pr = backend.diag(w_tensor)
    else:
        var = backend.asarray(cdata['var'], dtype=backend.float64)
        W_pr = backend.diag(1.0 / backend.sqrt(var))


    W = backend.to(W_pr, device)

    # ensure cdata pr and dop are tensors on the correct device
    pr_tensor = backend.asarray(cdata['pr'], dtype=backend.float64)
    pr_tensor = backend.to(pr_tensor, device)

    try:
        while iter < maxiter and backend.linalg_norm(dp) > 0.001:
            psr, H = pseudorange_observe_func(
                p, backend.dot(p_t_mask, p_t), cdata['satpos'][:,:3], cdata["sdt"][:,0],
                cdata['I'], cdata['T'], cdata['sagnac'], cdata['sys'], enable_torch=enable_torch,
                device=device
            )


            p_residual = pr_tensor - psr - b
            residual = p_residual


            # solve least squares
            W_H = backend.dot(W, H)
            W_r = backend.dot(W, residual)
            result = backend.linalg_lstsq(W_H, W_r, rcond=None)
            dp = result.solution if hasattr(result, 'solution') else result[0]

            # update
            p = p + backend.squeeze(dp[:3])
            p_t = p_t + backend.squeeze(dp[3:])
            iter += 1

    except Exception as e:
        zero_pos = backend.zeros(4, dtype=backend.float64)
        zero_pos = backend.to(zero_pos, device)
        return {
            "status": False,
            "pos": zero_pos,
            "msg": str(e),
            "data": {},
            "solve_data": cdata,
            "raw_data": raw_data
        }

    if iter >= maxiter or backend.linalg_norm(dp) > 1e-3:
        zero_pos = backend.zeros(4, dtype=backend.float64)
        zero_pos = backend.to(zero_pos, device)
        return {
            "status": False,
            "pos": zero_pos,
            "msg": "not converge",
            "data": {},
            "solve_data": cdata,
            "raw_data": raw_data
        }

    # store to cache
    cache_data[o_id][0] = backend.to_numpy(p)
    cache_data[o_id][1] = backend.to_numpy(p_t)


    # optional: return residuals and H
    if return_residual:
        psr, H = pseudorange_observe_func(
            p, backend.dot(p_t_mask, p_t), cdata['satpos'][:,:3], cdata["sdt"][:,0],
            cdata['I'], cdata['T'], cdata['sagnac'], cdata['sys'], enable_torch=enable_torch
        )

        pr_tensor = backend.asarray(cdata['pr'], dtype=backend.float64)
        pr_tensor = backend.to(pr_tensor, device)

        residual = pr_tensor - psr - b
        residual_info = {
            "residual": residual,
            "H": H,
            "W": W
        }
    else:
        residual_info = {}

    return {
        "status": True,
        "pos": p,
        "cb": p_t,
        "msg": "ok",
        "data": data,
        "solve_data": cdata,
        "raw_data": raw_data,
        "residual_info": residual_info
    }

wls_pnt_pos_vel(o, nav, use_cache=True, return_residual=False, enable_torch=False, wp=None, wv=None, b=None, device='cpu')

Performs Weighted Least Squares (WLS) positioning using GNSS observations.

This function implements an iterative WLS algorithm to solve for receiver position, velocity, and clock parameters. It supports caching for performance optimization and PyTorch backend for gradient-based optimization.

Key Features: - Uses caching to accelerate repeated calls with the same observation data - Supports PyTorch backend for gradient propagation, enabling use in neural network optimization - Handles multiple GNSS constellations with separate clock bias parameters - Includes atmospheric and relativistic corrections

o (obs_t): GNSS observation data structure for one epoch. nav (nav_t): Navigation data structure with ephemerides and satellite biases. use_cache (bool, optional): Whether to use cached preprocessing results. When True, if the same observation object is processed multiple times, the preprocessing results (satellite positions, atmospheric corrections, etc.) are cached and reused, significantly speeding up repeated calls. Default is True. return_residual (bool, optional): Whether to return residuals, Jacobian matrix, and weight matrix. When True, the returned dictionary includes a "residual_info" key containing: - "residual": Observation residuals vector - "H": Design matrix (Jacobian) - "W": Weight matrix These can be used to compute Dilution of Precision (DOP) metrics. Default is False. enable_torch (bool, optional): Whether to use PyTorch backend for computations. When True, the function uses PyTorch tensors and operations, allowing gradients to flow through the computation graph. This enables the use of this function in neural network training, where weights (wp, wv) and bias (b) can be optimized using gradient descent. Default is False. wp (array-like, optional): Weight matrix for pseudorange observations. If enable_torch=True, gradients will propagate through wp to the position solution, allowing wp to be optimized in neural networks. Default is None (uses inverse of observation variance). wv (array-like, optional): Weight matrix for Doppler observations. If enable_torch=True, gradients will propagate through wv to the position solution. Default is None (uses wp*10). b (array-like, optional): Bias vector to be subtracted from pseudorange observations. If enable_torch=True, gradients will propagate through b to the position solution, allowing b to be optimized in neural networks. Default is None (zero vector). device (str, optional): Device to run PyTorch computations on ('cpu' or 'cuda'). Default is 'cpu'.

dict: A dictionary containing positioning results, status, and additional information with keys: - "status" (bool): True if positioning succeeded, False otherwise - "pos" (array): Receiver position [x, y, z] in ECEF coordinates - "velocity" (array): Receiver velocity [vx, vy, vz] in ECEF coordinates - "cb" (array): Receiver clock bias for each GNSS system - "cd" (array): Receiver clock drift - "msg" (str): Status message - "data" (array): Processed observation data - "solve_data" (dict): Preprocessed data used in solving - "raw_data" (dict): Raw observation data - "residual_info" (dict, optional): Residuals and Jacobian matrix if return_residual=True

Source code in tasgnss/core.py 
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
def wls_pnt_pos_vel(o, nav, use_cache=True, return_residual=False, enable_torch=False, wp=None, wv=None, b=None, device='cpu'):
    """
    Performs Weighted Least Squares (WLS) positioning using GNSS observations.

    This function implements an iterative WLS algorithm to solve for receiver position, velocity,
    and clock parameters. It supports caching for performance optimization and PyTorch backend
    for gradient-based optimization.

    Key Features:
    - Uses caching to accelerate repeated calls with the same observation data
    - Supports PyTorch backend for gradient propagation, enabling use in neural network optimization
    - Handles multiple GNSS constellations with separate clock bias parameters
    - Includes atmospheric and relativistic corrections

    Parameters:
    o (obs_t): 
        GNSS observation data structure for one epoch.
    nav (nav_t): 
        Navigation data structure with ephemerides and satellite biases.
    use_cache (bool, optional): 
        Whether to use cached preprocessing results. When True, if the same
        observation object is processed multiple times, the preprocessing results (satellite positions,
        atmospheric corrections, etc.) are cached and reused, significantly speeding up repeated calls.
        Default is True.
    return_residual (bool, optional): 
        Whether to return residuals, Jacobian matrix, and weight matrix.
        When True, the returned dictionary includes a "residual_info" key containing:
        - "residual": Observation residuals vector
        - "H": Design matrix (Jacobian)
        - "W": Weight matrix
        These can be used to compute Dilution of Precision (DOP) metrics. Default is False.
    enable_torch (bool, optional): 
        Whether to use PyTorch backend for computations. When True,
        the function uses PyTorch tensors and operations, allowing gradients to flow through the
        computation graph. This enables the use of this function in neural network training, where
        weights (wp, wv) and bias (b) can be optimized using gradient descent. Default is False.
    wp (array-like, optional): Weight matrix for pseudorange observations. If enable_torch=True,
        gradients will propagate through wp to the position solution, allowing wp to be optimized
        in neural networks. Default is None (uses inverse of observation variance).
    wv (array-like, optional): Weight matrix for Doppler observations. If enable_torch=True,
        gradients will propagate through wv to the position solution. Default is None (uses wp*10).
    b (array-like, optional): Bias vector to be subtracted from pseudorange observations. If
        enable_torch=True, gradients will propagate through b to the position solution, allowing
        b to be optimized in neural networks. Default is None (zero vector).
    device (str, optional): Device to run PyTorch computations on ('cpu' or 'cuda'). Default is 'cpu'.

    Returns:
    dict: A dictionary containing positioning results, status, and additional information with keys:
        - "status" (bool): True if positioning succeeded, False otherwise
        - "pos" (array): Receiver position [x, y, z] in ECEF coordinates
        - "velocity" (array): Receiver velocity [vx, vy, vz] in ECEF coordinates
        - "cb" (array): Receiver clock bias for each GNSS system
        - "cd" (array): Receiver clock drift
        - "msg" (str): Status message
        - "data" (array): Processed observation data
        - "solve_data" (dict): Preprocessed data used in solving
        - "raw_data" (dict): Raw observation data
        - "residual_info" (dict, optional): Residuals and Jacobian matrix if return_residual=True
    """
    maxiter = 20
    o_id = id(o)
    if use_cache and o_id in cache_data:
        p, p_t, v, v_t, data, cdata, raw_data = cache_data[o_id]
    else:
        p, p_t, v, v_t, data, cdata, raw_data = preprocess_obs(o, nav, use_cache)
    iter = 0

    backend = Backend(enable_torch)

    def ensure_array(x, shape, dtype=backend.float64):
        if x is None:
            return backend.zeros(shape, dtype=dtype)
        return backend.asarray(x, dtype=dtype).reshape(shape)

    # initalize variables
    p = ensure_array(p, (3,))
    p_t = ensure_array(p_t, (len(SYS_NAME),))
    v = ensure_array(v, (3,))
    v_t = ensure_array(v_t, (1,))

    # ensure all variables are on the correct device
    p = backend.to(p, device)
    p_t = backend.to(p_t, device)
    v = backend.to(v, device)
    v_t = backend.to(v_t, device)

    dp = backend.asarray(np.array(100.0), dtype=backend.float64)
    dp = backend.to(dp, device)

    # construct p_t_mask
    idx = backend.array([list(SYS_NAME).index(s) for s in data[:,2]], dtype=backend.int64)  # 索引用 int64
    eye_matrix = backend.eye(len(SYS_NAME), dtype=backend.float64)
    p_t_mask = eye_matrix[idx] 
    p_t_mask = backend.to(p_t_mask, device)

    # process b
    if b is None:
        b = backend.zeros_like(backend.asarray(cdata['pr']), dtype=backend.float64)
    else:
        b = backend.asarray(b, dtype=backend.float64).reshape(-1,1)
    b = backend.to(b, device)

    # process Wpr
    if wp is not None:
        if isinstance(wp, (int, float)) and wp == 1:
            W_pr = backend.eye(len(cdata['pr']), dtype=backend.float64)
        else:
            wp_tensor = backend.asarray(wp, dtype=backend.float64)
            W_pr = backend.diag(wp_tensor)
    else:
        var = backend.asarray(cdata['var'], dtype=backend.float64)
        W_pr = backend.diag(1.0 / backend.sqrt(var))

    # process Wdop
    if wv is not None:
        wv_tensor = backend.asarray(wv, dtype=backend.float64)
        W_dop = backend.diag(wv_tensor)
    else:
        W_dop = W_pr * 100.0 

    # combine W
    W = backend.block_diag(W_pr, W_dop)
    W = backend.to(W, device)

    # ensure cdata pr and dop are tensors on the correct device
    pr_tensor = backend.asarray(cdata['pr'], dtype=backend.float64)
    dop_tensor = backend.asarray(cdata['dop'], dtype=backend.float64)
    pr_tensor = backend.to(pr_tensor, device)
    dop_tensor = backend.to(dop_tensor, device)

    try:
        while iter < maxiter and backend.linalg_norm(dp) > 0.001:
            psr, H_psr = pseudorange_observe_func(
                p, backend.dot(p_t_mask, p_t), cdata['satpos'][:,:3], cdata["sdt"][:,0],
                cdata['I'], cdata['T'], cdata['sagnac'], cdata['sys'], enable_torch=enable_torch,
                device=device
            )
            dop, H_dop = doppler_observe_func(
                v, v_t, p, cdata['satpos'][:,:3], cdata['satpos'][:,3:],
                cdata["sdt"][:,1], cdata['sys'], enable_torch=enable_torch,
                device=device
            )


            p_residual = pr_tensor - psr - b
            d_residual = dop_tensor - dop
            residual = backend.vstack((p_residual, d_residual))
            H = backend.block_diag(H_psr, H_dop)


            # solve least squares
            W_H = backend.dot(W, H)
            W_r = backend.dot(W, residual)
            result = backend.linalg_lstsq(W_H, W_r, rcond=None)
            dp = result.solution if hasattr(result, 'solution') else result[0]

            # update
            p = p + backend.squeeze(dp[:3])
            p_t = p_t + backend.squeeze(dp[3:-4])
            v = v + backend.squeeze(dp[-4:-1])
            v_t = v_t + backend.squeeze(dp[-1:])
            iter += 1

    except Exception as e:
        zero_pos = backend.zeros(4, dtype=backend.float64)
        zero_pos = backend.to(zero_pos, device)
        return {
            "status": False,
            "pos": zero_pos,
            "msg": str(e),
            "data": {},
            "solve_data": cdata,
            "raw_data": raw_data
        }

    if iter >= maxiter or backend.linalg_norm(dp) > 1e-3:
        zero_pos = backend.zeros(4, dtype=backend.float64)
        zero_pos = backend.to(zero_pos, device)
        return {
            "status": False,
            "pos": zero_pos,
            "msg": "not converge",
            "data": {},
            "solve_data": cdata,
            "raw_data": raw_data
        }

    # store to cache
    cache_data[o_id][0] = backend.to_numpy(p)
    cache_data[o_id][1] = backend.to_numpy(p_t)
    cache_data[o_id][2] = backend.to_numpy(v)
    cache_data[o_id][3] = backend.to_numpy(v_t)

    # optional: return residuals and H
    if return_residual:
        psr, Hp = pseudorange_observe_func(
            p, backend.dot(p_t_mask, p_t), cdata['satpos'][:,:3], cdata["sdt"][:,0],
            cdata['I'], cdata['T'], cdata['sagnac'], cdata['sys'], enable_torch=enable_torch
        )
        dop, Hd = doppler_observe_func(
            v, v_t, p, cdata['satpos'][:,:3], cdata['satpos'][:,3:],
            cdata["sdt"][:,1], cdata['sys'], enable_torch=enable_torch
        )

        pr_tensor = backend.asarray(cdata['pr'], dtype=backend.float64)
        dop_tensor = backend.asarray(cdata['dop'], dtype=backend.float64)
        pr_tensor = backend.to(pr_tensor, device)
        dop_tensor = backend.to(dop_tensor, device)

        residual_p = pr_tensor - psr - b
        residual_d = dop_tensor - dop
        residual = backend.vstack((residual_p, residual_d))
        H = backend.block_diag(Hp, Hd)
        residual_info = {
            "residual": residual,
            "H": H,
            "W": W
        }
    else:
        residual_info = {}

    return {
        "status": True,
        "pos": p,
        "velocity": v,
        "cb": p_t,
        'cd': v_t,
        "msg": "ok",
        "data": data,
        "solve_data": cdata,
        "raw_data": raw_data,
        "residual_info": residual_info
    }

xyz2enu(pos, deg=True)

Convert ECEF coordinates to ENU rotation matrix Parameters: pos: tuple or array of (lat, lon) in degrees or radians deg: if True, pos is in degrees, else in radians Returns: E: 3x3 rotation matrix from ECEF to ENU

Source code in tasgnss/core.py 
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
def xyz2enu(pos, deg=True):
    """
    Convert ECEF coordinates to ENU rotation matrix
    Parameters:
        pos: tuple or array of (lat, lon) in degrees or radians
        deg: if True, pos is in degrees, else in radians
    Returns:
        E: 3x3 rotation matrix from ECEF to ENU
    """
    E = prl.Arr1Ddouble(9)
    _pos = prl.Arr1Ddouble(2)
    if deg:
        _pos[0] = pos[0] * prl.D2R
        _pos[1] = pos[1] * prl.D2R
        prl.xyz2enu(_pos, E)
    else:
        _pos[0] = pos[0]
        _pos[1] = pos[1]
        prl.xyz2enu(_pos, E)
    return np.array(E).reshape(3, 3)