This repository has been archived by the owner on Nov 22, 2021. It is now read-only.
-
-
Notifications
You must be signed in to change notification settings - Fork 9
/
easy6.m
261 lines (236 loc) · 9.72 KB
/
easy6.m
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
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
%EASY6 We compute a baseline from C/A code and phase observations.
% The code does not handle
% 1. cycle slips and
% 2. outliers.
% In contrast to EASY5, we now use an extended Kalman filter for the
% estimation. The present code is no real RTK code as all computational
% steps do not happen on an epoch-by-epoch basis
%Kai Borre 27-07-2002
%Copyright (c) by Kai Borre
%$Revision: 1.1 $ $Date: 2005/01/20 $
% Initial computations of constants
v_light = 299792458; % vacuum speed of light m/s
f1 = 154*10.23E6; % L1 frequency Hz
f2 = 120*10.23E6; % L2 frequency Hz
lambda1 = v_light/f1; % wavelength on L1: .19029367 m
lambda2 = v_light/f2; % wavelength on L2: .244210213 m
% Read RINEX ephemerides file and convert to internal Matlab format
rinexe('site247j.01n','eph.dat');
Eph = get_eph('eph.dat');
% We identify the master observation file and open it
ofile1 = 'site247j.01o';
fid1 = fopen(ofile1,'rt');
[Obs_types1, ant_delta1, ifound_types1, eof11] = anheader(ofile1);
NoObs_types1 = size(Obs_types1,2)/2;
% We start by estimating the master position
[time1, dt1, sats1, eof1] = fepoch_0(fid1);
NoSv1 = size(sats1, 1);
m = NoSv1;
obs1raw = grabdata(fid1, NoSv1, NoObs_types1);
i = fobs_typ(Obs_types1,'C1'); % We use C/A pseudoranges
[X_i, el] = recpo_ls(obs1raw(:,i), sats1, time1, Eph);
[phi_i,lambda_i,h_i] = togeod(6378137,298.257223563,X_i(1),X_i(2),X_i(3));
% We close all files to ensure that the next reading starts
% at the top of the observation files
fclose all;
% Finding columns in Eph for each SV
for t = 1:m
col_Eph(t) = find_eph(Eph,sats1(t),time1);
end
% Computation of elevation angle to all SVs.
all_sats1 = sats1;
% Delete Sv with elevation smaller than 10 degrees
sats1(el<10) = [];
del_sat = setdiff(all_sats1,sats1);
no_del_sat = [];
for t = 1:length(del_sat)
no_dels = find(del_sat(t) == all_sats1);
no_del_sat = [no_del_sat; no_dels];
end
No_del_sat = length(no_del_sat);
% Selecting reference SV. We take the SV with largest elevation
[y,ind] = max(el);
refsv = sats1(ind);
fid1 = fopen(ofile1,'rt');
ofile2 = 'site24~1.01o';
fid2 = fopen(ofile2,'rt');
% We start reading both observation files
[Obs_types1, ant_delta1, ifound_types1, eof11] = anheader(ofile1);
NoObs_types1 = size(Obs_types1,2)/2;
obsstr = ['P1';'P2';'L1';'L2'];
Oc = [];
for t = 1:4
oc = strmatch(obsstr(t,:),strvcat('C1','P1','P2','L1','L2'),'exact');
Oc = [Oc oc];
end
[Obs_types2, ant_delta2, ifound_types2, eof12] = anheader(ofile2);
NoObs_types2 = size(Obs_types2,2)/2;
% Computation of covariance matrix Sigma for double differenced observations
m1 = m-No_del_sat; % original number of SVs - deleted SVs due to low elevations
D = [ones(m1,1) -eye(m1) -ones(m1,1) eye(m1)];
Sigma = D*D';
X = zeros(3+2*m1,1); % coord.diff., N1, N2
N = zeros(3+2*m1,3+2*m1); % initialization of normals
rs = zeros(3+2*m1,1); % initialization of right side
X_a = [];
X_j = X_i(1:3,1);
refrow = find(refsv == sats1);
% We process three epochs for estimating ambiguities; the present data evidently
% need three or more epochs for getting reliable estimates of the float ambiguities
for q = 1:6
X_j = X_i(1:3,1)+X(1:3,1);
[time1, dt1, sats1, eof1] = fepoch_0(fid1);
[time2, dt2, sats2, eof2] = fepoch_0(fid2);
if time1 ~= time2
disp('Epochs do not correspond in time')
break
end;
time = time1;
NoSv1 = size(sats1,1);
NoSv2 = size(sats2,1);
obsm = grabdata(fid1, NoSv1, NoObs_types1);
obsr = grabdata(fid2, NoSv2, NoObs_types2);
obs1 = obsm(:,Oc); % P1 P2 Phi1 Phi2
% Reordering of rows in obsr to correspond to obsm
for s = 1:NoSv1
Ind = find(sats1(s) == sats2(:));
obs2(s,:) = obsr(Ind,Oc);
end
% Computing rho for refsv
[tcorr,rhok_j,Xk_ECF] = get_rho(time, obs2(refrow,1), Eph(:,col_Eph(refrow)), X_j);
[tcorr,rhok_i,Xk_ECF] = get_rho(time, obs1(refrow,1), Eph(:,col_Eph(refrow)), X_i);
tt = 0;
A1 = [];
t0 = 1:NoSv1;
t1 = setdiff(t0,no_del_sat); % we delete the low satellites
for t = t1
tt = tt+1;
[tcorr,rhol_j,Xl_ECF] = get_rho(time,obs2(t,1), Eph(:,col_Eph(t)), X_j);
[tcorr,rhol_i,Xl_ECF] = get_rho(time,obs1(t,1), Eph(:,col_Eph(t)), X_i);
A0 = [(Xk_ECF(1)-X_j(1))/rhok_j - (Xl_ECF(1)-X_j(1))/rhol_j ...
(Xk_ECF(2)-X_j(2))/rhok_j - (Xl_ECF(2)-X_j(2))/rhol_j ...
(Xk_ECF(3)-X_j(3))/rhok_j - (Xl_ECF(3)-X_j(3))/rhol_j];
A1 = [A1; A0];
Phi1 = (obs1(refrow,3)-obs1(t,3)-obs2(refrow,3)+obs2(t,3))*lambda1;
Phi2 = (obs1(refrow,4)-obs1(t,4)-obs2(refrow,4)+obs2(t,4))*lambda2;
b(tt,:) = Phi1-lambda1*X(3+tt,1);
b(m1+tt,:) = Phi2-lambda2*X(3+m1+tt,1);
bk(tt,:) = rhok_i-rhok_j-rhol_i+rhol_j;
bk(m1+tt,:) = rhok_i-rhok_j-rhol_i+rhol_j;
end;
A_modi = eye(m1); % modified coefficient matrix
col = find(refsv == sats1); % find column for reference PRN
A_modi(:,col) = -ones(m1,1);
A_aug = [A1 lambda1*A_modi 0*eye(m1); A1 0*eye(m1) lambda2*A_modi];
N = N+A_aug'*kron(eye(2),Sigma)*A_aug;
rs = rs+A_aug'*kron(eye(2),Sigma)*(b-bk);
end %q
PP = pinv(N);
% X contains the three preliminary baseline components and the float ambiguities
X = PP*rs;
% Estimation of ambiguities by means of the Lambda method
[a,sqnorm,Sigma_afixed,Z] = lambda(X(4:4+2*m1-1,1),PP(4:4+2*m1-1,4:4+2*m1-1));
% Correcting baseline vector as consequence of changing float ambiguities to fixed ones
X(1:3,1) = X(1:3,1)-PP(1:3,4:4+2*m1-1)*inv(PP(4:4+2*m1-1,4:4+2*m1-1))*...
(X(4:4+2*m1-1,1)-a(:,1)); %select first set of candidates
X(4:4+2*m1-1,1) = a(:,1);
fprintf('\n N1 for PRN %3.0f: %3.0f',[sats1(t1)'; a(1:m1,1)'])
fprintf('\n')
fprintf('\n N2 for PRN %3.0f: %3.0f',[sats1(t1)'; a(m1+1:2*m1,1)'])
% We close and reopen all files in order to start reading at a known position
fclose all;
ofile1 = 'SITE247J.01O';
fid1 = fopen(ofile1,'rt');
ofile2 = 'SITE24~1.01O';
fid2 = fopen(ofile2,'rt');
% Setting covariances for the Kalman filter; the state vector contains (x,y,z)
P = eye(3); % covariances of state vector
Q = 0.05^2*eye(3); % covariances of system
R = 0.005^2*kron(eye(2),inv(Sigma)); % covariances of observations
% In ofile2 we substitute empty observations with NaN's to obtain 22 valid epochs
qend = 22;
% Preliminary estimate of baseline components
x = X(1:3,1);
x_acc = [];
delta_x = zeros(3,1);
for q = 1:qend
X_j = X_i(1:3,1)+x;
[phi_j,lambda_j,h_j] = togeod(6378137,298.257223563,X_j(1),X_j(2),X_j(3));
[time1, dt1, sats1, eof1] = fepoch_0(fid1);
[time2, dt2, sats2, eof2] = fepoch_0(fid2);
if time1 ~= time2
disp('Epochs do not correspond in time')
break
end;
time = time1;
NoSv1 = size(sats1,1);
NoSv2 = size(sats2,1);
obsm = grabdata(fid1, NoSv1, NoObs_types1);
obsr = grabdata(fid2, NoSv2, NoObs_types2);
obs1 = obsm(:,Oc); % P1 P2 Phi1 Phi2
% Reordering of rows in obsr to correspond to obsm
for s = 1:m
Ind = find(sats1(s) == sats2(:));
obs2(s,:) = obsr(Ind,Oc);
end
% Computing rho for refsv
[tcorr,rhok_j,Xk_ECF] = get_rho(time, obs2(1,1), Eph(:,col_Eph(1)), X_j);
[tcorr,rhok_i,Xk_ECF] = get_rho(time, obs1(1,1), Eph(:,col_Eph(1)), X_i);
tt = 0;
A = zeros(2*m1,3);
for t = t1
tt = tt+1;
[tcorr,rhol_j,Xl_ECF] = get_rho(time,obs2(t,1), Eph(:,col_Eph(t)), X_j);
[tcorr,rhol_i,Xl_ECF] = get_rho(time,obs1(t,1), Eph(:,col_Eph(t)), X_i);
A0 = [(Xk_ECF(1)-X_j(1))/rhok_j - (Xl_ECF(1)-X_j(1))/rhol_j ...
(Xk_ECF(2)-X_j(2))/rhok_j - (Xl_ECF(2)-X_j(2))/rhol_j ...
(Xk_ECF(3)-X_j(3))/rhok_j - (Xl_ECF(3)-X_j(3))/rhol_j];
A(tt,:) = A0;
A(m1+tt,:) = A0;
% Tropospheric correction using standard meteorological parameters
%[az,el_ki,d] = topocent(X_i(1:3),Xk_ECF-X_i(1:3));
%[az,el_li,d] = topocent(X_i(1:3),Xl_ECF-X_i(1:3));
%[az,el_kj,d] = topocent(X_j(1:3),Xk_ECF-X_j(1:3));
%[az,el_lj,d] = topocent(X_j(1:3),Xl_ECF-X_j(1:3));
%el_ki, el_li, el_kj, el_lj
%t_corr = tropo(sin(el_lj*pi/180),...
% h_j*1.e-3,1013,293,50,0,0,0)...
% -tropo(sin(el_li*pi/180),....
% h_i*1.e-3,1013,293,50,0,0,0)...
% -tropo(sin(el_kj*pi/180),...
% h_j*1.e-3,1013,293,50,0,0,0)...
% +tropo(sin(el_ki*pi/180),...
% h_i*1.e-3,1013,293,50,0,0,0);
Phi1 = (obs1(refrow,3)-obs1(t,3)-obs2(refrow,3)+obs2(t,3))*lambda1; %-t_corr;
Phi2 = (obs1(refrow,4)-obs1(t,4)-obs2(refrow,4)+obs2(t,4))*lambda2; %-t_corr;
b(tt,:) = Phi1-lambda1*a(tt,1);
b(m1+tt,:) = Phi2-lambda2*a(m1+tt,1);
bk(tt,:) = rhok_i-rhok_j-rhol_i+rhol_j;
bk(m1+tt,:) = rhok_i-rhok_j-rhol_i+rhol_j;
end; % t
%Extended Kalman filter, see pages 509--510 in Strang & Borre (1997): Linear
% Algebra, Geodesy, and GPS, Wellesley-Cambridge Press
P = P+Q;
K = P*A'*inv(A*P*A'+R);
x = x+K*(b-bk);
P = (eye(3)-K*A)*P;
fprintf('\nx: %8.3f m, y: %8.3f m, z: %8.3f m\n',x(1),x(2),x(3))
x_acc = [x_acc x];
end %q
% Transformation of geocentric baseline coordinates into topocentric coordinates
for i = 1:qend
[e(i),n(i),u(i)] = xyz2enu(phi_j,lambda_j,x_acc(1,i),x_acc(2,i),x_acc(3,i));
end
fprintf('\n\nBaseline Components\n')
fprintf('\nX: %8.3f m, Y: %8.3f m, Z: %8.3f m\n', ...
x_acc(1,qend),x_acc(2,qend),x_acc(3,qend))
fprintf('\nE: %8.3f m, N: %8.3f m, U: %8.3f m\n',mean(e),mean(n),mean(u))
plot(1:qend,[(e-e(1))' (n-n(1))' (u-u(1))']*1000,'linewidth',2)
title('Estimates of Baseline Components using Kalman Filter','fontsize',16)
ylabel('State Vector, Changes Relative to Initial Epoch [mm]','fontsize',16)
xlabel('Epochs [1 s interval]','fontsize',16)
legend('East','North','Up')
set(gca,'fontsize',16)
legend
print -depsc easy6
%%%%%%%%%%%%%%%%%%%%%% end easy6.m %%%%%%%%%%%%%%%%%%%