Professional Documents
Culture Documents
Lecture 7
research group of Astronomy and Geomatics
Carrier-based Differential
Positioning.
Ambiguity Resolution
Techniques
Contact: jaume.sanz@upc.edu
gAGE/UPC
Authorship statement
The authorship of this material and the Intellectual Property Rights are owned by
J. Sanz Subirana and J.M. Juan Zornoza.
This authorship statement must be kept intact and unchanged at all times.
gAGE/UPC
22 Jan 2015
- LAMBDA Method
Double difference
(•) rujk ≡ ∇∆ (•) rujk = (•) kru − (•) ruj = (•)uk − (•) rk − (•)uj − (•) rj
P jk
ru = ρ jk
ru +T jk
ru +I jk
ru +ν jk
P ru
Station
• Receiver clock
The same for carrier : • Receiver code instrumental delays
Lrujk = ρ rujk + Truj − I rujk + λ ωrujk + λ N rujk + ν L rujk • Receiver carrier instrumental
delays
Master of Science in GNSS
Carrier ambiguities N are integer
@ J. Sanz & J.M. Juan
4
gAGE
Single-Difference of measurements
(corrected by geometric range!!)
research group of Astronomy and Geomatics
PRN06
PRN06
PLAN-GARR: 15km
PRN30
∆( L1 − ρ ) ≡ L1ru
sat
− ρ rusat PRN30 ∆ ( P1 − ρ ) ≡ L1ru
sat
− ρ rusat
constant
Dif. Tropo. and Iono. :
Small variations
Master of Science in GNSS @ J. Sanz & J.M. Juan
6
gAGE
Linear model for Differential Positioning
Single difference (•) ruj ≡ ∆(•) ruj =•
( )uj − (•) rj
research group of Astronomy and Geomatics
where:
ρ=j
ru ρ0 ruj − ρˆ 0uj ⋅ ∆rru − ρˆ 0 ruj ⋅ ε site + ρˆ 0 ruj ⋅ ε eph
j
Double difference
(•) rujk ≡ ∇∆ (•) rujk = (•) kru − (•) ruj = (•)uk − (•) rk − (•)uj − (•) rj
Exercise:
research group of Astronomy and Geomatics
where ρ ru= ρu − ρ r
j j j
For short baselines (e.g. up to 10 km) and if the reference ∆rru ≡ ∆ru − ∆rr
station coordinates are accurately known, we can assume:
Trujk 0 ; I rujk 0 ; ωrujk 0 Note for baselines up to 10 km
the range error of broadcast
ρˆ 0 ru
j
⋅ ε eph
j
0 orbits is less than 1cm
ε site 0 ⇒ ∆rru ∆ru (assuming ε eph j
10 m ).
Remark: Pru − ρ 0 ru = Pu − ρ 0 u − Pr − ρ 0 r
Master of Science in GNSS jk jk jk jk jk jk
( ) @ J. Sanz & J.M. Juan
9
gAGE
Differential code and carrier positioning
As with the SD, the left hand side of previous equations can be spitted
research group of Astronomy and Geomatics
in two terms: one associated to the reference station and the other to
the user. Then, the differential corrections can be computed for code
and carrier as: jk
PRCP jk
ρ 0 r − Pr ;
= jk jk jk
PRCL jk
ρ 0 r − Lr
=
• The user applies this differential correction to remove/mitigate common
errors:
Pu jk − ρ 0 ujk + PRCP jk = −ρˆ 0 ujk ⋅ ∆ru + ν P rujk
The previous system for navigation equations is written in matrix notation as:
Pref P R ,1 0u
(
− ρˆ 1 − ρˆ R T
0u
) 0 0
∆ru
R ,1 ( )
T
− ρ0u − ρ0u
ˆ 1
ˆ R
1 0
Pref L
λ R ,1
= N
ru
− ρˆ nu−1 − ρˆ Ru
( )
R , n −1
Pref P
T
0 0
λ N ru
R , n −1
Pref R ,n −1 0 0
( )
L T
− ρ
ˆ n −1
− ρ
ˆ R
0 1
0u 0u
gAGE/UPC
• The single difference (SD) and its covariance matrix can be computed as:
X uk
X ruk 1 −1 0 0 X rk Thence, if the measurements are
j =
− j ; uncorrelated, so are they in single
X ru 0 0 1 1 Xu
j differences, but the noise is twice!
X r
σ X2 0 0 0 1 0
1 −1 0 0 0 σ X2 0 −1 0
gAGE/UPC
0
PX SD = 2σ 2
I
0 0 1 −1 0 0 σX2
0 0 1 X
2
0 0 0 σ X 0 −1
computed as:
X k
X ru 1 −1 0 j
jk ru
jl = X ru ;
X ru 0 −1 1 l
X ru
2σ X2 0 0 1 0
1 −1 0 2σ 2 2 1
PX DD =
0 −1 1 0 2σ 2
X 0 −1 −1 X 1 2
2
0 0 2σ
X 0 1
•
∇
• ≡ •−
k R
•
∆≡ •rov − •ref ∆∇
≡ ∆ k −∆ R =
Receiver errors affecting SIS errors affecting both
both satellites are removed receivers are removed = ∇
rov −∇ ref
(e.g. Receiver clock) (e.g. Satellite clocks,...)
Receiver errors common for all satellites do not affect positioning (as
they are assimilated in the receiver clock estimate). Thence:
gAGE/UPC
- Only residual errors in single differences between sat. affect absolute posit.
- Only residual errors in double differences between sat. and receivers affect
relative positioning.
Master of Science in GNSS @ J. Sanz & J.M. Juan
15
gAGE
research group of Astronomy and Geomatics
Single and double differences of receivers/satellites
•
∇
• ≡ •−
k R
•
∆≡ •rov − •ref ∆∇
≡ ∆ k −∆ R =
Receiver errors affecting SIS errors affecting both
both satellites are removed receivers are removed = ∇
rov −∇ ref
(e.g. Receiver clock) (e.g. Satellite clocks,...)
- LAMBDA Method
ρ rujk = − ( ρˆ uk − ρˆ uj ) ⋅ rru =
ρ ruk − ρ ruj = −ρˆ ujk ⋅ rru
ρ u
j
ρ rj
research group of Astronomy and Geomatics
ρ rujk = − ( ρˆ uk − ρˆ uj ) ⋅ rru =
ρ ruk − ρ ruj = −ρˆ ujk ⋅ rru ρ=j
ru ρuj − ρ rj
Lrujk = ρ rujk + Truj − I rujk + λ ωrujk + λ N rujk + ν L rujk Lrujk =ρ rujk + λ N rujk +ν L rujk
ru u P ru
baseline, without needing an
−ρˆ ujk ⋅ rru + λ N rujk + ν L rujk
Lrujk = accurate knowledge of the
reference station coordinates.
Master of Science in GNSS Note: but, at least the approximate coordinates of @ J. Sanz & J.M. Juan
reference station or user are needed to compute ρ 19
gAGE
research group of Astronomy and Geomatics Relative Positioning
−ρˆ ujk ⋅ rru + ν P rujk
Prujk =
where ρˆ u jk ≡ ρˆ uk − ρˆ u j
−ρˆ ujk ⋅ rru + λ N rujk + ν L rujk
Lrujk =
The previous system for navigation equations is written in matrix notation as:
− ( ρˆ 1 − ρˆ R )T 0 0 Baseline
PrR,u,1 u u
R ,1 rru vector
Lr ,u − ( ρ
ˆ u
1
− ρ
ˆ u )
R T
1 0 R ,1
= λ N
ru
Satellite-j
PrR,u, n −1 − ( ρˆ n −1 − ρˆ R )T 0 0 ρuj
λ N ru
R , n −1
R , n −1 u u
ρ rj
Lr ,u − ρˆ n −1 − ρˆ R T 0 1
(u u ) ρˆ uj
rrur= r −r
ru= ruu− rr r
gAGE/UPC
Reference
Station
DD Code and
Carrier
Carrier
ambiguities ru rr
measurements ρ= ρuj − ρ rj
j
ru
∇∆L1 ≡ L1ru
sat
∇∆P1 ≡ P1ru
sat
Satellite-j
∇∆L1 ≡ L = − ρ
jk
ru
jk
ru +T jk
ru −I jk
ru +λω +λ N
jk
ru
jk
ru +ν jk
L ru
ρuj
ρ rj
Station
Line-Of-Sight vector
Master of Science in GNSS @ J. Sanz & J.M. Juan
21
gAGE
research group of Astronomy and Geomatics
Double-Difference of measurements
∇∆L1 ≡ L1ru
sat
∇∆P1 ≡ P1ru
sat
IND1-IND2: 7 m IND1-IND2: 7m
PRN06-PRN30 PRN06-PRN30
gAGE/UPC
∇∆L1 ≡ L1ru
sat
∇∆P1 ≡ P1ru
sat
Thence, the user can form the double differences of its own
measurements with those of the reference receiver, satellite by
satellite, and estimate its position relative to the reference receiver.
RRC dL1
d ρ up to ~800 m/s
clock adjust
dL1 dL1
300Km=1ms
dL1 ≈ d ρ + dtrec
d ρ up to ~800 m/s
( )
2 2
2ρuj + rru eˆ j ⋅ rru= ρ rj − ρuj = User Reference Station
= (
ρ rj − ρuj )( ρ j
r + ρuj )
(
= ( ρ rj − ρuj ) ρuj + rru + ρuj )
Then:
− (ω j eˆ j ) ⋅ rru
ρuj − ρ rj =
gAGE/UPC
- LAMBDA Method
Let us consider again the problem of relative positioning for short baselines.
research group of Astronomy and Geomatics
δ L12ru ( u
− δ −δ u )
ρ
ˆ 2
ρ
ˆ 1 T
13
− (δ ρˆ uk − δ ρˆ uj ) ( )
T
δ Lrujk = ⋅ rru + δν jk δ
⇒ ru
L − δ ρ
ˆ 3
− δ ρ
ˆ 1
rru + ν
= L ru
u u
1K
δ
ru − δ ρˆ K − δ ρˆ 1
( u u)
L
gAGE/UPC
T
where:
δ Lrujk ≡ Lrujk (t2 ) − Lrujk (t1 ) For simplicity, we assign (j=1) to the
reference satellite
Master of Science in GNSS @ J. Sanz & J.M. Juan
31
gAGE
The Role of Geometric Diversity: Triple differences
This is from [RD-3]
δ L12ru ( u u)
1 T
research group of Astronomy and Geomatics
− δ ρ
ˆ 2
− δ ρ
ˆ
13 1 T
− (δ ρˆ uk − δ ρˆ uj )
δ Lrujk = ⋅ rru + δν ⇒ ru ( u
δ − δ ρ − δ ρ u)
3
jk L ˆ ˆ
= L ru rru + ν
1K
δ Lru
( ˆ uK − δ ρˆ 1u )
T
− δ ρ
from [RD-3]
Let us now consider a simple model for the estimation of the relative
position vector from SD carrier measurements, assuming short baselines
(e.g. <10km):
Lruj = ρ ruj + c δ tru + λ N ruj + bru + ν L ruj Lruj =−ρˆ uj ⋅ rru + d ru + λ N ruj + ν L ruj
ρ ruj =ρuj − ρ rj =−ρˆ uj ⋅ rru where d ru ≡ c δ tru + bru
L1ru ( u )
−ρˆ 1 T
1 Satellite-j
λ N ru
1
ρuj
2
Lru ( −ρˆ u )
2 T
1 rru λ N ru2 ρ rj
= + +ν
d ru ρˆ uj
K K
gAGE/UPC
rru= ru − rr
Lru −ρˆ K T λ N ru
( u ) 1
Reference
Station
ρ=j
ru ρuj − ρ rj
ru rr
Master of Science in GNSS @ J. Sanz & J.M. Juan
To Earth Centre 33
gAGE Estimation of position and change in position:
the role of Geometric Diversity This is
research group of Astronomy and Geomatics
from [RD-3]
Previous system can be arranged as:
( −ρˆ 1 )T 1 Considering now differences between two
L 1 u
λ N ru
1
ru
epochs t0 and t1 , and assuming no cycle-slips:
=
L
2
( −ρˆ )
2 T
1 rru λ N ru2
+ν
+
ru u
d ru rru (t1 ) rru (t0 )
K
Lru −ρˆ K T
K
λ N ru
L ru (t1 ) − L= (t
ru 0 ) G (t1
) − G (t 0
) + ν
( u )
1 ru 1
d (t ) ru 0
d (t )
L ru G δ rru=
(t1 ) rru (t1 ) − rru (t0 ) −G (t1 ) + G (t1 )
− ( ρˆ 1 (t ) − ρˆ 1 (t ) )T 0
u 1 u 0
δ r (t ) r (t ) ˆ1
− ( ρu (t1 ) − ρˆ 1u (t0 ) )
T
L ru (t0 ) G (t1 ) ru 1 + ( G (t1 ) − G (t0 ) ) ru 0 + ν
L ru (t1 ) −= 0
G (t1 ) − G (t0 ) =
δ d ru (t1 )
d ru (t0 )
− ( ρˆ 1 (t ) − ρˆ 1 (t ) )T 0
u 1 u 0
d ru (t0 ) cannot be
the geometry matrix matrix at time t1 . This would estimated at all!
at time t1 . This can be poor determined if such
be well determined. change is not significant.
Master of Science in GNSS @ J. Sanz & J.M. Juan
34
gAGE
research group of Astronomy and Geomatics
Contents
1. Linear model for DGNSS: Double Differences
1.1. Differential Code and carrier based positioning
1.2. Precise relative Positioning
1.3. The Role of Geometric Diversity
2. Ambiguity resolution Techniques
2.1. Resolving ambiguities one at a time
- Single-frequency measurements
- Dual-frequency measurements
- Three-frequency measurements
2.2 . Resolving ambiguities as a set: Search techniques
- Least-Squares Ambiguity Search Technique.
gAGE/UPC
- LAMBDA Method.
ωrujk 0 q
ru
σ L ≈ 0.5cm
q
σ L ≈ 1cm
jk
measurement error.
q
AsMaster
commented before, the ambiguity terms are integer numbers, and@we
of Science in GNSS can take benefit
J. Sanz & J.M. Juan
of this property to fix such ambiguities applying integer ambiguity resolution techniques.36
gAGE
research group of Astronomy and Geomatics
Contents
1. Linear model for DGNSS: Double Differences
1.1. Differential Code and carrier based positioning
1.2. Precise relative Positioning
1.3. The Role of Geometric Diversity
2. Ambiguity resolution Techniques
2.1. Resolving ambiguities one at a time
- Single-frequency measurements
- Dual-frequency measurements
- Three-frequency measurements
2.2 . Resolving ambiguities as a set: Search techniques
- Least-Squares Ambiguity Search Technique.
gAGE/UPC
- LAMBDA Method.
P= jk
ρ jk
+ ν jk
L1jk − P1 jk
1 P1
→ L − P1 = λ1 N +ν →
jk jk jk jk
N1 =
ˆ jk
L1jk =ρ jk + λ1 N1jk +ν Ljk1
1 1 P1
λ1 roundoff
λ1 20 cm
σ P ≈ 1m 1
1
jk
σ Nˆ σP ≈ 5
λ1
jk jk
σ L ≈ 1cm
jk
1 1
1
N +1
Fail Too much error (5 wavelengths)!
N + 1/ 2 Note that, assuming a Gaussian
Good N̂
N distribution of errors, σ Nˆ jk 1/ 2
guarantee only the 68% of success
1
N − 1/ 2
N −1 As the ambiguity is constant (between
cycle-slips), we would try to reduce
gAGE/UPC
- LAMBDA Method.
= N1 − N 2
NW
P= jk
ρ jk +ν Pjk1 f1 P1 jk + f 2 P2jk λW =
c
86.2 cm
P= = ρ jk +ν PjkN
1 jk
f1 − f 2
f1 + f 2
N
P=
2 ρ +ν
jk jk jk
P2
ρ + λ1 N +ν
jk
L =
jk jk jk jk N 1
f1 − f 2 σ L ≈ 6 σ L ≈ 6 cm
1 1 L1 W
jk jk
ρ + λ2 N +ν
W 1
L =
jk jk jk jk
2 2 L2
↓
L − P = λW N +ν
jk jk jk jk
→ LWjk − PNjk
W N W PN Nˆ jk
=
W
λW roundoff
Fixing N1 (after fixing NW)
1 71cm
σ Nˆ ≈ σP 0.8
L − L = λ1 N
jk
1
jk
2 1
jk
− λ2 N 2
jk
+ν jk
L1 − L2
jk
W λW N
jk
86.2cm
( λ1 − λ2 ) N1jk + λ2 NWjk +ν Ljk1 − L2
=
Now, with uncorrelated
λ1 = 19.0 cm
L1jk − L2jk − λ2 Nˆ Wjk measurements from 10
gAGE/UPC
λ2 = 24.4 cm Nˆ jk
= epochs will reduce noise
λ2 − λ1 =
1
λ − λ roundoff
5.4 cm 1 2
up to about ¼.
σ L ≈ 1cm
jk
1
1 1.4cm
σ Nˆ ≈ 2σ jk 1/ 4
Nˆ=jk
Nˆ 1jk in−GNSS
Nˆ Wjk 1
jk
λ1 − λ2 L1 5.4cm
Master
2 of Science @ J. Sanz & J.M. Juan
42
gAGE
research group of Astronomy and Geomatics
gAGE/UPC
Once the integer ambiguities are known, the carrier phase measurements
become unambiguous pseudoranges, accurate at the centimetre level (in DD),
gAGE/UPC
or better.
Thence, the estimation of the relative position vector is straightforward
following the same approach as with pseudoranges.
Master of Science in GNSS @ J. Sanz & J.M. Juan
44
gAGE
Exercises:
research group of Astronomy and Geomatics
Hint:
LW= λW φW ; φW = φ1 – φ2
- LAMBDA Method.
for short baselines (e.g. < 10 km) Ionosphere, troposphere and wind-
up differential errors cancel.
=
2
f1 − f 2
γ +1
f 2 L2 − f5 L5 σL = 25 σ 33,3 cm
LEW = ρ + λEW N EW +ν LEW
=
EW
γ 25 − 1 L 1
N= N − N W 1 2
f 2 − f5
gAGE/UPC
f1 P1 + f 2 P2
γ 12 + 1 = N 2 − N5
N EW
σP = σ 0, 712 m
P= = ρ +ν PN N
γ 12 + 1 P
1
f1 + f 2
N
γ +1 Exercise:
f P + f5 P5 σP = 25 σ 0, 707 m
= 2 2
PEN = ρ +ν PEN γ 25 + 1 P Justify the previous
f 2 + fin5 GNSS
EN 1
expressions for σ. 47
Master of Science @ J. Sanz & J.M. Juan
We still consider the above problem of relative positioning in DD
gAGE for short baselines (e.g. < 10 km) Ionosphere, troposphere and wind-
up differential errors cancel .
research group of Astronomy and Geomatics
L1 =ρ + λ1 N1 +ν L1 σ L ≈ σ L ≈ 1cm
GPS Frequency Wavelengths Combinations 1 2
=γ 12 (=
f1 / f 2 )
2
L − PEN 1 0.71m (77 / 60) 2
Nˆ EW = EW σ Nˆ ≈ σP 0.12
λEW = γ 25 (=
f 2 / f5 )
2
λ EW roundoff
EW EN
5.861m (24 / 23) 2
γ 12 + 1
σP = σ 0, 71m
N
γ 12 + 1 P 1
γ 12 + 1
gAGE/UPC
σL = σ 5, 7 cm
L1 − L2 − λ2 Nˆ W 1 1.4cm
W
γ 12 − 1 L 1
N1 =
ˆ σ Nˆ ≈ 2σ 1/ 4
λ1 − λ2 γ 25 + 1
λ1 − λ2 5.4cm
1 L1
roundoff σL = σ 33,3 cm
EW
γ 25 − 1 L 1
=γ 12 (=
f1 / f 2 )
2
L − PEN 1 (77 / 59) 2
Nˆ EW = EW σ Nˆ ≈ σP [ ]
λEW =γ 23 (=
f 2 / f3 )
2
λ EW roundoff
EW EN
(118 /115) 2
γ 12 + 1
σP = σ [ ]
N
γ 12 + 1 P 1
λEW Nˆ EW − ( LEW − LW ) 1 γ 25 + 1
NW =
ˆ σ Nˆ ≈ σL [ ] σP = σ [ ]
λ W roundoff W λW EW EN
γ 25 + 1 P 1
γ 12 + 1
σ [ ]
gAGE/UPC
σL =
L1 − L2 − λ2 Nˆ W 1
W
γ 12 − 1 L 1
N1 =
ˆ σ Nˆ ≈ 2σ [ ]
λ1 − λ2 γ 25 + 1
λ1 − λ2
L1
σ [ ]
1
roundoff σL =
EW
γ 25 − 1 L 1
=γ 12 (=
f1 / f 2 )
2
L − PEN 1 0.71m (77 / 59) 2
Nˆ EW = EW σ Nˆ ≈ σP 0.07
λEW = γ 23 (=
f 2 / f3 )
2
λ EW roundoff
EW EN
9.768 m (118 /115) 2
γ 12 + 1
σP = σ 0, 71m
N
γ 12 + 1 P 1
γ 12 + 1
gAGE/UPC
σL = σ 5.4 cm
L1 − L2 − λ2 Nˆ W 1 1.4 cm
W
γ 12 − 1 L 1
N1 =
ˆ σ Nˆ ≈ 2σ 1/ 4
λ1 − λ2 γ 25 + 1
λ1 − λ2 5.8cm
1 L1
roundoff σL = σ 54.9 cm
EW
γ 25 − 1 L 1
- LAMBDA Method.
Prefit-residual
ρ=
jk
(ti ) ρ jk
(ti ) − ρˆ 0 jk (ti ) ⋅ ∆r (ti ) G(t)
0 y(t)
The problem can be easily reformulated for the kinematic case. Kalman filtering
can be applied as well.
For an observation span relatively long, e.g. one hour, the floated ambiguities
would typically be very close to integers, and the change in the position
solution from the float to the fixed solution should not be large.
1. Initialization time
2. Reliability (or, correctness) of the integer estimates
• Estimate the floated solution N and its uncertainty (e.g. Nˆ =2502347.74 cycles, σ Nˆ = 0.6 cycles)
(
• Define as a volume to be searched e.g. ± 3σ Nˆ ±2 cycles and )
evaluate the cost function (the RMS residuals) over the 6 ambig.: 2502345, , 2502350
The integer ambiguity solution corresponding to the smallest RMS residuals is used
to select the candidate. However if two or more candidates give roughly similar
values of RMS, the test can not be resolute.
A ratio test (of 2 or 3, depending of the algorithm) between the two smallest RMS
gAGE/UPC
- LAMBDA Method
Residual of real-valued
(
floated solution ∆rˆ , N
ˆ )
[*] Remember that DD measurements are correlated, as already
Master of Science in GNSS
seen.
@ J. Sanz & J.M. Juan
58
gAGE LAMBDA Method
Thence, we have to find ∆r a 3-vector of real numbers, and N a (K-1)-vector
research group of Astronomy and Geomatics
of integers minimizing:
2 2
y − G ∆r − λ A N = y − G ∆rˆ − λ A N + ∆r − ∆rˆ (N) + λ2 N − N
2 2
ˆ ˆ
Wy Wy W∆rˆ ( N ) WNˆ
This term is irrelevant for This term can be This term must
minimization since it does made zero for be minimized
not depend on ∆r and N any N over the integers
Float solution and covariance matrix:
2
min N − N
ˆ → N
WNˆ
∆rˆ ∆rˆ P∆rˆ
ˆ ; Cov =
P∆rˆ ,Nˆ
ˆ ˆ −1
∆r =∆r (N) =∆r − W∆rˆ ,Nˆ WNˆ N − N
ˆ ( )
N Nˆ P ˆ PNˆ
∆rˆ ,N
gAGE/UPC
( ) ( )
2 T
c(N) =N −N
ˆ =N − N
ˆ WNˆ N − N
ˆ
WNˆ = PNˆ−1
WNˆ
• A diagonal WN matrix would mean that the integer ambiguity estimates are
uncorrelated.
• If the weight WN matrix is diagonal, the minimizing of the cost function is trivial. The
best estimate is the float ambiguity rounded to the nearest integer.
N̂ 2
( ) ( )
2 2
1/ σ ˆ ˆ
2
0 = N1 − N ˆ
1 N 2 − N ˆ
2
c ( N ) +
WNˆ =
N1N1
σ N2ˆ Nˆ σ N2ˆ Nˆ
0 1/ σ 2
N̂1
1 1 2 2
Nˆ 2 Nˆ 2
Ellipse parallel to coordinate axes
The idea is to apply a transformation that decorrelates the ambiguities so that the matrix
W becomes diagonal. W is a positive definite matrix and thence, can be always
diagonalized (as a real-valued matrix) with orthogonal eigenvectors. But the problem here
is that the integer ambiguities N must be transformed preserving its integer nature!
Thence, we are looking for an “integer-valued” transformation matrix Z that makes the
matrix W as close as possible to a diagonal matrix (decorrelating as much as possible
the ambiguities) and with similar axes (spherical).
Moreover, the inverse of transformation
matrix Z-1 must be also integer, to transform
N′ = Z N back the results after finding the ambiguities
PNˆ ′ = Z PNˆ ZT
ˆ ′ = ZN
N ˆ Note that Z, Z −1 ∈ ⇒ det( Z) = 1
(i.e. it is a volume-preserving transformation)
Z
gAGE/UPC
Pictures
from
[RD-6]
Exercise:
Show that:
−1
Z, Z ∈ ⇒ det(Z) =
1
Gauss manipulation over matrix P=W-1 can be applied to find-out the matrix Z.
PNˆ =
1 2
α i = − int pNˆ Nˆ / pNˆ Nˆ
pNˆ1Nˆ 2 pNˆ Nˆ
2 2 1 α 2
1 2 i i
Step 1:
1 −1 We transform first the element with
Z2 = − int [38.4 / 28.0] =
α2 = −1 largest variance (in this case N1)
0 1
1 −1 53.4 38.4 1 0 4.6 10.4 The half,
=PNˆ ′ Z=P Z T
0 1 38.4 28.0 =
at most!
−1 1 10.4 28.0
2 Nˆ 2
In general,
Step 2: to increase the
1 0
gAGE/UPC
1 0 1 −1 1 −1
=Z Z= −2 1 0 =
1 −2 3
Z
1 2
gAGE/UPC
1 −1 53.4 38.4 1 −2 4.6 1.2
=
PNˆ ′′ Z=
PN ZT 38.4 28.0 −=
−2 3 1 3 1.2 4.8
Master of Science in GNSS @ J. Sanz & J.M. Juan
65
Example:
gAGE 1.05 53.4 38.4 4.6 1.2
N=
ˆ
PNˆ = PNˆ ′′ =
1.30 38.4 28.0 1.2 4.8
research group of Astronomy and Geomatics
1 −1
Z=
−2 3
−1 0 3 1 0 2
=N Z= 2 2 =
1 2 2
1 N2
p p12 ( p11 + p22 + w ) N 2′
research group of Astronomy and Geomatics
P = 11 λ1=
p22 N1′
2
p12 1
λ2= ( p11 + p22 − w ) λ1
2 λ2 φ
w= ( p11 − p22 ) 2 + 4 p122 N1
λ 0
P′ = 1
0 λ2 tan 2φ =
2 p12
p11 − p22
Example:
53.4 38.4
PNˆ = ˆ = 1.05
N 1.30
38.4 28.0
=λ1 =
81.14 9.0
gAGE/UPC
81.14 0
′
PNˆ = λ2 =
=
0.25
0.25 0.5
0 tan 2φ= 3.02 ⇒ φ= 3585
BACKUP
Master of Science in GNSS @ J. Sanz & J.M. Juan
67
gAGE Consider again the previous problem of estimating ∆r , a 3-vector of real
numbers, and N a (K-1)-vector of integers, which are solution of
y = G ∆r + λ A N + ν
research group of Astronomy and Geomatics
• Integer rounding
Comment:
• Integer bootstrapping In principle, the previous transformation Z
is not required by the estimation concept;
• Integer Least-Squares
it is only to achieve considerable gain in
• ….. speed in the computation process [RD-5].
−0.25 0
=N′′ int
= 1.80 2
1. We start with the most precise ambiguity (here we will assume Nn)
2. Then, the remaining float ambiguities are corrected taking into account
their correlation with the last ambiguity.
N n = int Nˆ n Using the triangular
decomposition
N n −1 = ˆ
int N n −1|n =
ˆ
n−1 n
−2
(
int N n −1 − σ Nˆ , Nˆ σ Nˆ N n − N n
n
ˆ
) PNˆ = LT D L
lij = σ Nˆ σ N−ˆ2
( )
n
N1 = ˆ int N1 − ∑ σ Nˆ , Nˆ σ Nˆ N i|I − N i
int N n|I = ˆ −2 ˆ ˆ
j , N i|I i|I
i =2
1 i|I i|I
gAGE/UPC
Nˆ i|I Stands for the i-th ambiguity obtained through a conditioning of the
previous I= {i + 1, , n} sequentially rounded ambiguities.
Float
N2
Conditional
estimate
N̂1|2
N1
gAGE/UPC
= =
N 2 nint Nˆ 2 0
N= 1 nint ˆ =
N
1|2 nint
N1 N 2 , N1 N 2
2 (
ˆ − σ ˆ ˆ σ −ˆ2 Nˆ − N =
2
1
)
1. The target to find the integer vector N which minimizes the cost function
research group of Astronomy and Geomatics
( ) ( ) WNˆ = PNˆ−1
2 T
c(N) =N − N
ˆ =N − N
ˆ PNˆ−1 N − N
ˆ
PN−ˆ 1
2. The integer minimiser is obtained through a search over the integer grid
points on the n-dimensional hyper-ellipsoid:
( ) ( )
T
ˆ ≤ χ2
ˆ P ˆ−1 N − N
N −N N
( N − Nˆ ) ( ) ( N − N ) ( )
T
ˆ ≤ χ2
L−1 D−1 L−T N − N
T
≤ χ2
D−1 N − N
Defining:
(N ) + (N ) (N )
gAGE/UPC
2 2 2
1 − N1 2 − N2 n − Nn
=N − L−T (N − N
N − N =(N
ˆ ) → LT N (
ˆ − N)
= )
c( N ) + + ≤ χ2
d1 d2 dn
But N i depends on Nˆ i +1 , , Nˆ n .
N n = Nˆ n
(
− N = (N
LT N ˆ − N) )
∑ (N )
n
N i =Nˆ i + j − Nˆ i l ji ; i =n − 1, n − 2, ,1
j = i +1
( ( ) ) ( ( ) )
2 1/ 2 2 1/ 2
N n −1 − d n1/−21 χ 2 − N n − Nˆ n dn ≤ N n −1 ≤ N n −1 + d n1/−21 χ 2 − N n − Nˆ n dn
1/ 2 1/ 2
2 n 1/ 2
( ) ( )
n
χ − ∑ N j − N j d j ≤ N1 ≤ N n + d1 χ − ∑ N j − Nˆ j
2 2
N 1 − d1/ 2
1
ˆ 2
dj
j 2= j 2
gAGE/UPC
• A small χ2 may result in a ellipsoidal region that fails to contain the solution.
• A too large value for χ2 may result in high time-consuming for the search process.
• If p candidates are requested, the values of cost function c(N) are ordered in
ascending order and χ2 is chosen equal to the p-th value.
If more than n+1 candidates are requested, the volume of the search ellipsoid can be used
Master of Science in GNSS
([RD-6]).
@ J. Sanz & J.M. Juan
74
gAGE
Search with shrinking technique: practical example
This is an alternative to the previous strategy, based on shrinking the search
ellipsoid during the process of finding the candidates.
research group of Astronomy and Geomatics
4th nearest is
out of the
region.
No more
candidates at
gAGE/UPC
this step
Thence, the best 6 candidates are found (in the ISL sense). The one
with
Master thein GNSS
of Science smallest cost function c(N) value is the actual ISL@ J.solution.
Sanz & J.M. Juan
76
gAGE
Acceptance Test
research group of Astronomy and Geomatics
However if two or more candidates give roughly similar values of RMS, the
test can not be resolutive.
( ) ( )
T
RMS =N − N
ˆ = N −N
ˆ PNˆ−1 N − N
ˆ
gAGE/UPC
PN−ˆ 1
Note:
This document uses
the transposed
matrix ZT, but the
gAGE/UPC
principle is the
same.
load large Q, a
imagesc(Q) imagesc(Qz)
colorbar colorbar
Z
gAGE/UPC
[L,D] = ldldecom(Q)
research group of Astronomy and Geomatics
Qz=Lz’*diag(Dz)*Lz
Q=L’*diag(D)*L
-3 2 3 10 -8 -2 -5 0 -4 1 -4 0
-1 6 8 -1 2 1 2 7 3 -2 6 1
-8 7 -8 3 -6 -1 1 0 0 3 -1 -1
8 1 6 -3 5 4 -5 -3 0 -0 1 -3
round(a)
[ - 28491 65753 38830 5004 -29196 -298 -22201 51236 30258 3899 -22749 -159]
Decorrelation + bootstrapping
[Qz,Zt,Lz,Dz,az,iZ] = decorrel (Q,a)
azfixed=bootstrap(az,Lz);
gAGE/UPC
afixed=iZ*azfixed
[ -28451 65749 38814 5025 -29165 -278 -22170 51233 30245 3916 -22725 -144]
afixed=iZ*azfixed
[ -28451 65749 38814 5025 -29165 -278 -22170 51233 30245 3916 -22725 -144]
-28727 65935 39032 4844 -29337 -178 -22385 51378 30415 3775 -22859 -66 33.9
-28546 66062 39027 4998 -29228 -83 -22244 51477 30411 3895 -22774 8 34.5
-28229 65518 38583 5197 -29056 -500 -21997 51053 30065 4050 -22640 -317 34.7
-28365 65586 38683 5084 -29124 -418 -22103 51106 30143 3962 -22693 -253 35.5
afixed=iZ*azfixed
[ -28451 65749 38814 5025 -29165 -278 -22170 51233 30245 3916 -22725 -144]
-28727 65935 39032 4844 -29337 -178 -22385 51378 30415 3775 -22859 -66 33.9
-28546 66062 39027 4998 -29228 -83 -22244 51477 30411 3895 -22774 8 34.5
-28229 65518 38583 5197 -29056 -500 -21997 51053 30065 4050 -22640 -317 34.7
-28365 65586 38683 5084 -29124 -418 -22103 51106 30143 3962 -22693 -253 35.5
[RD-7] XW. Chang, CC. Paige, L. Yin, 2005, Code and carrier phase based
short baseline GPS positioning: computational aspects. GPs Solutions,
2005 (2005) 9:72-83, DOI 10.1007/s10291-004-0112-8 Springer.
Master of Science in GNSS @ J. Sanz & J.M. Juan
84
gAGE
research group of Astronomy and Geomatics Least-Squares Ambiguity Search technique
1
y = G ∆r + λ N + ν N=
ˆ ( y − G ∆rˆ )
λ
K-1 equations with Then , given the 3-D position, the (K-1) integer
3+(K-1) unknowns. ambiguities can be resolved automatically.
Thence, That is, the (K-1) integer ambiguities are
K-4 equations are redundant constrained to three degrees of freedom.
gAGE/UPC
• The tracked satellites into two groups: 4 sat. (with good DOP) + N-4 sat.
research group of Astronomy and Geomatics
• The primary group of 4 sat., is used determine the possible ambiguity sets N ( k )
(given an initial position estimate and its associated uncertainty ∆rˆ ± σ )
∆rˆ ± σ → N (=
k)
1
λ
( y − G ( ∆rˆ + δ ) )
(k ) → N(k )
(G G )
−1
N(k ) =
→ ∆rˆ( k ) T
G y − λ N ( k )
- LAMBDA Method