Part III : Positioning with GNSS

Positioning in multi-GNSS mode

2
 GNSS compatibility
• The emergence of new GNSS and the modernization of already existing
GNSS entails discussion over compatibility and interoperability.
• “Compatibility refers to the ability of global and regional navigation satellite
systems and augmentations to be used separately or together without
causing unacceptable interferences and/or other harm to an individual
system or services”.
• Compatibility means than GNSS systems do not interfere with each other.
• All the existing GNSS have been designed in order to be compatible.

3
 GNSS interoperability 1
• “Interoperability refers to the ability of global and regional navigation
satellite systems and augmentations and the services they provide to be
used together to provide better capabilities at the user level than would be
achieved by relying solely on the open signals of one system”.
• Interoperability is considered at system level or at signal level.
• Interoperability at system level: The capability of all systems to provide the
same solution standalone within the system accuracy.
• Interoperability at signal level : The different GNSS transmit signals allowing
to combine them in a “simple” receiver to give a combined solution
(improved performances obtained from combination).

4
 GNSS interoperability 2
• Interoperability at signal level depends on :
• Reference frames:
• is the difference between reference frames below target accuracy ?
• Time reference:
• same problem; the time difference between GNSS can also be estimated as and additional
unknown.
• Signal (carriers, codes, navigation messages, ..)
• The use of the same carrier frequencies has a high impact on receiver cost and allows
improved accuracy.
• Common centre frequencies (carriers) are required to have signal interoperability.

5
GNSS metrics

6
 GNSS metrics
• There are several ways to measure the performances of a GNSS.
• Accuracy
• Which accuracy level can be expected ?
• Depends on the positioning technique, the equipment used, …
• Robustness (service availability and continuity)
• Am I able to use the system when I need it ?
• Integrity
• Am I sure that the position estimate is accurate enough to meet my requirements ?
• Crucial for Safety-of-Life applications (plane landing).
• Will the system send me warnings when the accuracy is out of tolerance ?
7
Accuracy and Precision 1

8
 Accuracy and Precision 2
• If the “true” station coordinates are available, the accuracy (errors) can be
obtained by comparing each “instantaneous” solution to the true
coordinates.
 rError  rTrue  rSolution

• If the “true” station coordinates are not available, the precision (residuals) can
be obtained by comparing each “instantaneous” solution to the mean
coordinates.
 rRe sidual  rMean  rSolution

9
Positioning modes

10
 Different positioning modes 1
• Different positioning modes can be distinguished.
• Code- or phase-based positioning:
• Code or phase observables are both a measure of receiver-to-satellite distance.
• Code-based positioning usually gives an accuracy of a few meters using cheap
receivers and simple data processing techniques as code measurements are not
ambiguous.
• Phase-based positioning gives an accuracy ranging from a few decimetres to a few
millimetres:
• Phase measurements are more precise but they are ambiguous.
• It is mandatory to solve ambiguities before computing precise positions.
• The ambiguity resolution requires more sophisticated data processing techniques.
• Receivers able to make phase measurements are more expensive.
• Code and phase observable can also be “mixed” in the data processing procedures.
11
 Different positioning modes 2
• Absolute, differential or relative positioning:
• Absolute positioning:
• A standalone receiver computes its absolute position in the ECEF reference frame using
only its own GNSS measurements (code and/or phase).
• Single Point Positioning (SPP) = code-based absolute positioning.
• Precise Point Positioning (PPP) = phase-based absolute positioning.
• Differential positioning: it is a special case of absolute positioning; a standalone
receiver computes its absolute position in the ECEF reference frame using its own
GNSS measurements AND so-called differential corrections coming from reference
stations.

12
 Different positioning modes 3
• Absolute, differential or relative positioning:
• Relative positioning:
• The receiver (station B with unknown position 𝑅𝐵 ) processes its own GNSS measurements
together with the GNSS measurements collected by a reference station (station A whose position
in the ECEF frame 𝑅𝐴 is known).
• The baseline vector 𝑏𝐴𝐵 between the receiver and the reference station is obtained from “mixed”
data processing ( = station A + station B together).
• Then, using the known reference station position and the computed baseline, the receiver’s
absolute position can be obtained.

𝑅𝐵 = 𝑅𝐴 + 𝑏𝐴𝐵

13
 Different positioning modes 4
• Real time or post processing:
• Real time: the user needs to compute his position immediately after the code/phase
measurements have been made.
• Post processing: the user collects data during a given time interval and stores the data
which are processed with a delay wrt the time of measurement.
• Static or kinematic:
• Static: receiver with fixed coordinates.
• Kinematic: moving receiver.
• Kinematic positioning versus dynamic positioning: In dynamic positioning,
accelerations are measured in order to reconstruct the user’s trajectory. In kinematic
positioning with GNSS, the successive positions of the user are measured at a high rate
so that the trajectory can be reconstructed.

14
 Different positioning modes 5

• Examples:
• Mass market = car navigation: Real time absolute kinematic positioning with code
measurements (accuracy between 5 and 10 m during favourable conditions).
• Surveyors: Real time differential kinematic positioning with code and phase
measurements (accuracy of a few centimetres).
• Geophysicists/Geodesists : Post processed static relative positioning with code and
phase measurements (accuracy of a few millimetres).

15
 Chapter 1

Absolute positioning
 Single Point Positioning 1
• Originally, GPS has been designed to perform Single Point Positioning (SPP).
• SPP uses code measurements to perform absolute positioning, in real time
or in post processing, in kinematic or static mode.
• The observation model can be written:

Ppi  D pi  Tpi  I pi  M pi , g  c  t i   t p   pi , g 
• The receiver-to-satellite geometric distance :
X   Y   Z 
2 2 2
D 
i
p
i
Xp i
 Yp i
Zp

With (𝑋𝑝 , 𝑌𝑝 , 𝑍𝑝 ) and (𝑋 𝑖 , 𝑌 𝑖 , 𝑍 𝑖 ), the receiver “p” and satellite “i” coordinates. 17
 Single Point Positioning 2

• Usually, the receiver coordinates and clock error are computed from code
pseudorange measurements using the least square technique.
• To apply the least square technique to the observation model, it is mandatory
to have a linear relationship between the observations 𝑃𝑝𝑖 and the unknowns
(𝑋𝑝 , 𝑌𝑝 , 𝑍𝑝 , Δ𝑡𝑝 ).
• Therefore, the observation model must be linearized wrt to the unknowns
𝑋𝑝 , 𝑌𝑝 , 𝑍𝑝 (it is already linear in Δ𝑡𝑝 ).

18
 Single Point Positioning 3

• In order to simplify our discussion, we assume that:

• Multipath is negligible.
• The tropospheric and ionospheric errors have been computed using models giving the
𝑖 𝑖
values 𝑇𝑝,𝑚𝑜𝑑 et 𝐼𝑝,𝑚𝑜𝑑 .
• The satellite clock error is obtained from the model contained in the broadcast
𝑖
ephemeris giving the value ∆𝑡𝑚𝑜𝑑 .
• The satellite coordinates (𝑋 𝑖 , 𝑌 𝑖 , 𝑍 𝑖 ) are obtained from the broadcast ephemeris.

19
 Single Point Positioning 4
• The model becomes :
X   Y   Z 
2 2 2
P 
i
p
i
Xp i
 Yp i
Zp  c  t p  Tpi ,mod  I pi ,mod  c  t imod

• 𝐷𝑝𝑖 = 𝑓(𝑋𝑝 , 𝑌𝑝 , 𝑍𝑝 ) has to be linearized wrt (𝑋𝑝 , 𝑌𝑝 , 𝑍𝑝 ).

• We assume that we have an a priori value (𝑋𝑝,0, 𝑌𝑝,0, 𝑍𝑝,0) of (𝑋𝑝 , 𝑌𝑝 , 𝑍𝑝 ) around
which 𝐷𝑝𝑖 will be linearized : X p  X p ,0   X p
Yp  Yp ,0   Yp
Z p  Z p ,0   Z p

• We define the a priori receiver-to-satellite distance :

     
2 2 2
D pi ,0  X i  X p ,0  Y i  Yp ,0  Z i  Z p ,0

 f X p ,0 , Yp ,0 , Z p ,0  20
 Single Point Positioning 5
• Then:
f f f

f  X p , Yp , Z p   f X p ,0 , Yp ,0 , Z p ,0   X p  Yp  Z p
X p Yp Z p
X p , 0 , Yp , 0 , Z p , 0 X p , 0 , Yp , 0 , Z p , 0 X p , 0 , Yp , 0 , Z p , 0

• With : f  X i
 X p ,0 
 
X p D pi ,0
X p , 0 , Yp , 0 , Z p , 0

f
 
 Y i
 Yp ,0 
Yp D pi ,0
X p , 0 , Yp , 0 , Z p , 0

f
 
Z i
 Z p ,0 
Z p D pi ,0
X p , 0 , Yp , 0 , Z p , 0

• The linearized geometric distance :

D D
i i

 X i
 X p ,0   X  Y i
 Yp ,0  Y   Z i
 Z p ,0  Z 21
p p ,0 i p i p i p
D p ,0 D p ,0 D p ,0
 Single Point Positioning 6
• The model becomes : Ppi,0

P i

 X i
 X p ,0   X  Y i
 Yp ,0  Y   Z i
 Z p ,0  Z  c  t  D i
 T i
 I i
 c  t i
p p p p p p ,0 p ,mod p ,mod mod
D pi ,0 D pi ,0 D pi ,0

• We define the a priori (modelled) pseudorange observation 𝑃𝑝,0

𝑖
:

Ppi,0  D pi ,0  Tpi ,mod  I pi ,mod  c  t imod

• We replace it into the model :

P P
i i

 X i
 X p ,0  X 
 Y i
 Yp ,0  Y 
 Z i
 Z p ,0  Z  ct p
p p ,0 p p p
D pi ,0 D pi ,0 D pi ,0

4 unknowns 22
 Single Point Positioning 7

• Lets remark that the receiver clock error changes continuously with time and
must therefore be estimated at each epoch (i.e. for each time of measurement).
• We define :
• n = the number of available observations
• u = the number of unknowns.
• If n < u : no solution.
• If n ≥ u : the system can be solved using least square technique.

23
 Single Point Positioning 8
• For example, we assume that we track 4 satellites (i=1,2,3,4) at a given epoch:

P P
i i

 X i
 X p ,0  X 
 Y i
 Yp ,0  Y 
 Z i
 Z p ,0  Z p  c  t p
p p ,0 p p
D pi ,0 D pi ,0 D pi ,0

 l1 
 2
l  3 
We define l
Applied to the 4 equations In matrix form
l 
l i  Ppi  Ppi,0  4 
l 1  d 1  X p  e 1  Yp  f 1  Z p  c  t p l 
X i
 X p ,0 
d i 
D pi ,0 l 2  d 2  X p  e 2  Yp  f 2  Z p  c  t p l Ax
e i

Y i
 Yp ,0  l 3  d 3  X p  e 3  Yp  f 3  Z p  c  t p
D pi ,0
l 4  d 4  X p  e 4  Yp  f 4  Z p  c  t p d1 e1 f1 c  X p 
 Z  Z p ,0
i
  2
e2 f2 c
 


A 3
f i  d Y
x  
p
i
D p ,0 d e3 f3 c   Zp 
 4   
d e4 f4  c    t p 
24
Direction cosines of the
receiver-to-satellite vectors
 Single Point Positioning 9
• In the case of an identity weight matrix (all observation have same precision), the estimated
unknown vector 𝑥ො obtained by least squares :

xˆ  ( AT A) 1 AT l

• The 4 unknowns are computed based on code pseudorange measurements which are
“indirect measurements” .
• In this case, the precision of the computed unknowns is obtained by applying error
propagation to the observation model.
• In practice, the relationship between individual code pseudorange accuracy 𝜎𝑜𝑏𝑠 and
the position accuracy 𝜎𝑝𝑜𝑠 is given by:

 pos  PDOP obs

25
 Single Point Positioning 10
• The factor “PDOP” or “Position Dilution of precision” is obtained from the
cofactor matrix of the unknowns :
PDOP  q XX  qYY  q ZZ

With:

 q XX q XY q XZ q Xt 
 
 
1 q qYY qYZ qYt 
Qx  A A T Qx   YX
 qZX qZY qZZ qZt 
 
 qtX qtY qtZ qtt 
Contains the direction cosines of
the receiver-to-satellite vectors

26
 Single Point Positioning 11
• In practice, the PDOP depends on satellite geometry (i.e. satellite distribution
in the user’s sky).
• PDOP can be geometrically interpreted as being inversely proportional to the
volume of the tetrahedron formed by the 4 receiver-to-satellite directions:
• Larger volume  smaller PDOP  better position accuracy.

27
 Single Point Positioning 12
• Recall of the 2D navigation case where we demonstrated the influence of the
emitter “geometry” on the position error.

x
 Single Point Positioning 13
• PDOP and number of observed GPS satellites in Liège:

29
 Precise Point Positioning 1
• Basically, Precise Point Positioning (PPP) is based in the same principle as SPP
except that phase pseudoranges are used instead of code pseudoranges :


 ip (t )  D pi  Tpi  I pi  M pi ,  c  t i (te )   t p (t )   N pi 
• 𝑖
As in the SPP case, we neglect multipath, compute model values 𝑇𝑝,𝑚𝑜𝑑 𝑖
, 𝐼𝑝,𝑚𝑜𝑑 𝑖
, ∆𝑡𝑚𝑜𝑑
and 𝐷𝑝𝑖 is linearized:

 
i i

 X i
 X p ,0   X  Y i
 Yp ,0  Y   Z i
 Z p ,0  Z  c  t p   N pi
p p ,0 p p p
D pi ,0 D pi ,0 D pi ,0

• If we define :

l i   ip   ip ,0
30
 Precise Point Positioning 2
• Again, if 4 satellites are tracked at a given epoch, the model can be rewritten in
matrix form:

l Ax
 Xp 
 

 p 
Y
Z 
 l1   d1 e1 f1  0 0 c 
0  p

 2  2   Np 
1
e2 f2 0  0 c 
l  3 
l
A 3
d 0
l  x 2 
d e3 f3 0 0  0 c   Np 
 4   4   3 
l  d e4 f4 0 0 0  c   Np 
 4 
 Np 
t 
 p 

31
 Precise Point Positioning 3
• As compared to SPP, the use of ambiguous phase measurements brings
additional unknowns (1 unknown per observed satellite).
• In our example, 4 satellites are tracked; let’s further assume that the receiver
coordinates are fixed :
• Epoch 1 : 4 equations (measurements) and 8 unknowns  No solution.
• Epoch 2 : 8 equations and 9 unknowns (additional receiver clock error)  No solution.
• Epoch 3 : 12 equations and 10 unknowns (additional receiver clock error)  A solution
exists.
• In practice, PPP requires “external” information (precise orbits and satellite
clocks).
• Gives few cm (static) to dm level (kinematic) accuracy after a rather long
“convergence” time (about 1 hour) necessary to compute ambiguities
32
 Chapter 2

Differential positioning
 Differential Positioning 1

• Differential positioning also called DGNSS is a special case of absolute

positioning.
• It relies on the availability of reference station(s) or even networks of
reference stations whose positions are accurately known.
• The user determines his position based on the measurements collected by his
receiver (code or/and phase pseudoranges) in addition to so-called
differential corrections provided by one or several reference stations.
• Differential corrections allows to improve the accuracy of the computed
positions.

34
 Differential Positioning 2
• There are 3 different approaches to compute differential corrections.
• Solution-domain approach :
• The reference station computes its position in real time using its own GNSS receiver.
• The difference between the computed position and the reference (“real”) position is
transmitted to the user receiver often called rover receiver.
• The rover applies the position correction to its own computed position.
• This method requires that :
• User and rover receivers are tracking the same satellites (influence of PDOP !).
• User and rover receivers are close to each other.

35
 Differential Positioning 3
• Measurement-domain approach:
• Satellite coordinates can be obtained from broadcast ephemeris and reference station
coordinates are know.
• Therefore, the reference station can compute the “reference” receiver-to-satellite
distance.
• This reference distance is compared to receiver-to-satellite distances measured by code
and phase pseudoranges at the reference station.

36
 Differential Positioning 4
• Measurement-domain approach:
• The difference between measured pseudorange and reference range (= differential
range correction) which is due to errors (ionosphere, troposphere, …) is transmitted to
the rover receiver which can correct its own code and phase pseudoranges.
• This approach is valid as long as reference station and rover receiver are affected by the
same errors.
• In practice, the validity of differential corrections decreases with increasing distance
between reference station and rover receiver.
• Only one reference station is necessary: it is called Local Area Differential GNSS
(LADGNSS).

37
 Differential Positioning 5
• Space-state approach:
• A network of reference stations is used to derive models of different errors over the
whole network which are transmitted to the users.
• For example, a model of ionospheric vertical delays can be computed at grid points
above the network.
• It is the most effective and sophisticated approach which allows to apply DGNSS on
longer distances.
• It is called Wide Area Differential GNSS (WADGNSS)

38
 DGNSS with codes 1
• We consider a reference station (index “r”) and a user receiver (index “u”)
tracking a common satellite (index “i”) at epoch tk.
• The reference station code pseudorange model is given by:
P  t   D  t   T  t   I t   M t   c   t t     t t    
r
i
k r
i
k r
i
k r
i
k
i
r,g k
i
k
i
r ,g r k
i
r,g

• Based on know reference station and satellite coordinates obtained from

broadcast ephemeris, an approximate value of 𝐷𝑟𝑖 called (𝐷𝑟𝑖 )𝑒𝑝ℎ :

 
Dri  Dri
eph
 Eri

• 𝐸𝑟𝑖 represents the error in the computation of 𝐷𝑟𝑖 due to errors in broadcast ephemeris.

39
 DGNSS with codes 2
• The model for the reference station becomes :
  t   E t   T t   I t   M t   c   t t
Pr i  tk   Dri
eph k r
i
k r
i
k r
i
k
i
r ,g k
i
k  
  ri , g   tr  tk    ri , g

• In the same way, the model for the rover receiver can be written :
  t   E t   T t   I t   M t   c   t t
Pui  tk   Dui
eph k u
i
k u
i
k u
i
k
i
u,g k
i
k   i
u,g 
  t u  t k   i
u,g

Cannot be estimated as rover coordinates are unknown.

• The differential correction (range correction) is defined as :

  t   P t 
DC  tk   Dri
eph k r
i
k

  
  Eri  tk   Tr i  tk   I ri  tk   M ri, g  tk   c  t i tk   ri , g   tr  tk    ri , m 
40
 DGNSS with codes 3
• A corrected code pseudorange 𝑃෢ is obtained by applying the differential
𝑖
𝑢
correction to the user code pseudorange measurement :
Pui  tk   Pui  tk   DC  tk 

• It gives :

   t   E t   T t   I t   M t   c   t t    t t    
Pui  tk   Dui
eph k
i
ur k
i
ur k
i
ur k
i
ur,g k r k u k
i
u r ,m

• With : iu r ui ir

41
 DGNSS with codes 4
• The differential correction computed at the reference station using the
measurements collected at time tk does not reach the user “instantaneously”.
• The “latency” is defined as the time interval between tk and the time when the
rover receives the differential correction and applies it to its pseudorange
measurements.
• Typically, the latency is of the order of 3-4 s but latencies up to 10 s are
acceptable.
• To take the latency into account, the reference station also transmits the
differential correction rate of change (range rate) DCR so that:

DC  t   DC  tk   DCR  t  tk 

42
 DGNSS with codes 5

• The quality of differential corrections depends on the correlation (or the

absence of correlation) between errors affecting the reference stations and the
rover : high correlation  better correction.
• Correlation with distance : The shorter the distance between reference and
rover, the smaller the residual errors 𝐼𝑢𝑟
𝑖 , 𝑇 𝑖 , 𝐸𝑖 .
𝑢𝑟 𝑢𝑟

• Correlation with time : All errors (including 𝐼 𝑖

𝑢𝑟
𝑖 , 𝐸 𝑖 ) become uncorrelated
, 𝑇𝑢𝑟 𝑢𝑟
when latency increases but the satellite clock error gives the main contribution
to the error budget through the term   t i  tk  ri  tk    t i t  ui t  

43
 DGNSS with codes 6
• Uncorrelated errors : Multipath and receiver clock errors are not correlated.
• Multipath is a local effect; multipath at reference station and at rover combine in the term
𝑖
𝑀𝑢𝑟, 𝑔.

• Therefore, it is important to install reference stations in an environment with low multipath

to avoid to “contaminate” the user pseudoranges due to multipath contained in the
differential correction.
• The reference station and user receiver clocks combine in the term (∆𝑡𝑟 − ∆𝑡𝑢 ). This is not
a problem: the user estimates (∆𝑡𝑟 − ∆𝑡𝑢 ) instead of ∆𝑡𝑢 as clock error term.

44
 DGNSS with codes 7
• Example of DGNSS result (GPS C/A code, distance = 80 km) : depending on
distance between reference station and rover, an accuracy ranging from 0.7 m
to 3 m can be expected.
Erreur de positionnement en Latitude
Dourbes jour 120 en 2003

Sans corr dgps Brus

Avec corr dgps Brus
20

10

Erreur de positionnement (mètres)

0

-10

-20 45
0 5 10 15 20 25
Temps (heure)
 DGNSS with phases 1
• DGNSS based on phase measurements allows to reach an accuracy of a few
cm in real time both in static or in kinematic modes.
• The differential corrections are computed in the same way as for code
pseudoranges :

   
PDC  tk    Eri  tk   Tr i  tk   I ri  tk   M ri,  tk   c  t i tk   ri ,   tr  tk    N ri   ri ,  tk 

• The corrected user phase pseudorange Φ෢ 𝑖

𝑢 is :

  tk    ui  tk   PDC  tk 
i
u

• It can be rewritten :
   t   E t   T t   I t   M t   c   t t    t t     N
  tk   Dui   ui r ,  tk 
i
i i i i i
u
eph k ur k ur k ur k u r , k r k u k ur

46
 DGNSS with phases 2
• DGNSS with phase measurements can offer an accuracy of a few
centimetres only on short distances: the distance between the reference
station and the rover should not exceed 10 to 15 km:
• On short distances, the residual errors 𝐼 , 𝑇 , 𝐸 usually remain small.
𝑖
𝑢𝑟
𝑖
𝑢𝑟
𝑖
𝑢𝑟

• Multipath effect on phase pseudorange is usually small (cm level).

• Nevertheless, during a so-called “initialization phase”, the user must solve
the combined phase ambiguity before being able to measure positions.
• The Real Time Kinematics (RTK) positioning technique is based on the
principle of DGNSS with phase.
• As far as surveyors are concerned, RTK is the preferred positioning
technique.
47
 Chapter 3

Relative positioning
 Relative positioning 1
• Given a (fixed) station A whose position 𝑅𝐴 is known and a station B whose
position 𝑅𝐵 is unknown, the principle of relative positioning is to determine
the baseline vector separating A and B 𝑏𝐴𝐵 in order to compute 𝑅𝐵

RB  RA  bAB

• The basic observable used in relative positioning is the double difference of

code or phase measurements.
• In general, only phase measurements are used due to the high code observable
noise.
49
 Relative positioning 2
• A double difference is obtained from 2 single differences.
• Given 2 receivers A, B and a satellite i tracked by both receivers, a single
difference is the difference of 2 phase measurements made simultaneously
by receivers A and B on satellite i :
 AB
i
 t    Ai  t    Bi  t 

 Ai  t  
f
c
     
DAi  t   TAi  t   I Ai  t   M Ai ,  t   f  t i t   Ai ,   t A  t   N Ai   Ai ,  t  Receiver A

 Bi  t  
f
c
     
DBi  t   TBi  t   I Bi  t   M Bi ,  t   f  t i t   Bi ,   t B  t   N Bi   Bi ,  t  Receiver B

 AB
i
t  
f
c

i
DAB 
 t   TABi  t   I ABi  t   M ABi ,  t   f   t B t    t A t    N ABi   AB
i
,  t 

50
With  i
AB  
i
A
i
B
 Relative positioning 3
• If the 2 phase measurements are simultaneous :
 t  t     t  t    
i i
A,
i i
B , 0

• Single difference properties :

• The atmospheric errors 𝑇𝐴𝑖 , 𝑇𝐵𝑖 and 𝐼𝐴𝑖 , 𝐼𝐵𝑖 are correlated with baseline length. In other
𝑖 𝑖
words, the shorter the baseline, the smaller 𝑇𝐴𝐵 and 𝐼𝐴𝐵 .
• The satellite clock error is removed if phase measurements are simultaneous.
• 𝑖
Multipath errors are not removed but are combined in 𝑀𝐴𝐵 .
• Nevertheless, the receiver A and B clock errors are still present.

51
 Relative positioning 4
• The double difference is the difference of 2 single differences made
simultaneously by 2 receivers A, B on 2 satellites i, j :
 AB
ij
 t    ABi  t    ABj  t 

 AB
i
t  
f
c

i
DAB 
 t   TABi  t   I ABi  t   M ABi ,  t   f   t B t    t A t    N ABi   ABi , t 

 j
AB
c
 
 t   DABj  t   TABj  t   I ABj  t   M ABj ,  t   f   t B t    t A t    N ABj   ABj , t 
f

 AB
ij
t  
f
c
ij
DAB  
 t   TABi j  t   I ABi j  t   M ABi j ,  t   N ABi j   ABi j ,  t 

With  AB
ij
 AB
i
 AB
j 52
 Relative positioning 5
• On short distances (10-15 km), one assumes that residual errors are
negligible: ij
TAB 0
ij
I AB 0
ij
M AB 0
• In this case, neglecting the noise, the double difference model becomes:
 AB
ij
 DAB
ij
  N AB
ij

• In order to compute station B coordinates, the term 𝐷𝐴𝐵

𝑖𝑗
must be linearized.
ij
DAB  DAi  DAj  DBj  DBi

53
 Relative positioning 6
• Again, we assume that we have an a priori value (𝑋𝐵,0 , 𝑌𝐵,0 , 𝑍𝐵,0 ) of (𝑋𝐵 , 𝑌𝐵 , 𝑍𝐵 ) :
i or j
X B  X B ,0   X B
     
2 2 2
YB  YB ,0   YB DB*,0  X *  X B ,0  Y *  YB ,0  Z *  Z B ,0
Z B  Z B ,0   Z B

• In the same way as in the case of SPP and PPP :

D D
j j

 X j
 X B ,0   X  Y j
 YB ,0  Y   Z j
 Z B ,0  Z
B B ,0 B B B
DBj,0 DBj,0 DBj,0

D D
i i

 X i
 X B ,0  X 
 Y i
 YB ,0  Y 
 Z i
 Z B ,0  Z
B B ,0 B B B
DBi ,0 DBi ,0 DBi ,0

54
 Relative positioning 7
• The double difference model becomes:

 D D D D 

 X i  X B ,0  

X j  X B ,0    X 
  
 Y i  YB ,0 Y j  YB ,0    Y   
 Z i  Z B ,0

Z j  Z B ,0    Z   N AB
B   B  
ij i j j i ij
AB A A B ,0 B ,0
 DBi ,0 DBj,0   DBi ,0 DBj,0   DBi ,0 DBj,0  B
     

We define :
𝑖𝑗 𝑗 𝑗
𝛷𝐴𝐵,0 = 𝐷𝐴𝑖 − 𝐷𝐴 + 𝐷𝐵,0 − 𝐷𝐵,0
𝑖

𝑋 𝑖 − 𝑋𝐵,0 𝑋 𝑗 − 𝑋𝐵,0
𝑑 𝑖𝑗 = 𝑖
− 𝑗
𝐷𝐵,0 𝐷𝐵,0 The model is rewritten :
𝑌 𝑖 − 𝑌𝐵,0 𝑌 𝑗 − 𝑌𝐵,0
𝑒 𝑖𝑗 = 𝑖
− 𝑗 l i j  d i j  X B  e i j  YB  f ij
 Z B   N AB
ij
𝐷𝐵,0 𝐷𝐵,0
𝑍 𝑖 − 𝑍𝐵,0 𝑍 𝑗 − 𝑍𝐵,0
𝑓 𝑖𝑗 = 𝑖
− 𝑗
𝐷𝐵,0 𝐷𝐵,0
𝑖𝑗 𝑖𝑗
𝑙 𝑖𝑗 = 𝛷𝐴𝐵 − 𝛷𝐴𝐵,0 55
 Relative positioning 8
• If 4 satellites are tracked during 2 epochs, the system becomes :

l Ax

 l i j  t1    d i j  t1  e i j  t1  f i j  t1   0 0  XB 
 ik   ik   
 l  t1    d  t1  e i k  t1  f i k  t1  0  0   YB 
 l il  t1    d il  t1  e il  t1  f il  t1  0 0    ZB 
l  i j  A i j  x  i j 
  2 
l t  d  t2  e i j  t2  f i j  t2   0 0  N AB 
 l i k t    d i k t  e i k  t2  f i k  t2  0  0  N AB
ik 

 il 2   il 2   il 
 l t    d t  e il  t2  f il  t2   
 2   2 0 0  N AB 

56
Influence of residual errors
 Influence of station A position accuracy 1
• Relative positioning is based on the “known” station A position.
• What is the influence of station A position accuracy on the baseline
determination ?
• This can be estimated using the following rule of thumb:

Error on baseline length dbAB dRA Error on station A position


Baseline length bAB RAi Receiver-to-satellite distance

58
 Influence of station A position accuracy 2
• Example: If we want to have an accuracy of 1 cm on a 20 km baseline, what is
the minimal accuracy on station A position ?

dbAB  1 cm  102 m

bAB  20 km  2.104 m dRA  11 m

RAi  22 000 km  2, 2.107 m

59
 Influence of multipath
• Multipath is a local effect: therefore, its influence on double differences does
not depend on baseline length.
• The multipath contained in the 4 individual phase measurements (also called
“one-way” measurements) combine in double differences.
• In theory, as the maximum multipath effect is about 5 cm on one-way
measurements, the maximum effect on double differences is 20 cm but the
probability have very high multipath at the same time on the 4 observables
is very low.
• Under normal conditions, the multipath effect is of the order of 1-2 cm on
one way phase pseudoranges giving a total effect of a few cm in double
differences.
• Nevertheless, when possible, users should avoid sites with reflecting objects.
60
 Influence of the ionosphere 1
• Under normal circumstances, the residual ionospheric effect on double
differences can be considered as negligible on distances up to 10-15 km.
• For longer baselines, double differences of the ionosphere-free combination
have to be used which makes the ambiguity resolution more difficult.
• Nevertheless, 2 main types of ionospheric disturbances can be the origin of
local variability in TEC even on short distances (a few km):
• Noise-like structures in TEC : they are usually observed during disturbed Space Weather
conditions (geomagnetic storms).
• Travelling Ionospheric Disturbances (TIDs) which are waves in the electron density (and
as a consequence in TEC) usually propagating from North to South; they are observed
during all periods of the Solar activity cycle, even under quiet ionospheric activity
conditions; they are more frequent during winter time between 10h00-16h00 local
time.
61
 Influence of the ionosphere 2
• We show the influence of a TID on a double difference.
• Therefore, we choose 2 stations A and B both with precisely known
coordinates and we compute:
 AB
ij
 DAB
ij


• If our assumptions are valid (i. e. we have assumed that residual errors in
double differences are negligible on short baselines), then this quantity is
constant over time; indeed:

 AB
ij
 DAB
ij
 N AB
ij
Constant integer

62
 Influence of the ionosphere 3
• Baseline Uccle-Saint Gilles (!4 km !) on DOY 359/2004 (quiet ionospheric activity).

Doubles Differences L1 Doubles Differences L1

Day : 359 Year : 2004 Day : 359 Year : 2004
1064158.8 -568733.2
BRUS - ST GILL BRUS - ST GILL
satellites 29-26 satellites 17-21

1064158.4 -568733.6
Distance (cylces)

Distance (cylces)
1064158 -568734

1064157.6 -568734.4

1064157.2 -568734.8

2 3 4 5 6 7 8 11.6 12 12.4 12.8 13.2 13.6

Universal Time (30 sec) Universal Time (30 sec)

63
No ionospheric disturbance Effect of a Travelling Ionospheric
Disturbance
 Influence of the troposphere
• On short baselines, the tropospheric residual error is usually negligible: in
general, the tropospheric error variability is much smaller than the ionospheric
error variability.
• Nevertheless, the passage of a weather front, a violent storm, … can be the
origin of increased local variability in water vapour which can give rise to non
negligible tropospheric residual errors.
• On baselines longer than 20 km, other techniques must be applied in order to
mitigate the residual tropospheric influence.

64
 Real Time Kinematics (RTK) 1
• RTK is the surveyor’s preferred positioning technique.
• It is a kinematic positioning technique which provides an accuracy of a few
cm in real time.
• It requires a network of reference stations and can work both in differential
and in relative modes.
• The relative mode require a “heavier” data transfer from the reference
station to the user as the reference station measurements are processed
together with the user measurements (double differences).

65
 Real Time Kinematics (RTK) 2

• Nevertheless, the relative mode gives the best accuracy mainly due to the fact
that, the receiver clock error does not influence double differences while it must
be estimated (= 1 additional unknown) in differential mode.
• In practice, users can start to measure positions after a short (static) initialization
phase during which ambiguities are solved to the correct integer.

66