You are on page 1of 138

Terrestrial Navigation

Chapter 3:
Basic principles of terrestrial navigation

Priv.-Doz. Dr.-Ing. habil. Michael Meurer


German Aerospace Center

Email: Michael.Meurer@tum.de
24.10.2012

DLR und TU-München, Kommunikation und Navigation


Outline and Structure

3. Basic principles of terrestrial navigation

3.1 Dead Reckoning

3.2 Proximity Systems

3.3 Distance based Navigation

3.4 Distance difference based Navigation

3.5 Distance ratio based Navigation

3.6 Angle-of-arrival based Navigation

3.7 Signature based Navigation

3.8 Cooperative navigation in radio and sensor networks


24.10.2012

DLR und TU-München, Kommunikation und Navigation


2
Dead Reckoning

 Dead reckoning is the process of x m ,3


estimating one's current position based
upon a previously determined position, or
fix, and advancing that position based x m ,2
2
upon known speed, elapsed time, and
course.
1
 Measurements of speed: odometer, e.g. MS
wheelsensor or inertial sensor
(accelerometer) x2 x m ,1
 Measurement of course: compass or
inertial sensor (gyroscope) ( x3 ) x1
 Dead reckoning has been used in many  v cos  i 
applications like e.g. marine navigation, x m ,i 1  x m ,i   t  i 
v sin
 i i 
air navigation 
24.10.2012

displacement vector
DLR und TU-München, Kommunikation und Navigation
3
Dead Reckoning – Principle of Inertial Navigation
Today’s application of dead reckoning: inertial navigation systems

 Initial position, velocity & heading

 Sense Rotation

 Sense Acceleration

 Dead Reckoning: Finally you have all you need


24.10.2012

DLR und TU-München, Kommunikation und Navigation


4
Dead Reckoning – Inertial Sensor Platforms

 Full six-degree-of-freedom inertial sensor platform


– 3 Accelerometers
– 3 Gyroscopes
 Strapdown: Virtual alignment with attitude computer
 Gimbaled: Platform is kept aligned to navigation frame

Strapdown Gimbaled
24.10.2012

DLR und TU-München, Kommunikation und Navigation


5
Dead Reckoning – Inertial Sensor Performance

 Inertial navigators are known to drift. The drift arises due to


– Initial alignment errors
– Sensor imperfections (noise, biases, scale factors errors, axis misalignment,
nonlinearities, vibration & temperature sensitivity)
– Earth model errors (Earth gravity model errors)
– Computational errors (Numerical errors, e.g. due to series truncation,
numerical integration or round off)

 Drift Errors are motion dependent and nonlinear


24.10.2012

DLR und TU-München, Kommunikation und Navigation


6
Dead Reckoning – Example: NavShoe (1)

 Low-Cost MEMS-based multi-sensor


platform (Price 1.600 € in 2005)
 Size 5.7 x 4.6 x 1.1 cm
 Idea: NavShoe (Foxlin 2005)
– Mounting of INS on foot
– Classical INS computer
(Accelerometers & Gyros)
Foot mounting
– Drift compensation via step detection
& velocity zero pseudo measurement
when foot rests (ZUPT)

Foot at rest

24,8 % 20,5 % 38,0 % 16,7 %


24.10.2012

Stride Typical acceleration profile


DLR und TU-München, Kommunikation und Navigation
7
Dead Reckoning – Example: NavShoe (2)

Meeting
Room

Stride Rest phase DLR IKN building TE02


24.10.2012

DLR und TU-München, Kommunikation und Navigation


8
Proximity Systems

Proximity positioning systems are also


known as signpost or beacon systems,
the position estimates for an object are
determined from its proximity to fixed
detection devices

beacon
 Examples for detection devices: magnetic
sensors, radio transmitter, …
 Simplest positioning system
 Either self- or remote positioning
24.10.2012

DLR und TU-München, Kommunikation und Navigation


9
Proximity Systems –
Example Cell-ID Positioning

Today’s mobile radio systems are cellular:


A cell is served by a base station (BS) which has a unique identifier (cell id)
Cells usually don’t overlap. Geographic areas served by a base station and
location of each base station are known quite accurately.
Each mobile receives broadcast messages of all base station in reach and can
identify the “strongest” (probability the closest)

Positioning of a mobile station by


 determining the closest base station and
 using the reference position of the
respective cell as position estimate.

Possible reference positions: location of


base station, center of gravity of cell,..
Problem: correct determination of closest
BS, poor accuracy due to size of cells
24.10.2012

DLR und TU-München, Kommunikation und Navigation


10
Recall: Indirect Positioning

estimation geometry
of based x̂ m
M̂ geometric position
quantities estimator
Indirect Position Estimator

Examples for geometric quantities:


distances,
distance differences,
distance ratios
24.10.2012

DLR und TU-München, Kommunikation und Navigation


11
Recall: Indirect Positioning

estimation geometry
of based x̂ m
M̂ geometric position
quantities estimator
Indirect Position Estimator

Examples for geometric quantities:


distances,
distance differences,
distance ratios
24.10.2012

DLR und TU-München, Kommunikation und Navigation


12
Distance Based Positioning

1 distance ̂ 1 

estimator distance-
based
x̂ m
position
ˆ KB  distance ˆ 
KB 
estimator
h
estimator
MS position estimator

1. distance est.: 2. distance-based position est.:


ˆ  kB   kB  ˆ 1   ˆ
KB 
h 
ˆ x̂ m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


13
Position Determination by Multilateration

Positions of anchor points and distances between MS and APs are known:

AP 1
Anchor point (AP)


Mobile station (MS)

Observation area 
MS
AP KB

Position of MS can uniquely be deter-
mined, if distances to at least 3 APs
AP 2 are known (in 2D)
24.10.2012

DLR und TU-München, Kommunikation und Navigation


14
3D Position Determination by Multilateration

Positions of anchor points and distances between MS and APs are known:

 Anchor point (AP)



Mobile station (MS)
Possible Positions

Position of MS can uniquely be deter-


mined, if distances to at least 4 APs
 are known (in 3D) or additional
information can be used to exclude
one of the solutions
24.10.2012

DLR und TU-München, Kommunikation und Navigation


15
Trilateration –
Simple Mathematical Approach

 xMS  x1    y MS  y 1   12
2 2
1
AP 1
 MS 2   MS 2    22
2 2
2 x  x  y  y

 MS 3   MS 3    32
2 2
 3 x  x  y  y

MS
1 - 2 2 x MS  x 2  x1   2 y MS  y 2  y 1 
AP K3
 12   22  x12  x 22  y 12  y 22

1 - 3 2 x MS  x 3  x1   2 y MS  y 3  y 1 

AP 2  12   32  x12  x 32  y 12  y 32

xMS , y MS : Unknown position MS


Position determination by solving set of
xkB , y kB : Known position AP kB linear equations and cross-checking three
 k : Known distance MS ↔ AP kB original quadratic equations
24.10.2012

DLR und TU-München, Kommunikation und Navigation


16
Tri- und Multilateration

Perfect distance knowledge: Imperfect distance knowledge

AP 1 AP 1

̂1 ̂1

ˆ K B
ˆ K B

MS MS
AP K B AP K B

̂ 2 ̂ 2

AP 2 AP 2

Probabilistic geometry based approach for


positioning necessary (see next chapter 4)
24.10.2012

DLR und TU-München, Kommunikation und Navigation


17
Distance Based Positioning

1 distance ̂ 1 

estimator distance-
based
x̂ m
position
ˆ KB  distance ˆ 
KB 
estimator
h
estimator
MS position estimator

1. distance est.: 2. distance-based position est.:


ˆ  kB   kB  ˆ 1   ˆ
KB 
h 
ˆ x̂ m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


18
Distance Estimation
 kB
distance ˆ

estimation
distance estimator

ˆ ˆ 1 

ˆ
M  h ,..., h
K  B

 M 0 
ˆ  ˆ 1 ,..., ˆ  KB 
0 

ˆ  gˆ 1 ,..., gˆ  K 
M B

time based path-loss phase difference
approach based approach based approach
c0
 kB     k   
ˆ kB gˆ  ˆ kB ˆ0 B ˆ kB
k  k
hˆ Tˆ B B
distance distance
t estimation estimation
Tˆ kB
24.10.2012

DLR und TU-München, Kommunikation und Navigation


19
Single Trip Delay - Basic Idea

1. MS transmits a test signal at time instant

observation area  tT
AP 1
2. AP 1 receives the test signal at time instant
1
tR

MS
AP KB
1 1

 M  c0 t R  tT
ˆ 
AP 2
self and remote positioning possible !
24.10.2012

DLR und TU-München, Kommunikation und Navigation


20
Single Trip Delay - Inaccurate Clocks
Clock error MS: Clock error AP:
1
nC,MS nC,AP
observation area 
AP 1
1. MS transmits a test signal at time instant
tT reported as tT  nC,MS
2. AP 1 receives the test signal at time instant
1 1 1
MS tR reported as tR  nC,AP
AP KB
1 
ˆ M 
1 
 
1 
 c 0  t R  n C(1),A P  t T  n C ,M S 
 
1  1 
 c 0  t R  t T   c 0  n C ,A P  n C ,M S 
   
          
 M1  n M1 
AP 2
clocks in MSs and in APs have to
24.10.2012

be perfectly synchronized
DLR und TU-München, Kommunikation und Navigation
21
Round Trip Delay – Basic Idea
1. MS transmits a test signal at time instant
1 ()
1 1
tT considered as tT  nC,MS
observation area  2. AP 1 receives the test signal at time instant
1 1 1
AP 1 tR considered as tR  nC,AP

MS
AP KB

AP 2
24.10.2012

DLR und TU-München, Kommunikation und Navigation


22
Round Trip Delay – Basic Idea
1. MS transmits a test signal at time instant
1 ()
1 1
tT considered as tT  nC,MS
observation area  2. AP 1 receives the test signal at time instant
1 1 1
AP 1 tR considered as tR  nC,AP
2
3. AP1 transmits a test signal at time
1  2  1 2
instant tT  t R   ()
known
MS 4. MS receives the test signal at time instant
AP KB  2  2
tR considered as tR  nC,MS


ˆ M
1
 c0
 2
  1
t R  nC,MS  tT  nC,MS     c0
 2,1 1
t R  tT  
AP 2 2 2
MS and APs have to transmit a test signal
24.10.2012

DLR und TU-München, Kommunikation und Navigation


23
Time Based Approach –
Summary & Pros/Cons
Basic Idea of Time-Based / Time-of-Arrival (TOA) approach:
 Measurement of propagation time between APs and MS
 Knowledge of propagation speed of radio wave (speed of light in vacuum)
 Assumption of LOS to calculate distance between APs and MS

Pros:
 Quite accurate
 Moderate complexity

Cons:
 Clocks of all APs and MS have to be accurately synchronized if single trip
measurements are used
 Round trip measurements consum double the number of radio ressources
 In NLOS scenarios measured propagation time is too long and does not allow to draw
conclusions about the required geometric distance
24.10.2012

DLR und TU-München, Kommunikation und Navigation


24
Distance Estimation
 kB
distance ˆ

estimation
distance estimator

ˆ ˆ 1 

ˆ
M  h ,..., h
K  B

 M 0 
ˆ  ˆ 1 ,..., ˆ  KB 
0 

ˆ  gˆ 1 ,..., gˆ  K 
M B

time based path-loss phase difference
approach based approach based approach
c0
 kB     k   
ˆ kB gˆ  ˆ kB ˆ0 B ˆ kB
k  k
hˆ Tˆ B B
distance distance
t estimation estimation
Tˆ kB
24.10.2012

DLR und TU-München, Kommunikation und Navigation


25
Path-loss based Approach – Basic Idea (1)

Determine path loss a(ρ) by measuring received power Pr(t) for known
transmitted power Pt:

ˆ 4,367m observation area 


AP 1 Pr  t 

a
 t 
MS
Pt AP K B

Pt
a   
Pr  t  self and remote positioning possible !
24.10.2012

DLR und TU-München, Kommunikation und Navigation


26
Path-loss based Approach – Basic Idea (2)
Determine estimate ̂ of distance  due to relationship between distance  and a :
10log10 a    dB
Random Distribution
due to Slow Fading observation area 
AP 1 Pr  t 

a
Mean attenuation
 t 
am    dB MS
10 dB per decade Pt AP K B
 m
10 100 1000 10000

2
  c  
am    dB   g m    dB   10 log10  gt  g r      10 log10 

    4 f 


g m,0 dB  const
24.10.2012

DLR und TU-München, Kommunikation und Navigation


27
Path-loss based Approach – Distance Estimation (1)

Inverse of attenuation between MS and AP k B : gain g  k B 

mathematically: Due to shadowing effect gain g  k B  is a random


variable with a specific PDF (here: Lognormal-distribution)
   g ( k B)  ( k B)  
2

10 1 1

p g ( k B) ;  ( k B)  
ln 10  2  a g (kB)
exp   2 
 2 a 
10 log10 
g ( k B)   

  m ,0  

Estimate the distance ˆ  k B , which is a parameter of the PDF,


based on N independent measurements of the gain g n k B

24.10.2012

DLR und TU-München, Kommunikation und Navigation


28
Path-loss based Approach –
Distance Estimation (2)
g n kB 
Observation:
Gain (in dB) is
Gn
 kB 
 10 log 10 g n  ,  kB 

Gaussian distributed n  1 N

 kB  1 N
  kB 
G 
N
G
n 1
n

suboptimal
two-step approach
 kB    kB 
  g  kB 
m,0 10 G 10

ˆ  kB 
24.10.2012

DLR und TU-München, Kommunikation und Navigation


29

Path-loss based Approach – PDF of 

50
N  1 30
40  a  8 dB
 4
30
 
p    20

10

0
0 0.5 1 1.5 2
 

24.10.2012

DLR und TU-München, Kommunikation und Navigation


30
Path-loss Based Approach –
Summary & Pros/Cons
Basic Idea of Path-loss Based / Signal Strength Approach:
 Measurement of received power Pr of signal exchanged between APs and MS
 Knowledge of transmitted power Pt of signal
 Calculation of instantaneous path loss / gain using Pr and Pt
 Calculation of mean path loss / gain to eliminate the influence of slow fading
k 
 Assumption of LOS and propagation model incl. its parameters  and g m,0B to
calculate distance between APs and MS
Pros:
 Low complexity

Cons:
 Sensitive to slow and fast fading as well as model errors
 Low accuracy, accuracy decreases with absolute distance to be estimated
 Averaging over fading only possible if sufficient movement of MS
k 
 Parameters  and g m ,0B of the propagation model have to be known
 Transmitted power Pt has to be known which is critical if power control is applied
 In NLOS scenarios measured path loss does not allow to draw conclusions about the
24.10.2012

required geometric distance


DLR und TU-München, Kommunikation und Navigation
31
Distance Estimation
 kB
distance ˆ

estimation
distance estimator

ˆ ˆ 1 

ˆ
M  h ,..., h
K  B

 M 0 
ˆ  ˆ 1 ,..., ˆ  KB 
0 

ˆ  gˆ 1 ,..., gˆ  K 
M B

time based path-loss phase difference
approach based approach based approach
c0
 kB     k   
ˆ kB gˆ  ˆ kB ˆ0 B ˆ kB
k  k
hˆ Tˆ B B
distance distance
t estimation estimation
Tˆ kB
24.10.2012

DLR und TU-München, Kommunikation und Navigation


32
Phase Difference Based Approach – Basic Idea (1)

Transmit test signals of two different frequencies f1 and f2 simultanously and


exploit initial phase difference 0 at receiver:

observation area 
   
s  t ;    A cos  2  ft    AP 1 0
   
 t 
MS
AP K B

Tx Rx
f2 >f1
distance
f1
24.10.2012

 t 
DLR und TU-München, Kommunikation und Navigation
33
Phase Difference Based Approach – Basic Idea (1)

Transmit test signals of two different frequencies f1 and f2 simultanously and


exploit initial phase difference 0 at receiver:

    observation area 
s  t ;    A cos  2  f1t   
  1   AP 1 0

     t 
s  t ;    A cos  2  f 2 t    MS
  2  
AP K B

Tx Rx
f2 >f1
distance
f1
24.10.2012

 t 
DLR und TU-München, Kommunikation und Navigation
34
Phase Difference Based Approach – Basic Idea (2)

Transmit test signals of two different frequencies f1 and f2 simultaneously and


exploit initial phase difference 0 at receiver:

  observation area 
 0,1   2 
1  AP 1 0
   0   0,1   0,2
 
 0,2   2  t 
2  MS
AP K B

Tx Rx
f2 >f1
 0
distance
f1
24.10.2012

 t 
DLR und TU-München, Kommunikation und Navigation
35
Phase Difference Based Approach – Basic Idea (3)

Transmit test signals of two different frequencies f1 and f2 simultaneously and


exploit initial phase difference 0 at receiver:

 1 1  observation area 
  0  2   
  2 1  AP 1 0

2 f
 
f 2  f1   2

 t 
c c MS
 f 1 /  res AP K B

Tx Rx
f2 >f1
 0
distance
f1
24.10.2012

 t 
DLR und TU-München, Kommunikation und Navigation
36
Phase Difference Based Approach – Basic Idea (4)

Transmit test signals of two different frequencies f1 and f2 simultaneously and


exploit initial phase difference 0 at receiver:

observation area 

  0  2 AP 1 0
 res
c
 f   f 2  f1   res   t 
MS
f
AP K B

Tx Rx
f2 >f1
 0
distance
f1
24.10.2012

 t 
DLR und TU-München, Kommunikation und Navigation
37
Phase Difference Based Approach – Basic Idea (4)

Transmit test signals of two different frequencies f1 and f2 simultaneously and


exploit initial phase difference 0 at receiver:

observation area 
0
ˆ   res AP 1 0
2
c
 f   f 2  f1   res   t 
MS
f
AP K B

Tx Rx
f2 >f1
 0
distance
f1
24.10.2012

 t 
DLR und TU-München, Kommunikation und Navigation
38
Phase Difference Based Approach –
Ambiguity Problem (1)
Since test signals are periodic, zero Phase difference can not be determined
absolute but only „modulo 2“:

observation area 
   k0
2
ˆ   resˆ   res
0 , k AP 1 0
2 2
c
 f   f 2  f1   res   t 
f MS
AP K B

Tx Rx
f2 >f1
 0
distance
f1
24.10.2012

 t 
DLR und TU-München, Kommunikation und Navigation
39
Phase Difference Based Approach –
Ambiguity Problem (1)
Since test signals are periodic, zero Phase difference can not be determined
absolute but only „modulo 2“:

observation area 
0
ˆ   res  k  res , k AP 1 0
2
c
 f   f 2  f1   res   t 
f MS
AP K B

Tx Rx Rx
f2 >f1
 0
distance
f1
24.10.2012

 t   t 
DLR und TU-München, Kommunikation und Navigation
40
Phase Difference Based Approach –
Ambiguity Problem (2)
Since test signals are periodic, zero Phase difference can not be determined
absolute but only „modulo 2“:

observation area 
0
ˆ   res  k  res , k AP 1 0
2
c  t 
 f   f 2  f1   res  MS
f AP K B

Tx Rx Rx Rx‘
f2 >f1
 0
distance
f1
24.10.2012

 t   t  res
DLR und TU-München, Kommunikation und Navigation
41
Phase Difference Based Approach –
Ambiguity Problem (3)
Since test signals are periodic, zero Phase difference can not be determined
absolute but only „modulo 2“:

observation area 
0
ˆ   res  k  res , k AP 1 0
2
c
 f   f 2  f1   res   t 
f
MS
AP K B

Determination of unique position not possible!


Coarse position is required as starting point which can
be refined by phase difference estimation
24.10.2012

DLR und TU-München, Kommunikation und Navigation


42
Phase Difference Based Approach –
Imperfect Time Synchronization (1)
Clock of AP (receiver) and MS (transmitter) might be imperfectly synchronized:
Clock error MS:

    t

s  t ;    A cos  2  f1  t t    
observation area

  1   AP 1 0

   
s  t ;    A cos  2  f 2  t t      t 
  2   MS
AP K B

Tx Rx
f2 >f1
distance
f1
24.10.2012

 t 
DLR und TU-München, Kommunikation und Navigation
43
Phase Difference Based Approach –
Imperfect Time Synchronization (2)
Clock of AP (receiver) and MS (transmitter) might be imperfectly synchronized:
Clock error MS:
t
  observation area 
 0,1 2  2 f1t 
1  AP 1 0
  0  0,1 0,2

 0,2 2  2 f 2 t   t 
2  MS
AP K B

Tx Rx
f2 >f1
 0 distance
f1
24.10.2012

 t 
DLR und TU-München, Kommunikation und Navigation
44
Phase Difference Based Approach –
Imperfect Time Synchronization (3)
Clock of AP (receiver) and MS (transmitter) might be imperfectly synchronized:
Clock error MS:
2 t
 0  f 2  f1   2  f 2  f1  t observation area 
c   AP 1 0
f
f
 2    ct   t 
c MS
1/ res AP K B

Tx Rx
f2 >f1
 0 distance
f1
24.10.2012

 t 
DLR und TU-München, Kommunikation und Navigation
45
Phase Difference Based Approach –
Imperfect Time Synchronization (4)
Clock of AP (receiver) and MS (transmitter) might be imperfectly synchronized:
Clock error MS:
t
observation area 
0
ˆ   res  c  t AP 1 0
2
c
 f   f1  f 2   res   t 
MS
f
AP K B

Tx Rx
f2 >f1
 0 distance
f1
24.10.2012

 t 
DLR und TU-München, Kommunikation und Navigation
46
Phase Difference Based Approach –
Summary & Pros/Cons
Basic Idea of Phase Difference Based Approach :
 Exchange of two-tone signal between APs and MS
 Measurement of zero-phase difference between both tones at receiver
 Time synchronization between APs and MS
 Calculation of distance based on zero-phase difference

Pros:
 High accuracy
 Medium complexity
 Robust to fluctuations in amplitude of received signal (level fading)

Cons:
 Signals of two frequencies necessary
 Ambiguity of distance estimate  coarse distance estimate necessary as input
 Sensitive to imperfect time synchronization
 In NLOS scenarios measured path length does not allow to draw conclusions about
the required geometric distance
24.10.2012

DLR und TU-München, Kommunikation und Navigation


47
Recall: Indirect Positioning

estimation geometry
of based x̂ m
M̂ geometric position
quantities estimator
Indirect Position Estimator

Examples for geometric quantities:


distances,
distance differences,
distance ratios
24.10.2012

DLR und TU-München, Kommunikation und Navigation


48
Distance Difference Based Positioning

1  ̂
 2 ,1 
ĥ distance-
difference
Distance difference based
estimator x̂ m
position
ˆ KB   ˆ
KB ,1
h estimator

MS position estimator
 k B ,1 
    k B    1 
1. distance difference est.: 2. distance-difference based position est.:
ˆ  kB   k B ,1  ˆ
 2 ,1 
  ˆ  K B ,1
h  ˆ x̂ m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


49
Distance Difference Estimation
 k B,1
Distance difference  ˆ

estimation
Distance difference estimator


ˆ  hˆ 1 ,..., hˆ  K 
M
B

time based approach


c0 ̂1
ˆh kB Tˆ
k 
  
ˆ kB,1
B

ˆ kB
t
T
24.10.2012

DLR und TU-München, Kommunikation und Navigation


50
Single Trip Delay Difference - Basic Idea

1. MS transmits a test signal at time instant

observation area  tT
AP 1
2. AP 1..KB receive the test signal at time
instants
tR  tR B 
1 K

MS
AP KB 
ˆ M
1 

 c0 t R  tT
1


 ˆ M
K B ,1
0 
  c t
R
KB  1 
 tR 
AP 2
self and remote positioning possible !
24.10.2012

DLR und TU-München, Kommunikation und Navigation


51
Single Trip Delay Difference - Inaccurate Clocks
Clock error MS: Clock error APs:
nC,MS nC,AP
observation area 
AP 1
1. MS transmits a test signal at time
tT reported as tT  nC,MS
2. AP 1..KB receive the test signal at times
k 
tR  tR B 
1 K
MS reported as tR B  nC,AP
AP KB
k 
ˆ M
B
 k 
 
 c 0  t R  n C ,AP  t T  n C ,M S 

B

 
k 
 c 0  t R B  t T   c 0  n C ,AP  n C ,M S 
       
       
 M k B  nM
AP 2
24.10.2012

DLR und TU-München, Kommunikation und Navigation


52
Single Trip Delay Difference - Inaccurate Clocks
Clock error MS: Clock error APs:
nC,MS nC,AP
observation area 
AP 1
1. MS transmits a test signal at time
tT reported as tT  nC,MS
2. AP 1..KB receive the test signal at times
k 
tR  tR B 
1 K
MS reported as tR B  nC,AP
AP KB
k 
 c 0  t R   t T   c 0  n C ,AP  n C ,M S 
k
ˆ M B B

 
 
 
   
   

 M k B  nM

 k B ,1 
 c 0  t R B   t R  
k 1
 ˆ M
 
AP 2
clocks in MSs and in APs do not have
24.10.2012

to be perfectly synchronized !
DLR und TU-München, Kommunikation und Navigation
53
Distance Difference Based Positioning

1  ̂
 2 ,1 
ĥ distance-
difference
Distance difference based
estimator x̂ m
position
ˆ KB   ˆ
KB ,1
h estimator

MS position estimator
 k B ,1 
    k B    1 
1. distance difference est.: 2. distance-difference based position est.:
ˆ  kB   k B ,1  ˆ
 2 ,1 
  ˆ  K B ,1
h  ˆ x̂ m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


54
Locations of constant distance difference

P
 k B ,1 
    k B    1 
 


AP 1 AP kB

Locations of constant distance


AP1P -APk B P    
k B ,1 difference form a hyperbola,
foci of hyperbola are APs !
24.10.2012

DLR und TU-München, Kommunikation und Navigation


55
Hyperbola – Special Case of a Conic Section

Source: „conic sections“, en.wikipedia , 2008


24.10.2012

DLR und TU-München, Kommunikation und Navigation


56
Position Determination by Hyperbolic Location

Positions of anchor points and distance differences between MS and APs are known:

AP 2 Anchor point (AP)

MS Mobile station (MS)

AP 1 Observation area 
AP 3

Position of MS can uniquely be deter-


mined, if differences of distances to
AP KB
at least 4 APs are known (in 2D)
24.10.2012

DLR und TU-München, Kommunikation und Navigation


57
3D Position Determination by Hyperbolic Location

Positions of anchor points and distance differences between MS and APs are known:

Anchor point (AP)

Mobile station (MS)

Position of MS can uniquely be deter-


mined, if distances to at least 5 APs
are known (in 3D) or additional
information can be used to exclude
one of the solutions
24.10.2012

DLR und TU-München, Kommunikation und Navigation


58
Hyperbolic Positioning –
Direct Mathematical Approach

 xMS x2  yMS y2   xMS x1  yMS y1  21 2,1
2 2 2 2

AP 2
 xMS x3  yMS y3   xMS x1  yMS y1  31 3,1
2 2 2 2

MS

 xMS x4  yMS y4   xMS x1  yMS y1  41 4,1
2 2 2 2
AP 1

AP 3 2
a1xMS 2b1xMS yMS c1yMS
2
2d1xMS 2e1yMS f1  0
2
a2 xMS 2b2 xMS yMS c2 yMS
2
2d2 xMS 2e2 yMS f2  0
AP K B 2
a3 xMS 2b3 xMS yMS c3 yMS
2
2d3 xMS 2e3 yMS f3  0

xMS , y MS : Unknown position MS


xkB , y kB : Known position AP kB Position determination by solving
set of quadratic equations
 kB ,1 : Known distance difference
24.10.2012

MS ↔ AP kB - MS ↔ AP 1
DLR und TU-München, Kommunikation und Navigation
59
Hyperbolic Positioning

Perfect distance diff. knowledge: Imperfect distance diff. knowledge

AP 2 AP 2

MS MS

AP 1 AP 1

AP 3 AP 3

AP K B AP K B

Probabilistic geometry based approach for


positioning necessary (see next chapter 4)
24.10.2012

DLR und TU-München, Kommunikation und Navigation


60
Time Difference Based Approach
– Summary & Pros/Cons
Basic Idea of Time-Difference / Time-Diff. Of. Arrival (TDOA) approach:
 Measurement of propagation time between APs and MS
 Knowledge of propagation speed of radio wave (speed of light in vacuum)
 Assumption of LOS to calculate distance differences between APs and MS
 Clocks of MS does not have to be accurately synchronized to the clocks of APs,
single trip measurements are possible

Pros:
 Quite accurate
 Moderate complexity

Cons:
 More APs necessary than in distance based approaches
 In NLOS scenarios measured propagation time is too long and does not allow to draw
conclusions about the required geometric distance
24.10.2012

DLR und TU-München, Kommunikation und Navigation


61
Recall: Indirect Positioning

estimation geometry
of based x̂ m
M̂ geometric position
quantities estimator
Indirect Position Estimator

Examples for geometric quantities:


distances,
distance differences,
distance ratios
24.10.2012

DLR und TU-München, Kommunikation und Navigation


62
Distance Ratio Based Positioning

ĝ 1 ˆ   2 ,1 
distance-
ratio
Distance ratio based
estimator x̂ m
position
gˆ  KB  ˆ KB ,1 estimator

MS position estimator
  kB 
   k B ,1 
 1 
1. distance ratio est.: 2. distance-ratio based position est.:

gˆ  kB  
ˆ  k B ,1 ˆ   2 ,1   ˆ 
 K B ,1 
x̂ m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


63
Distance Ratio Estimation

M̂ Distance ratio ˆ kB ,1


estimation
distance ratio estimator


ˆ  gˆ 1 ,..., gˆ  K 
M B

path-loss based approach

ˆ B 
k ,1
gˆ 
k
B
Distance ratio
estimation
24.10.2012

DLR und TU-München, Kommunikation und Navigation


64
Path-loss based Approach – Basic Idea
Determine estimate ̂ of distance  due to relationship between distance  and a :
10log10 a    dB
Random Distribution
due to Slow Fading observation area 
AP 1 Pr  t 

a
Mean attenuation
 t 
am    dB MS
10 dB per decade Pt AP K B
 m
10 100 1000 10000

2
  c  
am    dB   g m    dB   10 log10  gt  g r      10 log10 

    4 f 


g m,0 dB  const
24.10.2012

DLR und TU-München, Kommunikation und Navigation


65
Path-loss based Approach – Distance Estimation
g n kB 
Observation:
Gain (in dB) is
Gn
 kB 
 10 log 10 g n  ,  kB 

Gaussian distributed n  1 N

 kB  1 N
  kB 
G 
N
G
n 1
n

suboptimal
two-step approach
 kB    kB 
  g  kB 
m,0 10 G 10

ˆ  kB 
24.10.2012

DLR und TU-München, Kommunikation und Navigation


66
Path-loss based Approach –
Estimation of Distance ratios

 kB    kB 
 kB   kB 

 G

10


g m,0 10

g m,0
10
G 
  kB   1
 G 10
1 1
  1
g m,010
 1
 G 10 g m,0

 kB 
g t g r   4 f 
1 2
g r 
1
g m,0 c
 
  4 f 
1
g r
 kB  2 kB 
g m,0 gt g r
c

Tx antenna gain, Tx power and attenuation not necessary


min. 4 BSs for 3 independent distance ratios required
24.10.2012

DLR und TU-München, Kommunikation und Navigation


67
Distance Ratio Based Positioning

ĝ 1 ˆ   2 ,1 
distance-
ratio
Distance ratio based
estimator x̂ m
position
gˆ  KB  ˆ KB ,1 estimator

MS position estimator
  kB 
   k B ,1 
 1 
1. distance ratio est.: 2. distance-ratio based position est.:

gˆ  kB  
ˆ  k B ,1 ˆ   2 ,1   ˆ   K B ,1 x̂ m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


68
Locations of constant distance ratio

P  kB 
 k B ,1  
 
 
 1 


AP 1 AP kB

Locations of constant distance


APk B P
  
k B ,1 ratio form a circle, the center
AP1P of the circle is NOT the APs!
24.10.2012

DLR und TU-München, Kommunikation und Navigation


69
Locations of constant distance ratio - example

1  2  1
1  2  1 2 1  2  2

AP1 AP2
24.10.2012

DLR und TU-München, Kommunikation und Navigation


70
Position Determination by Multilateration

Positions of anchor points and distance ratios between MS and APs are known:

AP 2 Anchor point (AP)

Mobile station (MS)


MS
AP 1 Observation area 
AP 3

Position of MS can uniquely be deter-


mined, if distances ratios to at least 4
AP KB
APs are known (in 2D)
24.10.2012

DLR und TU-München, Kommunikation und Navigation


71
Trilateration – Simple Mathematical Approach

 xMS  x2    yMS  y 2 
2 2
2 (2,1)
AP 1   
 xMS  x1    yMS  y1 
2 2 1

 xMS  x3    yMS  y 3 
2 2
 3 (3,1)
  
MS  xMS  x1    yMS  y1 
2 2 1
AP K3
  xMS  x4    yMS  y 4 
2 2
 4 (4,1)
  
 xMS  x1    yMS  y1 
2 2 1
AP 2

xMS , y MS : Unknown position MS


xkB , y kB : Known position AP kB
  k
B ,1 
: Known distance ratio
24.10.2012

MS ↔ AP kB / MS ↔ AP 1
DLR und TU-München, Kommunikation und Navigation
72
Trilateration – Simple Mathematical Approach

 xMS  x2   yMS  y2    xMS  x1   yMS  y1


(2,1)2 (2,1)2
2 2 2 2
1
AP 1
 xMS  x3   yMS  y3    xMS  x1   yMS  y1
(3,1)2 (3,1)2
2 2 2 2
2
  xMS  x4   yMS  y4  (4,1)  xMS  x1 (4,1)  yMS  y1
2 2 2 2 2 2
3

MS
AP K3 1  x 22  y 22 2 x 2 xMS 2 y 2 y MS  1 ( 3 ,1)
2


- x 3
2
 y 3 2 2 x 3 xMS 2 y 3 y MS 1   ( 2 ,1) 2
 
2
 ( 2 ,1) 2
   ( 3 ,1)
2
 x 1
2
 y 12  2 x 1 x M S  2 y 1 y M S 
1 
AP 2 -
3
xMS , y MS : Unknown position MS Position determination by solving set of
xkB , y kB : Known position AP kB linear equations and cross-checking three
  k
B ,1
: Known distance ratio original quadratic equations
24.10.2012

MS ↔ AP kB / MS ↔ AP 1
DLR und TU-München, Kommunikation und Navigation
73
Distance Ratio Approach – Summary & Pros/Cons

Basic Idea of Path-loss Ratio Based / Signal Strength Ratio Approach:


 Measurement of received power Pr at APs of signal exchanged between APs and MS
 Interpretation of received power as instantaneous path loss / gain
 Calculation of mean path loss / gain to eliminate the influence of slow fading
 Assumption of LOS and propagation model incl. its parameters  to calculate the
distance ratios between APs and MS
Pros:
 Low complexity
 No information about MS antenna gain and/or MS transmission power Pt are necessary if
measurement in uplink
Cons:
 Sensitive to slow and fast fading as well as model errors
 Low accuracy, accuracy decreases with absolute distance to be estimated
 Averaging over fading only possible if sufficient movement of MS
 Parameters  of the propagation model has to be known
 In NLOS scenarios measured path loss does not allow to draw conclusions about the
required geometric distance
24.10.2012

DLR und TU-München, Kommunikation und Navigation


74
Outline and Structure

3. Basic principles of terrestrial navigation

3.1 Dead Reckoning

3.2 Proximity Systems

3.3 Distance based Navigation

3.4 Distance difference based Navigation

3.5 Distance ratio based Navigation

3.6 Angle-of-arrival based Navigation

3.7 Signature based Navigation

3.8 Cooperative navigation in radio and sensor networks


24.10.2012

DLR und TU-München, Kommunikation und Navigation


75
Position Determination by Angulation

Positions of APs and Angles-of-Arrival (AoAs) between MS and APs are known:

AP 1
Anchor point (AP)

Mobile station (MS)

Observation area 
MS
AP KB

 Position of MS can uniquely be deter-


mined, if AoA for at least 2 APs
AP 2 are known (in 2D)
24.10.2012

DLR und TU-München, Kommunikation und Navigation


76
Position Determination by Angulation –
Self- and Remote Positioning

Remote Positioning: Self Positioning:



AP 1 AP 1


 MS
MS AP K B
AP K B


Anchor point (AP) AP 2
AP 2
Mobile station (MS)
Observation area 
24.10.2012

DLR und TU-München, Kommunikation und Navigation


77
Angle of Arrival Based Positioning

1 angle of arrival ̂ 1 



estimator AoA-
based
x̂ m
position
ˆ KB  angle of arrival ˆKB  estimator
h
estimator
MS position estimator

1. Angle of Arrival (AoA) est.: 2. AoA-based position est.:


 kB  ˆ 1   ˆ
KB 
ˆ
h ˆ  k B  x̂ m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


78
Angulation – Simple Mathematical Approach

 x MS   x1   cos 1 
AP 1 1  y    y   1  sin  
 MS   1   1 

 x MS   x 2   cos  2 
2  y    y   2  sin  
 MS   2   2 

MS !
AP K B 1  2 1, 2
 Substitution of parameters 1, 2
y in 1 and/or 2
AP 2
x
xMS , y MS : Unknown position MS
Position determination by solving
xkB , y kB : Known position AP kB
set of linear equations
k : Known AoA MS ↔ AP kB
24.10.2012

DLR und TU-München, Kommunikation und Navigation


79
Perfect and Imperfect Knowledge of AoA

Perfect AoA knowledge: Imperfect AoA knowledge:

AP 1 AP 1
 


MS MS
AP K B  AP

 

AP 2 AP 2

Probabilistic geometry based approach for


positioning necessary (see next chapter 4)
24.10.2012

DLR und TU-München, Kommunikation und Navigation


80
Hybrid Distance / AoA based Approach

Positions of anchor points, distance and AoA between MS and APs are known:

Anchor point (AP)


AP 1
Mobile station (MS)

Observation area 

MS
 AP KB Position of MS can uniquely be deter-
 mined, if AoA and distance for at least
one AP are known (in 2D)
AP 2
Example: Radio Detection and Ranging (RADAR)
24.10.2012

DLR und TU-München, Kommunikation und Navigation


81
Position Determination –
Simple Mathematical Approach
AP 1

 x MS   x kB   cos  kB 
 y    y    kB  
 MS
AP K B
 MS   kB   sin  kB 


y
AP 2 Position determination by simple
x
arithmetics
x M S , y M S : Unknown position MS
x k B , y k B : Known position AP kB
 k : Known AoA MS ↔ AP kB
B

 k : Known distance MS ↔ AP kB
24.10.2012

B
DLR und TU-München, Kommunikation und Navigation
82
Angle of Arrival Based Positioning

1 angle of arrival ̂ 1 



estimator AoA-
based
x̂ m
position
ˆ KB  angle of arrival ˆKB  estimator
h
estimator
MS position estimator

1. Angle of Arrival (AoA) est.: 2. AoA-based position est.:


 kB  ˆ 1   ˆ
KB 
ˆ
h ˆ  k B  x̂ m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


83
Determination of Angle of Arrival

Directive Antennas: Multi-Antennas:

RP Rx

KRx

Example: RADAR Example: Positioning in


Mobile Radio Communication
Systems
24.10.2012

DLR und TU-München, Kommunikation und Navigation


84
Determination of Angle of Arrival
using Multi-Antennas (1)

Wavefront  RX 
 l Rx
k  Rx 
 l Rx
k  RX 
 cos    Rx  k

 kRX  2  kRX 
 Rx   l Rx

  kRx 
a Rx  RX 
 exp j  Rx  k

 
T
1  K Rx 
a Rx  a Rx  a Rx
24.10.2012

DLR und TU-München, Kommunikation und Navigation


85
Determination of Angle of Arrival
using Multi-Antennas (2)
h R P  
1
h
1
  h
k RX 
   h RP   a Rx
k
RX 

  h 1   
 
RP Receiver h        a Rx h RP  
  K RX  
 h  

h
K Rx 
 
KRx
Angle of Arrival can be determined by

 l R xR X   l R xR x   c o s    R xR X 
k k k
 1) measuring h   ,
2 2) determining a Rx and
 R kx RX 
  l R xR X 
k

 3) estimating 
a R xR x   e x p
k
 j    kRX
Rx

Probabilistic geometry based approach for


 a    a   
T
1 K Rx
a Rx
positioning necessary (see next chapter 4)
24.10.2012

Rx Rx

DLR und TU-München, Kommunikation und Navigation


86
Angle of Arrival based Approach
– Summary & Pros/Cons
Basic Idea of Angle of Arrival (AoA) based Approach:
 Measure radiated signals at either MS or APs with multi-antenna(s) or directive
antenna(s)
 Determine AoA of radiated signals at either MS or APs
 Assuming direct radio propagation along the LOS use geometrical relationship and
known position of APs and AoA to determine position of MS

Pros:
 Only 2 or even 1 (hybrid) APs necessary for positioning
 No accurate synchronization / timing necessary
 Not sensitive to slow or fast fading, no influence of fading on positioning accuracy

Cons:
 In NLOS scenarios measured AoA does not allow to draw conclusions about real
geometry
 Utilization of multi-antennas / directive antennas at APs and/or MS necessary
 Medium accuracy, accuracy decreases with absolute distance between MS and APs
24.10.2012

DLR und TU-München, Kommunikation und Navigation


87
Outline and Structure

3. Basic principles of terrestrial navigation

3.1 Dead Reckoning

3.2 Proximity Systems

3.3 Distance based Navigation

3.4 Distance difference based Navigation

3.5 Distance ratio based Navigation

3.6 Angle-of-arrival based navigation

3.7 Signature based navigation

3.8 Cooperative navigation in radio and sensor networks


24.10.2012

DLR und TU-München, Kommunikation und Navigation


88
Radio propagation - Example: Urban Scenario
• multipath propagation! • no Line of Sight (NLOS) !

AP
MS

Building

Source: J. Maurer, W. Wiesbeck et al., Universität Karlsruhe (TH)


24.10.2012

DLR und TU-München, Kommunikation und Navigation


89
Example for Signature:
Channel impulse response between AP and MS
1,0
h  ; x m  0,9
0,8
0,7
0,6
0,5
0,4
0,3
0,2
0,1
0,0
0 2 4 6 8 10 12 14 16 18 20
/s
channel impulse response (CIR) is a nearly unique characteristic
M =  h  ; x m   (signature) for the location x m of an MS
24.10.2012

DLR und TU-München, Kommunikation und Navigation


90
Example for Signature: Time discrete
Channel impulse response between AP and MS
1,0
hw  xm  0,9
Time discrete channel impulse resp.:
0,8
h( xm )   h1  xm hW  xm  
T
0,7
0,6
0,5
0,4 Channel is
0,3 band limited !
0,2
0,1
0,0
0 2 4 6 8 10 12 14 16 18 W
w
time discrete channel impulse response (CIR) h( xm ) is a nearly unique
characteristic M =  h  x m   (signature) for the location x m of an MS
24.10.2012

DLR und TU-München, Kommunikation und Navigation


91
Signature Based MS Localization / Fingerprinting

localization domain kr

CIR measured on-line at


MS position is
compared to CIR at
reference points
 determined off-line and
stored in database

localization error :
distance between real
and estimated MS
position

observation domain
24.10.2012

DLR und TU-München, Kommunikation und Navigation


92
Signature Based Localization: Principle

offline:
• store signatures for sufficiently many reference points of
the scenario in a data base

online:
• comparison of currently measured signature with entries
in data base
• „best match“ gives the most probable position of the MS
24.10.2012

DLR und TU-München, Kommunikation und Navigation


93
Choice of Signature: CIR

CIR changes for small movements


in order of /2 due to fast fading

very small RP spacing required

prohibitively big database

??? how to reduce database size ???


24.10.2012

DLR und TU-München, Kommunikation und Navigation


94
Better Choice for Signature

RP kr

MS

chose signature so that only the wave fronts


impinging from different directions are relevant, but
not the way they superpose
24.10.2012

DLR und TU-München, Kommunikation und Navigation


95
Localization Using
Covariance Matrices (CMs)

offline comparison online


determined and for each RP kr measured
stored in database
for each RP kr

R
( kr )
E 
hl hl  H
 R MS  E 
hl hl
H

measurement
MS position l
position l

RP kr
k10
24.10.2012

DLR und TU-München, Kommunikation und Navigation


96
Comparison of
Covariance Matrices (CMs)

Physical effects may change scaling


( kr )
of CMs R and R MS !

scaling invariant comparison of normalized CMs:


(k )
R r R MS
(k )
  min
R r 2
R MS 2 2

maximization of cross correlation coefficient (CCC)

 (kr ) 

vec R 
( kr ) H
vec  R MS 
vec  R ( k r )  vec  R MS 
24.10.2012

Exercise: Derivation of maximization of CCC


DLR und TU-München, Kommunikation und Navigation
98
summary: localization based on
covariance matrices (CM)

1. determine and store R ( k r ) for each reference


point kr

2. determine R MS and compare to each reference


point using cross-correlation coefficient (CCC)

 (kr ) 

vec R 
( kr ) H
vec  R MS 
vec  R ( k r )  vec  R MS 

3. find “best match“: ˆ  (k r' ) 


k r  arg max 
k r' 1... K r 
 
24.10.2012

DLR und TU-München, Kommunikation und Navigation


99
localization based on
CMs using several APs

multi AP scenario: CMs for each AP kB:


( k r ,k B )  kB 
AP 1 R , R MS
1
h MS,l
CCC for each AP kB:
h MS,B l
K
MS   kB 
 kr 
h MS,l
 AP K B
2

maximization of total CCC:


KB
1
  kr     kB 
 kr 
AP 2
KB k B 1
24.10.2012

DLR und TU-München, Kommunikation und Navigation


100
exemplary scenario:
downtown Karlsruhe
N

W E

S
AP2

AP1 path of the MS

observation domain AP3

100m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


101
examplary scenario:
CCC  (kr) vs. assumed MS position
550 1.0
1.0

500 0.9
0.9

450
0.8
0.8

CCC  (kr)
400
0.7
0.7

350
0.6
0.6
x2 / m

300
0.5
0.5

250
0.4
0.4

200
0.3
0.3
buildings:
150

100 0.2
0.2

0.1
0.1
true pos.:
50
0.0
0.0
100 200 300 400 500 600 700
x1 / m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


102
results: localization error

W E

S
BS2

scenario:

• KB  3

•    dB
BS1

observation domain BS3

100m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


103
CCDF of localization error :
parameter number of APs 
0
10
forbidden region (E911)

scenario:
Prob{}

 = 1 • K B  1 ...3
-1
10
 = 2
•    dB

 = 3

-2
10
0 25 50 75 100 125 150 175 200 225 250 275 300
/m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


104
model and quality measure

ˆ  kB 
model for h , corruption by white Gaussian noise:
n  kB 

 kB 
h kB  ˆ
h

  k B   k B H  2
E n n   2 I
 

signal-to-noise-ratio:
kB  2
h
  k B  2
2
24.10.2012

DLR und TU-München, Kommunikation und Navigation


105
CCDF of localization error :
parameter SNR of CIRs 
0
10
forbidden region (E911)

scenario:
Prob{}

 = 6 dB
-1
• KB  3
10
 = 12 dB

 =  dB

-2
10
0 25 50 75 100 125 150 175 200 225 250 275 300
/m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


106
multi-antennas: space-time
CIR between AP and MS
directional CIR: antenna specific CIRs:
|h(1) ()| 1.0
n antenna 1
amplitude 0.8

0.6

0.4

azimuth 0.2
60 30
90 0.0

/s
0 0 5 10 15
120
330
|h(2) ()| 1.0

antenna 2
150 n
300
0.8
180
270
210 240 0.6
delay
0.4

0.2

0.0

/s
0 5 10 15

space-time channel impulse response (CIR) is a nearly


unique characteristic (signature) for the location of an MS
24.10.2012

DLR und TU-München, Kommunikation und Navigation


107
exemplary scenario:
downtown Karlsruhe
N
multi antenna
W E

S
AP2

antenna config.:
K A 1
AP1 path of the MS
l
K A 2 2

l
K A 4 2
observation domain AP3
antenna element
reference point
100m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


108
examplary scenario:
CCC  (kr) vs. assumed MS position

550 1.0
1.0

500 0.9
0.9

450
0.8
0.8

CCC  (kr)
400
0.7
0.7

350
0.6
0.6
x2 / m

300
0.5
0.5

250
0.4
0.4

200
0.3
0.3
buildings:
150

100 0.2
0.2

0.1
0.1
true pos.:
50
0.0
0.0
100 200 300 400 500 600 700
x1 / m KA 1
24.10.2012

DLR und TU-München, Kommunikation und Navigation


109
examplary scenario:
CCC  (kr) vs. assumed MS position

550 1.0
1.0

500 0.9
0.9

450
0.8
0.8

CCC  (kr)
400
0.7
0.7

350
0.6
0.6
x2 / m

300
0.5
0.5

250
0.4
0.4

200
0.3
0.3
buildings:
150

100 0.2
0.2

0.1
0.1
true pos.:
50
0.0
0.0
100 200 300 400 500 600 700
x1 / m KA  2
24.10.2012

DLR und TU-München, Kommunikation und Navigation


110
examplary scenario:
CCC k (kr) vs. assumed MS position

550 1.0
1.0

500 0.9
0.9

450
0.8
0.8

CCC  (kr)
400
0.7
0.7

350
0.6
0.6
x2 / m

300
0.5
0.5

250
0.4
0.4

200
0.3
0.3
buildings:
150

100 0.2
0.2

0.1
0.1
true pos.:
50
0.0
0.0
100 200 300 400 500 600 700
x1 / m KA  4
24.10.2012

DLR und TU-München, Kommunikation und Navigation


111
CCDF of localization error D:
parameter number of AP antennas KA
0
10
forbidden region (E911)
Prob{}

perfect CMs
-1  = 1 KB = 3
10
γ = 6 dB
 = 2
 = 4

-2
10
0 25 50 75 100 125 150 175 200 225 250 275 300
/m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


112
Signature based Approach – Summary & Pros/Cons
Basic Idea of Signature based Approach / Fingerprinting:
 Characterize Observation area with respect to signature (e.g. CM) and setup
database (offline step)
 Measure current signature for positioning (online step)
 Compare measured signature with database to determine position of MS

Pros:
 Well suited for scenarios with multi-path environment even if no LOS is available
 CMs are especially unique in multi-path non line-of-sight (NLOS) scenarios
 CCC is well suited for comparing CMs
 achievable localization accuracy quite promising even if considering a single BS
 robust with respect to disturbed estimates of CIRs

Cons:
 For accuracte positioning large database necessary
 Setup of large database and comparision with large database complex
 Setup of database needs accurate models or complicated measurement campaigns
 Impact of changes in the environment / weather on database has to be compensated
24.10.2012

DLR und TU-München, Kommunikation und Navigation


113
Outline and Structure

3. Basic principles of terrestrial navigation

3.1 Dead Reckoning

3.2 Proximity Systems

3.3 Distance based Navigation

3.4 Distance difference based Navigation

3.5 Distance ratio based Navigation

3.6 Angle-of-arrival based navigation

3.7 Signature based navigation

3.8 Cooperative navigation in radio and sensor networks


24.10.2012

DLR und TU-München, Kommunikation und Navigation


114
Radio- and Sensor Networks

 A sensor network is a network of


several similar tiny nodes which
communicate with each.
 The communication takes place
without additional infrastructure and
is ad-hoc

Mobile Station (MS), Node in the ‚  Examples: measurement systems,


Network, Position apriori not known monitoring, military applications
Mobile Station (MS), Beacon Node in (smart dust), …
the Network, Position known
 Positions of some Nodes may be
known (Beacon Nodes, here: APs),
positions of other nodes are
24.10.2012

a priori not known


DLR und TU-München, Kommunikation und Navigation
115
classic geometry
based localization (multi-lateration)

observation area 
AP 1



MS
AP KB


AP 2
24.10.2012

DLR und TU-München, Kommunikation und Navigation


116
multi MS scenario

observation area 
AP 1

MS 3
MS 2
MS 4

MS 1
MS 5 AP KB
MS K

AP 2
24.10.2012

DLR und TU-München, Kommunikation und Navigation


117
SPIDER:
Smart Position Identification Rationale

observation area 
AP 1

MS 3
MS 2
MS 4

MS 1
MS 5 AP KB
MS K

AP 2
24.10.2012

Reference: M. Meurer, P. W. Baier, T. Weber, C. A. Jötten, S. Heilmann, “SPIDER: Enhanced distance based localization of mobile
radio terminals”, in Proc. IEEE 60th Vehicular Technology Conference (VTC’04-Fall), Los Angeles, 2004.
DLR und TU-München, Kommunikation und Navigation
118
number of observables

no. of unknowns no. of observables


per MS per MS
conv.
2 KB
2  B
1 K 1  K
novel
2
24.10.2012

DLR und TU-München, Kommunikation und Navigation


119
increase of number of observables

„information gain“ 9
no. of obs. SPIDER 8
no. of obs. conventional
7

5
K B  3, 4,...,13
4

1
1 2 4 6 8 10 12 14 16
K / KB
24.10.2012

DLR und TU-München, Kommunikation und Navigation


120
joint distance-based position estimation:
SPIDER

coupled problems of MS localization

positions of the KB APs

noise joint estimate x̂ m of the


distances 
for all MSs ̂ distance- MS positions
based
position
estimator
position estimator
24.10.2012

DLR und TU-München, Kommunikation und Navigation


121
scenario
3000

2. coordinate /m observation area 


AP 2
2000

mobile station (MS)


MS K (randomly placed in  )
1000
MS 1
Anchor point (AP)

0 AP 1
MS 2

-1000
MS 3
KB  3
-2000
AP KB

-3000
-3000 -2000 -1000 0 1000 2000 3000
1. coordinate /m
24.10.2012

DLR und TU-München, Kommunikation und Navigation


122
simulations:
performance vs. number of MSs

0 parameter:
10 forbidden region (E911)
SPIDER • perfect knowledge
• KB  3
conventional
• K  2 64
-1
10 (random positions)
• white Gaussian
noise
K =2,4,8,
16,32,64 • Rn   2 I
-2
10
0 20 40 60 80 100 120 140 160 180 200 •   78 m
/m • .
24.10.2012

DLR und TU-München, Kommunikation und Navigation


123
centralized vs. distributed localization

For SPIDER Localization has to be done in a


central unit (CU) using all measurements !

• special CU required (drawback in sensor / Ad-hoc networks)


• signaling required to make measurement for CU available

modify SPIDER in order to allow


decentralizedSPIDER-S:
/ distributed localization
Smart Position Identification Rationale - Separated
24.10.2012

DLR und TU-München, Kommunikation und Navigation


124
distributed localization with SPIDER-S

observation area  assumption:


AP 1
each MS knows the positions of the
MS 2 APs and measurements of
• distances between this MS and
other MSs and
MS 1
AP KB
• distances between this MS and
a sufficient number of APs.

AP 2
24.10.2012

DLR und TU-München, Kommunikation und Navigation


125
distributed localization with SPIDER-S
step 1:
observation area 
AP 1 each MS estimates its position
using distances to APs only !
MS 2
• distances between this MS and
other MSs and
MS 1
AP KB • distances between this MS and
sufficient number of BSs.

AP 2
24.10.2012

DLR und TU-München, Kommunikation und Navigation


126
distributed localization with SPIDER-S
step 1:
observation area 
AP 1 each MS estimates its position
MS 2 • using distances
distances to APs
between thisonly
MS !and
other MSs and

MS 1
• distances
roughbetween
positionthis MS and
estimate
AP KB sufficient number of MS
of each BSs.

broadcast of rough positions


AP 2 estimates to other MSs
24.10.2012

DLR und TU-München, Kommunikation und Navigation


127
distributed localization with SPIDER-S
step 2:
observation area  assumption:
AP 1 each
each MS re-estimates its
MS knows measurements ofposition
using:
MS 2 • broadcasted (rough) positions of
• distances
other MSs,between this MS and
• other MSs to
distances and
APs and to MSs
MS 1
AP KB • distances between this MS and
sufficient number of BSs.

AP 2
24.10.2012

DLR und TU-München, Kommunikation und Navigation


128
distributed localization with SPIDER-S
step 2:
observation area  assumption:
AP 1 each
each MS re-estimates its
MS knows measurements ofposition
using:
MS 2 • broadcasted (rough) positions of
• distances
other MSs,between this MS and
• other MSs to
distances and
APs and to MSs
MS 1
AP KB • distances between this MS and
sufficient number
improved of BSs.
position estimate
of each MS

AP 2 broadcast of improved positions


estimates to other MSs
24.10.2012

DLR und TU-München, Kommunikation und Navigation


129
distributed localization with SPIDER-S
step :
observation area  assumption:
AP 1 each
each MS re-estimates its
MS knows measurements ofposition
using:
MS 2 • broadcasted (rough) positions of
• distances
other MSs,between this MS and
• other MSs to
distances and
APs and to MSs
MS 1
AP KB • distances between this MS and
sufficient number
improved of BSs.
position estimate
of each MS

AP 2 broadcast of improved positions


estimates to other MSs
24.10.2012

DLR und TU-München, Kommunikation und Navigation


130
partial knowledge of
distance measurements

full
knowledge of observation area 
measure- AP 1
ments of
distance to all MS 3
MS 2
APs and all MS 4
MSs MS 1
MS 5 AP KB
MS K

AP 2
24.10.2012

DLR und TU-München, Kommunikation und Navigation


131
partial knowledge of
distance measurements

only
knowledge of observation area 
distance to AP 1
APs and to
MS 3
adjacent MSs MS 2
MS 4

MS 1
MS 5 AP KB
MS K

AP 2
24.10.2012

DLR und TU-München, Kommunikation und Navigation


132
partial knowledge of
distance measurements

only
knowledge of observation area 
distance to AP 1
adjacent AP
MS 3
and adjacent MS 2
MSs MS 4

MS 1
MS 5 AP KB
MS K

Is it still possible to uniquely localize the MSs ?


graph theoretical
AP 2 mathematical question !
24.10.2012

DLR und TU-München, Kommunikation und Navigation


133
possibility of localization:
necessary and sufficient conditions (1)

necessary & sufficient counter example:


conditions possible localization:
observation area 

(1) For each MS at least 3 distances to


other MSs or APs have to be known!
24.10.2012

DLR und TU-München, Kommunikation und Navigation


134
possibility of localization:
necessary and sufficient conditions (2)

necessary & sufficient counter example:


conditions possible localization:
observation area 
(2) In the graph of measured/known
distances each MSs has to be
connected directly or indirectly
to at least 3 different APs!
24.10.2012

DLR und TU-München, Kommunikation und Navigation


135
possibility of localization:
example with fulfilled conditions
example with 2 MSs which can be localized:

(1) For each MS at least 3 distances to


other MSs or APs have to be known! observation area 

(2) In the graph of measured/known


distances each MSs has to be
connected directly or indirectly
to at least 3 different APs!
24.10.2012

DLR und TU-München, Kommunikation und Navigation


136
modified scenario
3000

2. coordinate /m observation area 


AP 2
2000

mobile station (MS)


MS K (randomly placed in 
)
1000
MS 1
Anchor point (AP)

0 AP 1
MS 2

-1000
MS 3
KB  3
-2000
AP KB

-3000
-3000 -2000 -1000 0 1000 2000 3000
1. coordinate /m

new: each MS measures only distance to K B AP and to  other MSs !


24.10.2012

DLR und TU-München, Kommunikation und Navigation


137
simulations:
performance vs. no. of distances 

parameter:
0
10 forbidden region (E911)
• partial knowledge
of MS-MS distances,
  K  10   0 K  1
-1 without noise • KB  3
10 no knowledge
parameter est.
• K  7
of MS-MS dist.
with noise (random positions)
parameter est.
full knowledge of all • white Gaussian
=0,78,156m
-2 MS-MS dist. noise
10
0 20 40 60 80 100 120 140 160 180 200 • Rn   2 I
/m •   78 m
.
24.10.2012

DLR und TU-München, Kommunikation und Navigation


138
Cooperative Positioning – Summary & Pros/Cons
Basic Idea of Cooperative Positioning:
 Use information of distances between nodes in a network to determine position of
nodes
 Use beacon node for absolute alignment and initialization
 Improve position accuracy by redundant measurements

Pros:
 High positioning accuracy in large networks
 Usable in partly connected networks
 Centralized and Decentralized positioning possible

Cons:
 Communication Ressources quite large
 In NLOS scenarios measured distances do not allow to draw conclusions about real
geometry
24.10.2012

DLR und TU-München, Kommunikation und Navigation


139

You might also like