principles of wireless sensor networks - kth
TRANSCRIPT
![Page 1: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/1.jpg)
Principles of Wireless Sensor Networkshttps://www.kth.se/social/course/EL2745/
Lecture 10
Positioning and Localization
Piergiuseppe Di MarcoEricsson Researche-mail:[email protected]
http://pidm.droppages.com/
Royal Institute of TechnologyStockholm, Sweden
September 24, 2015Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 1 / 37
![Page 2: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/2.jpg)
Course content
Part 1
I Lec 1: Introduction to WSNsI Lec 2: Introduction to Programming WSNs
Part 2
I Lec 3: Wireless ChannelI Lec 4: Physical LayerI Lec 5: Medium Access Control LayerI Lec 6: Routing
Part 3
I Lec 7: Distributed DetectionI Lec 8: Static Distributed EstimationI Lec 9: Dynamic Distributed EstimationI Lec 10: Positioning and LocalizationI Lec 11: Time Synchronization
Part 4
I Lec 12: Wireless Sensor Network Control Systems 1I Lec 13: Wireless Sensor Network Control Systems 2
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 2 / 37
![Page 3: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/3.jpg)
Previous lecture
How to estimate phenomena from noisy measurements?
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 3 / 37
![Page 4: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/4.jpg)
Today’s learning goals
Which measurements are used for estimating the position of a node?
How to estimate the position of a node?
What is the effect of measurement errors?
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 4 / 37
![Page 5: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/5.jpg)
Outline
Introduction
Specific sources of measurements
Estimation of the position
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 5 / 37
![Page 6: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/6.jpg)
Outline
Introduction
Specific sources of measurements
I Time of arrival
I Time difference of arrival
I Received signal strength
I Angle of Arrival
Estimation of the position
I Angle of arrival + velocity
I Triangulation
I Trilateration
I Iterative and collaborative multilateration
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 6 / 37
![Page 7: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/7.jpg)
Estimation
phenomena
1 2 . . . N
Fusion Center
Centralized Estimation:no intelligence on sensors
phenomena
1 2 . . . N
Fusion Center
Distributed Estimation:some intelligence on sensors
phenomena
1 2 . . . N
Completely distributed(or decentralized) Estimation
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 7 / 37
![Page 8: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/8.jpg)
Positioning and localization
Localization is defined as a technique to estimate the positions of nodes
It can be categorized into
I Centralized, where a central node estimates the position of the nodes
I Distributed, where many nodes help each other to find their own positions
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 8 / 37
![Page 9: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/9.jpg)
Outline
Introduction
Specific sources of measurements
I Time of arrival
I Time difference of arrival
I Received signal strength
I Angle of Arrival
Estimation of the position
I Angle of arrival + velocity
I Triangulation
I Trilateration
I Iterative and collaborative multilateration
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 9 / 37
![Page 10: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/10.jpg)
Time of arrival
t
t|t1
|t2
node i
node j
message
Assuming that the nodes are synchronized the distance measurement is
dij ' (t2 − t1) · v
where v is the propagation speed of the message
Uncertainty occurs when t2 ≈ t1
Problems
I Packet lossesI MAC delaysI CPU delay
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 10 / 37
![Page 11: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/11.jpg)
Outline
Introduction
Specific sources of measurements
I Time of arrival
I Time difference of arrival
I Received signal strength
I Angle of Arrival
Estimation of the position
I Angle of arrival + velocity
I Triangulation
I Trilateration
I Iterative and collaborative multilateration
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 11 / 37
![Page 12: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/12.jpg)
Time difference of arrival
t
t|t1
|t2
|t3
|t4
node i
node j
processing time
The distance measurement is
dij '(t2 − t1) + (t4 − t3)
2· v
where v is the propagation speed of the message
Problems
I Packet lossesI MAC delaysI CPU delay
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 12 / 37
![Page 13: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/13.jpg)
Outline
Introduction
Specific sources of measurements
I Time of arrival
I Time difference of arrival
I Received signal strength
I Angle of Arrival
Estimation of the position
I Angle of arrival + velocity
I Triangulation
I Trilateration
I Iterative and collaborative multilateration
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 13 / 37
![Page 14: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/14.jpg)
Received signal strength (RSS)
•TX
i
•RX
j
Ideal propagation
Pr = PtGtGrPLλ2
(4πdij)2 ⇒ dij '
λ
4π
√PtGtGrPL
Pr
Problems: multi-path, noise, etc, affect the propagation characteristics⇒ Channel estimation needed
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 14 / 37
![Page 15: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/15.jpg)
Outline
Introduction
Specific sources of measurements
I Time of arrival
I Time difference of arrival
I Received signal strength
I Angle of Arrival
Estimation of the position
I Angle of arrival + velocity
I Triangulation
I Trilateration
I Iterative and collaborative multilateration
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 15 / 37
![Page 16: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/16.jpg)
Angle of arrival
Example: Sound propagation by multiple microphones, amplitude and phase
Odometry sensor
•TX
• RX
θ
Problems: Errors in measurements of θ (e.g., due to magnetic fields of earth),estimation accuracy decreases with the distance.
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 16 / 37
![Page 17: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/17.jpg)
Outline
Introduction
Specific sources of measurements
I Time of arrival
I Time difference of arrival
I Received signal strength
I Angle of Arrival
Estimation of the position
I Angle of arrival + velocity
I Triangulation
I Trilateration
I Iterative and collaborative multilateration
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 17 / 37
![Page 18: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/18.jpg)
Estimation of the position
We now examine the specific techniques that are used to estimate the sensors’ positionbased on the measurements that were previously presented
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 18 / 37
![Page 19: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/19.jpg)
Outline
Introduction
Specific sources of measurements
I Time of arrival
I Time difference of arrival
I Received signal strength
I Angle of Arrival
Estimation of the position
I Angle of arrival + velocity
I Triangulation
I Trilateration
I Iterative and collaborative multilateration
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 19 / 37
![Page 20: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/20.jpg)
Angle of arrival + velocity
θ (k) = θ (k) + nθ (k)
θ (k) the angle measurement at discrete time k given by, e.g., odometry sensor
θ (k) the real angle value
nθ (k) the additive noise
v (k) = v (k) + nv (k)
v (k) the velocity measurement at discrete time k given by, e.g., accelerometer
v (k) the real velocity value
nv (k) the additive noise
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 20 / 37
![Page 21: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/21.jpg)
Angle of arrival + velocity
Node movingwith velocity v
•v
Let Xr (k) =
[x (k)y (k)
]be the true position of the node that we want to estimate
Then
x (k + 1) = x (k) + v (k) · T · cos θ (k)y (k + 1) = y (k) + v (k) · T · sin θ (k)
where T is the sampling time
A problem is that the resulting estimator is biased
E {x (k + 1)} = E {x (k)}+ E {v (k)} · T · E{cos θ (k)
}=
= E {x (k)}+ v (k) · T · cos θ (k) · e−σ2θ2 6= x (k + 1)
We need to estimate the bias as well.
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 21 / 37
![Page 22: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/22.jpg)
Outline
Introduction
Specific sources of measurements
I Time of arrival
I Time difference of arrival
I Received signal strength
I Angle of Arrival
Estimation of the position
I Angle of arrival + velocity
I Triangulation
I Trilateration
I Iterative and collaborative multilateration
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 22 / 37
![Page 23: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/23.jpg)
Triangulation
Anchors: nodes with fixed position used for determining the unknown position of anode
•x1, y1
•xr, yr
•x3, y3
•x2, y2θ1 θ2
θ3
Let Xr =
[xryr
]be the true position of the node that we want to estimate
The angle measurements are
θi = θi (Xr) + ni i = 1, 2, 3
where θi (Xr) = arctanxr − xiyr − yi
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 23 / 37
![Page 24: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/24.jpg)
Triangulation
Setting
n =
n1
n2
n3
as the vector of noises with E{n · nT
}= R =
σ21 0 00 σ2
2 00 0 σ2
3
θ (Xr) =
θ1 (Xr)θ2 (Xr)θ3 (Xr)
as the vector of true node position and
Y =
θ1θ2θ3
as the vector of angle measurements
we can estimate Xr from
Y = θ (Xr) + n
where θ (Xr) is a non-linear function
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 24 / 37
![Page 25: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/25.jpg)
Triangulation
Error covariance (cf. slide 19 in Lecture 8)
C(Xr)=[θ(Xr)− Y
]T·R−1 ·
[θ(Xr)− Y
]=
3∑i=1
(θi(Xr)− θi
)2σ2i
Goal: Choose Xr that keeps oscillations of θ(Xr)
around Y
Therefore
minXr
C(Xr)≡
dC(Xr)
dxr= 0
dC(Xr)
dyr= 0
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 25 / 37
![Page 26: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/26.jpg)
TriangulationWe can use iterative methods for solving the above system of non-linear equations, e.g.,
Newton-Gauss methodConsider the non-linear equation
g (x) = 0 g : RN → RN x ⊂ RN
Iterative method xt+1 = xt − δt · g (xt)
where δt = ∇−1g (xt)
limt→∞
xt → x∗ : g (x∗) = 0
∇ is the matrix of all first-order partial derivatives (Jacobian)
In this specific case, we use the iterative method to solve
g (x) = ∇xC(Xr)= 0
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 26 / 37
![Page 27: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/27.jpg)
Outline
Introduction
Specific sources of measurements
I Time of arrival
I Time difference of arrival
I Received signal strength
I Angle of Arrival
Estimation of the position
I Angle of arrival + velocity
I Triangulation
I Trilateration
I Iterative and collaborative multilateration
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 27 / 37
![Page 28: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/28.jpg)
Trilateration
•x1, y1
•Xr =
[xryr
]•
xN , yN
•x2, y2
d1 d2
dN
Let Xr =
[xryr
]be the true position of the node that we want to estimate
The distance measurements are
di = di + ni i = 1, .., N
From the trigonometry
(x1 − xr)2 + (y1 − yr)2 = d21...
(xN − xr)2 + (yN − yr)2 = d2N
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 28 / 37
![Page 29: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/29.jpg)
Trilateration
After substracting these N equations we arrive at the following system of equations
A ·Xr = Y
where A ∈ R(N−1)×2, Xr ∈ R2, Y ∈ R(N−1)×1
A = 2 ·
(xN − x1) (yN − y1)...
(xN − xN−1)
...(yN − yN−1)
Y =
d21 − d2N − x21 − y21 + x2N + y2N
...
d2N−1 − d2N − x2N−1 − y2N−1 + x2N + y2N
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 29 / 37
![Page 30: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/30.jpg)
Trilateration
From the previous lecture we considered a measurements model Y = HX + n, but wenow have
Y = AXr
A is not a square matrix and there is no explicit noise (all is included in Y ).Xr is constant
We can apply the Linear Minimum Mean Squared Estimator (LMMSE) with
Xr = L · Ywhere
L =(ATA
)−1
AT
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 30 / 37
![Page 31: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/31.jpg)
Trilateration
We can redefine Xr as a least square solution.
This is equivalent to define the following cost function
C (Xr) = (AXr − Y )T (AXr − Y )
and search for Xr that minimizes C (Xr)
dC (Xr)
dXr= 2AT
(AXr − Y
)= 0⇒ ATAXr = ATY ⇒ Xr =
(ATA
)−1
ATY
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 31 / 37
![Page 32: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/32.jpg)
Trilateration
Note that Xr is a random variable due to Y being noisyTaking the expected values,
E{Xr}=(ATA
)−1ATE {Y }
E {Y } =
E{d21
}− E
{d2N
}− x21 − y21 + x2N + y2N
...
E{d2N−1
}− E
{d2N
}− x2N−1 − y2N−1 + x2N + y2N
di = di + ni ⇒ d2i = d2i + n2
i + 2dini ⇒ E{d2i
}= d2i + E
{n2i
}= d2i + σ2
i
where σ2i = σ2
0ekσdi
The result is that Xr is a biased estimator since
E{Xr}=(ATA
)−1
ATE {Y } 6= Xr
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 32 / 37
![Page 33: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/33.jpg)
Outline
Introduction
Specific sources of measurements
I Time of arrival
I Time difference of arrival
I Received signal strength
I Angle of Arrival
Estimation of the position
I Angle of arrival + velocity
I Triangulation
I Trilateration
I Iterative and collaborative multilateration
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 33 / 37
![Page 34: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/34.jpg)
Iterative and collaborative multilateration
In lateration techniques, at least three anchor nodes are required for estimatingposition
Iterative and collaborative multilateration are extension of lateration that do notrequire three neighbouring anchors
In iterative multilateration
I Any node can become anchor after estimating its position and send anchormessages on the network
I In this way, all nodes can estimate their positions after several iterations
For example, in the figure below
I Gray node estimates its position by three black anchorsI Blue node estimates by gray and two black anchor nodesI White node is localized by blue and two black anchor nodes
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 34 / 37
![Page 35: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/35.jpg)
Iterative and collaborative multilateration
It is possible that nodes cannot have three anchors even after several iterations
In that case collaborative multilateration is used in which
I A graph of participating nodes is constructedI Participating nodes are the ones that are either anchors or have at least three
participating neighboursI This gives set of over-constrained quadratic equations relating distance among
nodes and their neighboursI These equation are solved to estimate positions
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 35 / 37
![Page 36: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/36.jpg)
Summary
We have studied the basic of localization for sensor networks
Localizing the nodes consists in applying estimation techniques
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 36 / 37
![Page 37: Principles of Wireless Sensor Networks - KTH](https://reader031.vdocuments.net/reader031/viewer/2022012300/61e0a1f914a721225b043fbb/html5/thumbnails/37.jpg)
Next lecture
Application of estimation and detection to synchronization
Piergiuseppe Di Marco (KTH) Principles of Wireless Sensor Networks September 24, 2015 37 / 37