Three Beacon Sensor Network Localization through Self

Download Report

Transcript Three Beacon Sensor Network Localization through Self

Three Beacon Sensor Network
Localization through Self
Propagation
You are here
Mohit Choudhary
Under Guidance of: Dr. Bhaskaran Raman
Sensor Network Defined
• Collection of communicating sensing
devices.
• collator
The base station is a master node.
• Data sensed by the network nodes
sent to the base station
collator
Central processing,
& rapid response
collator
Localization in Sensor Networks
• Spatial localization - determining
physical location of a sensor node in
the network.
• Localization is an essential tool for:
– The development of low-cost sensor
networks for use in location-aware
applications
– Geographic routing
Aim of the Thesis
• Distributed algorithm for a sensor
network localization in which nodes
have the ability to assimilate and
utilize the connectivity and
negativity information provided by
their nearby nodes.
– Presence of three beacon nodes that
are aware of their position apriori.
Problem Difficulty
• Often insufficient data to compute a
unique position assignment for all
nodes.
• Information itself may be erroneous
due to presence of noise and gray
areas
• Algorithm has to scale with the size
of the network
Contribution of the Thesis
• A novel distributed algorithm
• Evaluation using simulations for a
wide range of scenarios to establish
an optimal conditions required to
achieve effective localization with
the proposed scheme.
• Real life experiments with in-house
developed sensor nodes (based on
Atmel AT89C52 microcontroller and
radiometrix TX/RX)
Related Work
• Use of GPS.
• Use of multiple static
reference beacons.
– Use of signal-strength
estimation.
– Use of acoustic ranging.
– Use of connectivity
information.
• Use of Mobile Beacon.
– Expensive (cost, size, energy)
– Only works outdoors, on Earth
– Technology additional to the
requirement of the application.
– Technology additional to the
requirement of the application.
– Complex setup procedures
– Terrain uncertainties
– Expensive
– Not possible for ad hoc, sensor
networks
– Additional Technology
– Terrain uncertainties
Specific Related Work
• Doherty et al. - Centralized technique
using convex optimization.
• Bulusu et al. - Distributed algorithm
based on connectivity and multiple
beacons.
• Savarese et al.- Two-phase approach,
connectivity for initial position estimates
and trilateration for position refinement.
(multiple beacons but less in number.)
• Patwari et al. - One-hop multilateration
from reference nodes [received signal
strength (RSS) and time of arrival (ToA)]
Preliminaries
• Connectivity based RF Localization
Bounding Box - {[A, B, C, D]}.
Negativity based RF Localization
Bounding Box- {[A,B,C,G](K,J)}
Localization point and
Localization Error
• Let Bounding Box- {[B1, B2, B3, B4] (derived
points)} then LP is:
1. If (segments B1B3 and B2B4 intersect) LP = point of
intersection.
2. Else if (center points of both B1B3 and B2B4 lie inside
the polygon) LP =center point of B1B3 or B2B4 that lies
closer to the centroid of the polygon represented by
{[B1,B2,B3,B4](derivedpoints)}.
3. Else LP = center point of B1B3 or B2B4 that lies inside
the polygon represented by
{[B1,B2,B3,B4](derivedpoints)}.
• The localization error, εA - Length of the line segment
MLP .
Proposed Algorithm- Terminology
•
•
•
•
Beacons- Know their position apriori.
Non-Beacon- Location unknown at start.
Unique Node-ID – NID
Message ID - MID
– MID=1 is used by beacon nodes.
– MID=2 is used by pseudo-beacon nodes.
– MID=3 is used by other nodes.
• Negativity constraint set (NCS) – A set of
negativity constraints, each consisting of a 3tuple (NID, a, b).
• Message transmitted - A 5-tuple (NID, a, b,
MID, NCS).
Proposed Algorithm-Example Run
B2
Y
F
B3
A
X
E
B1
Step-2:
InBeacon
the second
stage,
the pseudo-beacon
localizes
Step-1:
nodes
transmit
their
location. We
assume
that
Step-2 (contd.):
Nodes
which
hear
a transmission
from
a itself,
there
and
transmits
is at least
itsthus
one
location
other
to
node
its neighbours.
which
can hear
This
all
transmission
three
pseudo-beacon
immediately
localize
themselves.
beacons.the
includes
A node
negativity
which constraint
hears all three
of allbeacon
the three
nodes
beacon
is termed
nodes.
as athis
For
pseudo-beacon
reason, this transmission
node.
is very expressive.
Valid Set of Messages
• A set of threshold number of
messages of MID=3, with the same
x and y values, but different NID’s
constitutes a valid set. These
different messages would be sent
by different non-beacon nodes
which localize themselves to the
same point.
• Also, a single message with MID=1
constitutes a valid set.
Example Run Contd.
F
A
Y
X4
X1
E X3
X2
X
Step-3: In the third stage, the algorithm proceeds in a distributed
Let us assume X1, X2, X3 and X4 are the nodes
fashion. Nodes which have localized themselves transmit their
in region E
Threshold
Threshold
=
1 2any negativity constraints
location to their neighbours,
along=with
as appropriate. This in turn helps in the localization of further
nodes.
Tuning
• In our simulations, we have found that
the use of threshold=2 usually works well.
• A node which does not receive two valid
sets of messages, but only one valid set,
temporarily localizes itself with just this
information and sends a message. On
receiving the second set, it further refines
its position, and sends a second message
announcing this information.
Sources of Error Possible
• There is error inherent in the connectivitybased and negativity-based constraints.
• A negative constraint may be false, due to
non-receipt of a valid message
• The presence of gray areas and nonuniform radio connectivity can cause
further errors in calculations.
• Some nodes which are "far away" from
other nodes may not receive enough
information to localize themselves.
Simulation Setup
• Platform- Visual Sense that builds on
and leverages Ptolemy II.
• Two sets of simulations for various region
of operation.
– TX radius of 30mts/ node
– TX radius of 50mts/ node
• Study the effect of varying the value of
threshold(1,2 or 3) in each case.
• Nodes within a localization error of 20%
of transmission range turn green, those
within 10% turn red.
Example Simulation Run
Simulation Results
Area [0,300]x[0,300]
Node radius: 50 mts
Threshold: 1
Avg. Error: 7.306
50
47
45
43
40
Number of Nodes
49
39
35
33
30
26
25
20
15
13 14
10
5
0
1
5
3
2
2
3
5
4 5 6 7 8 9 10 11 12
Localization error (mts)
Simulation Results
Area: [0,300]x[0,300]
Node radius : 50 mts
Threshold: 2
Avg. Error: 5.4693
50
49
48
45
45
Number of Nodes
40
37
35
30
25
20
19
15
14
10
6
5
3
1
0
1
2
3
4
5
6
7
8
Localization error (mts)
9
10
Simulation Results
Area: [0,300]x[0,300]
Node radius : 50 mts
Threshold: 3
Avg. Error: 4.64
50
45
Number of Nodes
40
35
34
31
30
25
23
20
15
15
10
7
5
3
1
0
1
2
3
4
5
6
7
8
Localization error (mts)
9
10
Factors
• Node Density (Number of Nodes
deployed / Region of Operation)
• Node TX Range
• Value of Threshold
• We define:
Coverage=(Node Density) x (Node Tx Range)
Compiled Simulation Results
Node TX
Radius
Threshold
1
30
50
Region of
Operation
Coverage
[0,120]x[0,120]
9.8125
% of Nodes
localized
Avg. Localization
Error
72
6.0
82
5.6
2
[0,120]x[0,120]
3
[0,120]x[0,120]
100
3.9
1
[0,180]x[0,180]
82
5.022
2
[0,180]x[0,180]
98
3.97
3
[0,180]x[0,180]
74
3.81
1
[0,240]x[0,240]
54
5.206
2
[0,240]x[0,240]
32
5.17
1
[0,200]x[0,200]
82
8.44
2
[0,200]x[0,200]
86
7.06
3
[0,200]x[0,200]
100
5.437
1
[0,300]x[0,300]
86
7.306
2
[0,300]x[0,300]
98
5.469
3
[0,300]x[0,300]
68
4.64
1
[0,400]x[0,400]
58
7.096
2
[0,400]x[0,400]
38
5.952
4.36
2.452
9.8125
4.358
2.452
Analysis
• Coverage ≈ 4.36
• Value of threshold=2 is good both in terms of
percentage of nodes localized and the Avg.
Localization error.
• We now put forward the following questions
that we aim to answer:
– How well does the algorithm scale with
increased area of operation?
– Can the algorithm be used in harsh
environments such as indoors considering the
multipath effect and gray areas produced in
RF communication?
– What is the effect of unreliable
communication on the proposed scheme and
what are the measures needed to ensure
reliability in communication?
Large Area of Operation
Total Nodes: 272
Total nodes : 555
Area: [0,700]x[0,700]
Node radius : 50 mts
Threshold: 2
Avg. Error: 5.76
Area: [0,1000]x[0,1000]
Node radius : 50 mts
Threshold: 2
Avg. Error: 6.08
300
256
454
238
Number of Nodes
Number of Nodes
250
523
500
267269
200
156
150
100
98
76
404
400
321
300
200
178
123
100
50
16
4
0
1
2
65
29
9
10
26
7
0
3
4
5
6
7
8
Localization error (mts)
472
1
2
3
4
5
6
7
8
Localization error (mts)
9
94% nodes localize
10
Gray areas and multipath effect
• Zhao J et al. “Understanding Packet
Delivery Performance in Dense Wireless
Sensor Networks.”
– Extent of gray areas is less for lower transmission
ranges.
– Extent of gray areas varies greatly with the
environment (indoor, out door, habitat) even for
the same transmission range.
– For a given transmission range the extent of gray
area is the maximum for indoor.
– the extent
– of gray areas in indoor environment measures up
to 1/3rd of the transmission range.
Gray areas and
Multipath effect
50
45
40
35
30
25
20
15
10
5
0
Area: [0,300]x[0,300]
Node radius : 35-50 mts (random)
Threshold: 2
Avg. Error: 6.097
50
50
45
40
42
40
Random
Ideal
24
12
0
1
1
4
Number of Nodes
Number of Nodes
Area: [0,300]x[0,300]
Node radius : 50-65 mts (random)
Threshold: 2
Avg. Error: 6.4
35
30
25
24
20
17
15
11
10
7
41
39
11
7
5
4
3
0
2 3
4 5
6 7
8 9
Localization error (mts)
1
1
2
3
4
5
6
7
8
Localization error (mts)
9
10
Effect of unreliable
communication
Number of Retransmissions/ node: 1
Area: [0,300]x[0,300]
Node radius : 50 mts
Threshold: 2
Avg. Error: 5.46
Area: [0,300]x[0,300]
Node radius : 50 mts
Threshold: 2
Avg. Error: 5.309
50
50
45
45
40
37
35
34
30
29
25
21
20
15
12
10
2
0
1
3
4
5
6
7
8
Localization error (mts)
9
10
39
30
25
24
20
15
12
9
5
2
2
37
35
10
7
5
45
40
38
Number of Nodes
Number of Nodes
49
4
3
0
1
2
Probability of transmission error : 0.2
3
4
5
6
7
8
Localization error (mts)
9
10
Experiment Setup
• Area of operation of [0,100] X
[0,100]
• Sensor nodes are made up of
onboard:
Atmel AT89C52 microcontroller.
– A radiometrix BiM-433-F radio
TX/RX. operating at 433 MHz
frequency.
– A voltage stabilization circuit .
– A 9V battery.
–
Block Diagram
Features:
1) Maximum Data Rate: 40 kbps.
2) Frequency: 433 MHz.
3) Output power: 6 dBm, Receiver sensitivity: -107dBm.
4) Rapid RX power up ( <1ms )
5) Simple UART interface
• Features
– 8K Bytes of In-System Reprogrammable
Flash Memory
– Endurance: 1,000 Write/Erase Cycles
– 256 x 8-bit Internal RAM
– 32 Programmable I/O Lines
– Three 16-bit Timer/Counters
– Eight Interrupt Sources
– Programmable Serial Channel
– Low-power Idle and Power-down Modes
– One Serial Interface
Circuit Diagram
Communication Protocol Details
• Microcontrollers configured to communicate
with the radio in UART mode 1 at 9600
Bauds.
Byte Transfer
UART Mode 1
• Microcontroller used to select the radio to
transmit or receive.
• CD (carrier detect) of the radio is connected
to INT0 to decide changeover.
Packet Structure
• Preamble: 3ms Preamble mandatory
for the receiver in the radiometrix chip
to stabilize.
• Control: We used a unique Byte pattern
(0xAA) to indicate the start of message
• Data: The data as used in the actual
implementation is as shown in Figure6.4.
• CRC: 16 Bit Checksum of control - data
fields has been used by the decoder to
verify the integrity of the packet.
Spatial Profile
Experiment Setup
• The experiment setup has been considered keeping in mind the
results obtained after simulation
Threshold=2
(Coverage = 4.358) ,
Coverage= [(Node density) X (TX Area / Node)]
For the transmission radius of 30 meters following node density is
desired
Node Density = 4.358/ (3.14x30x30)
Node Density = 0.00154
Thus, for area of 10,000 sq meters (100x100), the number of
nodes required would be
Number of nodes required = (Node density) x (Area)
Number of nodes
= 0.00154 x 10000
Number of nodes
= 15.4
Experiment Details
• Four sensor nodes available.
• Requirement of 15 nodes and three
beacons.
• Technique of rotating the nodes that had
once achieved localization.
Experiment Results
Total nodes: 15
Area: [0,100]x[0,100]
Node radius : 30 mts
Threshold: 2
Avg. Error: 4.266
Area: [0,100]x[0,100]
Node radius : 30 mts
Threshold: 2
Avg. Error: 3.56
15
14
14
12
Number of Nodes
Number of Nodes
12
14
10
8
6
4
4
3
3
2
8
6
4
4
4
2
1
1
0
1
10
2
3
4
5
6
7
8
Localization error (mts)
9
10
0
0
1
2
3
4
5
6
7
8
Localization error (mts)
9
10
Conclusions
• Proposed method of localization is a complete and
reliable process in itself.
• Applications based on sensor networks requiring
reasonable localization accuracy can effectively utilize the
proposed method to cut down on additional cost of
localization otherwise incurred.
• As a few numbers of nodes in each region localize to a
common point by the proposed method, this information
can be further utilized by the applications to either
perform power management by utilizing a very small
number of such nodes at one time or perform easy
clustering of all such nodes.
Future Work
• Analyzing the effect of using other radio models
such as a octagon instead of a square on the
proposed scheme.
• Attempt to extend the methodology to 3D
networks.
• On the hardware level as a first step attempt to
integrate the developed node with temperature
sensors in form of DS 1631 (Dallas semiconductor
digital thermometer chip).
• As a second step attempt to integrate RF-id reader
to the node as onboard sensor.
Thank you