1
Perpendicular Intersection: Perpendicular Intersection: Locating Wireless Sensors Locating Wireless Sensors with Mobile Beaconwith Mobile Beacon
Zhongwen Guo, Ying Guo, Feng Hong, Xiaohui Yang, Yuan He, Yuan Feng, Yunhao Liu
Ocean University of China
Hong Kong University of Science and Technology
2
Outline
Introduction Observations on RSSI Our Scheme: PI (Perpendicular Intersection) Design of Optimal Trajectory Experiments Conclusion and Future Work
3
Introduction
Locating sensor nodes is a crucial issue in WSN applications. OceanSense
https://www.cse.ust.hk/~liu/Ocean http://osn.ouc.edu.cn
4
Existing Approaches (1)
Range-Based Approaches TOA, TDOA, AOA
Require additional hardware support Expensive in manufactory cost and energy consumption
RSSI-based (Received Signal Strength Index) Easy to implement; Based on the log-normal shadowing model Rely on absolute RSSI values (unstable & irregular) Inaccurate due to channel noise, interference, attenuation,
reflection, and environmental dynamics
00
( ) ( ) 10 logT
dRSSI d P PL d X
d
Environment-dependent !
5
Existing Approaches (2)
Range-Based Approaches Mobile-assisted
Avoid cumulative errors of coordinate calculations and unnecessary communication overhead
Still rely on absolute RSSI values
Range-Free Approaches Rely on connectivity measurements (e.g. hop-count) Accuracy and precision affected by node density and
network conditions
6
More about OceanSense
Sparsely deployed restricted floating sensors Unstable wireless communications Varying network connectivity
The existing localization approaches cannot well support such a practically complex scenario.
7
Design Goals
A localization approach which is easy to implement in practice which has better accuracy, especially under dynamic and
complex environments. which is time and energy efficient in locating a network of
wireless sensors.
8
Observations on RSSI
The closer a node is to the signal sender, the larger RSSI value it perceives.
-15 -10 -5 0 5 10 15210
215
220
225
230
235
240
245
250
Distance Interval (m)R
SS
I
10m20m30m40m50m
Outdoor observation
9
Observations on RSSI
The closer a node is to the signal sender, the larger RSSI value it perceives.
Indoor observation
-8 -6 -4 -2 0 2 4 6 8200
205
210
215
220
225
230
235
240
245
250
Distance Interval (m)R
SS
I
4m8m12m16m
A
-8
-4
-2
-1
1
2
4
8(m)
Static Sensor Node
O
Mobile Node A’ PositionTrajectory of Node A
16(m)1284
10
PI (Perpendicular Intersection)
N
P1(x1,y1)
H’(x’, y’)
H’’(x’’, y’’)
(x, y)
start
stop start
stop M
Virtual Triangle (VT)
P2(x2,y2)
P3(x3,y3)
No longer absolute RSSI values!
11
Theoretical Estimation Error
Given the velocity of the mobile beacon as V and the broadcast frequency as F, the distance between two beacon points is V/F.
or
12
Optimal Mobile Trajectory (1)
The optimal mobile trajectory to locate a network of sensor nodes satisfies the following requirements: All the sensor nodes can be located.
The optimal trajectory consists of multiple joint VTs, which cover the entire deployment area.
It is the shortest trajectory. The mobile beacon traverses the entire area in the shortest time and
consumes the minimum energy cost.
The localization latency of a sensor node is minimized. The node should be located as soon as the mobile beacon traverses along
the two sides of the VT.
13
Optimal Mobile Trajectory (2)
The node is in the VT which has largest sum of RSSI values. Side length of a VT = R (transmission range)
P1
H
P2 N1
N1'
P3
P4
L
14
Optimal Mobile Trajectory (3)
Trajectory length:
Localization latency:
P1
H
P2 N1
N1'
P3
P4
L
15
Analysis on Overhead
Communication Cost Zero communication cost among the sensor nodes to be located. The number of beacon messages received by a sensor node when the
mobile beacon traverses one VT side is FR/V. A sensor node receives the beacon messages from at most 6 sides.
The upper bound of communication cost of a sensor node is 6FR/V . Computation Overhead
The computation overhead on a sensor node is O(FR/V). Storage Overhead
PI only needs to store at most 14 vertices with their corresponding RSSI values, which cost 70 bytes.
16
Experiments (1)
A prototype system with 100 TelosB motes Various environments:
library hall, laboratory, racket court, parking lots, and the sea. The mobile beacon is also a TelosB mote. A base station is deployed to collect the localization results. PI is compared with two existing approaches
TRL [1]: a range-based approach using trilateration. (3 vertices) BI [2]: a mobile-assisted localization approach that exploits Bayesian
inference to improve the estimation accuracy. (3 vertices + 3 random)
[1] J. Hightower and G. Borriello. Location systems for ubiquitous computing. IEEE Computer, 34(8):57 – 66, August 2001.
[2] M. Sichitiu and V. Ramadurai. Localization of wireless sensor networks with a mobile beacon. Proceedings of IEEE MASS, 2004.
17
Experiments (2)
Hall Experiment In a hall of our library, which is about 450m2. The side length of a VT R=15m. The moving velocity V=0.1m/s The broadcast frequency F=1time/sec.
18
Experiments (3)
Hall Experiment – Group 1 14 sensor nodes are deployed randomly.
RSSI values perceived by node N6
Ave. Estimation error: PI 1.2175m BI 2.4921m TRL 3.3631m
Deployment areaBase Sation
Unknown node
A Pillar
Wall
N1
N2 N3
N4
N5
N6
N7
N8
Mobile beacon trajectory
Mobile beaconN9
N10
N11
N12
N13
N14
19
Experiments (4)
Hall Experiment – Group 2 35 sensor nodes are deployed randomly.
Deployment areaBase Sation
Unknown node
A Pillar
Wall
Mobile beacon trajectory
Mobile beacon
Ave. Estimation error: PI 1.3213m BI 1.6978m TRL 3.6617m
20
Experiments (5)
Laboratory Experiment In the laboratory of computer software center, a room of 324m2 with
120 computers and desks inside. People sit, stand, or walk in the room. R=9m. V=0.1m/s. F=1time/sec.
21
Experiments (6)
Deployment area
Base Sation
Unknown node
Wall
N1N2 N3
N4
N5
N6
N7
Mobile beacon trajectory
Obstacle
Door
Mobile beacon
N9
N8
N12
N10
N11
Laboratory Experiment – Group 1 12 sensor nodes are deployed.
Ave. Estimation error: PI 1.7655m BI 2.9775m TRL 4.0880m
22
Experiments (7)
Laboratory Experiment – Group 2 100 sensor nodes are deployed.
Deployment area
Base station
Sensor node
Wall
Obstacle
Door
Mobile beacon
Trajectory ofMobile beacon
Ave. Estimation error: PI 2.5645m BI 3.5829m TRL 4.7299m
23
Experiments (8)
Racket Court Experiments R=15m. V=0.1m/s. F=1time/sec.
Ave. Estimation error: PI 1.2174m BI 2.2313m TRL 3.6942m
24
Experiments (9)
Parking Lots Experiments R=15m. V=0.1m/s. F=1time/sec.
Ave. Estimation error: PI 1.0952mBI 2.2079m TRL 3.7324m
25
Experiments (10)
Offshore Experiment V=1.5m/s. F=1time/sec.
0 20 40 60 80 100 1200
20
40
60
80
100
120
Deployment area (m)
Dep
loym
ent a
rea
(m)
1
2
1 20
5
10
15
Node ID
Loca
tion
Err
or (
m)
PI ErrorBI ErrorTrilateration ErrorAverge PI ErrorAverge BI ErrorAverge Trilateration Error
Ave. Estimation error: PI 7.3391 m BI 7.9094 m TRL 9.1620 m
26
Experiments (11)
Impact of Different Factors Evaluated with the hall experiments. V is set at 0.05m/s, 0.1m/s, 0.2m/s
and 0.4m/s, F is set at 0.5, 1, 2, and 4 times per second, respectively.
The intersection point of the two curves in the right figure represents a good setting in practice, which sets appropriate trade-off between the localization accuracy and the communication cost.
27
Conclusion and Future Work
Conclusion We propose a mobile-assisted localization approach: Perpendicular
Intersection (PI). Easy to implement Accurate in dynamic and complex environments Time and energy efficient
We examine the performance of PI by implementing a prototype system.
Further Work Improve the prototype system, introducing an automatic mobile
beacon. Large-scale field tests on the OceanSense platform. Extend PI in the underwater acoustic sensor networks.
28Thanks !