Transcript RSSI.Based.Tracking...

RSSI Based Tracking Algorithms for
Wireless Sensor Networks: Theoretical
Aspects and Performance Evaluation
Relatore: Dr. Michele Rossi
Correlatore: Dr. John Woods
Laureando: Riccardo Masiero
Corso di laurea specialistica in Ingegneria delle Telecomunicazioni A.A. 2006-2007
Contents
>
>
>
>
>
>
>
>
>
Objective of the Thesis
Organization of the Thesis
State of the Art: Overview
State of the Art: Tool Collection
Simulation Tests: Organization
Simulation Tests: Positioning Results
Experimental Tests: Organization
Experimental Tests: Refinement Results
Conclusion
Objective
> Investigation of suitable localization algorithms
- inexpensive (RSSI based);
- simple (low complexity).
> Taking the first steps towards a software
implementation for low cost wireless sensors
developed by the University of Essex.
Objective
Block diagram of a sensor node used during the
experiments.
Organization
> Localization and Tracking : State of the Art
- literature review;
- tool collection.
> Simulation tests
- simulation set-up;
- algorithm comparison.
> Experimental tests
- simulation validation;
- algorithm feasibility.
State of the Art:
Overview
Localization algorithm structure:
> Ranging (RSS, ToA, AoA, Connectivity/BER);
> Positioning (one-hop or multi-hop solutions;
absolute, relative or local);
> Refinement (function optimization, filtering,
cooperative organization).
State of the Art:
Tool Collection
> Ranging – Received Signal Strength (RSS)
(voltage measured by onboard circuitry - RSSI);
> Positioning - LATERATION, MINMAX, ROCRSSI+
(by now one-hop, distributed solutions with relative
localization);
> Refinement - KALMAN FILTERING
(suggested for tracking applications).
State of the Art:
Tool Collection
LATERATION
Mobile nodes are
localized using
overlapping
circles. The
circumference
radii are equal the
estimated
distance among
nodes.
State of the Art:
Tool Collection
MINMAX
Mobile nodes are
localized using
squares regions.
The square sides
equal twice the
estimated
distance among
nodes.
State of the Art:
Tool Collection
ROCRSSI+
Mobile nodes are
localized using
patchworks-like
region slices. The
region slices are
drawn comparing
the power
received by the
nodes.
State of the Art:
Tool Collection
KALMAN FILTERING
It is an optimal recursive data processing algorithm.
It allows the estimation of a measurable quantity
from both measurements and using a priori knowledge
about the observed phenomenon.
State of the Art:
Tool Collection
KALMAN FILTERING
Simulation Tests:
Organization
IMPLEMENTED CODE
Simulation Tests:
Positioning Results
BIAS
> LATERATION is
without error;
> MINMAX and
ROCRSSI+
inherently lead to
estimation errors;
Estimation error [cm],
path loss only.
> the bias becomes
smaller as the anchor
nodes increase.
Simulation Tests:
Positioning Results
Estimation errors [cm]
ALGORITHMS
> MINMAX achieves
the best performance;
> ROCRSSI+
performance is
adequate.
TOPOLOGY
performance becomes
better:
> as the anchor nodes
increase;
> arranging the anchor
nodes around the
space of interest.
Experimental Tests:
Organization
Experiment position-patterns
Experimental Tests:
Refinement Results
ALGORITHMS
> MINMAX and
ROCRSSI+ are
both robust;
> MINMAX bias
is difficult to
correct in noisy
conditions;
> LATERATION
Experiment along a square shaped path.
is useless.
Visualization of estimated trajectories.
Experimental Tests:
Refinement Results
SIMULATION
> KALMAN
FILTERING can
be used to track
a mobile
satisfactorily;
> the nodes must
provide further
knowledge wrt
plain RSSI
Experiment along a square shaped path.
measurements.
Estimation errors [cm].
Conclusion
ALGORITHMS
> MINMAX is the best choice for the used sensor nodes both
for its simplicity and performance;
> KALMAN FILTERING is suitable for the used sensor nodes.
The assumption of a linear process, made to simplify the
implementation, can be adopted;
FUTURE WORK
> the nodes have to be equipped with a velocity sensor and
magnetic compass;
> the SIMULATION set-up realized can be powerfully used
for future research.
Conclusion
THANK YOU FOR YOUR ATTENTION!
ANY QUESTIONS ?