Design of Distributed

Observers over Multiple

Sensors
By: Maryam Asad

Department of Electrical Engineering (DEE)

Pakistan Institute of Engineering and Applied Sciences (PIEAS)
 Supervisor
Dr. Muhammad Tufail
Professor and Deputy Chief Scientist at PIEAS-PAEC

Co-Supervisor
Dr. Muhammad Rehan
Professor and Principal Scientist at PIEAS-PAEC
 Breakdown of Thesis Title

Distributed
Cooperative Control problem

Observer Design Multiple Sensors

Design of State Estimators Wireless Sensor Networks
 Contents of This Presentation

Wireless Sensor
01 Motivation 02
Networks
A futuristic application to explain Explanation of the concept
the motivation of thesis

Distributed State Conclusion and

03 04
Estimation Future Work
Explanation of the concept Concluding remarks and
future work.
 01
Motivation
 Earthquake in
California in the
year 2050
How Wireless Sensor Networks and
Distributed State Estimation Prevented a
Disaster
 Location Of Earth Quake
Mojave Desert
Lancaster Veagas

Los Angeles

Victorville

Southern California
Disaster Strikes at 08:01 am
• The earth began to rupture several miles under the surface of
an uninhabited part of the Mohave desert.

• Decades of pent-up energy was violently released, sending

huge shear waves speeding toward greater Los Angeles, home
to some 38 million people.

• The potential for epic disaster might be only seconds away.

• The quake was enormous, even by California standards, as its

magnitude surpassed 8 on the Richter scale.
 The Timeline for Rescue

8:00:00:00
• The few dozen
sensors close to
08:00:00:10 08:00:30:00
the epicentre • Alarms sounded
• The sensors process the • Building de-tuned
detect unusual
patterns in the 08:00:00:03 data • Evacuation Orders
depths of desert • Sensors communicate
• Inaccessible states are 08:00:04:00 Sounded
estimated • Shock absorbers
anomalous behaviour • Measurements and • The information
from the wireless of internal beams
• Noise is ruled out estimated communicated
sensor network actuated.
• Consensus is reached • Data Fusion done
reaches the cities. • Traffic signals turn
• Dormant sensors • Consensus reached on red
activated estimated states and final
estimates communicated
to control centres in city
Chronology of Scientific Advancements that prevented the disaster

02 04
01
Equipped miniature
sensors for data
processing, fusion,
Intelligent storage and Distributed
Wireless
Sensors communication
Observers
Communication
Made possible Solved the
pervasive sensing problem of

03
in remote areas Made possible estimating the
without requiring decentralized control system’s
and achieving physically
infrastructure Consensus collective goals inaccessible full
Control through local state signal.
interaction only
 02
Wireless Sensor
Networks
 Introduction to WSNs
What are WSNs? Applications of WSNs Core Challenges to WSNs

A Wireless Sensor • Environment Monitoring • Energy Consumption

Network (WSN) refers to • Underground Mining • Communication and
a large assembly of • Target Tracking Networking
sensor nodes whose • Human Activity • Self organization and
spatial deployment Monitoring management of Network
serves the purpose of • Structural Health • Concern of Security and
monitoring a physical Monitoring Privacy
environment with • Supply Chain • Aggregation and Fusion of
effective cooperation via Management Data
wireless communication • Precision Agriculture
 03
Distributed
State Estimation
 Key Concepts
Multi Agent Systems Cooperative Control Consensus
To model the nodes in To control multi agent To model the synchronization
WSNs systems using behaviour in nature for control
decentralized approach problems

Distributed State
State Estimation Graph Theory
Estimation
To solve the problem of To decentralize the state To model the
inaccessible full state signal estimation for multi agent communication topology
of a dynamical system systems of WSNs
 Multi Agent Systems

● A multitude of autonomous dynamical systems that collaborate to achieve

synchronization in order to meet desired group behaviour characteristics or
desired group goals.

● Each system is characterized by an endowment of its own state variables

and dynamic characteristics and may be either continuous time (CT) or
discrete time (DT).
 Cooperative Control
● Cooperative control has the objective that an assemblage of autonomous dynamical
systems or MASs work in effective collaboration to meet desired group behavioural
goals relying only on local communication and information flow between constituent
agents

● The notion of cooperative control became popular owing to the technological leaps
and bounds in low cost, energy efficient and miniature size electronics,
communication and storage, sensing and actuation devices

● The brilliance of cooperative control of MASs derives from its providence of

redundancy, flexibility, cost reduction and increased efficiency in contrast to the
complicated control of a single agent.
 Consensus
● Consensus is a fundamental and important control problem in the field of cooperative
control of MASs. It refers to the states of all agents in the MAS reaching the same final
value. In other words, state of each and every agent gaining consensus and
synchronization

● When all the autonomous dynamical systems that serve as agents in a MAS have the
same role and place in the communication topology then the states of these agents
reaching consensus is technically referred to as Leaderless Consensus

● When one or more than one dynamical autonomous agent in a MAS is given the
authority to generate or command a target trajectory for the other agents to follow
then the synchronization of such a MAS with leader/s is termed as leader-following
consensus.
 Distributed State Estimation
● Distributed state estimation is where the agents in a MAS locally estimate the
full state signal of the dynamical system

● The local estimates are the states of the agents and should reach consensus
for the computation of the overall estimate of the dynamical system’s full state.

● Provided a local sensor measurement, the objective is the design of

distributed observers so that each observer locally estimates the state of the
system under the effects of uncertainty, disturbance and noise to meet a
predefined performance index.
● Two possible structures with respect to Fusion Centre.
 Graph Theory

● A discrete structure, G described as

G = (V, E) where V is a non-empty set
of vertices and E is a non empty set
comprising edges is termed as a
Graph in discrete time mathematics.
An edge has at least one and at most
two vertices connected to it.
 Challenges and
concerns in distributed
state estimation
1 Uncertainties and non-linearities in the dynamical system’s model.

2
The contamination of the system with external disturbances that are
unknown

3 Noise corruption in the measurements of the sensors.

Network induced phenomenon in the information transmission channels of

4 WSN. These network-induced phenomena include sensor saturation,

communication delays, quantization errors, packet drop outs and
measurement uncertainties
Information Flow between Distributed Estimators
● Distributed state estimator has two parts; the estimates of the estimator
itself and the estimates or measurements or both from neighbouring
estimators

● What information is shared depends on design constraints and

performance requirements

● Freedom and flexibility in the design of observers

 Communication Constraints
● Wireless communication is the backbone of WSNs and provide
flexibility, scalability and manipulability.

● Communication constraints include quantization errors,

transmission delays, data losses in the form of packet dropouts
and measurement uncertainties, time varying communication
topologies, competition between nodes, aperiodic sampling and
fading network channels etc.
 Noise Modelling
● Inevitability of noise contamination in system state and sensor
measurement equation
● Additive and Multiplicative Noise
● Noise modelling necessary to avoid destabilization and performance
degradation of state estimator
● Noise models in literature; random noise with known statistical properties,
unknown non-random noise with bounded energy and unknown, non-
random noise with bounded magnitude also referred to as UBB noise.
 Performance Criterion
● For the evaluation of the response quality of the error dynamics in the
context of distributed state estimation

● Decided based upon estimator structure and type, noise type and its
model and the model of the dynamic system itself.

● The most common performance criteria in literature at present constitute

Bounded Error Covariance, H∞ performance and Set Membership
performance.
 04
Conclusion and
Future Work
 Proposed Problems to be investigated
● Discrete Time System to be considered
𝑥 𝑘 + 1 = 𝐴𝑥 + 𝑓 𝑥 + 𝐵𝑢 + 𝐵𝑤 𝑤
𝑦 𝑘 = 𝐶𝑥 + 𝑔 𝑥 + 𝐷𝑤 𝑤
f and g are non-linear functions and w is UBB noise
● Scalable Filter Design under state and output non linearity
● Event Triggered State Estimation problem for locally Lipchitz models
subject to UBB noise and set membership performance criterion.
The aim of this thesis is rooted in the design of distributed observers
for WSNs borrowing mathematical aid from cooperative control of
MASs.

The future work in this thesis will be devoted to

• the study of relevant literature
• problem formulation
• investigation of existence of solution
• verification of mathematical results.
Thank You!