Sesh Commuri School of Electrical and Computer Engineering University of Oklahoma, Norman Research...

Sesh CommuriSchool of Electrical and Computer Engineering

University of Oklahoma, Norman

Research in Intelligent Systems-Wireless Sensor Networks-Reconfigurable Computing

-Applications

December 08, 2006

SWAN 2006University of Texas, Arlington, TX

Coverage Issues in WSNs

• Optimum Coverage in 2D and 3D spaces

• Border Coverage and Detection of Holes

• Self Healing Networks

• Tracking using Sensor Networks

What is a Wireless Sensor Network?

A Wireless Sensor Network (WSN) is an ad-hoc network composed of densely-populated tiny electronic sensing devices, distributed over an area or volume. The basic function of the network is to observe some phenomenon by using the sensors in the tiny devices and communicate it to a central point called the sink.

Introduction

Wireless sensor networks (WSN) consists of group of sensor nodes to perform distributed sensing task using wireless medium.

1. Characteristics low-cost, low-power, lightweight densely deployed prone to failures two ways of deployment: randomly, pre-determined

2. Objectives Monitor activities Gather and fuse information Communicate with global data processing unit

Maximal Breach Path , Distance to

the closest sensor is

maximized (Worst Case)

Maximal Support Path , Distance to

the closest sensor is minimized

(Best Case)

The longer the exposure time , the greater the

sensing ability is.

Sensing Region

Coverage The sensing region of C= {S1,…Sn} can

be expressed as :

A set of sensors is said to cover the region Ri if and only if

1

n

isi

C A

1 2, ,... }i a a amC S S S

i iR C

Coverage in WSNs Problem 1(Optimal deployment problem):

Given a region R, to be monitored, what is the minimum number of sensors needed in order to completely cover R?

Problem 2(Coverage problem): Given a dense deployment of sensor nodes, find a minimum subset of active nodes that gurantee full coverage of the required region R.

Optimal Coverage Theorem: Let D be the deployment of the sensor nodes

at the vertices ‘L’ of a body centered cubic (bcc) lattice spanning a region. Let be the distance between adjacent vertices of L and be the sensing radius of each sensor node. Then the deployment D is optimal if the lattice spacing .

Definition 5.3: The measure of optimality of a WSN is the ratio of the number of active nodes in the network to the minimum number of nodes that can completely

cover the same region.

1.118sR

Determination of Redundancy in Coverage

Consider a sensor node S0. The neighbor set of S0,N(S0)={S1,S2,..Sn}. Let Ck=A0∩Ak .

S0 is completely covered by its neighbors iff all Cks are covered by its neighboring sensing regions.

A circle C0 is completely covered by spheres if all the intersection points Ci∩CJ in D0 are covered by one or more adjacent spheres

0 100 200 300 400 500 600 700 800 900 10000

10

20

30

40

50

60

70

80

90

100

Grid Points

Num

ber

Of

Sen

sors

Cov

erin

g E

ach

Grid

Poi

ntBefore Running the Coverage Algorithm

After Running the Coverage Algorithm

1000 Grid points were tested, Sensor Radius=2, Grid Size =1, Occupancy Matrix - Initial Deployment (top) and after running the algorithm (bottom).

Minimum and maximum thickness before and after running the algorithm

Min. Thickness Max. Thickness

Before 8 89

After 1 5

The effect of number of queries on the coverage lifetime of a self healing WSN with 800 nodes –(i) Initial deployment (Original); (ii) Reduced deployment (Algorithm).

The effect of number of queries on the coverage lifetime of a self healing WSN with 1600 nodes –(i) Initial deployment (Original); (ii) Reduced deployment (Algorithm).

Coverage Hole Problem

Given a dense deployment of sensor nodes in a region R, find the subset of active nodes that lie on the boundary of the coverage holes.

Key Idea

If the circumference of the sensor’s circle is completely covered by its neighbors then it is not a border sensor and therefore could be deactivated without affecting the overall border coverage.

Successor

Key Theorem

Border Coverage

250 sensor nodes with sensing radius 0.75 units are randomly deployed in a 10x10 region.

31 sensor nodes are on the boundary cover of the 6 bounded coverage hole.

Optimal Sensor Placement for Border Perambulation

Sensor Network issues are addressed from the standpoint of optimal deployment of sensors nodes for boundary coverage. Lower bounds on the number of sensors needed for surveillance and target intrusion detection is presented and a “Border Perambulation” technique where each border node can swap between working and sleeping modes and the network only maintains a subset of working nodes

Optimal 2D Deployment

System Lifetime

FPGA Implementation of Dynamic Run-Time Behavior Reconfiguration in Robots


GoalDesign an appropriate architecture and implementation methodology for the implementation of Intelligent Robots.

- The architecture must allow for the integration of the various

functional modules and subsystems into the overall system.

- The methodology must address issues such as the network

connectivity, latency, bandwidth, reliability, and inter-module

communications that affect the system performance.


The architecture must address issues such as:

the ability to dynamically configure and retask individual robots,

the ability of the robots to identify and accommodate system faults,

distributed /centralized control of the robot teams, information and resource sharing between multiple

robots, and the efficiency of fault tolerance and learning algorithms.

Hardware Reconfiguration

What is Reconfiguration ? Ability of a system to change its execution path as

commanded by an application

Types of Reconfiguration Static Reconfiguration Dynamic Reconfiguration

Advantage FPGA’s can be reconfigured any number of times to

implement a hardware circuit

Static Reconfiguration

To upgrade/install new configuration for a system System has to be taken offline Reload with new set of data Reboot

Negative Impacts System Downtime

Dynamic Reconfiguration

Changes or modifications can be made while the system is under operation

System down time is avoided Flexibility to upgrade standards/features

Types of Dynamic Reconfiguration Full Reconfiguration Partial Reconfiguration

Full Reconfiguration

System starts with a Built In Self Test

Checks the proper functionality of the connected components

Reconfigures itself and resumes normal operations

Partial Reconfiguration

System modifications can be done on the fly

Reconfigures a portion of the system leaving the operational units untouched

Adds flexibility to the system

Use Case View of System Functionality

Logical Diagram of Layer 2 Entities and

Interfaces to Layer 1

Logical Diagram of Layer 2 Entities and

Interfaces to Layer 1

d d ix u y u cos ; sin ;

Controller for Wall Following

090 ( ) ( )2

f rh s h f r d

d du K K d d K d

Controller for Leader Follower

212 1

12 1 2 112

1( ')

dd

b b

kd k

N

k k x xd

Internal Architecture of Xilinx FPGA

Architecture of Prototype Test Bed

Creation of Interface Circuits for Sensors

Wall Following Using IR Sensors

Implementation of Reconfigurable Controllers on FPGAs

Module Occupied Slices (total : 4928) Equivalent Gate Count

Slice Count Percentage

Static Modules only 2862 58.1% 1,967,380

Static Modules + Wall Following Behavior Module

2983 60.5% 2,061,154

Static Modules + Wall Following + Leader/Follower

Behavior Modules

3273 66.5% 2,192,918

Static Modules + Wall Following +Leader/Follower + Obstacle Avoidance Behavior Modules

3463 70.3% 2,380,516

Overall Design without using partial reconfigurable

FPGAs

3509 71.2% 2,412,137

Implementation of Reconfigurable Controllers on FPGAs

Module Occupied Slices (total : 4928) Equivalent Gate Count

Slice Count Percentage

Static Modules only 2862 58.1% 1,967,380

Static Modules + Wall Following Behavior Module

3001 61% 2,062,930

Static Modules + Wall Following+Leader/Follower

Behavior Modules

3162 64.1% 2,170,551

Static Modules + Wall Following +Leader/Follower + Obstacle Avoidance Behavior Modules

3188 64.7% 2,190,868

Overall Design using partially reconfigurable FPGAs

3205 65.04% 2,202,382

Dynamic Reconfiguration based Fault Accommodation

MAGNETs: Mobile Adhoc Grid Networks

MAGNETs are a set of mobile computing devices that form a MANET which can be used for coordinated computation and communication.

Motivation and Objectives: A distributed set of mobile devices could be

observing a phenomenon and/or computing and communicating.

A user A would use these devices to perform analysis X, and a user B would use the same of subset of these devices to perform analysis Y.

Need for a adaptively reconfigurable architecture with protocols for to allow multi user application execution using a set of mobile communication and computational devices.

Develop a framework to manage resources in MAGNET.

Develop a high-level language specification using which a user can configure and use MAGNETs.

Layered MAGNET Architecture

Layered architectures allow of adaptively configuring and reconfiguring MAGNETs that suit the application needs.

Fabric Layer: data collection, storage, network resources, resident code for processing data, and computational resources

Connectively Layer: connection management functions, reliable transport of data, code fragments, and output resulting from distributed execution, authentication services for establishing user-based trust relationships.

Resource Layer: access to the resources at a node in the grid, API Interfaces, mobile code management

CollectiveGrid Resource Manager

ApplicationRoute Guidance and Planning

ResourceNode Resource Manager

ConnectivityLocal Communication Protocols

FabricNode Resources

Layered

MA

GN

ET

Arch

itecture

• Collective Layer: manages resources of the entire MAGNET.

• Application Layer: Execution of Mobile code on the behalf of the user.

Augmentation of WSNs using Mobile Reconfigurable Cluster Heads (RCHs)


Query based reconfiguration in the RCH


0

2

4

6

8

10

12

0 10 20 30 40 50 60

Number of Sensor Nodes in the Network

Pro

ce

ss

ing

Tim

e (

mic

ro s

ec

on

ds

)

Using SCH

Using RCH with DPR

Number of Sensor Nodes in the network (Vs) Query processing time (in a network that supports 5 data aggregation operations)

0

20

40

60

80

100

120

140

160

180

200

0 10 20 30 40 50 60Number of Sensor Nodes in the Network

Po

wer

Co

nsu

mp

tio

n (

mW

)

Using SCH

Using RCH with DPR

Number of Sensor Nodes in the network (Vs) power consumption (in a network that supports 5 data aggregation operations)

Conclusions

• The case studies demonstrate the performance enhancement using hardware / software reconfiguration. • The methodology that is proposed enables the designer to accurately capture the system requirements (using Use Cases), model the system, simulate the system to understand the interactions between the different objects, and then dynamically generate the code. • The design methodology enables the implementation of high risk/ changing modules in the FPGA while high performance / static modules can be implemented in custom circuitry or ASICs. • Finally, it is shown through the case studies that the requirements for intelligent robots can be met in an efficient manner using hardware / software reconfiguration. It is shown that systems can be designed to function under centralized control or for autonomous behaviors.

Advances in

Intelligent Compaction of AC Pavements Sesh Commuri

Associate ProfessorSchool of Elec. & Computer

Engineering

Musharraf ZamanAssociate Dean for Research

College of Engineering

What is Intelligent Compaction (IC)?

Intelligent Compaction (IC) is the process of continuously controlling the operational parameters of a vibratory compactor to optimize compaction and meet required conditions.

The IC controls the different compaction parameters for the roller such as the frequency and amplitude of the vibrations of the drum, and the roller speed.

Key components of IC are a system to measure the stiffness of asphalt, sensors to monitor the location (GPS) and the operational parameters of the roller, and devices to record and display these data to the user.

Advantages of Intelligent Compaction

• Higher efficiency and increased productivity

(More than 30% reduction in labor and fuel costs

Reduction in the number of conventional spot tests

Example: 75% reduction in QA testing (Sweden))

• Higher adaptability (thin/thick lifts, soft/stiff subgrades)

• Better quality resulting from uniform and optimum compaction

• Less crushing of the aggregates

• Complete coverage of compaction surface evaluation

• Dynamic measurement of soil stiffness

• Increased life of the equipment.

Disadvantages:

• It requires sophisticated equipment to survive in rugged, off-road conditions

• It requires operator training

• IC rollers are more expensive than conventional compactors.

From J.L. Briaud, J. Seo, Intelligent Compaction: Overview and Research Needs, December 2003.

http://www.webs1.uidaho.edu/bayomy/trb/afh60/IntCompaction_Briaud_September2004_.pdf

Spectral Analysis of Vibrations During Compaction

(a) Loose Mix Specimen (b) Pre-Compacted Specimen

Vibration Signature with Compaction Pressure of 758.4 kPA (110 PSI)compaction time = 60 s; Mix Temp.=152 ; 6.5 kgs (14.33 lbs)0C

-PR7: 8.2% air void

-PR8: 8.7% air void

Effect of Compacted Density on Vibration Characteristics

Field Compaction Results – Effect of Subgrade / Pavement Design

(a) Interstate I-35 2-inch (S3 – PG 64-22) (b) 3-inch (S3 – PG 64-22) on 6 inch Concrete base

(c) 4-inch (S3 – PG 64-22) on compacted subgrade (d) 2-inch (S4 – PG 70-28 OK) on 8 inch S3

Construction Site

Will Rogers International AirportConstruction July 20-28, 2006

Will Rogers International AirportConstruction July 20-28, 2006

1500 feet

860 feet

790 feet

61

8 fe

et

Spectrogram of Vibrations and Predicted Density – Pass 2

Distance (feet)

Fre

qu

en

cy (

Hz)

NN

Pre

dic

ted

Co

mp

act

ion

Le

vel

Spectrogram of Vibrations and Predicted Density - Pass 3

Distance (feet)

Fre

qu

en

cy (

Hz)

NN

Pre

dic

ted

Co

mp

act

ion

Le

vel

W

idth

(fe

et)

De

nsi

ty -

So

uth

La

ne

(% C

om

pa

ctio

n)

De

nsi

ty -

No

rth

La

ne

(% C

om

pa

ctio

n)

Loc. Core PQI Vicinity

1 93.2 93.3 92.7-93.7

2 93.5 93.2 92.6-93.2

3 92.5 92.8 92.1-92.8

4 92.1 93.1 92.4-93.6

As Built Density Map

As Built Density Map

feet ->

feet ->

feet ->

% D

ensi

ty%

Den

sity

IACA

Output

~ 90 %

~ 92 %

~ 94 %

Relief Features in Subgrade

NDP Control of Prosthetic Ankle

NDP Control of Prosthetic Ankle

Fully locked out

Slow walking

Fast walking

Jogging

Heel Strike

Mid Swing

Toe Off

Heel Off

Mid Stance

Foot Flat

Heel Strike

0

Damper resistance during each stage of gait for a variety of activities.

Sesh Commuri School of Electrical and Computer Engineering University of Oklahoma, Norman Research...

Documents

Transcript of Sesh Commuri School of Electrical and Computer Engineering University of Oklahoma, Norman Research...