a virtual infrastructure for mitigating typical challenges in sensor networks

71

Upload: michele-weigle

Post on 26-May-2015

658 views

Category:

Technology


2 download

DESCRIPTION

Hady Abdel-Salem's PhD Defense Slides Department of Computer Science Old Dominion University November 1, 2010 Note: You may need to download the file to see all of the animations.

TRANSCRIPT

Page 1: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks
Page 2: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Agenda Introduction. [5 min]

Motivation & Related work. [5 min]

Backbone Construction Protocol. [10 min]

Mitigating Network Challenges: Clustering & Sensor Localization. [5 min] Energy-Aware Task Management.

Workforce Selection [10 min]

Centralized Approach. Distributed Approach.

Sensor Sleep Scheduling [10 min]

Data Aggregation & Routing. [5 min] Energy Hole Problem. [5 min]

Conclusions & Future Work [5 min]

Page 3: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

• Collected data are locally aggregated then forwarded for further processing at sink nodes which in terms of energy, computation, and communication have more powerful capabilities compared to sensors.

Introduction• Sensor Networks are special type of Ad-Hoc networks which

include two types of nodes, “Sensor nodes’’& “ Sink nodes’’.

• Sensor nodes are tiny electronic devices with limited sensing, computational, and communicational capabilities.

• In a typical scenario, sensors are massively deployed in an area of interest to collect data that serve the mission of the network.

Page 4: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

• Sensor networks have many applications … some of them are military related:

battlefield surveillance. borders monitoring.

• Others are civilian: fire and habitat monitoring home automation. traffic control. target tracking. body sensor networks

Sensor Network Applications

Page 5: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Design Limitations

• Limited and non-renewable energy budget.• Short duty cycle.• Limited computing power.• Small transmission range.• High failure rate (unreliable).• Location unawareness (no GPS).

• Random nature of deployment.• Working unattended (autonomously).• Dynamic topology.

Sensor:

Network:

Page 6: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Motivation & Related Work• Several techniques have been proposed to address different

problems in WSNs. For Localization, we have APIT, DV-Hop, Centroid, Weighted Centroid, … For Routing, we have GPSR, GFS, MFR, … For clustering, we have Hierarchical clustering, LEACH, …

• A common drawback when using any of these techniques, they attempt to solve one problem in isolation from the others, hence protocol designers have to face the same common challenges again and again.

• This, in turn, has a direct impact on the complexity of the proposed protocols and on energy consumption.

Page 7: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Motivation• In this thesis, we present a different approach to address the problems mentioned

earlier.

• Specifically, In a one-time investment, we construct a lightweight network backbone which

provides some kind of an infrastructure that makes the network easier to manage.

On top of the constructed backbone, we provide solutions for the following problems: Sensor Localization Clustering Energy-Aware Task Management

Workforce Selection Sleep Scheduling.

Data aggregation & routing. Energy hole problem.

Page 8: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Backbone Construction Protocol

Page 9: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Construction Protocol• Initially, the deployment area is hypothetically tiled with identical hexagons starting

from the sink outward.

• The first hexagon is placed so its center coincides with the center of the sink node.• Hexagons are placed next to each other in the following directions:• The geometry of the gaps between the hexagons in any two consecutive directions allow

perfect tiling.• The size of the hexagon is determined by the distance between the centers of any two

neighbored hexagons which is chosen to be ≈ tx

Sensor

Sink

Tx

Page 10: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

• Hence, the deployment area is divided into six sectors.• We use a ternary coordinate system <sector, row, column> to uniquely identify hexagons.• The closest sensor to the center of each hexagon is determined (backbone sensor). These sensors collectively form the network backbone.

Construction Protocol (cont.)

Page 11: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Order of Backbone Sensors Selection

• Selection process goes recursively, backbone sensors in any row are responsible for selecting backbone sensors in the next row.

• Each backbone sensor with odd column coordinate selects 2 sensors in the next row. i.e Sensor <s, r, 2c-1> selects sensors <s, r+1, 2c-1> and <s, r+1, 2c >

• Backbone sensors in the first column in even rows select 3 sensors i.e Sensor <s,2r,1> selects sensors <s,2r+1,1> , <s,2r+1,2> and <s-1,r+1,r+1 >

• Backbone sensors can be selected in many ways that may result in different order. To reduce collisions, we propose the following rules.

• Selection of backbone sensors with the same row and column coordinates in different odd sectors (i.e. s=1,3,5) occurs in the same time. A similar rule applies for even sectors (i.e. s=2,4,6).

Page 12: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Order of Backbone Sensors Selection

Page 13: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

How backbone sensors are selected ?

• Backbone sensors are selected to be the closest sensors to the centers of the hexagons they represent.

• To determine these sensors, our approach requires sensors to be able to measure their angle to the sink node.

• Hence, we find it more appropriate to start by showing how a sensor can estimate it angle to the sink node.

Page 14: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Directional Transmission Pattern

The transmission pattern of directional antennas consists of a major lobe oriented towards the transmission direction and several minor and side lobes oriented in other directions.

Page 15: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Measuring Angles

Received Power

Initially, the sink uses it omnidirectional transmitter to send a sequence of wakeup messages to make sure sensors in its neighborhood are awake.While rotating its unidirectional antenna, the sink sends a sequence of beaconing messages attaching current transmission angle to each message.From received messages, a sensor can estimate its angle to the sink as the average of the received angles weighted by the received power pr.

θ

Sensor

Sink

Page 16: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Measuring Angles

Received Power

The process can be repeated for better accuracy.

θ

Page 17: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

• Sensors that receive the message, verify that the absolute difference between their measured angle and ϕ is less than a predetermined threshold, otherwise, they ignore the message.

• Sensor S1 estimates the angle ϕ between the sink node and the target hexagon and starts the search process by broadcasting a message that includes ϕ to its neighbors.

ϕ

S1

3

7

5

6

14

9

2

6

4

5

13

8

1

5

3

4

12

7

New Backbone Sensor

Selection of backbone sensors

• Sensors within range, estimate the square of their distance to the center of the target hexagon, and initialize a countdown timer to this value.

• The sensor that has the first timer to expire broadcast a message to its neighbors to announce itself as the newly selected backbone sensor.

Sink

Page 18: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Estimation of Distance e2

Case 1 Case 2

Page 19: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Common intermediate terms can be reused to evaluate the Sine and the cosine functions using 10 multiplication operations only, with an accuracy up to 4 decimal digits.

Page 20: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Problem of Voids

Page 21: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Problem of Voids (Cont.)

Void

Page 22: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Even Neighbor Replacement

• Recall that in our basic selection rules, only sensors with odd column coordinates are allowed to select.

• Note that, every even neighbor can receive selection messages transmitted from its two immediate odd neighbors.

Page 23: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Even Neighbor Replacement• If an odd sensor is missing due to voids, then all its

selection tree will be pruned.• Idea: allow immediate even neighbor sensors to replace

missing sensors in order to continue the selection chain.

Page 24: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Recovering From Voids(Even Neighbor Replacement)

Page 25: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Recovering From Voids Backward Selection• If selection of a backbone sensor was initiated by a sensor

other than the one determined by the initial rules. (e.g. Even neighbor replacement or backward selection),

Then the newly selected sensor should try to select the odd sensor that was supposed to select it.

Page 26: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Recovering From Voids (Backward Selection)

Page 27: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Backbone Switching

θ

Rotation Angle θ

• To balance the load on backbone sensors, we use alternative backbones.• Idea: rotate sector orientation angles by θ1, θ2…

Page 28: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Actual Hexagons obtained through Simulation

Page 29: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Actual vs. theoretical positions of backbone sensors

Page 30: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Sensor Localization & Clustering

During the selection of backbone sensors, sensors can use their knowledge about their angle to the sink along with the computed distance to the sink to fully localize themselves.

Moreover, the hexagonal structure of the backbone provides an implicit clustering mechanism in which each hexagon represents a cluster and the backbone sensor around its center represents its cluster head.

Hence, sensor localization & clustering are direct by-products of the backbone construction protocol.

Page 31: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Average Localization Error for Different Localization Protocols

Page 32: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Energy-Aware

Task Management

Page 33: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Workforce Selection

• Sensing tasks are issued to sensors through sink nodes.

• A task T identified by the tuple (x, y, w), where (x,y): is the position where data need to be collected. w : QoS requirements expressed in terms of number of sensors

participating in the task (we refer to them as task workforce).

• Obviously, only sensors whose sensing range cover the point (x,y) can join the workforce for the task T.

Page 34: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

• Assume that each of the tasks T1 and T2 requires 3 sensors.

• Case 2: assign < a, b, c> to T1 We can run T2 using < e, g, i >.• Case 1: assign < b, e, f > to T1 No way to run T2.

• Considering sensor remaining energy during workforce selection can enhance network reliability and durability.

• Assigning tasks to sensors improperly can consume sensor energy unevenly which in turn may result in many problems

e.g. reducing network density, creating energy holes, …

Example

Page 35: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Workforce SelectionCentralized Approach

Page 36: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

• A task starts when the sink sends a sequence of CTW to get the attention of enough number of sensors (including backbone sensors across routing path)

Workforce Selection

• CTW message should contain:• W: required workforce w,• (x,y): task position,• Emax: sink estimate of maximum

energy in target hexagon.• S: number of bidding slots in

first bidding round.

Page 37: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

10

14

14 12

12

13

14

13

• Candidate sensors whose sensing range cover the point (x, y) participate in the task with probability

• Obviously, the formula shown above gives sensors with higher energy higher chance to participate in the task than other sensors.

• Immediately, after the last CTW message, time line is divided into s time epochs (slots). Sensors willing to participate should choose a random slot and transmit a short message that contains its current energy level.

( For the example assume S = 8)

LastCTW 13, 14 14 10,12 13 14 12

Workforce Selection

Page 38: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

LastCTW 13, 14 14 10,12 13 14 12

• Immediately, after the last bidding slot, the task coordinator (the backbone sensor within the hexagon) transmits a message that contains the results of the last bidding round.

LastBidding Slot 0 0 1 0 2 0 3 4 Info

.

• While evaluating the bidding process, only single-bidder slots are considered, other slots are ignored. A temporary id is assigned to sensors that bid in any of these slots, this id determines the order by which sensors send their sensory data to the task coordinator during data aggregation.

• The info field in the bidding result message should include information needed for next bidding round (if any):

• S: number of slots ,• W: number or remaining sensors needed,• Emax: maximum energy among sensors.

Page 39: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

• The bidding process continues till the whole workforce is recruited or a maximum number of bidding rounds is reached.

• Task execution starts immediately after collecting required workforce. When task execution is completed, sensors start to send their sensory data to the

backbone sensor in the order assigned during workforce selection.

• Finally, the backbone sensor within the hexagon sends the locally aggregated sensory data to the sink node along with Emax, the maximum energy among sensors within this hexagon, hence the sink can attach this value with CTW messages transmitted for any upcoming tasks within this hexagon.

Page 40: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

• We provided mathematical derivation of all parameters needed for our workforce selection protocol, including

Number of slots in any bidding round

Expected number of candidate sensors collected by CTW messages

Awake probability of sensors.

Number of CTW messages needed

Page 41: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Workforce SelectionDistributed Approach

• Estimation of Maximum Energy• Workforce Selection

Page 42: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

• Assume that sensor energy can be quantized into 2n

levels, (i.e it can be encoded in a string of n bits).

Estimation of Maximum Energy

For illustration, we assume, n = 4

E1

Last CTW

E2 E3 E4

clock drift margin

Time line is divided into 4 time epochs

• Immediately after the last CTW message, time line is divided into n time epochs.

Page 43: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

• Sensors in the same sensing area transmit short messages that represent their energy levels bit by bit starting from the most significant bit.

• A value of 0 is not transmitted, while a value of 1 is transmitted.

• Sensors stop transmitting if they have a 0 and received a packet or detected a collision in the corresponding epoch.

Estimation of Maximum Energy (cont.)

For illustration, assume we have 3 sensors with the following energy levels, S1 [1110], S2 [1100], S3 [1011]

clock drift margin

E1

Last CTW

S2

S1 S1

S2

S1

S3

E2 E3 E4

Page 44: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

• Sensors that pick up the values transmitted should use the following disambiguation scheme: No packets are received: 0 is recorded. A packet is received or collision is detected: 1 is recorded.

Estimation of Maximum Energy (cont.)

S1 [1110], S2 [1100], S3 [1011]clock drift margin

E1

Last CTW

S2

S1 S1

S2

S1

S3

E2 E3 E4

Maximum: 1 1 1 0

Using appropriately long epochs, synchronization should not be a problem.

Page 45: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

1 10 0

1 01 0

1 110

0 11 1

# Channel Status Bit

4 <c> 1

3 Ambient 0

2 Collision <b,c> 1

1 Collision <a,b,c> 1

Maximum: 1101Estimation of Maximum Energy

Page 46: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Workforce Selection

• The sensing area is divided into k regions with equal size • Workforce selection goes in decision rounds, each round is

associated with an energy level (Emax, Emax-1, Emax-2,…etc).

• Each decision round has k slots corresponding to the regions determined above.

k

R2.

Max Estimation

Decision Round Emax

Decision Round Emax-2

Decision Round Emax-1

……

clock drift margin

S1S2 Sk

RoundE+1

RoundE-1

Page 47: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

• Based on task and sensor position, each sensor estimates its region and transmits a short message in the decision round/slot that matches its energy and region.

• Protocol terminates when required workforce is recruited.• Sensors keep track of number of recruited sensors as follows:

Collision is detected: increment by 2. No packets are received: 0 is recorded. One packet is received: increment by one.

S1

RoundE+1 S3

S2

S1

S2 S3 S4

RoundE-1

+2 +0 +1

S6

S5S4

+2

Workforce Selection

Page 48: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

# Sensors Wnew Wtotal

4 Ambient 0 3

3 c<S4 , S5> 2 3

2 Ambient 0 1

1 S2 1 1

Round-1 (E = 10)

Assume required workforce is 4

Page 49: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

# Sensors wnew Wtotal

1 S1 1 4

Round-2 (E = 9)

Workforce collected,

Protocol terminates

Page 50: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks
Page 51: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks
Page 52: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks
Page 53: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Sensor Sleep Schedule

Due to their limited energy budget, sensors spend most of their lifetime in a sleep mode.

Various deterministic and probabilistic schemes can be used to determine the schedule based on which sensors sleep and wake up.

Each scheduling scheme has its impact on :

The effective sensor density (ESD), defined as the density of awake sensors.

The coverage capability of the network.

Page 54: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Scheduling Schemes

Static Schemes

Dynamic Schemes

Energy-Aware Schemes

Sensor sleep for a random amount of time uniformly selected in the interval [Ts, TS], and stay awake for another random amount of time uniformly selected in the interval [Ta, TA].

Dynamic scheduling schemes which adaptively change the upper bounds TA and TS based on sensor remaining energy.

In general, we can classify scheduling schemes into,

Sensors sleep for Ts time units and stay awake for Ta time units.

Page 55: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

PASTA & Sleep Scheduling in WSN The PASTA property stands for Poisson Arrivals See Time Average.

We show that the PASTA property can be applied to effective sensor density in WSNs.

Specifically, we proved that fraction of events that occur in a certain area while k sensors are awake equals the fraction of time this area is under the surveillance of k sensors.

This result allowed us to use the time-invariant probability distribution of k-coverage to analyze different scheduling schemes.

PASTA has been used for too long in queuing systems and in general it implies the equivalence between the time average view seen by an internal observer who has been watching the system for long time and the view seen by external observers that arrive at the system according to a Poisson process.

Page 56: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Energy-Aware Scheduling We proposed a backbone-guided energy-aware scheduling scheme

designed to extend network reliable lifetime by balancing energy consumption among sensor nodes.

The main idea of our scheme is to continuously and probabilistically adjust sleep and awake times of sensors based on the differences in their remaining energy.

In other words, the proposed scheme prolongs the sleeping periods of sensors with relatively low energy and compensate for their absence by shortening sleeping periods of sensors with relatively high energy.

This goal has to be done without probabilistically affecting the effective sensor density.

Page 57: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Energy-Aware Scheduling We proved that if sensors independently updated their upper bounds for

TA & TS after the execution of each task according to the following equations, their energy consumption will almost be even.

The figure below show the Energy-Aware scheme in action

S9

S8

S7

S6

S5

S4

S3

S2

S1

Sens

ors

Sleep / Awake Cycle

Page 58: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks
Page 59: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks
Page 60: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Sink

Local Aggregation

Backbone Sensor 

Regular Sensor

Inter-backbone sensor aggregation through reversed selection rules

Selection rules

Data Aggregation & Routing

Page 61: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Energy Hole Problem

Energy Hole

Page 62: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Rendezvous-based Routing

Page 63: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Source 2

Δ c

Source 1

Destination 2

Δ r

Δ cΔ r

Destination 1

Δ𝑐=𝑐𝑑𝑠𝑡−𝑐𝑠𝑟𝑐

+ Δc anti-clockwise

- Δc clockwise

𝑐𝑠𝑟𝑐=(𝑠𝑠𝑟𝑐−1 )𝑟𝑚𝑖𝑛+𝑐𝑠 𝑟𝑐

𝑐𝑑𝑠𝑡=(𝑠𝑑𝑠𝑡−1 )𝑟𝑚𝑖𝑛+𝑐𝑑 𝑠𝑡

)

If (

Δ𝑐=6𝑟𝑚𝑖𝑛−Δ𝑐else If (

Δ𝑐=Δ𝑐−6𝑟𝑚𝑖𝑛

Hexagon to Hexagon Routing

Δ 𝑟=𝑟𝑑𝑠𝑡−𝑟𝑠𝑟𝑐 + Δr outward direction

- Δr inward direction

, ) >> , )

Page 64: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Source 2

Δ c

Source 1

Destination 2

Δ r

Δ cΔ r

Destination 1• Once Δr & Δc are known, each sensor can adaptively select its next hop as follows:

If (|Δr| > 0) & (|Δc| > 0) , whenever possible, select the hop which decrements both |Δc| and |Δr|. Otherwise,

If (routing outward i.e Δr > 0) select the next hop which decrements |Δc|

If (routing inward i.e Δr < 0) select the next hop which decrements |Δr|

Hexagon to Hexagon Routing , ) >> , )

Page 65: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Conclusions

In this work, we explored the construction and the advantages of having an infrastructure for WSNs.

The proposed backbone proved to be useful in mitigating many of the typical challenges inherent to WSN including:

Sensor Localization.

Data Aggregation & Routing

Energy-Aware Workforce Selection

The construction protocol initially tiles the deployment area around sink nodes using identical hexagons.

After that backbone sensors are selected to be the closest sensors to the centers of the hexagons they represent.

Clustering

Energy Hole Problem

Energy Aware Sleep Schedule

Page 66: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Future Research Directions

Despite the encouraging results, many important challenges and research questionsremained unanswered.

In the construction protocol, it would be useful to reduce the amount of inaccuracy due to RSS. One way around this difficulty would be to: Estimate the initial distance through several readings.

Continuously update distance estimates form different messages received throughout the network lifetime.

Given the limited on-board energy budget available to sensors, it would be of interest to see how far can one streamline the computational requirements of the construction protocol.

If backbone sensors in different columns can be selected in parallel, can we also select backbone sensors in different rows in parallel ?

Page 67: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Let the sink node estimate the positions of all the hexagon centers and broadcast them to sensors which can work in parallel to estimate the closest sensor to the center of each hexagon.

Future Research Directions (cont.)

Any Problems?

Sensor Synchronization is one of the most challenging problems in sensor networks. It would be of an interest to show, if possible, how our proposed backbone can be useful in simplifying the synchronization problem.

Network Security is something that we completely overlooked in our presented work. Revisiting the proposed protocols from a security point of view would definitely open new dimension of research challenges.

Finding more applications of the PASTA property in in sensor networks analysis.

Page 68: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Publications

Journals:

Conferences/Workshops:

Page 69: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Conferences/Workshops (cont.)

Page 70: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Conferences/Workshops (cont.)

Page 71: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks

Questions ?

Thank You !