Networking Research at UCI

1
Networking Research at UCI

• Tatsuya Suda
• Professor
• Information and Computer Science
• University of California, Irvine
• (email) suda_at_ics.uci.edu
• (web) netresearch.ics.uci.edu

2
Introduction
3
Who am I?

• Professor, with ICS/UCI since 84
• Program Director, Net. Research, NSF, 96 - 99
• Example Professional Activity
• IEEE Fellow
• Editorial Board, Encyclopedia of Electrical and
  Electronics Engineering, Wiley and Sons
  
• Editor, IEEE/ACM Trans. on Networking
• IEEE Tech. Committee on Comp. Commun., Chair,
  96-98

4
Network Research at Netgroup

• Goal to provide user-to-user flexible network
  services
  
• Two Research Foci
• High Speed Networks
• Middleware and Object Oriented Network
  Frameworks
  
• Two Research Approaches
• Theoretical Approach
• mathematical modeling and analysis
• Empirical Approach
• software design/implementation, measurements

5
ResearchApproaches
mathematical analysis
6
Research Approaches

• Empirical Approach
• Example ACE
• designed and implemented ACE (an object-oriented
  framework) with D. Schmidt
  
• adopted by
• Lucent, Bellcore, Cisco, DEC, Bell South Cellular
  Corp, Ericsson Radio, Motorola, Kodak,
  Boeing/McDonnell Douglas, and many more
  
• used by AudioActive at the 97 Grammy awards

7
Example Projects

• Networks
• Ad Hoc Networks
• Disaster Recovery Networks
• Real-time Video Multicast
• Internet Traffic Control
• Network Measurements
• QoS Mechanisms
• Object Oriented Network Frameworks
• Bio-Networking Architectures
• Peer to Peer Discovery Mechanisms

8
Researchers

• Network Group
• 5 ph.d students
• 3 visitors
• 15 or so under grad students
• Middleware/OO group
• 3 ph.d students
• 2 post doc researchers
• 5 under grad students

9
Funding Sources

• NSF
• DARPA
• ARO
• CA State
• Private industry

10
Bandwidth Efficient Multicast Routing Protocol
for Ad Hoc Networks
11
Introduction

• Ad Hoc Networks
• Highly dynamic
• With multihop wireless connections
• Limited resources (bandwidth, power)
• For ad hoc network applications
• Increasing needs for multicast services
• To reduce the resource consumption, bandwidth
  efficient multicast protocols is needed with
  
• Low communication overhead
• High multicast efficiency

12
Source
Receiver2
Receiver1
13
Proposed Bandwidth Efficient Multicast Protocol

• Route set up
• Route recovery
• Broadcast-multicast
• Local re-join
• Route optimization

14
Route Setup
Multicast Group
Source
Receiver
X
Try to find the nearest forwarding node
15
Route Setup
Multicast Group
Source
Receiver
X
Try to find the nearest forwarding node
16
Route Setup
Multicast Group
Source
Receiver
X
Try to find the nearest forwarding node
17
Route RecoveryBroadcast-Multicast
A
B
C
D
18
Route RecoveryBroadcast-Multicast
A
B
C
D
19
Route RecoveryBroadcast-Multicast
A
B
C
D
20
Route RecoveryBroadcast-Multicast
A
B
C
D
21
Route Recovery Local Rejoin
F
E
A
B
C
D
22
Route Recovery Local Rejoin
F
E
A
B
C
D
23
Route Recovery Local Rejoin
F
E
A
B
C
D
24
Route Recovery Local Rejoin
F
E
A
B
C
D
25
Route Recovery Local Rejoin
F
E
A
B
C
D
26
Route Optimization
A
B
C
D
E
27
Simulation Parameters

• Variables
• Other parameters
• Simulation Area 1Km x 1Km
• Channel Speed 2Mbps
• Transmission Range 200m
• Multicast Source 128 Kbps, CBR

28
Multicast Efficiency
High multicast efficiency at large group size
29
Communication Overhead
Lower communication overhead than other protocols
30
Conclusions

• A bandwidth-efficient multicast routing
  protocol for ad-hoc wireless networks was
  proposed.
  
• The proposed protocol can achieve high multicast
  efficiency with low communication overhead.

31
Peer to Peer Discovery
32
Peer to Peer Discovery

• Need for finding certain types of Objects
• information that soldiers collect in a combat
  situation
  
• information collected by fire fighters at the
  ground zero
  
• Under dynamic network changes
• Objects may move (soldiers move) (fire fighters
  move)
  
• Objects or links may become unavailable
• Military applications
• Crisis management applications

33
Outline

• Proposed Discovery Scheme
• Community
• Keyword strength
• Query forwarding
• Query Hit
• Simulation Results

34
Community Concept in Our Discovery Scheme

• Community
• Knowledge about network research
• Students ask me details of current projects

current project E
C
current project E
Net Research
A
D
D
35

• Community
• Robust to dynamic network changes

current project E
current project E
Net Research
A
C
D
D
36
Community Creation Using Reward

• Community Creation

E
C
B
A
D
D
37

• Community Creation

E
C
Query for E
B
A
D
D
38

• Community Creation

E
C
Query hit
Query for E
B
A
D
D
39

• Community Creation

Add E
E
Reward
C
Query hit
Query for E
B
E
A
D
D
40

• Community Creation

Add E
Add E
E
Reward
E
C
Query hit
Query for E
B
E
A
D
D
41

• Community Creation

Add E
Add E
Add E
E
Reward
E
C
Query hit
Query for E
B
E
A
D
D
42

• Community Creation

Add E
Add E
Add E
Add E
E
E
Reward
E
C
Query hit
Query for E
B
E
A
D
D
43
Keyword Strength Concept

• Keyword Strength (usefulness)

E
E
Reward
E
C
B
E
A
D
D
44

• Keyword Strength (usefulness)

Increase strength for E
Increase strength for E
E
E
Increase strength for E
Reward
E
C
B
E
A
D
D
45

• Keyword Strength (usefulness)

Increase strength for E
Increase strength for E
E
E
Increase strength for E
Reward
E
C
B
E
Increase strength for E
A
D
D
46
Query Forwarding

• Query forwarding

E
E
E
C
B
E
A
D
D
47

• Query forwarding

E
E
E
C
B
E
A
D
E
D
E
E
48

• Query forwarding
• Probabilistic, proportional to keyword strength

E
E
E
C
B
E
A
D
E
D
E
E
49

• Query forwarding
• Probabilistic, proportional to keyword strength

E
E
E
C
Query for E
B
E
A
D
E
D
E
E
50

• Query forwarding
• Probabilistic, proportional to keyword strength

Forward with higher probability
E
E
E
C
Query for E
B
E
A
D
E
D
E
E
51

• Query forwarding
• Probabilistic, proportional to keyword strength

E
E
E
C
Query for E
B
E
A
D
E
D
E
Forward with smaller probability
E
52

• Query forwarding
• Probabilistic, proportional to keyword strength
• Robust to dynamic network changes

E
E
E
C
Query for E
B
E
A
D
E
D
E
Forward with smaller probability
E
53

• Adjusting Keyword Strength

Forward a query
E
E
E
C
Query for E
B
E
A
D
E
D
E
E
54

• Adjusting Keyword Strength

Decrease strength
Forward a query
E
E
E
C
Query for E
B
E
A
D
E
D
E
E
55

• Adjusting Keyword Strength
• If a successful hit returns, increase keyword
  strength when reward is returned

E
E
E
C
B
E
A
D
E
D
E
E
56
Returning a Query Hit

• Who returns a query hit?

E
E

C
B
E
A

D
E
D
E
E
57
Returning a Query Hit

• Who returns a query hit

with higher probability
E
E

C
B
E
A
D
E
D
E
E
58
Returning a Query Hit

• Who returns a query hit

with higher probability
E
E

C
B
E
A
with lower probability
D
E
D
E
E
59
Outline

• Proposed Discovery Scheme
• Community
• Keyword strength
• Query Forwarding
• Query Hit
• Simulation Results

60

Community and User Satisfaction
Community Creation (Keyword strength)
of Queries (Time)
Degree of User Satisfaction (Proposed Method)
of Queries (Time)
61
User Satisfaction
User Satisfaction (Proposed Method)
Proposed Method
? 0.842
of Queries
User Satisfaction (Existing Method)
Existing method
? 0.554
of Queries
62
Homeland SecurityDisaster Recovery

• Netgroup, ICS, UCI
• (email) suda_at_ics.uci.edu
• (web) netresearch.ics.uci.edu

63
Homeland Security Conception

• The preparation for, prevention of, deterrence
  of, preemption of, defense against, and response
  to threats and aggressions directed towards US
  territory, sovereignty, domestic population, and
  infrastructure as well as crisis management,
  consequence management and other domestic civil
  support. Also called HLS.
• by Randy Larsen, Dave McIntyre, and Mark DeMier,
  ANSER Institute for Homeland Security (a
  nonprofit public-service research organization)
  

64
HLS Issues

• Homeland defense of the United States
• The protection of US territory, sovereignty,
  domestic population, and infrastructure against
  external threats and aggression. Also called
  HLD.
• Consequence management
• Dealing with the effects of attacks by weapons of
  mass destruction, including military support to
  civilian authority
  
• Crisis Management
• Measures to identify, acquire, and plan the use
  of resources needed to anticipate, prevent,
  and/or resolve a threat or act of terrorism.
  
• Immigration control and border security
• Computer network defense (network security)
• Disaster recovery
• As a issue related to HLS, disaster recovery
  should focus on disaster caused by terrorist
  attack
  
• But research in this field may also apply to
  natural disaster
  
• ETC.

65
Disaster Characteristics

• Disaster site characteristics
• Time is critical
• Full of danger
• Former information about the disaster site is not
  dependable

66
Disaster Characteristics

• Network related characteristics
• A large number of heterogeneous devices/users
  exist
  
• Sensors, personal communication devices (PDA,
  hand phone), devices brought in by rescue teams
  
• Mobile and non-mobile devices
• Intelligent and non-intelligent devices
• Power constraint/limited resources of sensors and
  personal communication devices
  
• Rescue team members, law enforcement officers,
  survivors
  
• Depend on extent of disaster
• Wireless network environment with or without
  fixed/wired gateways
  
• Wireless networks within a disaster area
• Wired network in ERD (Emergency Rescue Director)
  centers
  
• Communication links may not be reliable,
  available all the time

67
Example Scenarios

• A high rise building crashed from a terrorist
  attack
  
• The building was embedded with micro-sensors
• Micro-sensors monitor temperature, air pressure,
  etc.
  
• Some micro-sensors are still active
• People are trapped inside
• Conditions of a disaster site are unknown
• How many people died/are wounded/survived
• Degree of damages of the buildings and their
  facilities, such as power supply, gas lines,
  etc.
  
• Degree of dangers, such as electrical, fire,
  poison gas, etc.
  
• ERCs (Emergency Rescue Crew) form a rescue team
  and go into the site
  
• to find survivors and to rescue people trapped

68

• A rescue team goes into the 1st floor of the
  damaged building.
  
• Rescue team members start monitoring the
  environment using the devices they bring in and
  also locate active micro-sensors in the disaster
  site by sending query beacons.
• Some parts of the 1st floor are blocked by
  collapsed walls. Rescue team members dont know
  if there are survivors or potential danger on the
  1st floor.

69

• Rescue team sends out mobile robot sensors. The
  mobile robot sensors detect some active
  micro-sensors on the 1st floor, and it appears
  that the micro-sensor nets on the 1st floor are
  segmented.
• The mobile robot sensors act as relaying nodes to
  forward information between the rescue team and
  the segmented micro-sensor nets on the 1st
  floor.
• By examining the data collected, the rescue team
  finds that there are some people trapped on the
  1st floor and proceeds with a rescue plan.

70

• Rescue team has finished the 1st floor and tries
  to go to the 2nd floor. However, fire and smoke
  are blocking the stairways.
  
• Rescue team sends mobile robot sensors to the 2nd
  floor to assess the 2nd floor conditions.
  
• Mobile robot sensors autonomously move around,
  monitor the 2nd floor and reestablish
  connectivity among isolated islands of
  micro-sensor nets on the 2nd floor.

71

• Rescue team starts receiving sensor data from the
  2nd floor.
  
• Information collected may include survivor
  locations, fire distribution, etc.
  
• Rescue team finds that the temperature on the 2nd
  floor is very high, indicating possible explosion
  and also small probability of survivors.
  
• Rescue team also finds that active micro-sensors
  on the 3rd floor are sending information
  indicating possible survivors.

72

• Through analyzing collected sensor data, rescue
  team now knows fire and smoke distribution and
  finds a possible passage to the 2nd and 3rd
  floors (i.e., through a 2nd floor lounge by the
  stairs to 3rd floor).
• The rescue team then drill a hole on the lounge
  floor, goes up to the 2nd floor, and then, go to
  the 3rd floor by stairs.
  
• A number of survivors are gathering in the lounge
  on the 3rd floor (some are from the 2nd floor).

73

• Since micro-sensors power is limited, a
  micro-sensor net uses power-saving routing to
  maximize the life of the net, namely,
  micro-sensors in critical areas act as relay less
  often (to save power) than those in non-critical
  areas.

74
Network Requirements

• Need to maintain network coverage (R1)
• Need to extend micro-sensor network life (R2)
• Need to locate active micro-sensors and isolated
  micro-sensor network islands (R3)
  
• Need self-healing capability in micro-sensor
  networks (R4)
  
• A micro-sensor network re-arrange and re-organize
  micro-sensors to maintain communication in
  response to network disconnection.

75
General Requirements

• Data query and process (R5)
• How to find the desired information
• How to process the data in the sensor network
• Personal communication (R6)
• How survivors communicate to each other or send
  emergency signals in disaster situation
  
• Power saving
• Service provision
• Ad-hoc mode communication (R7)
• A good network architecture (R8)
• A large number of heterogeneous devices existing
• Intelligent software entities (R9)

76
Research Issues in Networks

• Power-saving sensor network researches (R1, R2)
• Power-saving
• Information gathering
• Direction/location based sensor network
  communication research (R3)
  
• Relative coordination establishment
• Relative direction acquirement
• Self-deploy and self-reorganization sensor
  network research (R4)
  
• Mobile sensors/robots
• Network coverage situation
• Mobile sensors/robots movement algorithm

77
General Research Issues

• Data query and process research (R5)
• Discovery
• Data aggregation
• Personal communication research (R6)
• Power-saving
• Service composition
• Ad-hoc network researches (R7)
• Routing technologies
• Power-saving
• Network architecture researches (R8)
• Intelligent software researches (R9)
• Middleware
• Soft agent

78
Existing Research

• Existing network technologies
• Wired network
• IP network
• Fixed topology
• Vulnerable, weak self-healing capability when
  devices damaged
  
• Wireless network
• GSM, CDMA
• Signal may be blocked because due to obstacles
  (e.g., collapsed walls/buildings)
  
• If a base station is damaged, wireless devices
  cannot communicate
  
• Not for supporting sensor devices, and cannot
  monitor environmental information (temperature,
  chemical poison, etc.)

79

• Ad-hoc network
• In disaster scenario, it can only be used for
  local, open area communication because
  communication range for ad-hoc network is
  relative short.
• Not every survivor carries an ad-hoc
  communication device
  
• GPS (global Position system)
• Useful in open areas, but cannot be used inside a
  building

80
Key Technologies Used in Our Researches

• Some key technologies that we assume in our
  research
  
• Progress in MEMS (Micro Electronics Mechanics
  System)
  
• Smaller, high capacity, cheap sensors
• Mobile sensors
• UWB (Ultra Wide Band) wireless technology
• Wide bandwidth10 to 1000 Mbps
• Low power consumption communications
  
• Short range wireless technology (e.g. Bluetooth)

81

• Directional antennas
• Global Positioning System (GPS)
• For open area rescue

82
Research Goal

• Our research goal is
• To investigate mechanisms to leverage these
  technologies for effectively achieving our
  general goals in disaster recovery (described
  above).

83
Current Research in Netgroup

• Power-aware sensor networks
• To achieve longer network life by power saving
  routing
  
• Self-deploying and self-organizing networks
• To maintain network connectivity

84
Network Connectivity Recovery

• Netgroup
• (email) suda_at_ics.uci.edu
• (web) netresearch.ics.uci.edu

85
Outline

• Problem Definition of Network Connectivity
  Recovery
  
• Scenario Description
• Possible Solutions and their Trade-Offs
• Use sensor type devices
• Use robot type devices
• Issues and modeling of using sensor type
  devices
  
• Issues and possible schemes of using robot type
  devices

86
Problem Definition

• Problem Definition
• Regain connectivity between disconnected
  sub-networks
  
• Regain connectivity to an existing
  (not-destroyed) network

87
Possible Scenarios

• Disconnected sub-networks
• In disaster scenarios, a network may be
  disconnected, resulting in multiple, disconnected
  islands of small sub-networks, because of
  
• Disconnection of physical lines
• Host becoming unavailable (broken, no power
  supply etc)
  
• Obstacles against wireless links

Disconnected Network
88

• Backbone network
• In disaster scenarios,
• emergency rescue teams may establish a backbone
  network
  
• some backbone network infrastructure may not be
  severely damaged and may still be available
  
• Backbone network
• may be able to broadcast (uni-directional) to
  simple devices and disconnected sub-networks
  
• may also be able to control movement of robot
  type devices (by, for instance, using a strong
  transmission power).

Disconnected Network
Backbone Network
89
Possible Solutions

• Re-establish connectivity by deploying wireless
  devices
  
• Very small, simple sensor type devices
• Very limited processing, memory capability
• Very limited power (transmission range)
• Intelligent mobile robot type devices
• Some processing, memory capability
• Some power (transmission range)

90
An Example of Small Sensors

• SMART DUST project at UCB http//robotics.eecs.ber
  keley.edu/pister/SmartDust/

91
Examples of Mobile Robots

• Weight 12.5 g
• Wing span 9 inch
• Flapping amp. 65 deg
• Flapping freq. 20 Hz
• Flight velocity 4 m/s
• Power required 2 W
• Power source Battery
• Propulsion Flapping Wings

92
Trade-Offs

• Very small, simple sensor type devices
• Devices are very small and simple
• inexpensive
• Devices can be deployed in non-safe areas where
  humans cannot go
  
• Need to deploy massive number of small devices
  into the target area
  
• E.g., scatter devices from an aircraft
• Because
• devices may not be able to move and and
  strategically place themselves to restore
  connectivity, and
  
• devices may have limited power (i.e., short
  transmission rage)
  
• In some situations, it is difficult to deploy
  large number of devices
  
• E.g. underground, in the corrupted building

93

• Intelligent mobile robot type devices
• Devices are complex and require some intelligence
  
  
• expensive
• fragile against severe conditions (e.g., high
  temperature, under water)
  
• Robots can explore the area where humans or
  static devices cannot reach
  
• No massive deployment
• of robots being deployed is likely to be small
• ? Self deployment
• Deploy several robots into the area.
• Robots autonomously explore the area,
  strategically place themselves and restore
  connectivity

94
Possible Approach(3) Let Robots to deploy devices

• Scenario
• Deploy several robots into the area. Each robot,
  while exploring the area, leaves small wireless
  devices, which establish links between
  disconnected sub-networks
• Benefits
• Not many robots are necessary
• Robots can explore the area where human or static
  device cannot reach
  
• Robots behavior can be specialized for network
  recovery
  
• Problems
• Robots are complex and expensive

95
Sensor Type DevicesResearch Issues

• Understanding distribution and density of
  deployed sensor devices and degree of
  connectivity regained
  
• Varying density and distribusion
• Uniform deployment, biased deployment
• Identifying and locating disconnected
  sub-networks
  
• How to identify and locate the disconnected
  sub-networks?
  
• Mathematical modeling

96
Sensor Type DevicesMathematical Modeling

• Boolean model can model this problem efficiently
• Central points are distributed according to the
  generalized Poisson point process.

97
Robot Type DevicesResearch Issues

• How many robot devices to deploy
• When and where a robot device should move
• Each robot autonomously determines, based on
  local information, when and where to move to
  explore the area
  
• Robots may be controlled by human, but robot may
  also autonomously behave
  
• Factors impacting robot behavior
• Topology information
• Location of isolated networks, backbone network
• Location of other robots
• Signal interference, transmission power control

98
Robot Type DevicesPossible Schemes

• Two possible schemes
• Expand the size (network coverage area) of a
  single sub-network until all the sub-networks are
  connected
  
• Search and locate disconnected sub-networks, then
  establish a link between them

Sub-net
Sub-net
99
Scheme 1 Expand a sub-net

• Robots start exploring the area from a single
  sub-network (source sub-network)
  
• E.g. start from Backbone Network
• Each robot seeks disconnected sub-networks while
  keeping connectivity to the source network
  
• When a disconnected sub-network is found, (which
  means that a new sub-network is connected to the
  source sub-network,) some robots remain at the
  current position to maintain connectivity.
• Then some robots explore further from the new
  sub-network.
  

!!
100

• To efficiently expand a sub-network without
  losing connectivity and to maximize the network
  coverage area, the concept of force may be
  used.
• Two types of forces between robots (repel force
  and recall force)
  
• Each robots makes its decision based on force.

101

• Repel force
• Is used to maximize coverage area of robots.
• Is a tuple
• function of the relative position between two
  robots.
  
• Defined for each neighboring robots.
• Power is inversely proportional to the distance
  between two robots
  
• Direction is the direction of the other robot
• Is set to NULL when a signal from a neighboring
  robot is discontinued.
  
• Two thresholds upper_threshold and
  lower_threshold
  
• When lower_thresholdremains static.
  
• When lower_threshold power, robot moves toward
  the other robot
  
• When upper_thresholdthe other robot

102

• To find the relative positions of robots, the
  following may be used
  
• Power measurement (to determine relative distance
  )
  
• Packet loss/error rate (to determine relative
  distance )
  
• Directional antenna (to determine relative
  direction )

103

• Example 1
• Robot 1 and Robot 2 are very close to each other.
  
  
• Robot 1 computes repel force R based on the
  direction of and distance to Robot 2.
  
• If R upper_threshold, then Robot 1 moves away
  from Robot 2.
  

-Repel.direction
Repel.powerupper_threshold
Repel.powerlower_t
hreshold
104

• Example 2
• Robot 1 and Robot 2 are far from each other.
• Robot 1 computes repel force R based on the
  direction of and distance to Robot 2.
  
• If RRobot 2.

Repel.direction
Repel.powerRepel.powerlower t
hreshold
105

• Recall force
• Is used to allow robot to self-reorganize when
  some robots become unavailable.
  
• Is a tuple
• function of the repel force
• If lower_threshold
  force power, and recall force direction repel
  force direction.
• If repel.power upper_threshold
  (power, direction) is set to NULL.
  
• If repel force NULL, but recall force ! NULL,
  Robot moves in recall force direction.

106

• Example 1
• Robot 1 and robot 2 both satisfy lower_threshold
  
  1 and robot 2 remain static (namely, it does not
  move). Robot 1 and robot 2 jointly connect the
  two disconnected sub-networks.
• Since the condition lower_threshold power
  recall force to its own repel force. Robot 2
  also sets its recall force to its own repel
  force.
• After some time, robot 2 failed, and two
  sub-networks are disconnected again. At this
  time, robot 1 stops receiving signal from robot
  2. Thus, the robot 1 sets its repel force (power,
  direction) to NULL. Recall force (power,
  direction) of robot 1 remains unchanged.

107

• Since the condition repel force NULL, but
  recall force ! NULL is met, robot 1 moves in
  the direction of recall force direction,
  connecting the two sub-networks again.

108
lower threshold robot 2 1 repel force power to sub-net 1 threshold set robot 1 recall force to robot 2 re
pel force to robot 2 set robot 1 recall force
to sub-net1repel force to sub-net1
Robot 1 repel force to robot 2NULL
lower threshold sub-net 1 to robot 2!NULL
lower threshold robot 2 1 repel force power to sub-net 1 threshold set robot 1 recall force to robot 2 re
pel force to robot 2 set robot 1 recall force
to sub-net1repel force to sub-net1
109
this slide will be updated later.
lower threshold robot 2 1 repel force power to sub-net threshold Let Recall1Repel1 Recall2Repe
l2
Repel1NULL lower thresholdr threshold
Recall1Last Repel1!NULL
Recall1.direction
lower thresholdlower thresholdLet Recall1Repel1 Recall2Repel2
110
Scheme 2 Search a Subnet and Establish a Link

• Robots start exploring the area from a single
  sub-network (source sub-network)
  
• E.g. start from Backbone Network
• Each robot explores the area without keeping
  connectivity to the source sub-network
  
• When a robot finds a disconnected sub-network, it
  comes back to the source sub-network and brings
  other robots to establish a link between two
  networks

!!
111
Weilin Christina, please reviese Location
Problem

• Weilin, in your scheme, we can obtain relative
  location. But, nodes in your scheme does not
  adjust their location based on the information.
  Here in this context, nodes (robots) adjust their
  location.
• Relative Coordinate System (refer to Weilings
  slides)
  
• Robot movement algorithm
• According to the relative position information,
  each mobile sensor node determines when and where
  to move.
  
• Describe possible algorithm

112
Robot Movement Back and Forth Mode

• IdeaRouting with mobile devices
• Robots can migrate between sub-networks to store
  and deliver packets
  
• New routing algorithm that also controls device
  movement can be designed
  
• Back and Forth Mode
• If there are not enough robots to establish a
  link, some robots move back and forth between two
  disconnected subnets to deliver packets.
  

113
Power Consideration in Micro-sensor Networks
114
Micro-sensor Networks

• Network of small/simple sensors
• Micro-sensors transmit sensor data to a data
  collection node either hop by hop or directly
  
• Assumptions
• Micro-sensors are power- and computation-capabilit
  y- limited
  
• Micro-sensors have adjustable signal power (i.e.,
  transmission range)
  
• Micro-sensors know the distance to other sensors
  through examining power of received signals
  
• Micro-sensors are distributed randomly in an
  area
  
• Micro-sensor sensing accuracy decreases with
  distance

115
Existing Research

• Power saving routing
• To route sensor data to a destination by
  consuming minimal power
  
• By choosing a closest next hop node to minimize
  the power for data transmission
  
• MTE (Minimum transmission energy) protocol,
  Timothy Jason Shepard, MIT

116

• Energy distribution
• To uniformly distribute power consumption over
  the network sensors
  
• Example solutions
• LEACH (Low-Energy Adaptive Clustering Hierarchy),
  Wendi R.H., Anantha C., Hari B., MIT
  
• Sensors are organized into clusters. Head of the
  cluster will be represented by members in turn.
  Only cluster head will participate data relaying

117

• PEGASIS (Power-efficient Gathering in Sensor
  Information System), Stephanie Lindsey, Cauligi
  S. R., The Areospace Corporation
  
• A optimization of LEACH, it uses a greed
  algorithm to form cluster by assuming each node
  have a global view of the network. Each node
  communicate only to a close neighbor and take
  turn to send data to data collection node.
• Energy Aware Routing, Rahaul C. Shah and Jan M.
  Rabaey, UC Berkley
  
• Multi paths are maintained between source and
  destination. The probability of a route being
  chosen depend on the energy metri of each route.
  Thus make the energy distribute more evenly among
  microsensors.

118
Goal of Our Project

• To maximize the life of a micro-sensor network
• A micro-sensor network life is from the time
  between its deployment to the time that it fails
  to cover the entire area.
  
• A micro-sensor net loses coverage because some
  sensors become out of power.

119
Longer Network Life
A sensor network composed of five sensors.
Red nodes sensing area is covered by the other
four black nodes. The black nodes can monitor the
area with enough accuracy. Option 1 all nodes se
nd data to the yellow collection node
Better option Each black sensor forwards data to
red node, and red node forwards data to the
yellow node. Black nodes can save power, and
still the green area is monitored with enough
accuracy.
Data collection node
120
A Difficulty

• Maximizing a network life is not same as
• choosing the closest next hop node in routing to
  minimize the power for data transmission

121
A microsensor network with choose the closest
next hop node routing (e.g., MTE)
data collection node
micro-sensors
122
data forwarding route based on MTE.
Green node forwards more traffic than other
nodes, and thus, its power decreases more quickly.
123
---- the node out of power.
Although there are a number of nodes still
active, network does not function as a network.
It fails to cover the entire area.
124
---- the node out of power.
Although there are still enough alive nodes, the
network should be considered dead since it cant
monitor the area circled by brown dotted line by
enough accuracy.
125
Difficulty 2

• Maximizing a network life is not same as
• Uniformly distributing power consumption over the
  network sensors

126
A microsensor network with distributing power
uniformly over the network policy (I.e. LEACH)
127
LEACH scenario ---- Current cluster head N
etwork are organized into clusters. The nodes in
the brown dotted circle is an example of cluster.

128
Each node in the cluster will forward the data to
head, and the head relays the data to the data
collection node. The algorithm achieves relative
fair energy consumption within a cluster.
129
If, for example, the coverage area of the green
node is covered by other nodes in the cluster, we
can consume green nodes power first to make the
network life longer.
130
Our Scheme Maximizing Network Life

• Sensors sensing value
• A sensors sensing value measures the
  contribution of the sensor in covering the area
  that the sensor covers
  
• When more number of other sensors cover the same
  area, the less the value of the sensor becomes
  
• The red sensor has a low sensing value, as its
  area is collectively covered by 4 other black
  nodes.
  
• The blue sensor has a high sensing value, as it
  is the only sensor that covers the area.

Data collection node
131
Our Scheme Maximizing Network Life

• Our scheme
• Sensing area of red node is covered by its
  surrounding nodes.
  
• If the red node dead, the sensing coverage
  doesnt change.
  
• So the red node just behaves a message forwarder
  of its surrounding nodes.

Data collection node
132

• Network energy distributes evenly based on area
  not on individual microsensor
  
• I.e. in the example in last slide, the red node
  will die soon, since it forwards lots of packets
  to the remote data collection node. After the red
  nodes death, the energy distribution looks more
  evenly over the area

133

• Calculating a sensing value
• Voronoi polygon
• All points in a nodes Voronoi polygon are closer
  to the node than any other nodes?????
  
• Since sensor accuracy depends on the distance,
  the node can monitor any point in a nodes
  Voronoi polygon more accurately than any other
  node?????

3
4
134

• Compute sensing value
• Acreage of Voronoi polygon as sensing value
• Larger acreage a nodes Voronoi is, higher
  sensing value it owns
  
• Larger acreage means the node can provide more
  accurate surveillance to a larger area than any
  other nodes in the network
  
• How to compute the acreage of a nodes Voronoi
  ploygon?
  

135

• Compute sensing value
• Acreage computation

The polygon includes four quadrangles, I, II, III
and IV. Let consider quadrangle I.
We knows the distance between nodes. So we can
know the angle ? by some triangle computation. We
can compute the quadrangles acreage. The acreage
of the Voronoi polygon is the total of acreage of
all the quadrangles.
2
1
IV
0
?
I
III
II
3
4
136
Details of Our Scheme

• Compute sensing value
• Surrounding sequence

The algorithm above sounds good. But it depend on
that a node knows the surrounding sequence of its
neighbors. It is a sequence of all of node 0s ne
ighbor nodes, in which any two consecutive node
numbers, supposing XY, exist no Z, which is
another neighbor node, that angle X-O-Y contains
angle X-O-Z. For example, 1234 and 3412 are vali
d surrounding sequence but 1324 is not.
A algorithm allowing a node decide surrounding
sequence is designed. Time complexity is O(n), n
is the number of neighbor nodes.
2
1
IV
0
?
I
III
II
3
4
137
How Our Scheme Works

• ---- nodes with data to transmit
• The green node locates in a very dense area
• Each microsensor determines its sensing value
  (Voronoi polygon acreage algorithm)
  
• Some election algorithm to select the
  transmission proxy (in this case, green node is
  selected as transmission proxy since it has
  lowest sensing value)

Transmission Proxy
Data collection node
138

• ----Data transmission
• when one sensor has data to transmit, it decides
  which transmission proxy to chose
  
• In this case, all blue nodes choose the green as
  transmission proxy to relay their data

139

• The green node then decides how to relay the data
  from blue nodes
  
• Here it forward the data directly to the data
  collection node

140

• ---- node out of power
• This process continues until the green node is
  out of power
  
• Some election algorithm will be executed to
  choose another transmission proxy
  
• Almost all microsensors deplete their energy at
  about same rate
  
• Some microsensors with low sensing value spend
  their energy quickly
  
• The entire network can cover larger part of area
  compared to other power saving mechanism

141
Methodology

• Mathematical analysis
• Network coverage problem
• Set cover problem (NP-complete problem)
• Some approximate algorithm needed
• Simulation