l23

Wireless Sensor Networks
1
Puzzle
Four integers: 1, 2, 3, 4
Three operators: +, -, *
What is the minimum positive integer that cannot be
generated using an expression containing all
(repetition not allowed) four of the above integers,
and three (repetition allowed) operators?





1 * 2 + 3 * 4: allowed
1 + 2 + 3 + 4 : allowed
1 * 2 + 3 + 1 : not allowed
1 + 2 * 3 : not allowed
12 + 3 + 4 : not allowed
2
Wireless Sensor Networks
Used for sensing!
Large number of sensor nodes densely
deployed either inside the phenomenon
of interest or close to it
Random deployment feasible due to
low cost nature of sensors
Sensor: sensing, processing,
communication
3
WSN Organization
Backbone
User/
Decision maker
Sink
Sensing Field
Sensors sense, process and give information
to sink
Sink propagates information back to the
user/decision maker
4
WSN Applications
Sensors – temperature, humidity, motion,
light, pressure, soil make-up, noise levels,
stress, etc.
Military – surveillance, targeting, damage
assessment, chemical/biological agent
detection
Environment – forest fires, biocomplexity
mapping, flood detection, precision
agriculture
Health – telemonitoring, tracking, drug
administration
Home – automation, smart environment
5
WSN vs Ad-hoc Networks
Larger number of nodes
Density of nodes
Failure proneness
Communication patterns
Limited Node capabilities
Lack of Global Identifiers
6
Key Factors in WSNs
Fault tolerance (survivability)
Scalability
Production costs
Hardware constraints
Topology management
7
WSN Protocol Stack
Layers





Application
Transport
Network
Data link
Physical
Planes



Power management
Mobility management
Task management
8
Application Layer
Time synchronization
Controlling ON/OFF decisions of
sensors
Querying & controlling sensor network
configuration
Security
Task assignment
Data collection
9
Transport Layer
Conventional transport layer protocols
(TCP, UDP, NORM, SRM, etc.) cannot be
uses in a sensor network



Point-to-point vs point-to-multipoint
(downstream) & multipoint-to-point
(upstream)
Upstream: Information reliability vs data
reliability
Downstream: Dimensions of reliability
10
Network Layer
Data centric routing and flooding


How many pedestrians do you observe in
region X?
Let me know if the temperature in your
local neighborhood is greater than 100F
Potential data processing on paths!
11
Directed Diffusion
A node requests data by sending interests
for named data

The request “How many pedestrians do you observe
in region X” is broadcasted to region X
Data matching the interest is then “drawn”
down towards the node

When a node in region X receives the request, it
activates its sensors, and returns sensed information
along reverse path of interest propagation
Intermediate nodes can cache, or transform
data

Combine reports from multiple sensors to more
accurately pinpoint pedestrian’s location
12
Elements of Directed Diffusion
Interests

Query of what the user wants
Data messages

Collected or processed information of a physical
phenomenon
Gradients

Direction state created in each node that receives
the interest
Reinforcements

Of one or a small number of the available paths
13
Naming
Attribute-value pairs
Example:

Vehicle detection task (query)
 (Type=wheeled, interval=20ms,
duration=10seconds, rect=[-100,100,200,400])

VDT (response)
 (type=wheeled,instance=truck,location=[125,2
20],intensity=0.6,confidence=0.85,timestamp=
01:20:40)
14
Interests
Can be initiated by the sink
Exploratory interest with a large
interval, followed by reinforcements

e.g. to detect any wheeled vehicles
Soft-state refreshing of interests

reliability & overheads
Each node maintains one entry per
interest in an interest-cache
15
Interests (contd.)
Each interest entry contains a gradient
(neighbor, report rate, and lifetime)
Interest entry possibly created upon
receipt of interest
Interest possibly forwarded to a subset of neighbors

e.g. based on cached data
16
Gradient Establishment
A generic notion
Can be implemented in several ways:
binary values, probabilistic forwarding,
load balancing
Gradients might be set-up differently
for different tasks
17
Data Propagation
Nodes in “rect” sense data
Propagates data according to the gradients
to the corresponding interest entry
If an intermediate node receives data, but
finds no interest entry, it drops the data
Gradients can change as data is being
forwarded

e.g. down-sampling : 100 events/second to 50
events/second
18
Reinforcement
Exploratory gradients vs. data gradients
Sink reinforces one (or a subset) of the
neighbors reporting back exploratory events
Data gradients can have a higher reporting
rate – positive reinforcement
Allows sink to reinforce selective paths and
reduce multi-path routing for the real heavy
data
19
Other Issues
MAC
Topology control (with sensing
reliability)
Sensor placement
Reliable transport
Congestion control
20
GARUDA
21
Puzzle
Three bulbs inside a room
Three switches outside the room
Room initially locked
You can initially play with the switches
You then need to enter the room and
be able to match switches to their
respective bulbs
Devise a strategy
22