Overview

• Home
• Research
• People
• Download
• Courses
• Contact

Wireless Sensor Networks

The Big Picture

Convergence of many contributing technologies happening now:

Development Trends

New Perspective

• In contrast to traditional wireless networks, wireless sensor networks:
  • Are deployed for a specific sensing application – not for only communication purposes
  • Energy consumption a primary issue to prolong network’s lifetime
    • Nodes deployed in harsh environments
    • Large number of nodes
  • Nodes collaborate to accomplish a common task
  • Generally low bandwidth data transmission
  • Dense deployment, redundancy in acquired data
  • Nodes prone to failure, topology may change frequently
  • Limited in power, computing and memory resources
  • Nodes collaborate (not compete) on resource allocation
  • Nodes may not have global ID – Need some sort of geographical/functional labeling
  • Optimization emphasis on energy efficiency (instead of QoS and other BW, throughput constraints)
  • Data flow mostly uni-directional (source to sink), often broadcasting
• Other characteristics
  • Need to self-configure
    • Large deployments
  • Scalable
    • Varying coverage area
  • Redundant
    • To handle node failures
  • Localized algorithms
    • Sharing data long distance too expensive
  • Local functional clusters
    • To coordinate sensing, processing, communication
  • Measurements prone to inaccuracies
    • Inexpensive devices, harsh environments
• Priorities and metrics are very different from other ad-hoc networks
• Cannot tune traditional methods of ad-hoc networks to a special case
• Need a design paradigm shift
• WSN Design Approach
  • Not a pure communication system design view
    • "Data will be properly processed at a powerful central computer"
    • "Maximize throughput and rate to get data there fast"
  • Not a pure algorithm design view
    • "All data will be made available in time"
    • "Find optimal solution in a global sense"
  • How to reconcile?
    • "No central unit to process"
    • "Limited bandwidth"
    • "Long-distance transmission expensive"
    • "Sub-optimal solution ok in many cases"
    • "Large correlation/redundancy in data"
    • "High bandwidth data, low bandwidth communication"

The Message

Novel WSN Trends

• How to go beyond what traditional sensor networks measure?
• Can high bandwidth data be brought into picture?
• Idea:
  • High BW data but low BW communication?
  • Can this be done via local, collaborative processing?
• Applications: Smart environments
  • Examples
    • Smart homes
    • Smart office
    • Smart factories
    • Smart theme parks
    • Smart roadways
  • Most of the needed technologies are available
  • How to make deployment less costly?
  • Intelligent network initialization is the key

New Design Paradigm

In Wireless Domain

Other Wireless Networks	Wireless Sensor Networks
Network's role: data transport	Network's role: information collection and dissemination
Network nodes compete for resources	Nodes collaborate on resource allocation
High data rates (e.g. full images transmitted)	Low data rates (e.g. image attributes transmitted)
Metric: maximize network throughput	Metric: Maximize network lifetime

In Processing Domain

Other Processing Networks	Wireless Sensor Networks
Few high-accuracy sensors	Many low-accuracy sensors
Raw data communicated	Data processed first
Centralized processing	Distributed processing
Relies on high accuracy of measurements	Application assumes inaccurate but many measurements
Closer to optimal solution	Sub-optimal solution, better energy and BW efficiency

• High-BW data, low-BW communication
• Calls for increased level of local data processing
  • On-board
  • Between neighbors in a cluster
• May need to communicate high-BW data infrequently
  • Bursty transmission protocols

Image Sensor Networks

• New development domain in WSN
  • Enabled by low-power image sensors, on-board processing
• High bandwidth data by nature
• Traditional methods based on transmit to processing center not possible
  • High energy consumption
  • Large bandwidth for image transmission
• Local processing reduces communication load
  • Too much of a challenge if:
    • Large frame size
    • Too many frames per second
    • Too many image processing operations
• Image sensors can add intelligence to networks
• Main requirements
  • Processing
    • Infrequent image capture
    • Simple image processing tasks, use small filter sizes and localized pixel operations
  • Communication
    • No or minimal image transmission
    • Nodes should not engage in transmitting significant amounts of data to each other or to a central processing unit
  • Deployment
    • Zero or minor camera calibration effort during the deployment or the operation of the network
  • Scalability
    • Techniques be readily scalable to large networks without much additional communication coordination requirements
  • Robustness
    • The failure of any node should not affect the performance of the technique in the rest of the network
    • Low precision time synchronization is adequate in many applications
• Types of applications
  • Applications that may need no or minimal transfer of images
  • Applications based on communicating attributes provided by local processing
  • Applications with networked low-resolution image sensors
  • Sensor fusion applications