VisualSense: A Modeling and Simulation Environment for Sensor Networks

1
VisualSense A Modeling and Simulation
Environment for Sensor Networks

• Yang Zhao
• Graduate Student
• UC Berkeley

2
Sensor Network Overview

• Sensor node
• Battery life poses limits on sensing,
  communication and computing.
• Individual sensor readings are of limited use.
  Need aggregation of data from a set of sensors.
• Large scale, high density, dynamic, data centric.
• An individual node is inherently unpredictable.
  On the other hand, the high redundancy may give
  us an opportunity to achieve a predictable
  overall behavior.

Sample applications
radio
computer
sensing element
battery
habitat sensing
hazardous environment exploration
structural health monitoring
military tracking
vehicle detection
3
Why a Simulation Environment?

• To understand the behavior of the system
• Complexity
• Dynamics
• To verify correctness/robustness of design
• To be used as a development and test platform
• Experiment on a real platform can be difficult
  and costly
• ex fire monitoring in a building
• Allow people to focus on their problems
• Automatically generate code
• Correctness
• Reduce time cycle

4
Challenges for Sensor Nets Simulation

• Efficiency and scalability
• Large scale systems
• Time is an important aspect in the system
• Processing and propagation takes time
• Interaction with the physical world
• Distributed and Asynchronous
• Various communication media with different
  properties
• radio
• sound
• optical

5
Challenges for Sensor Nets Simulation (cont.)

• Dynamics in the system
• Node mobility and failure
• Changing communication properties
• signal radii changing because power management
• Communication interference
• Heterogenous interaction
• Coordination at multiple levels of design
• Networking infrastructure
• Power management
• Application
• Composition of partial solutions

6
Principles of the Ptolemy II Architecture
Director from a library defines component
interaction semantics
Ptolemy II example
Component
Large, domain-polymorphic component library.

• Model of Computation defines
• Messaging schema
• Flow of control
• Concurrency

Key idea The model of computation is part of the
framework within which components are embedded
rather than part of the components themselves.
7
Discrete Event ModelsOur Basis for Sensor
Network Modeling
The DE domain uses an event queue to process
events in chronological order, as in VHDL,
Verilog, and a number of network simulation
languages (e.g. NS).
8
Representing Wireless Communication

• Traditional Ptolemy models have explicit
  connection topology.
• In sensor network, it is not natural to
  interconnect sensors point to point.
• In sensor networks, connections between sensors
  are conditional and dynamically changing.
• Multiple communication media (acoustic, optical,
  radio) may be used in combination.

9
VisualSense Modeling Sensor Networks

• Discrete event model for the system-level
  interaction.
• Sensor components are connected via wireless
  channels.
• Sensor components can be modeled in Java or using
  Ptolemy II hierarchical models.

10
Example the SmallWorld System
red nodes receive message from the Initiator in
one hop
green nodes receive message from the Initiator in
more than one hops
11
Example 2 Routing in the SmallWorld System
exponent 2.0
exponent 3.0
A short path to route a message to the base
component
12
Example 3 Sound Localization
In this example, a sound source is localized by
triangulation, fusing data from a field of
randomly located sensors. The plot above shows
the identified locations as the sound source
moves through the sensors.
13
Example4 MAC Protocol
The CSMA/CA protocol is modeled using an FSM.
14
Extensibility of the Framework

• Customized Sensor nodes
• Construct as a diagram using actor components
  from the library.
• Write your own actor in Java.
• Channels
• Provide a rich set of channel library.
• Easy to implement your own channel.

15
The Objective

• Provide an infrastructure for the sensor nets
  community to share models.
• Provide a framework for performing specialized
  experiments without having to build everything
  from scratch.
• E.g., can experiment with sound localization
  without having to implement well-known sensor
  localization strategies from scratch. Just use
  these from a library
• Provide a repository for channel models
• Propagation models
• Multipath effects
• Collisions
• Provide a repository for MAC protocols and
  routing strategies.
• Provide a repository for models of sensor
  platforms.
• E.g., the Berkeley Motes.