Habitat Monitoring: Application Driver for Wireless Communications Technology

1
Habitat Monitoring Application Driver for
Wireless Communications Technology

• CS294-1 Reading
• Aug 28, 2003
• Jaein Jeong

2
Outline

• Introduction
• Challenges for sensor network
• Habitat monitoring
• How to achieve low duty-cycle operation
• Frisbee Model
• Time Synchronization
• Tiered Architecture for scalability
• Hardware and software platforms

3
Introduction

• The rise of sensor/actuator network
• Thanks to the micro-sensor and low-power wireless
  communications.
• Challenges for sensor network
• Need for good physical models.
• Increased dynamics.
• Energy constraints.
• Scaling challenges.

4
Principles for habitat monitoring.

• Self-configuration.
• The ability to operate autonomously.
• Sheer number of nodes cant depend on manual
  configuration.
• Sensor nodes should adjust to the environment
  dynamics.
• Self-assembly / healing, localization, time
  synchronization, etc.
• Energy efficiency.
• Techniques are needed to reduce power consumption
  for longevity.
• Filtering and in-network processing reduces
  communication traffic (e.g. seismic sensor
  camera).
• Ability to wake-up nodes in response to
  interesting events.

5
Frisbee Model

• Motivation
• Not all the sensors need to be active.
• Save energy by waking up only the nececessry
  ones.
• Frisbee Model
• Only a region of sensors are active (frisbee).
• Active region can move.
• Power saving by making sensors asleep
  outsidethe frisbee

6
Frisbee Model Localized Algorithms for Defining
Frisbee Boundary

1. Nodes that have detected the target wake up
   neighbors.
2. Query the nodes already awake to determine speed
   and direction.
3. After a certain amount of time, the nodes time
   out and sleep.

7
Time Synchronization

• The clock of each sensor is synchronized within a
  bound e.

• Rationale
• Suppression duplication for data aggregation.
• Scheduled wake-ups (e.g. TDMA).
• Detection of speed and direction of phenomena.
• Application specific variables
• Precision of time synchronization.
• Frequency to fix.
• Sensor network density.

8
Time Synchronization Methods

• Universal Time
• A single time signal is broadcast to all nodes.
• Example WWVB, GPS
• Pros best for eliminating accumulated error.
• Cons requires more resources (HW, power).
  Depends on infrastructure (e.g. WWVB).
• Peer-to-peer time distribution.
• NTP establishes a federation of synced nodes.
• Works with no external time source.
• Works over existing comm. Infrastructure.
• Can be synced with external time source.

9
Tiered Architecture

• Idea
• Use different levels of hardware platforms to
  make an efficient system.
• Analogous to memory hierarchy.
• Sensors more expensive, larger, but
  fasterTags more limited, but smaller,
  cheaper
• Advantages
• Lower cost ? More densely deployable.
• Lower power ? Longevity
• Smaller form factor ? More easily deployable.

Sensors
L1 Cache
Tags
L2 Cache
Motes
Main Memory
Tiered architecture of sensors
Memory hierarchy of desktop systems
10
Hardware Platforms
COTS MoteUC Berkeley
Prototype radio controllerfor tag platform
PC104-based sensor node

• velcroable device
• Master, radio, sensor and DSP module
• Under development

• High end node compatiblewith desktop PCs.
• Lower cost w.r.t. laptops
• Fast µ-processor (66MHz),large memory (16MB),
  Radio.

• Smallest device in tiers
• Envisioned to be the size of dusts.

11
Software

• Radiometrix Device Drivers
• Linux device driver for RPC.
• Emlog
• Linux kernel module that displays the most recent
  output from a process.
• Useful for logging and debugging.
• Parapin
• A tool helps writing C code to control parallel
  port.

12
Discussion