Maintaining Optimal Sensing Coverage and Connectivity in Large Sensor Networks

1
Maintaining Optimal Sensing Coverage and
Connectivity in Large Sensor Networks

• Honghai Zhang

2
Introduction

• Sensor Network has two functionalities
• Sensing monitor environment, acoustic signal,
  or electronic signal
• Networking transmit the sensed signal to other
  nodes
• Additional Goal
• Use power efficiently, extend network lifetime
• Reduce message exchange for maintenance
• Assumptions
• Network is dense.
• Sensing range and radio range are different

3
Methods

• General Method
• Static scenario pick minimal number of working
  nodes to fulfill the coverage and connectivity.
• Dynamic scenario when a node is going to die,
  wake up some sleeping nodes.
• UCLA work (technical report), by Fan Ye, Gary
  Zhong, Songwu Lu and Lixia Zhang
• Ottawa work (WSNA02), by Di Tian and Nicolas D.
  Georganas.

4
UCLA approach

• Each node has three modes sleeping, wakeup, or
  working.
• A sleeping node wakes up after sleeping for some
  time.
• When a node wakes up, it broadcasts a message
  PRB.
• When a working node receives a PRB message and is
  within range r, it sends a reply.
• When a wakeup node doesnt get a reply within
  time Tw, it will start to work. Otherwise, it
  continues to sleep.
• Two design parameters
• Probing range r
• Wakeup rate ? -- decide how often should a node
  wake up

5
Designing probing range r

• Given a desired density ?, design r.
• ? 1/S(r) , S(r) is the average area occupied by
  each node.
• Given probing range r,
• In the densest case, Sd(r)
• In the sparsest case, Ss(r)
• Average density ? (1/Sd(r) 1/Ss(r) )/2

6
Designing probing range r (cont)
7
Designing wakeup rate ?

• Given desired aggregate wakeup rate ?d, set the
  individual wakeup rate ?.
• Working node estimate the aggregate wakeup rate
  ?e, rebroadcast to each of its wakeup neighbor,
• Wakeup node update its wakeup rate
• ? ? ?d/ ?e

8
Summary of UCLA work

• Concerns more on dynamic scenario
• Problems
• Assume all nodes have same sensing range
• Not ensure coverage,
• not ensure connectivity

9
Ottawa work

• Consider the coverage problem only (assume the
  radio range is larger than sensing range)
• Focus more on static scenario
• Ensure coverage
• Assume nodes may have different sensing range

10
Coverage Preserving

• When node i and js sensing range intersect, the
  maximal sector of i that is also covered by j is
  called the sponsored area by node j to i.
• If a nodes sensing area is covered by all its
  working neighbors sponsored area to it, then the
  node can be turned off.

11
What we plan to do

• Explore some properties when nodes are optimally
  chosen to cover the whole area.
• Design algorithms based on those properties.
• Explore the relationship between sensing range
  and radio range.
• More usage on those relationships

12
Property 1

• Minimal number of working nodes ? Minimal
  overlapping of working nodes, assuming working
  nodes covers the whole area.

13
Property 2

• Assuming very dense network, in order for minimal
  overlap, three nodes need to be placed with equal
  distance d and d sqrt(3) r.
• If the positions of two nodes are fixed, then the
  third node should have equal distance d to the
  other two nodes and d is only related to the
  distance between the first two nodes.

P
B
A
O
R
Q
C
14
Relationship between sensing range and radio range

• The necessary and sufficient condition for that
  whenever a convex area is covered then all nodes
  are connected is,
• Radio range gt 2 sensing range

15
Proposed Methods

• Assume radio range gt 2 sensing range.
• One or a few nodes act as the lead nodes that
  initiate the process of finding the working set.
• When a node hears the initiating message, it
  delays time t1 to decide to act as a working node
  and then broadcasts the initiating message
• When a node hears two initiating message, it
  delays time t2, and then broadcasts the
  initiating message.
• T1 and t2 are proportional to the nodes distance
  to the optimal point but with some random effect.

16
Problems

• When does a node think it can sleep?
• It is covered by some other nodes.
• The most recent intersecting point is beyond its
  sensing range.
• How to do it in a dynamic scenario?
• Round by round, (adopted by Ottowa)
• Each working node remembers its neighbors, at
  least the two closest intersecting neighbors, and
  replies with such information to the wakeup
  nodes.