Addons to Dynamic Topology Construction

1
Add-ons to Dynamic Topology Construction

• Yi Pan

2
Context

• Wireless sensor network
• Issues in power-saving topology construction
• In the research of ad hoc network topology
  construction, the following objectives are
  concerned
• Continuous connectivity with minimal active ad
  hoc nodes
• Shorten the delay caused by power-saving topology
  construction
• Alleviate the capacity decrease by less active
  nodes
• In more power stringent sensor network, I assume
• Non-traffic adaptive continuous connectivity is
  not required
• In traffic intensive area, continuous
  connectivity maintained by the topology is
  required
• In area with little traffic, continuous
  connectivity is not required all the time
• Adaptive scheme to allow the links in topology to
  be quickly turned on in continuous mode for
  ongoing traffic is needed to make the delay and
  bandwidth acceptable with even low power
  consumption

3
Context

• Wireless sensor network (contd)
• I also assume the sensor can be in multiple
  power-consumption levels
• In the research community of sensor/ad hoc
  network, multiple power-consumption levels have
  been already identified
• Ie. sleeping, receiving, and transmitting are
  samples of three different levels of power
  consumption
• Further levels of power consumption can be added
• For example, by restricting the transmission
  range of the node to several levels of distances,
  the power consumption in transmission mode can be
  further divided into sub-levels
• By controlling the power consumption levels of
  the sensors, we can define different roles the
  sensor playing in the topology construction and
  routing
• Thus, different power-consumption levels will
  allow the construction of local links and
  global links
• Different levels of links are not detected by all
  nodes in all power consumption modes

4
Basic Ideas

• Global link coordinators
• Some of the sensors should work at a high power
  consumption mode to be global link coordinators
• It maintains the topology which is seen by the
  routing protocols
• The always-continuous-connectivity is not the
  requirement
• Some of the global links may only be actually
  on for short intervals periodically if no
  traffic is directed through those links
• The routing protocol is still able to see the
  link without knowing the exact state of that link
• The time interval that the link is on
  automatically adapts to the traffic volume
  passing through
• The role of global link coordinator is taken by
  a set of sensors in the local area in turn
• If the traffic load becomes really heavy,
  increasing the network capacity by splitting the
  coverage area of the coordinator and create
  multiple active coordinators to balance the
  traffic load should be considered
• Splitting of the coverage area not only increases
  the bandwidth capacity dynamically, but also
  reduces each coordinators power consumption for
  each bit-transmitted

5
Basic Ideas

• Local link maintainer
• A portion of sensors are in the role of local
  link maintainer by listening to the global
  coordinator but transmitting packets only within
  the coverage of the coordinator
• Those sensors have a long distance downlink
  from the coordinators
• Multi-hop transmission may be needed for those
  sensors to forward data to the coordinator
  through uplink
• The local link maintainers forms a local SAN for
  the coordinator
• When the coordinator is in sleep mode, the local
  link maintainers can pick up the packets from
  neighboring global coordinators and store the
  packets locally
• When the coordinator wakes up, it first signals
  the local link maintainers to forward the stored
  data to itself
• Major issues are
• Is coordination between the local link
  maintainers needed?
• Should be local link maintainers behavior be
  adaptive to the traffic?
• Preferred answers no to both to save power
• Majority of sensors should be sleepers

6
Basic Ideas

• Routing perspective
• Consider the routing on top of the topology
  construction, I have
• Routing algorithm should be performed on the
  global link topology only
• To allow the network scale to large size, this
  feature is needed
• Routing algorithm should not need exact current
  link status of the global link
• To avoid frequent link updates overhead, this
  feature is also needed
• Inaccurate routing is preferred
• Instead of routing the packets according to the
  exact sensor ID, some abstract set/direction
  information about the destination is preferred to
  avoid large routing table size and computational
  overheads
• Instead of directing the packet to next hop
  which is the exact node ID, an abstract or only
  directional next hop information is preferred so
  that the update of global link coordinator may
  not always be needed

7
Basic Ideas

• Routing perspective (contd)
• Location-based routing
• If we have a rough idea/information of the
  destination by some location mechanism, a
  directiondistance information associated to
  the source should be enough without routing
  tables in intermediate nodes
• At the source node, it attaches the angle toward
  the destination and the rough distance to the
  packet and sends it toward the destination
• The intermediate node receiving the signal
  identifies the incoming nodes angle and
  distance, then adjust the outgoing direction and
  distance based on geometry calculation instead of
  routing entries
• The routing nodes can broadcast the packet to a
  range of angles to increase the reachability and
  reliability of the packet
• The broadcast range of angles can be adaptive to
  the distance to the destination. The shorter the
  distance is, the larger the broadcast range is
• Issues
• How can we get the rough direction and distance
  toward the destination?
• May be the application specifies it
• Some location service is required

8
Basic Ideas

• Illustration of directional broadcast

?
s
Adjusted ?
d