Algorithms in sensor networks

1
Algorithms in sensor networks

• By
• Raghavendra kyatham

2
What are sensor networks
3
What are sensor networks

• A sensor network is a collection of some
  (sometimes even hundreds thousands) smart
  sensor nodes which collaborate among themselves
  to form a sensing network.
• Smart sensors are wireless computing devices that
  sense information in many variety of environments
  to provide a multidimensional view of the
  environment.
• ex some sensors can sense light, some
  can sense temperature simultaneously.
• The main task of a sensor network can be divide
  into three categories. Sensing, processing and
  acting.
• After sensing the environment based on the query
  provided by the user, a sensor node can process
  the sensed data, may even sometimes aggregate it
  with other nodes data and send it to the base
  station.

4
What are sensor networks contd

• Based on the results provided by individual
  nodes, the network can act by providing the
  results to the user or to a node connected to the
  internet.
• Are smart sensors possible ?
• Recent advances in MEMS technology have led to
  the development of a new class of computing
  devices with wireless communication capabilities
  called smart sensors. That are low cost, low
  power, multifunctional miniature devices.
• A single smart sensor is limited in its
  capabilities, like restricted memory, restricted
  battery power etc. But when formed as a network
  with many sensors it can do some high
  computational tasks.

5
What are sensor networks contd
6
What are sensor networks contd
7
What are sensor networks contd

• sensor nodes are distributed randomly over an
  object of interest, to from a sensing network and
  monitor the environment. These nodes can group
  and self organize.
• Sensor network can provide access to information
  anywhere and at anytime by collecting,
  processing, analyzing and disseminating the data.
• what should be the number of nodes in the
  network?
• Dissemination of data in sensor networks is of
  two types, query driven and continuous update.
• Smart environments and ubiquitous computing is
  possible with sensor networks.

8
Examples of sensor networks

• Vigilant surveillances like security in a
  shopping mall, an air passengers behavior.
• Predetermining Environmental hazards, providing
  precision agriculture.
• Monitoring of computer server rooms.
• Monitoring of manufacturing plants.

9
Sensor network challenges

• The challenges faced by a sensor network depend
  on the transient nature of the nodes and the
  number of nodes in the network.
• The transient nature of the nodes is due to there
  limited capableness. And because of that there
  topology changes frequently.
• Many situation require ad hoc deployment of
  sensor nodes.
• As sensor nodes are limited in power, capacity
  and memory extending the lifetime of the system
  is difficult.
• The large number of sensor nodes in a network
  will also lead to many different challenges like,
  avoiding collisions, optimal routing etc

10
Requirements of the sensor network

• The requirements of a sensor network depends on
  the type of the application and the number of
  sensor nodes in the network.
• But almost all the networks have some common
  requirements.
• Aggregating its own sensed data with other nodes
  data.
• Self organization of the network.
• Provide queriying ability.
• Maximizing the lifetime of the system by
  appropriate utilization of the energy.

11
Sensor node architecture

• A general sensor node may consist of the
  following five components.
• A sensing hardware, Memory, Processor, Power
  supply, Transceiver.

12
Routing in sensor networks

• one basic operation of sensor networks is to
  gather the sensed data and transmit it to the
  base station, for further processing or as result
  to a given query.
• The general scenario in these networks is, during
  data gathering the intermediate nodes can
  aggregate the data in order to avoid redundant
  transfers.
• The order in which the data or the aggregated
  data is transmitted from the source node to the
  base station is the problem of routing.
• Why cant we apply the standard routing
  algorithms?
• Tree based routing is used when there are few
  number of nodes and hierarchical routing is used
  for large number of nodes.

13
Routing in sensor networks contd

• All the routing protocols should respect the
  energy constraints of the nodes.
• All the routing algorithms mentioned below
  consider sensor nodes to be static, homogeneous
  and energy constrained.
• Almost all the algorithms mentioned below will
  try to maximize the lifetime of the sensor
  network.
• The lifetime of the network can be described as
  the time till data can be transferred, before a
  sensor node gets completely drained of its
  energy.
• The effective utilization of energy is the
  typical measure of performance in sensor
  networks.

14
Sensor network topology with routing tree
overlays

• The most common way of routing in a sensor
  networks is routing trees (multi hop routing).
• A routing tree is a collection of sensor nodes
  with the base station as the root of the tree.
• Sensor A is the parent for sensors
• B and C.
• . Sensor nodes transmit all there
• results to there parent nodes only. It
• is the responsibility of the parent node
• for forwarding them to the base station
• . A child can keep track of several
• parent nodes, and depending on
• the power levels or the quality of the
• communication links a child node
• can change its parent node.

15
The Maximum Lifetime Data Aggregation Problem
(MLDA)

• Given a collection of sensors and a base
  station, together with their location and the
  energy of each sensor, find a data gathering
  schedule, where sensors are permitted to
  aggregate incoming data packets, with maximum
  lifetime.
• Routing structures such as routing trees is well
  suited when there are only a few number of nodes
  in the network.
• Managing the routing trees in such case will
  become infeasible and the overlaps in the routing
  trees can not be effectively utilized.
• A data gathering schedule is a way the data
  packets are collected from all the sensors and
  routed to the base station with maximum lifetime.

16
The Maximum Lifetime Data Aggregation Problem
(MLDA)

• The main assumption of this algorithm is that the
  location of the sensors, base station and energy
  values of the sensor nodes are known priori.
• In this model the lifetime of the system is
  intrinsically connected to the data gathering
  schedule.
• During each round a sensor will collect its own,
  neighbors data and possibly aggregate it and
  send it to the base station.
• If there is T such rounds and f be the total
  number of packets a sensor node i transmits to
  sensor node j .By respecting the energy
  constraints at each node, the data transferring
  schedule can be viewed as flow network G (V,E).
• Schedule S induces a flow network G ( V, E).

17
The Maximum Lifetime Data Aggregation Problem
(MLDA)

• By maintaining the flow conservation principle
  and the energy constraints of each sensor, an
  optimal admissible flow network is constructed
  i.e. a directed graph G having all the sensors as
  nodes and the base station as the root.
• Each directed tree rooted at the base station is
  considered as an aggregate tree, and schedule is
  a collection of such trees.
• The number of rounds the aggregation tree is used
  to transmit data is denoted by f and associating
  it with each one of the edges.
• The depth of a schedule is defined as max depth
  (v) v belongs to V.

18
The Maximum Lifetime Data Aggregation Problem
(MLDA)

• An iterative algorithm GETTREE is used to get an
  aggregation tree A with life time f from the
  admissible flow network. The running time of the
  below described algorithm is of polynomial time
  in the number of sensors.

19
The Maximum Lifetime Data Aggregation Problem
(MLDA)
20
max-min zPmin

• max-min zPmin is an approximation algorithm for
  online power aware routing.
• The goal of this algorithm is similar to that of
  the previous algorithm that we discussed, to
  maximize the lifetime of the network.
• Online routing refers to that there is no fixed
  schedule for routing the messages.
• In this algorithm the network is represented as a
  weighted graph G (V, E). The nodes in the network
  are the vertices of the graph with weights
  corresponding to there power levels. Edges
  correspond to the communication link between
  nodes and the edge weight as the cost of sending
  data between them.

21
max-min zPmin

• The max-min zPmin is defined as routing the data
  along a path with maximal minimal fraction of the
  remaining power in a sensor node after the data
  is transmitted i.e. max-min path and a path with
  minimal power consumption Pmin, with zPmin being
  the relaxed power consumption for sending the
  data.
• The algorithm runs the Dijkstra algorithm for at
  most E times to find the shortest path. The
  running time of this algorithm is O (E. (E
  V log V)).

22
Hierarchical routing

• All the above discussed algorithms tried to
  maximize the lifetime of the system by finding a
  routing path that uses less energy.
• This type of routing is known as multi hop
  routing or static clustering which has very
  serious limitations when the number of nodes in
  the network becomes very large.
• Static or multi hop routing protocols require the
  knowledge of the energy levels of the sensor
  nodes which may be difficult to obtain in large
  networks. One method of obtaining such
  information is through broadcasting. But ?
• In large network networks transmitting data
  through intermediate nodes may sometimes consume
  more than routing directly to the base station.
  So large networks are divided into zones are
  clusters.

23
Leach (low energy adaptive clustering hierarchy)

• Leach is also an energy efficient protocol for
  routing in sensor networks.
• Leach is based on the principle of clusters and
  is organized into rounds.
• In each round a self elected cluster head
  collects data from all other sensors in the
  cluster, aggregates it and transmits it directly
  to the base station.
• During the setup phase a predetermined fraction
  of nodes elect themselves as cluster heads. A
  threshold value T(n) is used to compare the
  random values generated by the node wanting to be
  the cluster head.
• If the value of a particular node is less than
  the threshold value, then it will act as the
  cluster head for the current round.

24
Leach (low energy adaptive clustering hierarchy)

• Once a cluster head is selected it broadcasts its
  ID to all other nodes in the cluster.
• A non cluster head may receive many broadcasts
  from different cluster heads it makes a
  selection among them by comparing the quality of
  the communication link with various cluster
  heads.
• On receiving the decision of the noncluster heads
  the cluster head creates a schedule and informs
  it to nodes in the cluster.
• In this way each node follows the schedule and
  transmits the data to the cluster head, and the
  head after aggregating the data transmits it to
  the base station directly.
• The key feature of leach when compared to the
  above discussed protocols is its localized
  coordination for cluster setup and operation.

25
Zone based max-min zPmin

• Zone based routing is a hierarchical approach to
  the max-min zPmin.
• The algorithm groups the nodes in the network
  structurally into geographical zones that can
  overlap, and organizes zones hierarchically to
  control routing across zones.
• The algorithm is divided into three main parts,
  first how the nodes in a zone collaborate to
  estimate the energy level of the zone. Second,
  how data is routed within a zone and third, how
  data is routed across zones.
• The energy estimation of the zones is done
  relative to the direction of data transmission.
• The zones are assumed to be squares with their
  neighbors being in north, east, west, and south
  directions.

26
Zone based max-min zPmin

• There is a controller node in each zone which
  estimates the energy level of the zone i.e.
  estimating the number of messages that can flow
  through the zone. The controller poles each node
  in the zone for its energy level and then runs
  the max-min zPmin algorithm. Then it simulates
  sending proportionate amount of data units, and
  repeats it until a node on the path gets
  saturated.

27
Zone based max-min zPmin
28
Zone based max-min zPmin

• After estimating the power level of each zone
  with respect to the directions of the other
  zones, the next thing is estimating a global path
  to route the data.
• The zones are represented as a K1 graph, where k
  vertices correspond to each data direction
  through the zone. The zone label vertex is
  connected to all the data direction vertices and
  the data direction vertices are connected to
  neighboring zone vertices if data can be
  transmitted in that way.

29
Zone based max-min zPmin

• The edges in this zone graph do not have weights,
  and a global route for sending data can be found
  as the max-min path in the zone graph.
• The path that is selected should be the path that
  goes through zones with maximum power levels i.e.
  a slight modification to the max-min zPmin
  algorithm.
• After a global path through the zones is found
  the next task is to find routes within a zone.
• For each node in the overlap region, the number
  of paths that can be locally routed through each
  node is computed during the energy level
  estimation.
• Finally only those nodes that have maximum data
  weight is selected to maximize the global flow
  between zones i.e. choosing nodes which can be
  useful in both local and global routing.

30
Zone based max-min zPmin

• The algorithm to find global path to route the
  data.