Docstoc

Energy Efficient_ Fault Tolerant and Adaptive Area Monitoring for Sensor Networks

Document Sample
Energy Efficient_ Fault Tolerant and Adaptive Area Monitoring for Sensor Networks Powered By Docstoc
					        National Conference on Role of Cloud Computing Environment in Green Communication 2012
                                                                                                                                            244




           Energy Efficient, Fault Tolerant and Adaptive
              Area Monitoring for Sensor Networks
                                           Shibitha Mol.Ka,Mrs.A.Sherly Alphonseb
                            a
                           PG Scholar ,Einstein College of Engineering, Tirunelveli-627012,Tamil Nadu
                        b
                         Senior Lecturer, Einstein College of Engineering, Tirunelveli-627012, Tamil Nadu

                                                                       circuits, architecture, algorithms, and protocols have to be
Abstract— For many sensor network applications it is necessary          energy efficient. Another problem is due to the potential
to provide full sensing coverage to a security-sensitive area. To       deployment in uncontrolled and harsh environments and due to
actively monitor the target area, the subset of sensors is
                                                                        the complex arch, wireless sensor networks are and will be
redundantly deployed. One of the major challenges in devising
such network lies in the constrained energy and to tolerate             prone to a variety of malfunctioning. Hence it is needed for the
unexpected failure to prolong the life span of the network. The         WSN should be fault tolerant, so that any failure of nodes will
proposed system rapidly restore the field monitoring, by                not affect the purpose of surveillance.
periodically refreshing and switching the cover to tackle                  In coverage area, sensors are redundantly deployed to
unanticipated failure in an energy efficient manner, because            improve the quality of field coverage, which leads to loss of
energy is the most critical resource considering the irreplaceable
                                                                        battery power (energy). To overcome this, nodes are put to
of batteries of the sensor nodes. In the same time it should
amenably support more than one sensor at a time with different          sleep periodically. For this two approaches were proposed.
degree in distributed approach that periodically selects the covers     They are MAC–Based approach and Application-Based
and switch between them to extend coverage time and tolerate            approach. The former approach uses duty cycle to schedule
unexpected failures at runtime. In this scheme sensor is an             node‟s sleep and wakeup periods based on expected traffic
autonomous system that has the authority to decide how to cover         pattern. In latter approach, a subset of sensors actively
its sensing range. It also incorporates a novel technique for offline
cover update (OCU) to facilitate asynchronous transition between
                                                                        monitors the field, rest are put to sleep. Here application-based
covers. This approach that is robust to failure pattern is not          approach is used as it can balance the load, controls the quality
uniform.                                                                of surveillance and adapt to changing network conditions. For
                                                                        the conservation of energy, distributed cover selection
   Index Terms—Backup cover, deployment, energy efficiency,             protocol is employed. We should also concentrate on adapting
fault tolerance, scheduling, wireless sensor networks.                  to non-uniform pattern of failure. To achieve fault tolerance,
                                                                        an approach of frequently refreshing covers for replacing the
                                                                        failed node by a new active node is used. But refreshing covers
                      I. INTRODUCTION                                   frequently is a network wide process. Selecting k-cover is an
                                                                        another approach which says that the target area is covered by
   A wireless sensor network is a system of tiny, wireless              at least k-sensors (k>1). But it is not energy efficient and also
communicating nodes which monitors a specific region where              it cannot adapt to different failure patterns.
each node is equipped with multiple components. In particular,             So we are going for cover selection protocol to tolerate
each node has a computation engine, communication and                   failure and to improve the lifetime of a network. In this,
storage subsystems, a battery supply, sensing, and in some              minimum set of sensors are selected to cover the target region
cases actuating devices. There are numerous different fields of         and also additional backup set that wakes up periodically to
application of sensor networks. For example, forest fires can           check the active set. Each active node has neighboring nodes
be detected by sensor networks so that they can be fought at an         which are also called as back up nodes. These back up nodes
early stage. Sensor networks can be used to monitor the                 are 1-hop neighbor of the active nodes. It should also have
structural integrity of civil structures by localizing damage for       adaptable coverage when the failure rate varies from one
example in bridges. Further, they are used in the health care           region to another region. These can then be scheduled for
sector to monitor human physiological data. In most of these            making the nodes to be in sleep/wakeup state to increase the
applications and many other anticipated applications,                   node lifetime.
following operating conditions and constraints must be taken               Previous research has not given enough attention to the
into consideration:                                                     tradeoff between fault tolerance and energy efficiency in field
   A wireless sensors node is typically battery operated and            coverage. Proposed system exploits the redundant node
therefore energy constrained. To maximize the lifetime of a             deployment and tolerate unexpected failure in an energy
sensor node after its deployment, other aspects including               efficient way.



Department of CSE, Sun College of Engineering and Technology
        National Conference on Role of Cloud Computing Environment in Green Communication 2012
                                                                                                                                         245


                      II. RELATED WORK                                coordinate for determining their working schedules. In [5], the
   Failure in wireless sensor networks is one of the major            authors addressed the problem of embedded sensor network
threats and fault tolerance has received significant attention. In    fault tolerance by proposing heterogeneous back-up scheme,
[11], the goal is deploy a chain of wireless sensors in the           where one type of resources is substituted with another.
barrier, which usually is a long belt region, to prevent mobile       Specifically, they introduced techniques that enable efficient
objects from crossing the barrier undetected. The multi-round         multimodal sensor fusion in presence of faults and errors.
sensor deployment splits sensor deployment into multiple                We focus on DCS approach that suits critical surveillance
rounds and can better deal with placement errors that often           applications. This exploits the benefits of having a fixed
accompany sensor deployment. This approach reduces the                working cover and schedules wakeup for selected backup
number of sensors needed to provide guaranteed barrier                nodes to achieve fault tolerance, while reducing the required
coverage. In [3], the authors proposed distributed                    energy cost.
Deterministic Cover Selection (DCS) technique to select
backup covers. The each node in sensor networks must self-
organize to monitor the target area as long as possible.                               III. OVERVIEW AND DESIGN
Optimizing energy consumption in area coverage, request
broadcasting, and data aggregation can significantly extend              Initially Sensor nodes are created using many parameters.
network life. The DCS approach relies on selecting a set of           The nodes are deployed randomly in specific region. At the
nodes that guarantees coverage of at least % of the field,           time of deployment, each node keeps a list of its 1-hop
where  ≤ 100. The DCS approach has the advantage of                  neighbors and periodically announces this list. In backup cover
guaranteeing the maximum coverage that the network can                selection, every node calculates the backup cover from the
offer. It is also not sensitive to how nodes are distributed in the   neighbor list in a distributed manner using Deterministic Cover
field and can achieve minimal covers. In PEAS [9],                    Selection algorithm (DCS). The DCS constructs S-cover that is
probabilistic sleep/wakeup was proposed to maintain network           as node-disjoint as possible. In scheduling backup cover stage,
connectivity under unexpected failures. In [2], they proposed         each node in S-cover scheduled to wake up in certain interval
centralized deterministic cover selection method to extend the        to verify whether the region is monitored by active set. The
sensor network life time by organizing the sensors into a             nodes, which monitoring the specific region assigned into
maximal number of set covers that are activated successively.         active set VA and it should remains active state for T time
Only the sensors from the current active set are responsible for      slots. Backup nodes in each cover activated for each M
monitoring all targets and for transmitting the collected data,       interval called tolerance interval. The node remains active for
while all other nodes are in a low-energy sleep mode. By              certain period. After selecting the backup covers, the active
allowing sensors to participate in multiple sets, this problem        node notifies the nodes in the backup covers of their roles and
formulation increases the network lifetime. In [7], they used         their sleep/wakeup schedule. Backup nodes can be assigned
“MAC-based” approach in which a node uses a duty cycle to             opportunistic sleep/wakeup schedule. In backup node probing,
schedule its sleep and wakeup periods based on the expected           each activated node can perform active probing by sending
traffic pattern. In sensor networks with light traffic load, duty     message and waiting for replies. When a backup node u wakes
cycling is a very useful technique for reducing the energy            up, it sends message to active nodes in its neighborhood. If u
consumption due to idle listening. They used k as a parameter         discovers that its current neighbors in VA completely cover its
that captures the duty cycling requirements of an application.        sensing region, it checks if any of them has assigned it a new
To achieve the requisite duty cycling, a sensor should be kept        sleep schedule and then goes back to sleep. If u‟s sensing
awake on an average for 1 k fraction of the time slots. If a          region is not completely covered, it declares itself a member of
node has to forward a packet to its neighbor, it can wake up at       VA and remains active throughout VA the remaining duration of
the active reception slot of that neighbor and transmit the           VA‟s operation. It also employs offline cover update technique
packet. This conserves energy of both the transmitting and the        (OCU) by which it recovers from failure by means of proxy
receiving node. It minimizes the communication latency by             selection.
providing energy efficient periodic sleep cycles for nodes in            This framework provides different degree of redundancy.
wireless sensor networks. It also reduces the energy                  Every node autonomously decides the degree of redundancy in
consumption under sustained load patterns. In [4], the authors        its neighborhood. This framework dynamically maintains 1-
modeled the coverage problem as a decision problem, whose             coverage of every point and tolerates unexpected failure. The
goal is to determine whether each location of the target sensing      steps that give the fault tolerant and energy efficient field
area is sufficiently covered or not. Rather than determining the      coverage are given as follows. The steps are also illustrated in
level of coverage of each location, the solutions are based on        the system architecture in Fig. 1.
checking the perimeter of each sensor‟s sensing range. In [8], a
differentiated surveillance service for sensor networks based
on an adaptable energy-efficient sensing coverage protocol has
been proposed. The protocol allows neighboring nodes to

Department of CSE, Sun College of Engineering and Technology
         National Conference on Role of Cloud Computing Environment in Green Communication 2012
                                                                                                                                         246


                                                                           1) The neighbor nodes should have least percentage of
                                                                           area in common
                                                                           2) Give preference to the node which does not belongs to
                                                                           any other backup cover
                                                                         After sorting the neighbor list of each node selects the top-
                                                                      listed neighbor to its backup cover, the DCS algorithm
                                                                      computes a maximum number of S backup cover sets that are
                                                                      as node-disjoint as possible. The DCS assigns each neighbor
                                                                      node in cover by verifying following parameters
                                                                            a) Least frequency
                                                                            b) Maximizes the covered area

                                                                       D. Backup Cover Scheduling

Fig.1. System Architecture
                                                                         The sleep–wake scheduling policy that is chosen by each
                                                                      node determines when the node wakes up. Each node in
                                                                      backup cover assigned with Opportunistic Sleep/Wakeup
  A. Deploying Sensor Nodes                                           Schedule. The nodes in active set VA remains in “ACTIVE”
   Sensor nodes are deployed using many parameters such as            state until its life time. The node belongs to VS put into
Channel type, Node size, Common Node Number, Common                   “SLEEPING” mode in order to save energy and to prolong the
Node Location, Topography instance, Initial Energy,                   network lifetime. The Opportunistic approach activates the
Transmission Range, Dissemination Type, Idle power and                nodes in backup cover for certain periodic interval. The
Sleep power. The nodes are deployed randomly throughout the           important node periodically updates the S backup cover. The
specified region. The region, which needs continuous                  node in first cover activated within M slots and the second
monitoring selected in random manner. The node comes under            cover nodes activated in 2M time slots and so on.
the specific region initially assigned into active set and                 Each node within cover activated with different interval
remains awake throughout the entire slot. Rests of the nodes          and remains in “PROBING” state for some minutes. It does
are put into asleep mode.                                             not overlap in activation period with each other node in same
                                                                      cover. The node belongs any other cover may have a same
  B. Selecting 1-Hop Neighbor                                         activation period and may be activated simultaneously.
       At the time of deployment, every node keeps a list of its
1-hop neighbors and periodically announces this list. This              E. Probing Backup Node
approach requires an active 1-cover be selected using any DCS              The probing node sends message to neighbor node which
algorithm. The active nodes are assigned into Active set VA.          belongs to Active node set (VA). If the current node region is
The remaining nodes are assigned to the tentative sleeping set        already monitored by any other active sensor node then the
VS.                                                                   node goes back to SLEEP mode. Or else it will actively
       The neighbor nodes for each node in VA are selected by         monitor the region throughout the total T time slots.
calculating distance between the active set node and other
deployed nodes in application region. If the distance is less
than the two nodes transmission range then the two nodes are                              IV. METHODOLOGIES
neighbors. Then each node maintains the other in its neighbor
                                                                        A. Neighbor node discovery
list. Initially, all the nodes participate in selecting VA. Through
periodic neighbor list announcements from nodes in VA, a                   After nodes are deployed, they need to discover their 1-
newly deployed node is made aware of the existence of                 hop neighbors. Neighbor discovery is a first step in the process
sleeping neighbors.                                                   of self-organization of a wireless ad-hoc network.
                                                                           The neighbor nodes for each node in VA are selected by
  C. Backup Cover Selection                                           calculating distance between the active set node and other
  This module employs distributed Deterministic Cover                 deployed nodes in application region. The neighbor set of
Selection Algorithm (DCS), which constructs S cover that are          node V, is defined as
node-disjoint as possible. A backup cover of node v as a set of                 Nb []= { n ϵ Ns| d (v, n) ≤ r, n  v }
v„s neighbors that covers the entire sensing range of node v.            Where,
Before selecting the backup cover each node updates the                    Ns are node set in the deployment region.
neighbor list and performs sorting on it based on following                d (v ,n) denotes the distance between node v and node n.
optimization parameter.



Department of CSE, Sun College of Engineering and Technology
        National Conference on Role of Cloud Computing Environment in Green Communication 2012
                                                                                                                                      247


                                                                  Nb is added to same backupcover and verifies the above. This
  Node V sorts N b [] according to optimization parameter         process continues until the S cover is constructed.
     1) Largest percentage of area in common                        Let us consider example is shown Fig.2. The node v has 8
     2) Preference to node does not belong to any other           sensor nodes in its neighbor list. All neighbor nodes has same
   cover set.                                                     sensing region. The nodes are named as A, B, C, D, E, F, G
                                                                  and H. From these neighbor nodes, 2-cover is constructed
                                                                  using DCS algorithm which is shown in Table I.
 B. DCS Algorithm
        The DCS algorithm constructs S backup covers. A
backup cover of a node V, as a set of V‟s neighbors that cover
the entire sensing range of V. If node V decides to be part of
active set then it computes a maximum of backup cover sets
that are as node-disjoint as possible.

  Nb = 1-hop neighbors of V
  For i=1 to S             // number of backup covers
  IsCoverDone = FALSE
  Nc = 1-hop neighbors of V
  While (isCoverDone == FALSE)
  IF not empty Nb
  Sort Nb based on an optimization parameter                      Fig.2. configuration containing node v and its neighbors
  Pick u: the node on top of Nb
  Add u to BackupCoverSet[i]                                                            TABLE I
  Nb = Nb – u                                                     SELECTING TWO BACKUP COVERS FOR v SHOWN IN FIG. 2
  Freq [u] = freq [u] +1
  If area(V) is covered
  IsCoverDone = TRUE
  ELSE
  Pick u: { u ϵ Nc & Nc BackupCoverSet [i] & u has the
  least Freq & u maximizes the covered area }
  IF u exists
  Add u to BackupCoverSet[i]
  Nc = Nc-u
  Freq [u] = Freq [u]+1                                                The numbers under each neighbor denote the frequency
  IF area(v) is covered                                           of occurrence of that neighbor in the S covers. Node G was not
  isCoverDone = TRUE                                              selected in S2 because it became redundant after selecting F. it
  ELSE                                                            shows that no partial covers of V are selected unless all full
  isCoverDone =TRUE                                               covers are.

   The algorithm determines the backup cover for node V in
                                                                     C. Scheduling
following way.
   The 1-hop neighbor for node V is determined by the                This method is one of the energy saving technique. The
distance measure and maintained in node V. Nb [] and Nc []        scheduling method translates the wireless sensor nodes to
holds the neighbor sensor nodes. S denotes the number of          alternate between Active and Sleep mode. This approach
backup cover.                                                     reduces the cost of energy and also increases the network
   Initially the coverdone is set as false. The set of neighbor   lifetime.
nodes in Nb ordered using following optimization parameters.         Different states of sensor nodes are shown in Fig. All nodes
             a) Largest percentage of area in common.             initially start in the “SLEEPING” state. If a node is selected in
             b) Preference to node does not belongs to any        VA, it moves to the “ACTIVE” state and remains there for T
           other cover set.                                       time slots. Otherwise, it remains in the SLEEPING state.
   The top listed sensor node in Nb is added to the backup        Among nodes in the SLEEPING state, some nodes are selected
cover and its frequency is incremented by one. Then the           by their neighbors to serve in backup covers. Backup nodes
algorithm verifies the entire V‟s region is covered by the        periodically wake up and go to the “PROBING” state. When a
selected backup cover. If it satisfies, the coverdone is set to   backup node discovers an unmonitored region, it joins V A and
true and 1-cover is constructed. Otherwise another node from      moves to the ACTIVE state. Non backup nodes in Vs have no


Department of CSE, Sun College of Engineering and Technology
          National Conference on Role of Cloud Computing Environment in Green Communication 2012
                                                                                                                                      248


role during the operation of VA and remain asleep throughout      action, they find that their sensing regions are completely
the slots.                                                        covered by active nodes and they go back to sleep.


                                                                                    V. PERFORMANCE EVALUATION

                                                                     Initially the sensor nodes are deployed in a random manner
                                                                  in given region. The set of active nodes are deployed in the
                                                                  region which needs continuous monitoring. And the active
                                                                  nodes are labeled as MS and other sensor nodes are labeled as
                                                                  S. The 1- hop neighbor for each MS node is calculated and
                                                                  listed. The neighbor sensor nodes are labeled as BS. Initially
                                                                  all MS nodes are in “ACTIVE” state and the BS nodes are in
                                                                  “SLEEPING” state.
                                                                        From the neighbor list, the k-backup cover is calculated
                                                                  using distributed deterministic cover selection algorithm. After
                                                                  selecting the backup cover, each node in backup cover
                                                                  scheduled to wake up in certain interval using Opportunistic
                                                                  Sleep/Wakeup schedule. The slot interval (ts) is 1 minutes. The
Fig.3. Different states of sensor nodes                           activated node remains in “PROBING” state for 0.2*ts
                                                                  seconds. If the important sensor node fails at any interval, the
                                                                  system checks whether any other nodes in backup cover is in
   This system employs the opportunistic sleep/wakeup             PROBING state. The probing node put into active state if it
schedule. Let Si be the ith backup cover of v, 1 ≤ i ≤ S .The     can monitor the maximum region. Otherwise the activated
opportunistic approach activates the nodes in S1 within M         node goes back to SLEEP mode.
slots, the nodes in S2 within 2M slots, and so on.Let VJ be the         The fig.5 shows the random failure of active node. The
neighbor of V. Initial wakeup time for the node VJ is max {       node 3 is failed at random time and the node 3 is replaced by
iM-nci+j , 0}. VJ can sleeps for min {iM ,T1} cycles.             its neighbor node 5 and it remains active for entire time slot to
                                                                  monitor the specific region.




Fig.4. Opportunistic Sleep/Wakeup Schedule for two covers
                                                                  Fig.5. simulated output of recovery from failure
      To demonstrate the operation of the opportunistic
approach, consider a node V that has two backup covers S 1=          Our metrics are fault tolerance and energy consumption.
{A, B, C} and S2= {D, E, F}. The sleep schedule of these          “Coverage Quality” is a measure of fault tolerance and is
covers is illustrated in Fig.4. When node V fails, the earliest   defined as the minimum fraction of area coverage at any time
backup nodes that wake up are A, D. Assume that A and D do        slot. Fig.6 and Fig.7 shows the performance of our framework
not reduce the size of the hole that has resulted from V‟s        in variable timeslots (M). We assume that energy is directly
failure, and thus they go back to sleep. Node B, however, can     proportional to the duration in which nodes are active. The
cover the entire hole and thus remains active after probing.      graph for total energy consumption is shown in Fig.6. Where
Now, when nodes C, E and F wake up after B‟s recovery             X axis represents the time slots and Y axis represents the
                                                                  energy. The simulation is performed for different node failure.

Department of CSE, Sun College of Engineering and Technology
         National Conference on Role of Cloud Computing Environment in Green Communication 2012
                                                                                                                                                       249


The total consumed energy is determined by calculating the           application-imposed constraints, such as response time with
active node duration and all other nodes probing duration.           mobile sensor devices ([12], [13]).


                                                                                                     REFERENCES
                                                                     [1]    M. Younis, M. Krunz, and S. Rama Subramanian, “Resilient Online
                                                                            coverage for surveillance Application” ACMTrans. Sensor Netw., vol.
                                                                            19, no. 1, Nov-2011
                                                                     [2]    M. Cardei, M. T. Thai, Y. Li, and W.Wu, “Energy-efficient target
                                                                            coverage in wireless sensor networks,” in Proc. IEEE INFOCOM, Mar.
                                                                            2005, vol. 3, pp. 1976–1984.
                                                                     [3]    J. Carle and D. Simplot-Ryl, “Energy-efficient area monitoring for
                                                                            sensor networks,” IEEE Comput., vol. 37, no. 2, pp. 40–46, Feb. 2007.
                                                                     [4]    C.-F. Huang and Y.-C. Tseng, “The coverage problem in a wireless
                                                                            sensor network,” in Proc. ACM WSNA, Sep. 2003, pp. 115–121.
                                                                     [5]     F. Koushanfar, M. Potkonjak, and A. S. Vincentelli, “Fault tolerance
                                                                            techniques for wireless ad hoc sensor networks,” in Proc. IEEE Sensors,
                                                                            2002, vol. 2, pp. 1491–1496.
Fig.6. Energy Consumption                                            [6]     L. Lazos and R. Poovendran, “Stochastic coverage in heterogeneous
                                                                            sensor networks,” ACMTrans. Sensor Netw., vol. 2, no. 3, pp. 325–358,
                                                                            Aug. 2006.
         The Fig.7 shows the coverage quality when time slots        [7]    G. Lu, N. Sandagopan, B. Krishnamachari, and A. Goel, “Delay
(M) varies from 1 to 14. In this experiment, sensing range of               efficient sleep scheduling in wireless sensor networks,” in Proc. IEEE
                                                                            INFOCOM, Mar. 2005, vol. 4, pp. 2470–2481
Rs = 75 m is used. K-cov is not sensitive to M. The fault            [8]    T. Yan, T. He, and J. Stankovic, “Differentiated surveillance for sensor
tolerance of our framework slightly deteriorates as M increases             networks,” in Proc. ACM SenSys, Nov. 2003, pp. 51–62.
because backup nodes to wakeup less frequently. In graph x           [9]    F. Ye, G. Zhong, S. Lu, and L. Zhang, “PEAS: A robust energy
axis represents the time slots and y axis represents coverage.              conserving protocol for long-lived sensor networks,” in Proc. IEEE Int.
                                                                            Conf. Distrib. Comput. Syst., 2003, pp. 28–37.
                                                                     [10]    W. Ye, J. Heidenmann, and D. Estrin, “An energy-efficient MAC
                                                                            protocol for wireless sensor networks,” in Proc. IEEE INFOCOM, New
                                                                            York, Jun. 2002, vol. 3, pp. 1567–1576.
                                                                     [11]   G.Wang and G. Qiao, “Multi-round sensor deployment for guaranteed
                                                                            barrier coverage,” cc. IEEE INFOCOM, 2010, pp. 1–9.
                                                                     [12]   J.Naskath,      Dr.K.G.Srinivasagan     and     S.Pratheema3”Coverage
                                                                            Maintenance using Mobile Nodes in Clustered Wireless Sensor
                                                                            Networks” in Proc. International Journal of Computer Applications
                                                                            (0975 – 8887). Volume 21– No.4, May 2011
                                                                     [13]   Theofanis P. Lambrou, Christos G. Panayiotou “Improving Area
                                                                            Coverage Using Mobility in Sensor Networks”.




Fig.7. Coverage Quality




                            VI. CONCLUSION
   In this work, we presented Energy efficient and Fault
tolerant framework that leverages cover-selection algorithms
to achieve adaptability, fault tolerance, and energy efficiency.
This framework can employ both area coverage and barrier
coverage techniques. It allows different degrees of redundancy
across the field, in addition to controlled speed of recovery of
failures. It also incorporates a novel technique for offline cover
update to facilitate asynchronous transition between covers.
We are currently working on static sensor devices. For future
work, we plan to study the performance of this system under


Department of CSE, Sun College of Engineering and Technology

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:10
posted:7/26/2012
language:
pages:6