Вы находитесь на странице: 1из 32


ABSTRACT Coverage of sensor nodes is an important research problem in Wireless Sensor Networks. Coverage problem is generally categorized into three types, full area coverage, barrier coverage and point coverage. This work deals with the point coverage problem, wherein we only need to monitor oints of !nterest " #!$ periodically. %s the coverage demand by each #! is time&variant, we can utilize small number of mobile sensors to cover a large number of #!s. C%'W#C proposes a distributed coordination algorithm for controlling the sensor mobility, to provide efficient and effective coverage. C%'W#C minimizes the delay in meeting the #! and also minimizes the energy consumed by mobile sensor node. This (uality of service factors are addressed by algorithms nature of providing optimal control decisions. This work ends with e)tensive simulation to study the performance of the proposed algorithm. This approach is compared for coverage efficiency and average #! delay with randomized scheme and it*s learnt that C%'W#C outperforms the latter significantly. KEYWORDS + oint of interest, Coverage problem, ,pidemic -outing. 1 INTRODUCTION % wireless sensor network "WSN$ consists of spatially distributed autonomous sensors to monitor physical or environmental conditions, such as temperature, sound, pressure, etc. and to cooperatively pass their data through the network to a main location. The more modern networks are bi&directional, also enabling control of sensor activity. The development of wireless sensor networks was motivated by military applications such as battlefield surveillance. today such networks are used in many industrial and consumer applications, such as industrial process monitoring and control, machine health monitoring, and so on.

/igure 0.0 Typical multi&hop wireless sensor network architecture

!n figure 0.0, a general view of WSN architecture is depicted. The WSN is built of nodes from a few to several hundreds or even thousands, where each node is connected to one "or sometimes several$ sensors. ,ach such sensor network node has typically several parts+ a radio transceiver with an internal antenna or connection to an e)ternal antenna, a microcontroller, an electronic circuit for interfacing with the sensors and an energy source, usually a battery or an embedded form of energy harvesting. % sensor node might vary in size from that of a shoebo) down to the size of a grain of dust, although functioning 1motes1 of genuine microscopic dimensions have yet to be created. The cost of sensor nodes is similarly variable, ranging from a few to hundreds of dollars, depending on the comple)ity of the individual sensor nodes. Size and cost constraints on sensor nodes result in corresponding constraints on resources such as energy, memory, computational speed and communications bandwidth. 'obile wireless sensor networks "'WSNs$ can simply be defined as a WSN in which the sensor nodes are mobile. 'WSNs are a smaller, emerging field of research in contrast to their well&established predecessor. 'WSNs are much more versatile than static sensor networks as they can be deployed in any scenario and cope with rapid topology changes. The advantage of allowing the sensors to be mobile increases the number of applications beyond those for which static WSNs are used. /or e)ample, sensors can be attached to people or animals for health monitoring or movement tracking. Sensors may also be attached to unmanned aerial vehicles "2%3s$ for surveillance in military situations or private security scenarios. The 2%3s may also be used for environment mapping in situations where human presence is not possible, such as irradiated buildings. Since there is no fi)ed topology in these networks, one of the greatest challenges is routing data from its source to the destination. WSN routing protocols provide the re(uired functionality but cannot handle the high fre(uency of topology changes. Whereas, '%N,T routing protocols are can deal with mobility in the network but they are designed for two way communication, which in sensor networks is often not re(uired. Currently there is no standard routing protocol for 'WSNs, so a popular alternative is the '%N,T routing protocol %#43 "%d hoc #n&4emand 4istance 3ector$. %nother popular routing techni(ue is to utilise location information from a 5 S module attached to the nodes. This can be seen in protocols such as 6one 7ased -outing "67-$, which uses the location information to aid a hierarchical method of routing. !n comparison, 5eographically #pportunistic -outing "5#-$, is a flat protocol that divides the network area into grids and then uses the location information to opportunistically forward data as far as possible in each hop. 'ultipath protocols provide a robust mechanism for routing and therefore seem like a promising direction for 'WSN routing protocols. The work proposed here, C%'W#C deals with the point coverage problems in

W'SN. C%'W#C basically proposes a localized mobile sensor node coordination algorithm inspired from ,- , which provides each sensor node an effective yardstick to control motion decisions. The prime motive of this approach is to increase the coverage efficiency of the predefined oints of !nterest " #!$. To be more elaborate, each #! will be given a pre&defined patrol period 8t* "which means the #! must be visited atleast once every 8t* seconds$. The motion control logic defined in this approach can enable the sensors to take optimal motion decisions, so as to meet the time&variant re(uirements of all mobile sensors effectively. C%'W#C addresses two main challenges. /irstly, the big challenge of propagating the local monitored information to the whole network is achieved with minimal overhead, by adapting the approach by ,- . Secondly, the challenge of covering mobile sensors with less lag and less ineffective control decisions is also addressed. This paper work highlights the recent works done in the field of node coordination and the challenges faced by researchers in this field in the Section 9. Section : illustrates the working and the modules involved in the C%'W#C algorithm. Section ; details the epidemic routing protocol ",- $, which forms the basis of this approach. Section < briefs the assumptions and implementation particulars of C%'W#C. Section = provides the performance analysis with the random scheme, and the report concludes with a chapter, holding the future relevant works in this field and the conclusion part. 2 LITERATURE SURVEY The deployment of sensors in the target field directly determines the performance and functioning of the network and further influence the coverage and efficiency of wireless sensor network. % good deployment approach will augment the degree of resource allocation in the network and enable a better performance on information gathering and communication. 4ifferent approaches are used for the deployment of sensors over the target field. The approaches are Computational 5eometry based, otential field based, robabilistic approach, Computational !ntelligence based and 7io inspired approaches. ,)isting approaches dealing with sensor deployment can be generally classified into two types+ hysics&based and 5eometric based >'ahboubi 9?0:@. !n physics based approach the sensor nodes are assumed as points subAect to attractive or repulsive forces like Newton*s Baw of forces while in Computational geometry based approach the sensor nodes are assumed as points on 94 or :4 plane. #ther approaches use different algorithms which are discussed in this section. 9.0 3irtual force based approach & !n virtual force based approach, the sensor nodes are modelled as points subAected to attractive or repulsive force according to the distance between each two sensors. 7y setting a threshold of desired distances between sensors, each sensor moves in accordance with the summation of the force vectors and eventually a uniform

deployment is achieved. 3irtual force method helps move sensors from high density area to low density areas, thereby minimizing sensing overlap. This approach is useful in self&deployment of sensor ensuring mobility of sensors. The forces are of type Newton*s Baw. ,nhanced form of 3irtual force method is proposed called as connectivity preserved virtual force method "C 3/$ in >5uang Tan 9??C@. The authors consider the obstacles during deployment. The given approach is used for mobile sensor network for the self&deployment. The authors overcome the limitations of previous approaches where large communication range, dense network and obstacle free field or full knowledge of the field layout have been assumed. 3irtual force method is used for sensor relocation proposed in >5aretoo 9??D@. 2pon occurrence of physical phenomenon the nodes relocate themselves so as to control the event, while maintaining network connectivity. %fter the end of the event all the nodes return to the monitoring configuration. %n algorithm to manage mobility using network dynamics is proposed in >6hang 9??E@. ,ach node periodically calculates the virtual force it receives from its neighbours based on the distance with all its neighbours. %ccording to the resulting virtual force a node determines the movement speed and direction in the ne)t interval. >Biu 9?0?@ propose a simple movement control algorithm for the problem of achieving ma)imized coverage of network, minimal total moving distance. This algorithm is also efficient in using only 0&hop information. This algorithm emulates the attractive and repulsive force in nature, such that each robot simply follows the resultant virtual force to move. They also derived bounds on the ma)imum coverage and the minimum moving distance. They developed the first physics model for movement control in WSN. >'ilan ,rdelA 9?0:@ proposes a distributed, autonomous approach of sensor node placement, aiming for an optimal coverage of the predefined #! with a connectivity constraint. !t employs -elative Neighbourhood 5raph "-N5$ "a reduced graph from initial configuration wherein an edge that connects two sensors is removed if there e)ist another sensor that is at a lower distance from both sensors$. The distance travelled by the mobile sensor is constrained by maintaining connectivity with its -N5 neighbours. This was the first work towards maintaining the connectivity all along the deployment procedure and therefore allows the tracking of mobile #!. 9.9 7io inspired approach The collective behaviour of biological species "e.g., ants$ provides a natural model for distributive problem solving without any e)tra central control or coordination. /or individuals, their behavioural capabilities are limited and their cognitive systems are not powerful enough to ac(uire global knowledge. Studies have shown that self& organization and stigmergy are two key ideas in swarm systems. The basic concepts of

self&organization include positive feedback, negative feedback, fluctuation amplification, and multiple interactions >!yengar 9??D@. %s an e)ample, consider the ant colony to illustrate these concepts+ the action of disposing pheromone is a positive feedback mechanism to recruit more ants such that more pheromones get disposed on the shorter path. however, the evaporation of pheromone is a negative feedback to reduce the pheromone level. in this way, the shortest paths to the food source can be found accordingly. 'oreover, stigmergy is defined as the indirect communication used by ants in nature to coordinate their Aoint problem solving activities. %nts achieve stigmergic communication by laying a chemical substance called pheromone, that induces changes in the environment which can be sensed by other ants. -ecent works highlights the fact that computer scientists were able to transform the models of collective intelligence of ants into the useful optimization and control algorithms. >Bee 9?09@ adapted the conventional ant&colony optimization "%C#$ algorithm for addressing ,nergy&efficient coverage ",,C$ problem. They found out that the performance of the %C# algorithm is determined by how it initializes the pheromone field and how it makes the construction graph reflect the characteristics of the problem. So, they applied a new initialization method for the pheromone field and the modified construction graph, which ultimately improved their algorithm performance. 9.: 'ovement assisted approach !n this type of sensor deployment, the mobile sensor move from an initial unbalanced state to a balanced state. 5enerally the initial distribution is meant to be random and then the sensor takes into account different parameters needed to be minimized. Such F#S parameters include total moving distance, convergence rate, total number of moves and communicationGcomputation cost. %fter analysis these parameters, the sensors take optimal control decisions such that at each stage F#S parameters are minimized. 'ovement assisted approach is highly useful in addressing communication hole problem, where some regions of the network have no deployed sensors. >Hang 9??D@ proposed a load balancing algorithm named S'%-T to eliminate holes in the network by achieving an even deployment of sensors with modest costs. The advantages of this work include less rounds re(uirement for load balancing and rapid balanced state achievement. Iowever the performance is drastically reduced when the number of deployed nodes is less than ;, where 8n* denotes the number of grids. They also neglected the usage of intra&cluster load balancing, which could have achieved high& resolution load balancing. >Bee 9??C@ proposed a faster hole detection and recovery algorithm. This work utilize voronoi diagram to discover the coverage holes. 9.; Computational geometry based approach

The primary goal of computational geometry is to develop efficient algorithm and data structure for solving problem stated in terms of basic geometrical obAects+ points, lines, segments, polygons, etc. The sensor nodes are assumed as points in 94 or :4 plane. The geometrical structures used are grid and polygons for modelling of sensors. The two common data structure used are 3oronoi 4iagram and 4elaunay Triangulation. The 3oronoi diagram has been reinvented, used, and studied in many domains. %ccording to >/. %urenhammer 0CC0@, it is believed that the 3oronoi diagram is a fundamental construct defined by a discrete set of points. !n 94, the 3oronoi diagram of a set of discrete sites "points$ partitions the plane into a set of conve) polygons such that all points inside a polygon are closest to only one site. This construction effectively produces polygons with edges that are e(uidistant from neighbouring sites. 3oronoi diagrams are efficient in finding coverage holes. !n >'ahboubi 9?0?@ and >'ahboubi 9?00@, multiplicatively weighted 3oronoi "'W&3oronoi$ diagram is used to partition the sensing field in a network of mobile sensors with non&identical sensing ranges. 4ifferent deployment strategies are subse(uently introduced to improve sensing coverage of such networks. 3ector&based and 3oronoi&based algorithms are proposed in >Wang 9??=@ to move the sensors in the field in such a way that network coverage increases. !n >Wang 9??D@, basic protocols and virtual movement protocols are introduced for sensor deployment to increase network coverage. 5radient&descent coverage algorithms are presented in >Jwok 9?0?@, which are distributed in the sense of the 4elaunay graph. !n >Cortes 9??<@, coordination algorithms are provided for sensor deployment and coverage, where a class of aggregate obAective functions is also considered based on the geometry of the 3oronoi partitions and pro)imity graphs. The main drawback with the above approaches is that they assume the coverage priority for different points in the field is uniform "i.e., the importance of coverage of every point in the field is the same$. While this is a realistic assumption in many real&world problems, sometimes certain areas in the field have higher priority as far as coverage is concerned, due, for e)ample, to safety considerations. The sensor deployment problem in a non& uniform field is considered in >Cortes 9??<, 7ullo 9??C and Cortes 9?0?@. Iowever, due to the computational comple)ity of the corresponding techni(ues, they may not be suitable for the case when the processing capability of the sensors in the network is limited. >Iamid 'ahboubi 9?0:@ proposed the first approach for coverage of prioritized field with non&identical sensors. !n this work, new distributed deployment strategies were introduced to increase coverage in a network of non&identical mobile sensors with a prescribed priority function for the sensing field. To this end, a priority function is assumed to be given which specifies the coverage priority of different points in the sensing area. The 'W&3oronoi diagram is used to partition the sensing field >4eza 9??C, #kabe 9??? and -eitsma 9??D@. This partitioning is then used to discover

coverage holes and relocate the sensors accordingly to minimize them, while taking into account the coverage priority of different points in the field. The main idea behind their work is to move each sensor iteratively in such a way that its weighted coverage increases. !n an iteration, if the local weighted coverage by none of the sensors is improved by a certain amount, then the algorithm is terminated "to ensure a finite number of iterations$.

3 EPIDEMIC ROUTING ,pidemic -outing supports the eventual delivery of messages to arbitrary destinations with minimal assumptions regarding the underlying topology and connectivity of the underlying network. !n fact, only periodic pair&wise connectivity is re(uired to ensure eventual message delivery. The ,pidemic -outing protocol works as follows. The protocol relies upon the transitive distribution of messages through ad hoc networks, with messages eventually reaching their destination. ,ach host maintains a buffer consisting of messages that it has originated as well as messages that it is buffering on behalf of other hosts. /or efficiency, a hash table inde)es this list of messages, keyed by a uni(ue identifier associated with each message. ,ach host stores a bit vector, called the summary vector that indicates which entries in their local hash tables are set. When two hosts come into communication range of one another, the host with the smaller identifier initiates an anti&entropy session with the host with the larger identifier. To avoid redundant connections, each host maintains a cache of hosts that it has spoken with recently. %nti&entropy is not re&initiated with remote hosts that have been contacted within a configurable time period. 4uring anti&entropy, the two hosts e)change their summary vectors to determine which messages stored remotely have not been seen by the local host. !n turn, each host then re(uests copies of messages that it has not yet seen. The receiving host maintains total autonomy in deciding whether it will accept a message. /igure :.0 depicts the message e)change in the ,pidemic -outing protocol. Iost % comes into contact with Iost 7 and initiates an anti&entropy session.

/igure :.0 The ,- when two hosts, % and 7, come into transmission range of one another. !n step one, % transmits it summary vector, S3% to 7. S3% is a compact representation of all the messages being buffered at %. Ne)t, 7 performs a logical %N4 operation between the negation of its summary vector, KS37, "the negation of 7*s summary vector, representing the messages that it needs$ and S3%. That is, 7 determines the set difference between the messages buffered at % and the messages buffered locally at 7. !t then transmits a vector re(uesting these messages from %. !n step three, % transmits the re(uested messages to 7. This process is repeated transitively when 7 comes into contact with a new neighbour. 5iven sufficient buffer space and time, these anti&entropy sessions guarantee eventual message delivery through such pair& wise message e)change. 4 Sys !" #!s$%& '&# '()*$ !) +(!

/igure ;.0. %rchitecture diagram !n /igure ;.0, all the components employed in C%'W#C algorithm are clearly displayed. The sensors are assumed to follow a s(uare mobility model. Ience the possible alterations in the directions are /#-W%-4 and -,3,-S,. The sensor nodes are also assumed to have a boolean disk coverage model, which is represented as a shaded circular regions. %t any time instant 8t*, when two such regions of different sensors have some area common on their intersection, then those sensors are defined to be neighbouring sensors at that instant 8t*.

/igure ;.9. Component diagram /igure ;.9 illustrates the key components of a mobile sensor. Sensor memory can consist of two data structures, atrol Table, which will be detailed later and #! positions, which is the storehouse of the static positions of the points of interest. %nother important block in the sensor is the decision control block, which tries to take optimal direction control decisions, so as to achieve ma)imum global coverage. 'ost of the previous works in this field use centralized node to direct the monitoring sensors. This approach is least beneficial and scalable when e)tended for highly dynamic and disconnected environment. Iowever, C%'W#C is devoid of such fallacies, as it*s a distributed algorithm with the sensor monitoring information propagating to whole network by e)change of some designated data structure named 8 atrolTable*. This e)change takes place when the sensor comes close to each other. ,ach sensor mutually updates themselves with the latest information from the neighbouring sensors. !n this way, the monitoring information gets disseminated and the sensors are provided with the best latest information. The available information helps a sensor choose an optimal direction to proceed further, so as to continuously increase the network coverage efficiency.

/igure ;.: Topology 4iagram /igure ;.: depicts the general topology diagram of C%'W#C approach. The bus point represents the #!, with each #! i having its own patrol period ti. The mobile device represents the mobile sensor, with its mobility defined by its model "in this approach, every mobile sensor is assumed a s(uare mobility although the algorithm can be readily implemented for any mobility ranging from circular to random mobility. /rom the figure ;.9, it*s clear that each sensor has its own atrol table, having three fields. The fields are #! !d i, which represents the #! sensed, atrol time ti, which depicts the time at which the #! is sensed and Sensor !d Si, that represents the identifier of the sensor, that monitored the sensor Si.

/igure :.; Sensor state diagram The sensor state diagram is diagrammatically represented in figure ;.;. The sensor enters update state in two cases. !n case 0, a sensor finds a #! i in its sensing range, so it either inserts a new entry into the patrol table "where i is not found before, meaning, it has not sensed i or it has not got sensing information pertaining to i from any other neighbouring sensors$ or updates the e)isting entry corresponding to i, with the new sensed time. The sensor S0 can enter e)change state, when it encounters any neighbouring sensor S9 in its sensing range. S0 accounts for a Aoin operation with S9 and vice&versa and all the changes are durable. !n case of entries being common with both S0 and S9 patrol tables, the latest sensed time will be taken into account. The sensor S0 can enter decide state periodically, whenever it tries to take motion decisions. !t checks the atrol tables and levies the most eligible #! needed to monitored ne)t by S0. %ppropriate distance computation and direction computation process is followed, and the sensor is directed to the computed path. Similarly the sensor enters clean&up state, for every m seconds and the out&dated entries "i.e #! entries whose deadline is invalid with respect to the current time$ are removed from the table. The value m is a tunable parameter. When the value of m is too small, then cleanup operation will take the most time and it entails high overhead. #n the other hand, when the m value is too large, then the sensor may have out&dated entries for the most time, which can have negative impact in the performance.

The arrangement of the modules needed for the C%'W#C algorithm can be viewed below.

class S!&s-(L function P!(.-("B!*'/$-("$L setCurrent osition"$. GGStore the current position of the sensor updateNearest #!"$. GG'ark any #!s in the vicinity as MC#3,-,48 'obilityControlBogicNSensor"s(uareNside,speed,initial 4irection$. selectNe)t #!"$. GGselect the most depriving #! checkforNeighbor%nd,)changeTables"$ clean2p"$. GGCleaning out&dated entries. O function I&$ B!*'/$-("$L sensor ositions.add"currentposition$. O O

Exchange table algorithm The main task of the module is to synchronize the patrol tables of the neighbouring sensors. This module gets initiated when there is an upcoming sensor in the sensing range of every sensor. function E0)*'&%!T'12!"Sensor S%, Sensor S7$ L GGSensor S% copies swept information from S7 /oreach" #! i in Sensor S7$ L !f" i is new in S%*s table$

L New,ntry"TableNS%, i,time$ O ,lse if" i already e)ists in the S%*s table$ L GGupdate the swept time of the found #! #ldSweptTimeP,)tractSweptTime"TableNS%, i$ 4elete,ntry"TableNS%, i$ New,ntry"TableNS%, i,time$ O O GGSensor S7 copies swept information from S% /oreach" #! i in Sensor S%$ L !f" i is new in S7*s table$ L New,ntry"TableNS7, i,time$ O ,lse if" i already e)ists in the S7*s table$ L GGupdate the swept time of the found #! #ldSweptTimeP,)tractSweptTime"TableNS7, i$ 4elete,ntry"TableNS7, i$ New,ntry"TableNS7, i,time$ O O

Send %CJ to the neighbouring sensors. O Cleanup operation function clean2p"$L GG7oth sensors perform the clean&up operation to delete the out&dated entries BistQ #!R outdatedentriesPnull. /oreach" #! pi in SensorTable S$ L ,lapsedTimeP i.SweptTime & CurrentTime !f" ,lapsedTime R i. atrol eriod$ #utdatedentries.add" i$ O 4elete all #! in out&dated entries from the patrol table. 3 Next poi selection algorithm The main task of the module is to decide the ne)t #! to serve. The challenge lies in the fact that it*s difficult to prioritize the urgent #!s. This module contains three se(uential cases. %t any time, when any case clicks right, the module e)ecution terminates. The primary case to be checked is to see any uncovered #! 0. This can be successful if we cannot find a #! in the patrol table. !n this case, the obvious #! that is returned is 0. !f the primary check fails, the ne)t check is to find the #! whose deadline is nearing. !f we have many #!s with same deadline, then the #! which is nearer to the parent sensor is chosen. !f the secondary check also fails, then an e)haustive search is started, which shall find a nearby #!, on a hop&by&hop basis. function S!2!) N!0 POI "Sensor S$ L

44CHECK 51 M-s !2$%$12! )'&#$#' ! $s *! POI &- 6(!s!& $& *! s7!!6 '12! foreach" #! p in #!Bist$ if"p is not in atrol Table >S@ $ return p. 44CHECK 52 F$&# *! POI 7*-s! #!'#2$&! $s &!'($&% foreach" #! p in atrol Table >S@ $ ne)t atrolPp.last atrolTime S poi atrol!ntervals>p@. ne)t4eadlinePne)t atrol T currentTime. /ind p with earliest ne)t4eadline if"p e)ist$ return p. 44CHECK 53 E0*'+s $/! s!'()* .-( &!'(1y POIs, $. '22 #!'#2$&!s .-( POIs )-+2# "! "! hopCountP0. 4ma)P'a)imum distance to a #! from current sensor position While" till eligible candidate is found out$ Select #! with minimum distance from current sensor position with nearest threshold defined by 4ma)UhopCount -eturn if such a #! e)ist hopCountSP0. return N#TN/#2N4. O

8 Sys !" $"62!"!& ' $-& C%'W#C is simulated in a Vava :4 robot simulator named 8Simbad*, as it directly supports dynamic movement modules unlike #mnetpp and NS9. This re(uires the following dependencies, simbad&0.;.Aar, which holds all the libraries, classes and interface in the Simbad framework and Aava:d&windows&amd=;, which installs the :4

environment into the system architecture. The simbad simulator is launched on Netbeans, a fully featured Vava !4,, which provides an application framework for Vava desktop applications and others.

C-&s '& 9+'& $ $!s Table <.0 list out all the constant (uantities of the C%'W#C algorithm. N'"! P+(6-s!

&!'(!s S!&s-(T*(!s*-2# % value to determine closeness of two sensors. &!'(!sPOIT*(!s*-2# N--.s!&s-(s N--.6-$s 6-$P' (-2I& !(/'2s:; 6-$P-s$ $-&s:; % value to determine closeness of sensor and #!. Total number of sensors. Total number of pois. %rray with the time&variant patrol re(uirements of each #! %rray with the default locations of #!s. Table <.0. Constant Fuantities V'($'12! 9+'& $ $!s Table <.9 list out all the variable (uantities of sensors in C%'W#C algorithm. N'"! s!&s-(P-s$ $-&s:; 6' (-2T'12!:S ($&%, 6E& (y; 6E& (y:S ($&%, $& , S ($&%; P+(6-s! %rray with the current locations of sensors. %rray for a specific sensor with entries for pois. % row per #! for a sensor, with last atrolTime and sensor id. #! id,

Table <.9. 3ariable Fuantities < P!(.-("'&)! '&'2ys$s The mobile sensor platform is raised on Simbad simulator, which is chosen as it directly allows to define dynamic mobility. /igure <.0 illustrates the sensor environment, deployed with 9; #!s and D sensors, in a world size :?/. The velocity of sensors is assumed to be ;mGs.

/igure =.0 Simbad environment !n figure =.0, the window consist of two panels, one named 8World* denotes the graphical layout of the sensor and the environment. %nother one named 8Control* provides some simulator controls to adAust the camera orientation, e)ecution speed.

/igure =.9 4ecision control algorithm at clock interval 9;:. !n figure =.9, we could see the program state behavior at clock interval time 9:. %t the instant 9:, the sensor employs selectNe)t #! method to choose the ne)t direction "/#-W%-4G -,3,-S,$. The selected #! is 0?, which needs no alteration of direction for sensor < as its current direction -,3,-S, shall cover it optimally.

The analysis of the coordination algorithm in terms of coverage efficiency is done by plotting the average coverage efficiency of all #!s against patrol period. /or a patrol period i, average coverage efficiency is given by e(uation <.0.

/igure =.: The patrol period and global average coverage of all #!s by -%N4 and C%'W#C scheme. "vP ; mGs$ /igure =.: depicts the patrol period versus coverage efficiency graph for different patrol periods, incremented by 0?, upto 09?. Iere the velocity of sensors is assumed to be ; mGs.

/igure =.; The patrol period and global average coverage of all #!s by -%N4 and C%'W#C scheme. "vP = mGs$ /igure =.; depicts the patrol period versus coverage efficiency graph for different patrol periods, incremented by 0?, upto 09?. Iere the velocity of sensors is assumed to be = mGs.

/igure =.< The patrol period and global average coverage of all #!s by -%N4 and C%'W#C scheme. "vP E mGs$ /igure =.< depicts the patrol period versus coverage efficiency graph for different patrol periods, incremented by 0?, upto 09?. Iere the velocity of sensors is assumed to be E mGs. %t the point p0, the -%N4 scheme "CCW$ overtakes the C%'W#C scheme "CE.<EW$. /rom the above discussion, we can learn that C%'W#C scheme is predominantly coverage efficient for smaller patrol period. !n case of larger patrol period "greater than E?$, the largest variation is 9.9DW that of -%N4 scheme.

/igure. =.=. %verage coverage efficiency versus velocity of mobile sensors. /igure =.= shows the velocity versus average coverage efficiency graph for C%'W#C and random schemes. This graph is Aust an aggregate of timing graphs. Thus it*s learnt that our algorithm significantly covers more region than that covered by random schemes. The second F#S parameter to be measured is the coverage delay. Iere itMs done by plotting the patrol intervals and average poi coverage delay for a specific patrol period. %s usual, C%'W#C approach is compared against the straight forward random scheme. The sensor velocities are varied from ;&E mGs, by 9 units and the patrol intervals is assumed to be 0??. /igure =.D and =.E illustrates the delay graph for the patrol period 0??ms and 09?ms respectively. /rom both the figures it*s clear that the average coverage delay is significantly less for C%'W#C scheme.

/igure =.D 4elay estimate "pP0??ms$

/igure =.E 4elay estimate "pP09?ms$

Table =.9 depicts the comparison of the F#S parameters of both the contrasting approaches.
vP;mGs E<.CEW RANDOM vP=mGs vPEmGs E=.00W C?.?;W avg. =>?3>@ vP;mGs EE.<W CAMWOC vP=mGs vPEmGs C0W C<.:W avg. A2@

C-/!('%! E..$)$!&)y D!2'y d.P0??ms d.P09?ms

C.C9ms 09.Dms

C.<ms 0?.;ms

E.E;ms E.:ms

A?4"s 1B?8"s

D.9ms D.9ms

D.Ems D.=ms

E.9ms <ms

>?>"s <?<"s

Table =.9 erformance remarks > CONCLUSION AND FUTURE WORKS atrol coverage is a new concept for sensor network monitoring. Such type of patrol inspection with mobile sensors is an efficient scheme for many environments surveillance applications with specified delay bounds. We define the concept of patrol coverage to model the re(uirements of periodically monitoring a set of #!s in such applications. C%'W#C work is primarily focussed on proposing a less overhead, localized approach adapted from ,- , to disseminate the local information globally. The prime advantage with this proposed approach is high scalability and the fact that the approach enables the mobile sensors to take less delay and less ineffective motion decisions, so as to satisfy the time variant #! re(uirements. Comparative analysis depicted the efficiency and effectiveness of the approach. The latest research is going towards the anchor based localization for the mobile sensors. /uture work in C%'W#C includes the e)tension of the approach to mobile #! and multiple mobile #!s. %lso a relation to determine the minimum number of sensors re(uired given the patrol coverage re(uirements, would be helpful while incorporating this algorithm in real time environment.



%kyildiz, !./, 'elodia, T., and Chowdhury, J.-. "9??D$ 8% survey on wireless multimedia sensor networks*, ,lsevier, Computer Networks, 3ol. <0 No. ;, pp C90&C=?.


%ndrey Savkin, 3., Vaved, /., and %le)ey 'atveev, S. "9?09$ 8#ptimal 4istributed 7lanket Coverage Self&4eployment of 'obile Wireless Sensor Networks*, !,,, Communications Betters, 3ol. 0=, No. =, pp C;C&C<0. 7ai, X., Hun, 6., Xuan, 4., Ten Bai, I., and Via, W. "9?0?$ 8#ptimal atterns for /our&Connectivity and /ull Coverage in Wireless Sensor Networks*, !,,, Transactions on 'obile Computing, 3ol. C, No. : , pp ;:<&;;E. 7artolini, N., Calamoneri, T., Thomas Ba orta, /., and Silvestri, S. "9?00$ 8%utonomous 4eployment of Ieterogeneous 'obile Sensors*, !,,, Transactions on 'obile Computing, 3ol. 0?, No. =, pp D<:&D==. 7atalin, '. "9??9$ 8'ulti&robot 4ynamic Coverage of a lanar 7ounded ,nvironment*, roc. !,,,G-SV !nt*l Conf. !ntelligent -obots and Systems. 7ulusu, N., Ieidemann, V., and ,strin, 4. "9???$ 85 S&less low&cost outdoor localization for very small devices*, !,,, ersonal Comm. 'agazine, 3ol. D, No. <, pp 9E&:;. Chang, C., Sheu, V., Chen, H. and Chang, S.W. "9??C$ 8%n #bstacle&/ree and ower&,fficient 4eployment %lgorithm for Wireless Sensor Networks*, !,,, Transactions on Systems, 'an and Cybernetics& art %+ Systems and Iumans, 3ol. :C, No. ;, pp DC<&E?=. Chang, C.T., Chang, C.H., and Bin, C.H. "9?09$ 8%nchor&5uiding 'echanism for 7eacon&%ssisted Bocalization in Wireless Sensor Networks*, !,,, Sensor Vournal, 3ol. 09, No. <, pp 0?CE&0000. Chen, V., 6hang, l., and Juo, H. "9?0:$ 8Coverage&,nhancing %lgorithm 7ased on #verlap&Sense -atio in Wireless 'ultimedia Sensor Networks*, !,,, Sensor Vournal, No. CC, pp 0<:?&0<:=. Cheng, C.5., and Tsai, J.T. "9?09$ 84istributed 7arrier Coverage in Wireless 3isual Sensor Networks With Y&Fo'*, !,,, Sensor Vournal, 3ol. 09, No. =, pp 0D9=&0D:<. 4an Wang, Viangchuan Biu and 5ian 6hang, "9?0?$ 8#n 'obile Sensor %ssisted /ield Coverage*, %C' Transactions on Computational Bogic, pp. 000&0;?. 4atta, S., Jinowski, C., and Jhale(ue, '.S. "9??=$ 84istributed Bocalization in Static and 'obile Sensor Networks*, roc. Second !nt*l Conf. Wireless and 'obile Computing, Networking and Comm., pp =C&D=.



<. =.








4oherty, l., ister, J.S.V., and 5haoui, B.,l. "9??0$ 8Conve) osition ,stimation in Wireless Sensor Networks*, roc. !,,, !N/#C#' , 3ol. : , pp 0=<<&0==:. ,rdelA, '., -azafindralambo, T., and Simplot&-yl, 4. "9?0:$ 8Covering oints of !nterest with 'obile Sensors*, !,,, Transactions on arallel and 4istributed Systems, 3ol. 9;, No. 0, pp :9&;:. 5iordano, 3., 7allal, #., Bewis, /., Turchiano, 7., and 6hang, V.7. "9??=$ 8Supervisory Control of 'obile Sensor Networks*, !,,, Transactions on Systems, 'an and Cybernetics& art 7 + Cybernetics, 3ol. :=, No. ;, pp E?=& E0C. 5uha, S., 'urty, -., and Sirer, ,.5. "9??<$ 8Se)tant+ % 2nified Node and ,vent Bocalization /ramework 2sing Non&Conve) Constraints*, roc. %C' 'obiIoc, pp 9?<&90=. Ieo, N., and 3arshney, . "9??<$ ,nergy&,fficient 4eployment of !ntelligent 'obile Sensor Networks*, !,,, Transactions on Systems, 'an and Cybernetics& art %+ Systems and Iumans, 3ol. :<, No. 0, ppDE&C9. Iorstmann, C.S. and Cornell, 5. "9??E$ 8Core Vava, 3olume 0Z9&%dvanced /eatures, EG,*. Iu, l., and ,vans, 4. "9??;$ 8Bocalization for 'obile Sensor Networks*, roc. %C' 'obiCom, pp ;<&;D. Iugues, l., and 7redeche, N. 8Simbad rogramming 5uide*, %vailable from http+GGsimbad.sourceforge.netGguide.php . accessed ?0 November 9?0:. Iugues, l., 7redeche, N., and /uturs, T.!. "9??=$ 8Simbad+ an %utonomous -obot Simulation ackage for ,ducation and -esearch*, SpringerMs Becture Notes in Computer Sciences G %rtificial !ntelligence series "BNCSGBN%!$. !yengar, S.S., Wu, I.C., 7alakrishnan, N., and Chang, S.H. "9??D$ 87iologically !nspired Cooperative -outing for Wireless 'obile Sensor Networks*, !,,, Systems Vournal, 3ol. 0, No. 0, pp 9C&:D. Vi Buo, 4an Wang, and Fian 6hang "9?09$ 8#n the 4ouble 'obility roblem for Water Surface Coverage with 'obile Sensor Networks*, !,,, Transactions on arallel and 4istributed Systems, 3ol. 9:, No. 0, pp 0;=&0<C. Jarakaya, '., and Fi, I. "9?09$ 8Coverage ,stimation for Crowded Targets in 3isual Sensor Networks*, %C' Transactions on Sensor Networks, 3ol. E,





0E. 0C. 9?. 90.




No. :, %rticle 9=, pp 9=&;D. 9<. 9=. Jay, -. "9??;$ 8The design space of wireless sensor networks*, !,,, Vournal on Wireless Communications, 3ol.00, No. =, pp <;&=0. Jhatib, #. "0CE<$ 8-eal&time obstacle avoidance for manipulators and mobile robots*, roc. !,,, !nternational Conference on -obotics and %utomation, 3ol. 9, pp <??&<?<. Batimer, 4., Srinivasa, S., Bee&Sue, 3., Sonne, S., Choset, I. and Iurst, %. "9??9$ 8Towards sensor based coverage with robot teams*, roc. %C' 'obiIoc, 3ol. 0, pp C=0&C=D. Bee, V.W., and Bee, V.V. "9?09$ 8%nt&Colony&7ased Scheduling %lgorithm for ,nergy&,fficient Coverage of WSN*, !,,, Sensor Vournal, 3ol. 09, No. 0?, pp :?:=&:?;=. Bee, S., Jim, /., Jim, C., and Jim, J. "9??C$ 8Bocalization with a 'obile 7eacon 7ased on 5eometric Constraints*, !,,, Transactions on Wireless Communications, 3ol. E, No. 09, pp <E?0&<E?<. Bi, X., /rey, I., Santoro, N., and StoAmenovic, !. "9?00$ 8Strictly Bocalized Sensor Self&4eployment for #ptimal /ocused Coverage*, !,,, Transactions on 'obile Computing, 3ol. 0?, No. 00, pp 0<9?&0<::. Bin, C.H., Chang, C.H., and Chang, C.T. "9?09$ 8Tone&7ased Bocalization for 4istinguishing -elative Bocations in Wireless Sensor Networks*, !,,, Sensor Vournal, 3ol. 09, No. <, pp 0?<E&0?D?. Biu, 7., 7rass, ., 4ousse,#., Nain, ., and Towsley, 4. "9??<$ 8'obility improves coverage of sensor networks*, roceedings of the =th %C' international symposium on 'obile ad hoc networking and computing, pp :??&:?E. Biu, 7., 4ousse, #., Nain, ., and Towsley, 4. "9?0:$ 84ynamic Coverage of 'obile Sensor Networks*, !,,, Transactions on arallel and 4istributed Systems, 3ol. 9;, No. 9, pp :?0&:00. Biu, I., Chu, X., Beung, H.W., and 4u, -. "9?0?$ 8Simple movement control algorithm for bi&connectivity in robotic sensor networks*, !,,, Vournal on Selected %reas in Communication, 3ol. 9E, No. D, pp CC;&0??<. 'a, J., 6hang, H., and Trappe., W. "9??E$ 8'anaging the 'obility of a 'obile Sensor Network 2sing Network 4ynamics*, !,,, Transactions on










arallel and 4istributed Systems, 3ol. 0C, No. 0, pp 0?=&09?. :=. 'a, '., and Hang, H. "9??D$ 8%daptive Triangular 4eployment %lgorithm for 2nattended 'obile Sensor Networks*, !,,, Transactions on Computers, 3ol. <=, No. D, pp C;=&C<E. 'ahboubi, I., Iabibi, V., %mir %ghdam, 5., and Sayrafian our, J. "9?0:$ 84istributed 4eployment Strategies for !mproved Coverage in a Network of 'obile Sensors With rioritized Sensing /ield*, !,,, Transactions on !ndustrial !nformatics, 3ol. C, No. 0, pp ;<0&;=0. Nishiyama, I., Ngo, T., %nsari, N., and Jato, N. "9?09$ 8#n 'inimizing the !mpact of 'obility on Topology Control in 'obile %d Ioc Networks*, !,,, Transactions on Wireless Communications, 3ol. 00, No. :, pp 00<E&00==. #u, C.I. "9?00$ 8% Bocalization Scheme for Wireless Sensor Networks 2sing 'obile %nchors With 4irectional %ntennas*, !,,, Sensor Vournal, 3ol. 00, No. D, pp 0=?D&0=0=. #u, C.I., and Ie, W.B. "9?0:$ 8 ath lanning %lgorithm for 'obile %nchor& 7ased Bocalization in Wireless Sensor Networks*, !,,, Sensor Vournal, 3ol. 0:, No. 9, pp ;==&;D<. oduri, S., and Sukhatme, 5. "9??;$ 8Constrained coverage for mobile sensor networks*, roc. !C-% M?; !,,, !nternational Conference on -obotics and %utomation, 3ol. 0, pp 0=<&0D0. oduri, S., attem, S., Jrishnamachari, 7., and Sukhatme, 5.S. "9??C$ 82sing Bocal 5eometry for Tunable Topology Control in Sensor Networks*, !,,, Transactions on 'obile Computing, 3ol. E, No. 9, pp 90E&9:?. -azafindralambo,T., and -yl, 4.S. "9?00$ 8Connectivity reservation and Coverage Schemes for Wireless Sensor Networks*, !,,, Transaction on %utomatic Control, 3ol. <=, No. 0?, pp 9;0E&9;9E. Shen, 6., Chang, H., Viang, V, Wang, H., and Han, 6. "9?0?$ 8% 5eneric /ramework for #ptimal 'obile Sensor -edeployment*, !,,, Transactions on 3ehicular Technology, 3ol. <C, No. E, pp ;?;:&;?<D. Sheu, V. ., Chen .C., and Su C.S. "9??E$ 8% 4istributed Bocalization Scheme for Wireless Sensor Networks with !mproved 5rid&Scan and 3ector&7ased -efinement*, !,,, Transactions on 'obile Computing, 3ol. D, No. C, pp 000?&009:.











Sheu, V. ., Iu, W.J., and Bin, V.C. "9?0?$ 84istributed Bocalization Scheme for 'obile Sensor Networks*, !,,, Transactions on 'obile Computing, 3ol. C, No. ;, pp <0=&<9=. Sheu, V. ., Bi, V.'., and Isu, C.S. "9??=$ 8% 4istributed Bocation ,stimating %lgorithm for Wireless Sensor Networks*, roc. !,,, !nt*l Conf. Sensor Networks, 2bi(uitous, and Trustworthy Computing, 3ol. 0, pp 90E&99<. Shiu, !.C., Bee, C.H. and Tang, C.S. "9?00$ 8The 4ivide&and&Con(uer 4eployment %lgorithm 7ased on Triangles for Wireless Sensor Networks* ,!,,, Sensor Vournal, 3ol. 00, No. :, pp DE0&DC?. Sichitiu, '.l., and -amadurai, 3. "9??;$ 8Bocalization of Wireless Sensor Networks with a 'obile 7eacon*, roc. !,,, !nt*l Conf. 'obile %d&Ioc and Sensor Systems "'%SS$, pp 0D;&0E:. Song, ., and Jumar, 3. "9??9$ 8% otential /ield 7ased %pproach to 'ulti& -obot 'anipulation*, roc. !C-% M?9 !,,, !nternational Conference on -obotics and %utomation, pp 090D&0999. Ssu, J./., #u, C.I., and Viau, I.C. "9??<$ 8Bocalization with mobile anchor points in wireless sensor networks*, !,,, Transactions on 3ehicular Technology, 3ol. <;, No. :, pp 00ED&00CD. Tan, 5. "9??C$ 8Connectivity&5uaranteed and #bstacle&%daptive 4eployment Schemes for 'obile Sensor Networks*, !,,, Transactions on 'obile Computing, 3ol. E, No. =, pp E:=&E;E. Teddy Cheng, '., and %ndrey Savkin, 3. "9??C$ 8% 4istributed Self& 4eployment %lgorithm for the Coverage of 'obile Wireless Sensor Networks*, !,,, Communications Betters, 3ol. 0:, No. 00, pp EDD&EDC. Tseng, H.C., Chen, .H., and Chen, W.T. "9?09$ 8k&%ngle #bAect Coverage roblem in a Wireless Sensor Network*, !,,, Sensor Vournal, 3ol. 09, No. 09, pp :;?E&:;0=. 3ahdat, %., and 7ecker, 4. "9???$ 8,pidemic -outing for artially Connected %d Ioc Networks*, Technical -eport CS&9????=, 4uke 2niversity. 3ikram 'unishwar, ., and Nael %bu&5hazaleh, 7. "9?0:$ 8Coverage algorithms for visual sensor networks*, %C' Transactions on Sensor Networks. 3oulgaris, S., and 'aarten 3an Steen "9??:$ 8%n ,pidemic rotocol for









<<. <=.


'anaging -outing Tables in 3ery Barge eer&to& eer Networks*, Self& 'anaging 4istributed Systems, Springer 7erlin Ieidelberg, pp 9E=D,;0&<;. <E. <C. Wang, 7. "9?00$ 8Coverage roblems in Sensor Networks % Survey*, %C' Computing Surveys, 3ol. ;:, No. ;, %rticle :9, pp :9&E;. Wang, 5., Cao, 5., 7erman, ., and Thomas Ba orta, /. "9??D$ 87idding rotocols for 4eploying 'obile Sensors*, !,,, Transactions on 'obile Computing, 3ol. =, No. <, pp <=:&<D=. Wang, W., Srinivasan, 3., and Chua, J.C. "9??D$ 8Trade&offs between mobility and density for coverage in wireless sensor networks*, roceedings of the 0:th annual %C' international conference on 'obile computing and networking, pp :C&<?. Wang, W., Srinivasan, 3., and Chua, J.C. "9??E$ 8Coverage in Iybrid 'obile Sensor Networks*, !,,, Transactions on 'obile Computing, 3ol. D, No. 00, pp 0:D;&0:ED. Wang, H., eng, W., and Tseng, H. "9?0?$ 8,nergy&7alanced 4ispatch of 'obile Sensors in a Iybrid Wireless Sensor Network*, !,,, Transactions on arallel and 4istributed Systems, 3ol. 90, No. 09, pp 0E:=&0E<?. Wilson, V.S. "9??<$ 8Sensor Technology Iandbook*, ,lsevier. Xu, ,., 4ing, 6., and 4asgupta, S. "9?0:$ 8Target Tracking and 'obile Sensor Navigation in Wireless Sensor Networks*, !,,, Transactions on 'obile Computing, 3ol. 09, No. 0, pp 0DD&0E=. Hang, S., Bi, '., and Wu, V. "9??D$ 8Scan&7ased 'ovement&%ssisted Sensor 4eployment 'ethods in Wireless Sensor Networks*, !,,, Transactions on arallel and 4istributed Systems, 3ol. 0E, No. E, pp 00?E&0090. 6avlanos, '., and appas, 5. "9??D$ 8 otential fields for maintaining connectivity of mobile networks*, !,,, Transactions on -obotics, 3ol. 9:, No. ;, pp E09&E0=. 6avlanos, '.', and appas, 5.V. "9??E$ 84ynamic %ssignment in 4istributed 'otion lanning With Bocal Coordination*, !,,, Transactions on -obotics, 3ol. 9;, No. 0, pp 9:9&9;9. 6avlanos, '.', -ibeiro, %. and appas, 5.V. "9?0:$ 8Network !ntegrity in 'obile -obotic Networks*, !,,, Transaction on %utomatic Control, 3ol. <E. 6hang X. "9?00$ 8%daptive Control and -econfiguration of 'obile Wireless




=:. =;.




=E. =C.

Sensor Networks for 4ynamic 'ulti&Target Tracking*, !,,, Transactions on %utomatic Control, 3ol. <=, No. 0?, pp 9;9C&9;;;. D?. 6hang, S., Cao, V., BiAun, C., and 4ao)u, C. "9?0?$ 8%ccurate and ,nergy& ,fficient -ange&/ree Bocalization for 'obile Sensor Networks*, !,,, Transactions on 'obile Computing, 3ol. C, No. =, pp ECD&C0?. 6hong, 6., and Ie, T. "9?09$ 8Sensor Node Bocalization with 2ncontrolled ,vents*, %C' Transactions on ,mbedded Computing Systems, 3ol. 00, No. :, %rticle =<, pp =<&C?. 6hu, H., 7ao, H., and Bi, 7. "9?09$ 8#n 'a)imizing 4elay&Constrained Coverage of 2rban 3ehicular Networks*, !,,, Vournal on Selected %reas in Communication, 3ol. :?, No. ;, pp E?;&E0D. 6ou, H., and Chakrabarty, J. "9??D$ 84istributed 'obility 'anagement for Target Tracking in 'obile Sensor Networks*, !,,, Transactions on 'obile Computing, 3ol. =, No. E, pp ED9&EED.