ENERGY EFFICIENCY ENHANCING TECHNIQUES IN WIRELESS SENSOR NETWORKS
Collection of Ph.D. Theses Zoltán Vincze by
Research Supervisor:
Rolland Vida, Ph.D.
Department of Telecommunications and Media Informatics
Budapest, Hungary, 2008
1 Introduction
Wireless sensor networks (WSN) represent one of the hottest current research topics, both because of the ever-widening range of related attractive applications, and the numerous challenging open issues that still have to be solved for these applications to operate reliably and eciently. These special wireless networks consist of tiny, low-cost sensor devices with limited resources (memory, processing power, energy); their usual task is to monitor some physical phenomenon and to send information about it to one, or several dedicated sink nodes. The sink can gather data from the network in three ways. In case of time-driven operation each sensor sends data periodically, while in case of event-driven operation only those sensors report to the sink which have sensed an event. The third approach is the query-driven operation, where the sensors send data to the sink only upon receiving a query.
There is a broad variety of sensor applications. The initial driving force for develop- ing WSNs were the military applications (e.g., intelligent mineeld [1], sniper localization [2]). However, civil applications (e.g., habitat monitoring[3], environmental monitoring[4]) have also got nowadays in the focus of ongoing research. In these scenarios the WSNs are usually deployed randomly in a wide area, and the network elements form a self-organizing ad-hoc multi-hop network. This makes it necessary to use ad-hoc communication algo- rithms; however, the existing solutions are not applicable directly because of some major dierences between WSNs and traditional ad-hoc networks. Thus, new communication al- gorithms have to be proposed, tailored to the special features and requirements of wireless sensor networks.
One of the most important issues to be aware of when handling wireless sensors is their energy consumption. Sensors are deployed in huge number to hardly accessible terrains;
thus their batteries are often impossible or impractical to replace or recharge. Therefore, energy should be spared, so as to be able to perform, for the longest possible period, the role they were deployed for. Energy is consumed by several tasks, but the most important energy consumer is the communication module. In small networks, sensors can send data directly to the sink node; in larger setups multi-hop communication is needed, i.e., sensors forward each other's data toward the sink. In both cases the energy consumption depends on the communication distance.
There were many proposals recently targeting energy eciency. Some approaches focused on energy conserving routing techniques, i.e., nding optimal routes in terms of consumed power, and balancing the energy consumption among all nodes [5, 6, 7, 8].
Others were based on topology control schemes, i.e., deploying sensor and sink nodes in an ecient way or reshaping the topology through dynamic power control of the participating sensors [9, 10, 11, 12, 13]. Clustering techniques are also widely employed; the network is divided into small clusters, a cluster head being responsible for aggregating and relaying towards the sink the information gathered from the sensors of its cluster [14, 15].
Since the energy demand of the communication depends on the communication dis- tance, it is straightforward to enhance energy eciency of network operation by reducing these distances. This can be done in many ways. One possibility is to deploy multiple sinks; this decreases the average number of hops a message has to pass through before being received and processed by a sink, as data will always be sent to the closest sink.
There are many solutions that build on this idea; however, they usually need global net- work information for the deployment, that is hard or impossible to assume in wide-area networks. Thus, in Thesis 1 I propose a novel local information-based multiple sink de- ployment algorithm.
Another solution is to have a mobile sink that during sensor data collection moves inside the monitored region, so as to reduce the communication distances. This solution has also the benet of distributing the high trac load in the neighborhood of the sink, caused by the multi-hop communication, among all the sensors in the network. In Thesis 2 I propose a novel adaptive sink mobility algorithm that assumes event-driven network operation, as opposed to existing solutions that address only the time-driven scenario.
2 Research objectives
The nodes in a wireless sensor network have scarce energy supplies that are very expensive or impossible to reload or replace. Thus, energy ecient operation has crucial importance.
Consequently, the goal of Thesis 1 was to deploy multiple sinks in time-driven wireless sensor networks in an energy ecient manner, prolonging the lifetime of the network. The existing solutions in the area proposed integer linear programming [16, 17, 18, 19, 20, 21], ow-based [22] or iterative clustering methods [23] to solve the deployment problem. Nev- ertheless, they have serious shortcomings, such as the lack of scalability or the requirement of having a priori global network information. As opposed to these, in Thesis 1 I propose a scalable local information-based solution that yields to a signicant reduction of the energy consumption.
Energy ecient operation and network lifetime elongation through the use of mobile sink nodes is also an approach that was addressed in numerous papers. The proposed so- lutions can be categorized in three groups: random [24, 25, 26, 27], predictable [28, 29], or controlled [30, 31, 32] mobility. Nevertheless, they always assumed that the network oper- ates in a time-driven manner, without considering event-driven wireless sensor networks.
There are however important applications (e.g., intrusion detection, seismic activity mon- itoring) that need event-driven operation. The ecient utilization of the energy reserves is very important in the case of these networks too. Thus, the objective of Thesis 2 was to propose adaptive sink mobility strategies for event-driven wireless sensor networks.
3 Methodology
In the following theses I rst formulate the sink deployment and the sink mobility issues as a mathematical problem, and present some numerical results about the optimal sink locations in time-driven networks, as well as total energy consumption and maximum trac load in event-driven networks. Based on these results, I propose a novel sink deployment algorithm and two sink mobility strategies. I also investigate the eciency and the applicability of the novel solutions through extensive simulations.
4 New Results
4.1 Deployment of Sink Nodes in Wireless Sensor Networks
In wireless sensor networks the ecient utilization of the energy reserves has crucial importance. The sensors spend most of their energy for the communication with the sink;
therefore, it is important to reduce the energy needs of the communication. This can be achieved by shortening the distances the packets have to travel, since the energy cost of transmitting a packet is proportional to the distance of the transmission. Deploying multiple sinks, and making each sensor communicate with the closest sink can reduce the communication distances very eciently.
T
HESES 1: [C1, C2, NC1] I proposed dierent strategies to deploy sink nodes in an energy ecient way in time-driven multi-hop wireless sensor networks.In my investigations I assumed a time-driven multi-hop wireless sensor network where the nodes do not use radio power adjustment and have a uniform communication range, rc. In my analytical model I considered a network with N sensor nodes and K sinks, each sensor communicating only with the closest sink. Since there is no radio power adjustment, sending a packet to one hop distance costs each sensor the same amount of energy, Ehop . Forwarding a packet to distance d has a cost of rdc ·Ehop; thus, the energy used for the communication in the network is minimized if the overall distance the packets have to travel is minimized. In order to achieve this, the sinks have to be deployed so as to minimize the average distance between the sensors and the closest sink.
First, by means of analytical investigations I determined the optimal locations of the sinks so as to minimize the communication energy cost in the network. Based on these results, I proposed then two sink deployment algorithms. The rst one uses global information about the network. However, in wide area sensor networks global network in- formation is usually not available. Therefore, I also proposed a sink deployment algorithm that is based solely on local information.
I will use the following notations in the theses: let sk =¡ xk, yk¢
denote the location of the kth sink and (xi, yi) denote the location of the ith sensor. The distance vector from the kth sink to the ith sensor is d(k)i = (xk−xi, yk−yi) and the distance between them is dki. The unit vector pointing from the kth sink towards theith sensor is e(k)i =d(k)i /dki and routeik is the set of the indices of the intermediate sensors on the route from the ith sensor to the kth sink. Ik is a boolean indicator function that has a value of 1 only if d(k)i = minld(l)i .
T
HESIS 1.1: [C2, NC1] By means of analytical investigations, I determined the optimal locations where sink nodes have to be deployed in a time-driven multi-hop wireless sensor network so as to minimize the energy spent for communication. In my network model this is equivalent with minimizing the average distance between the sensors and the closest sink that is obtained at the sink locations s1, . . . ,sK whereXN
i=1
Ike(k)i = 0, k = 1, . . . , K. (4.1) Thus, the optimal sink locations are determined not by the distance between the sinks
and the sensors, but by the orientation of the unit vectors pointing from the sink nodes towards the sensors that report to them. The solution can be applied for any convex-shaped area covered by arbitrarily distributed sensors.
T
HESIS 1.2: [C2, NC1] I proposed an iterative strategy called Global to deploy sink nodes in a time-driven multi-hop wireless sensor network to positions that approximate the optimal locations determined in Thesis 1.1. The algorithm uses global information about the locations of the sensors and the sinks. The owchart of the algorithm's operation is shown by Figure 1.First, the network is clustered based on the initial sink locations, each sensor belonging to the cluster of the closest sink. Next, the DETERMINE_CENTROIDS procedure is called iteratively, until each sink moves to the place inside its cluster where the resultant vector Rk is 0. The input parameter max_step denes the maximal possible step for a sink in one round. When all the sinks have reached the desired locations, the network is clustered again, based on the new sink locations. If the clusters do not change, then the algorithm nishes the sink deployment. Otherwise, the DETERMINE_CENTROIDS procedure is called again. The simulations have shown that the algorithm converges in a few iteration rounds.
The Global algorithm uses global information about the network. However, in wide- area wireless sensor networks usually it is impractical to assume such global knowledge, since it may either be impossible to gather those information, or it would introduce a too high overhead. Therefore, I proposed a local information based solution to eliminate the need for global knowledge.
T
HESIS 1.3: [C2, NC1] I proposed an iterative deployment strategy called 1_hop that approximates the optimal sink locations determined in Thesis 1.1 based only on "one- hop knowledge". The sink nodes have information only about the sensors inside their communication range - they know their orientation, and the number of remote sensors that use them as a last-hop node when talking to the sink. The orientations of the remote sensors are approximated using these information. The algorithm can work with any least cost routing solution. The owchart of the algorithm's operation is shown by Figure 2.This algorithm uses only local information for the sink deployment, since the sinks know only the orientation vectors of the sensors within one hop distance, and the number of remote sensors that use those sensors as last-hop node during transmission. First, the sinks determine the orientation vectors of the sensors within one hop distance. Then, they wait for a time period t, long enough to ensure that each sensor transmits during that period at least one packet to the sink. Using the source eld of the received packets, the sinks determine how many distant sensors forward data through each of the neighboring nodes. The sinks do not know the orientation vectors corresponding to distant nodes;
these are thus approximated by the orientation vector of the neighboring sensor that is used as the last-hop node in the communication. After calculating the resultant vector of these orientation vectors, each sink moves in the direction of the resultant vector;
the maximal possible step for a sink is given by the max_step input parameter. The iteration ends if the size of each sink's resultant vector is below a given threshold, set by thethresholdinput parameter. The simulations have shown that the algorithm converges
AlgorithmGlobal(number of sinks,max step)
Start
w=0, z=0, K=number of sinks;
Initializing sink locations: sw1, . . . , swK; Determining the initial clusters:
Ckz={i:d(k)i = minld(l)i sw1,...,swK
, i∈N}, k= 1. . . K;
proceduredetermine centroids fork=1toK
Rk= P
i∈Cz k
e(ik)
#Ckz
sw
1,...,swk
if |Rk|== 0 then sw+1k =swk
else sw+1k =swk +Rk·max step endfor
w=w+1 end
∃k∈ {1, . . . , K}:|Rk| 6= 0 Yes
Recalculating clusters using the new sink locations:
Ckz+1={i:d(k)i = minld(l)i |sw1,...,swK, i∈N}, k= 1. . . K.
z = z + 1
No
∃k∈ {1, . . . , K}:Ckz6=Ckz−1 Yes
No End
Figure 1: The owchart of the operation of the Global algorithm.
Algorithm1 hop(number of sinks,max step, threshold)
Start
w = 0, K = number of sinks, r c = communication range of a sensor;
Initializing sink locations: sw1, . . . , swK;
proceduresink relocation fork= 1 toK
Hk={i:d(k)i < rc|sw1,...,swK, i∈N};
Determiningei,∀i∈Hk; endfor
Collecting the messages for a time periodt;
fork= 1 toK foreachi∈Hk
nri= #{i∈routejk,∀j:d(k)j = minld(l)j |sw1,...,swk};
end Rk=
P
i∈Hke(
k) i ·nri
P
i∈Hknri ; if |Rk|< threshold
then sw+1k =swk;
else sw+1k =swk +Rk·max step;
endfor w =w+ 1;
end
∀k∈K:|Rk|< threshold? No
Yes End
Figure 2: The owchart of the operation of the 1_hop algorithm.
1 e1
2
e2 3
e3
4
e4
5
e5
6 e
6
7 e
7
8
e8 e 9
R 9
(a)
1 e1 2
3 e3
4 5
6
7 8
9 e9
R
(b)
Figure 3: Computing the resultant vectors in case of the Global (a) and the 1_hop (b) algorithm.
in a few iteration rounds.
The dierence in the calculation of the resultant vectors in case of the two algorithms is presented in Figure 3.
T
HESIS 1.4: [C2, NC1] By means of extensive simulations, I have shown that deploying the sinks in dense, homogeneous networks using the 1_hop strategy (Thesis 1.3) results in most of the simulation runs in less than 5% longer average route length than in case of the global information based iterative strategy (Thesis 1.2), that at its turn approximates the theoretically optimal solution (Thesis 1.1).In the simulations I deployed uniformly but randomly 1000sensors in a 800m×800m sized area. The communication range of the sensors was set to 80m. The max_step parameter was set to 400m, while the threshold parameter was set to 0.15. I ran the algorithms for 50dierent topologies, with2,3and 4sink nodes. The sinks were initially deployed in the same random places for the two methods. At the end of a run the average distance between the sensors and the closest sink nodes was calculated for both algorithms.
The x axis of Figure 4 shows the relative dierence of the achieved average distance by the 1_hop and the Global algorithms, while the y axis indicates the number of simulation runs corresponding to that result. It can be seen that in most of the cases the operation of the 1_hop algorithm results in less than 5% longer average route length than the operation of the Global approach. Thus, 1_hop can be used eciently for energy ecient sink deployment in network setups where global information is not available about the network topology.
T
HESIS 1.5: [C1] Assuming optimal shortest path routing in the network, I have shown that the 1_hop algorithm might result in signicantly shorter average communication dis- tances than the Global algorithm in network scenarios where, due to environmental eects, the communication paths are signicantly dierent from the line-of-sight line between the sinks and the sensors. In certain scenarios the average communication paths can be of up to 50% shorter.0 10 20 30 0
10 20
2 sinks
relative difference (%)
frequency
0 10 20 30
0 10 20
3 sinks
relative difference (%)
frequency
0 10 20 30
0 10 20
4 sinks
relative difference (%)
frequency
Figure 4: Histogram of the relative dierences in the achieved average route length be- tween the 1_hop and the Global algorithms.
In Thesis 1.1 and 1.2 I determined the optimal locations of the sinks in a network model where the communication paths follow closely the line-of-sight line between the sinks and the sensors. However, this assumption may fail because of several reasons: obstacles can obstruct direct communication, sparse network areas, depleted regions or regions of an irregular, concave shape can result in routing paths diverging from the line-of-sight. In such cases the Global algorithm is not guaranteed anymore to nd the optimal locations.
However, although the 1_hop algorithm only approximates the orientation of most of the sensor nodes, it results in shorter communication distances than Global. This is due to the fact that 1_hop takes into account the real orientation of the routes in the network, while Global does not.
T
HESIS 1.6: [C2, NC1] I showed that the sink deployment strategy presented in Thesis 1.3 can be used for adaptively moving multiple sinks in time-driven multi-hop wireless sensor networks not only during the deployment, but also during the operation phase.The decreasing energy level of some sensors, and the complete depletion of others, trigger load balancing routing algorithms to change the routes in the network. Thus, I proposed to adaptively move the sink nodes, based on the 1_hop algorithm, so as to minimize the average length of the newly established routes between the sensors and the closest sink. I called this strategy 1_hop relocation. By means of extensive simulations I have shown that this solution results in more than 100% longer network lifetime compared to the case of deploying the sink nodes statically.
For the simulations, I deployed 1000 sensors uniformly but randomly in a square- shaped, 800m × 800m sized network. The communication range of the sensors was 80m, they used multi-hop communication, and the messages were sent using the GOAFR (Greedy Other Adaptive Face Routing) [33] routing algorithm. Every sensor had initially 10000units of energy, and sending or receiving a packet consumed 1 unit of energy. The threshold parameter of 1_hop relocation was set to 0.15, the max_step parameter was 400m. The network was time-driven, i.e. in every round every sensor sent a packet to the closest sink. In the network I considered an ideal MAC protocol that results in very low packet drop rates; thus, I neglected packet retransmissions during the simulations.
The lifetime of the network was dened as the time until an operating sensor could not send its message anymore to the closest sink. Since the robust geographic routing used in
2sinks 3sinks 4sinks 0
500 1000
Lifetime(round)
1hop relocation random global fix global relocation
Figure 5: Network lifetime for the dierent strategies.
the simulation is able to get around holes in the network, the impossibility of delivering a message would necessarily mean that the sink became isolated from the operating sensors.
I ran the simulations for 50dierent topologies with 2,3 and 4 sinks.
Figure 5 shows the achieved network lifetime for the dierent strategies. In case of the random strategy the sinks were moving randomly, in case of global x the sink nodes were deployed statically using the Global algorithm, while in case of the global relocation and 1_hop relocation strategies the sinks were continuously moving based on the Global and the 1_hop algorithms. It can be seen that 1_hop relocation severely outperforms even the global relocation strategy. This is due to the fact that Global calculates the resultant vectors based solely on the locations of the sensors, while 1_hop can also take into account the orientations of the actual communication routes, through the approximation of the distant sensors' orientation. The homogenous spatial distribution of sensor depletion that can be usually observed around the sink nodes in a time-driven multi-hop wireless sensor network does not change the resultant vectors signicantly in case of the Global algorithm.
As opposed to this, in case of 1_hop the altering of the communication routes, caused by the depletions, will change the resultant vectors signicantly enough to make the sinks move to better locations.
4.2 Mobile Sink in Event-driven Multi-hop Wireless Sensor Net- works
T
HESES 2: [J1, J2, C3, C5, C3, C7, NC2] I proposed two sink mobility strategies for event-driven multi-hop wireless sensor networks that enhance energy eciency, and thus signicantly prolong network lifetime.In an event-driven multi-hop network the sensor nodes close to the sink suer from a high trac load, since they have to forward the packets of all the nodes that sense an event. This uneven load leads to the early depletion of those sensors, isolating the sink and ruining network operation. The depletion of only a small part of the sensors can thus
Z r0
∆ϕ
d A0
A1
h h
S
Figure 6: Approximating the maximal trac load.
render the WSN unoperational, even if other parts of the network still possess signicant energy reserves. This negative eect can be avoided by the dynamic relocation of the sink;
by doing so, the high trac load is distributed among all the sensors, and the network lifetime is prolonged. Moreover, since the energy demand of a packet transmission is proportional to the distance the packet has to travel, moving the sink closer to the sources of the packets results in energy savings that can also prolong network lifetime.
In order to propose ecient sink mobility strategies I rst investigated the trac load in the network. I determined the location of the most heavily loaded sensor and based on that I recommended two sink mobility strategies for the elongation of network operation time in case of event-driven WSNs. One of these startegies minimizes the overall energy consumption of the network, while the other one minimizes the maximum energy consumption of the most heavily loaded sensor.
T
HESIS 2.1: [J1, C3, C5] By means of analytical investigations, I have shown that the sensors suering from the highest trac load in a densely deployed, event-driven, multi- hop wireless sensor network with one sink are located near the sink, in the direction of the farthest current event.During the investigations I assumed a densely deployed and strongly connected multi- hop WSN. The network was built of N sensor nodes deployed uniformly in an area A, the sink S being located at coordinates (xS, yS). During the operation of the network an evolving event (z) inside the sensing range (r0) of a sensor activates that sensor, and the node starts sending data packets to the sink. Due to the multi-hop nature of the communication, one sensor can be requested to forward much trac even if it is far from any events to be sensed.
The energy requirement of the most loaded sensor node (Emax) can be approximated as follows. Sensors that are only one hop (h) away from the sink towards the event location (i.e., within areaA0 on Fig.6) have to forward packets generated within the sensing range of the event (i.e., inA1). Thus, the load on the last hop nodes is proportional to the ratio of A1/A0, i.e.,
Emax = A1
A0Ehop= 2dr0
h2 Ehop, (4.2)
where
A0 = ∆ϕ 2π
£(h+h/2)2−(h−h/2)2¤
π = ∆ϕh2, (4.3) A1 ≈ ∆ϕ
2π
£(d+r0)2−(d−r0)2¤
π= 2∆ϕdr0. (4.4)
Thus, Emax is a linear function of the distance dbetween the sink and the event location.
Let us assume now that there are more than one event at a time. In this case, using (4.2) we can identify for each event zi (i= 1, . . . , I) the most heavily loaded sensor with energy requirement Emaxi . By comparing these highly loaded sensors on the sensor eld, we get the highest energy requirement by
EmaxSN = max
1≤i≤IEmaxi . (4.5)
This energy load is on the sensor that is close to the sink and in the direction towards the most distant event zj, that is given by
j = arg max
1≤i≤Id(zi, S). (4.6)
Note, that here I neglected the fact that one sensor could take part in relaying messages of more than one event at a time. However, since the most loaded sensors are on the line between the event and the sink, I basically neglect only the case when there are two or more events directly behind each other.
After determining, through analytical investigations, the distribution of the trac load in an event-driven WSN, I proposed two sink mobility strategies. The goal of these strategies was to enhance the eciency of the network energy utilization, thus prolonging network lifetime. In order to simplify notation, let di denote the distance between the sink node and event zi, i.e, di =d(zi, S)
T
HESIS 2.2: [J1, J2, C3, C5, C7, NC2] I proposed two sink moving strategies that take into account the location of the current events, in order to optimize the energy utilization of the sensors based on dierent aspects.The rst strategy (called minmax), based on the results of Thesis 2.1, minimizes the highest trac load in the network, which is equivalent with minimizing the maximum event distance from the sink, i.e.,
1≤i≤Imaxdi →min. (4.7)
The second one (called mintotal) minimizes the overall energy utilization of the net- work, which is equivalent with minimizing the sum of the event distances from the sink, i.e.,
XI
i=1
max(h/2, di)→min, (4.8)
The evolution of events can lead to network setups where the trac load of the sensors is rather uneven, leading to the early depletion of the heavily loaded sensors. In order to avoid this problem, the transmission energy for the most heavily loaded sensor in the network is minimized in case of the minmax strategy. Hence, energy consumption will be more balanced. As the maximal trac load depends on the biggest event distance from the sink node (see (4.2) and (4.5)), this strategy is equivalent with minimizing the maximum event distance from the sink, i.e.,
1≤i≤Imaxdi →min. (4.9)
This minimization task is equivalent to the Minimal Enclosing Circle Problem, where the task is to nd the minimum radius circle that encloses all points of a point set on the plane. There are several algorithms to solve this problem. For example, it has been shown that it can be solved inO(n)time using the prune-and-search techniques for linear programming [34].
As opposed to this, the idea of the mintotal strategy is to place the sink node so as to minimize the overall energy consumption of the network.
Let Ehop be the energy needed to pass a packet at one hop distance (h), Ni be the number of sensors activated by an event zi, and ki be the average hop number necessary to send an alert about the event to the sink. The total energy needed to report event zi is
Etotali =NikiEhop. (4.10)
Since the average hop number can be well approximated by ki = ddi/he, the energy requirement of reporting an event is proportional to the event distance di. Raking now into account all theI events inside the sensor eld, the energy requirement of the whole network is given by
EtotalSN = XI
i=1
Etotali . (4.11)
Since the energy requirement of reporting an event is proportional to the event distance di (see (4.10)), this is equivalent to minimize the sum of event distances, i.e.,
XI
i=1
max(h/2, di)→min, (4.12) where the maximum means that there is no gain when moving closer to a particular event than the half of the hop length. Practically, this is the location that gives the minimal average distance from the events. There is no closed formula to nd this location, but the problem can be solved numerically.
As opposed to event-driven networks, in a time-driven WSN sensors do not report to the sink only when sensing an event, but they do so periodically. Luo and Hubaux have shown that in case of time-driven multi-hop wireless sensor networks the optimal strategy is to move the sink on the periphery of the network [29].
T
HESIS 2.3: [J1, C5, NC2] By means of extensive simulations, I have shown that in case of an event-driven wireless sensor network, with events moving according to a correlated random walk, the strategies proposed in Thesis 2.2 (mintotal, minmax) achieve 20% longer network lifetime compared to Luo's solution (circular). They also prolong the operation of the network compared to the case of a static (x), or a randomly moving sink (rwp).In the simulations I assumed that the covered region is a circular area of radius R = 1000m, in which I randomly distributed 10.000 sensors, using a uniform distribution model. The maximum communication range of each sensor was xed to 80m. At the beginning of a simulation run each sensor was loaded with 1000 units of energy. The
fix circular minmax mintotal rwp 0
500 1000 1500 2000
time(round)
Figure 7: Average network lifetime.
cost of receiving a packet was 1 unit. In the network I considered again an ideal MAC protocol that results in very low packet drop rates; thus, I neglected packet retransmissions during the simulations.
On the other hand, contrary to the assumptions made in Thesis 1, I assumed here that the sensor nodes can do power adjustment. This only enhances the possible energy gains obtained by reducing communication distances, even for direct communication between two neighboring nodes. I considered thus that the cost of sending one packet depends on the transmission distance d (ET x ∼ dα, α = 3); the transmission consumed 1 unit of energy for d = h = 80m. The multi-hop communication used the GOAFR [33] routing algorithm.
The network lifetime achieved by the dierent strategies can be seen in Figure 7.
The results show the average of 10 simulation runs. One can see that the two proposed strategies utilize the energy supplies more eciently, which results in longer network operation when compared to other solutions.
All the algorithms proposed in my dissertation and described in this thesis booklet are iterative in nature; thus, it would be useful to prove that they do converge. For this disseration no such investigations were conducted. However, even if this cannot be accepted as a proof, I must note that during the simulations the algorithms always converged in a relatively low number of steps. The analytical proof of convergence remains a topic of future work.
5 Application of New Results
The energy ecient operation has a crucial importance in case of wireless sensor net- works. Sensors have strongly limited energy supplies, but are supposed to operate in an unattended manner for months or even years. Replacing or recharging their batteries is usually impossible or impractical; thus, energy ecient strategies have to be developed in order to ensure long network operational times. In my theses I have proposed two novel solutions to that issue, both of them being applicable and very useful in a large number of possible use cases.
The rst solution is based on the use of multiple sink nodes; I proposed dierent strate- gies to nd the optimal sink positions in a time-driven scenario. There are several possible applications that can build on these results. Using multiple sink nodes is especially use- ful in cases where the monitored area is large. We can thus mention here applications such as environmental monitoring (monitoring the spread of an oil slick on the sea, a radioactive cloud or a forest re) or wildlife habitat monitoring. It is usually not realistic to assume a densely deployed sensor network over such large areas; moreover relaying reports over hundreds of highly error-prone wireless links in order to reach a remote sink several kilometers away from some of the sensors is not the best of ideas either. Also, some applications might require fast alert and reaction times, which cannot be ensured by using a single sink for the monitoring of large areas. Thus, employing multiple sink nodes is often a necessity. Nevertheless, it is very important where to deploy those sinks, so as to ensure the longest possible network operation. The deployment strategies that I proposed here might provide a valuable help in that matter.
On the other hand, in Thesis 4.2 I presented strategies to adaptively move the sink node in an event-driven scenario. There are also many applications that can make use of these solutions. The event-driven operation is specic to applications where a continuous data ow from sensors to the sink is not necessary; the sink is only interested in measure- ment results that provide information about something "unusual". Intrusion detection is a typical application that falls into this category. In the dissertation I explicitly addressed this case, and I even gave a method for a (multiple step) forecast on the intruder's tra- jectory, using a correlated random walk model. If the sink node has an ecient tool to predict the future evolution of the events, it can proactively move to locations that will be optimal when those future events occur.
Building mobile sinks introduces of course extra costs and complexity in the deploy- ment of WSNs. However, an easy and cheap way to produce a mobile sink is to create a network where the sink determines its new location e.g., on a daily basis, and a contracted worker puts the sink to the computed location. Some WSNs operate in areas that are not accessible for humans, but in which machines can move arbitrarily (e.g., chemically polluted areas). In these cases the sink has to be mounted on an unmanned vehicle (e.g., robot car, robot airplane). Building such sinks is of course a lot more expensive and com- plex; however, it is not unfeasible, since this kind of robots do exist, they only have to be integrated into WSNs. The constant improvement in the eld of robotics will also make the usage of robots a lot cheaper. There are also network scenarios where the monitored area is hardly accessible for robots too, making it impossible for them to move arbitrarily.
In this case a robot airplane ying over the area may be the alternative solution. In spe- cial applications like military ones, special conditions can be given on the movement of the sink (e.g., do not move closer to the target than 10 meters, because of safety reasons);
these will have to be taken into account during the control of the sink.
To demonstrate the viability of the sink mobility strategies, we decided to even build up such a mobile sink at the Department of Telecommunications and Media Informatics.
A small tracked-vehicle was equipped with a MicaZ mote and acted as a sink. The mote was connected to a magnetic eld sensor that served as a compass, and helped the positioning of the sink.
The data was collected from the network by the mobile sink and it was forwarded to a computer. The wireless sensor network was built from several MicaZ motes with temperature, light and acoustic sensors. It operated in event-driven mode, i.e. only those sensors that detected an event have sent data to the sink. As an example, a simple event can be the temperature going over a given threshold. The sink determined the ID of the sensor that detected the event, and it moved close to that sensor.
The operation of the mobile sink was presented in several successful demos (Mobil- Show 2006, MobilShow 2007, IST Mobile Summit). There were also two television shows presenting the application (RTL Klub Infomania, M2 Infogeneráció). Nevertheless, up until now the mobility of the sink node was quite limited, as there was no complex algo- rithm running on the sink to determine the optimal position; the sink just moved next to the alerting sensor over a hard-coded path. The future goal is to implement the sink mobility algorithms proposed in my theses. Right now a second version of the mobile sink is being developed that is equipped with a Lippert Motemaster. This mote has strong computational and storage capabilities, being thus able to run complex algorithms.
References
[1] W. M. Merrill, F. Newberg, K. Sohrabi, W. Kaiser, and G. Pottie, Collaborative networking requirements for unattended ground sensor systems, in Proc., IEEE Aerospace Conference 2003, Mar. 2003, pp. 21532165.
[2] G. Simon, M. Maroti, A. Ledeczi, G. Balogh, B. Kusy, A. Nadas, G. Pap, J. Sallai, and K. Frampton, Sensor network-based countersniper system, in Proc., 2nd international conference on Embedded networked sensor systems (SenSys'04), New York, NY, USA, Nov.
2004, pp. 112.
[3] A. Mainwaring, J. Polastre, R. Szewczyk, D. Culler, and J. Anderson, Wireless sensor networks for habitat monitoring, in Proc., WSNA 2002, Atlanta, USA, Sep. 2002.
[4] D. Culler, D. Estrin, and M. Srivastava, Overview of sensor networks, Computer, vol. 37, no. 8, pp. 4149, Apr. 2004.
[5] J.-H. Chang and L. Tassiulas, Energy conserving routing in wireless ad-hoc networks, in Proc., IEEE INFOCOM 2000, vol. 1, Tel Aviv, Israel, Mar. 2000, pp. 2231.
[6] G. Zussman and A. Segall, Energy ecient routing in ad hoc disaster recovery networks, in Proc., IEEE INFOCOM 2003, San Francisco, CA, USA, Apr. 2003, pp. 405421.
[7] K. Kar, M. Kodialam, T. V. Lakshman, and L. Tassiulas, Routing for network capacity maximization in energy-constrained ad-hoc networks, in Proc., IEEE INFOCOM 2003, San Francisco, CA, USA, Mar. 2003, pp. 673681.
[8] A. Sankar and Z. Liu, Maximum lifetime routing in wireless ad-hoc networks, in Proc., IEEE INFOCOM 2004, Hong Kong, Mar. 2004.
[9] J. Pan, Y. T. Hou, Y. S. L. Cai, and S. X. Shen, Topology control for wireless sensor networks, in Proc., 9th Annual International Conference on Mobile Computing and Net- working, San Diego, CA, USA, Sep. 2003, pp. 286299.
[10] N. Li and J. C. Hou, Topology control in heterogeneous wireless networks: Problems and solutions, in Proc., IEEE INFOCOM 2004, Hong Kong, Mar. 2004.
[11] R. Wattenhofer, L. Li, P. Bahl, and Y. Wang, Distributed topology control for wireless multihop ad hoc networks, in Proc., IEEE INFOCOM 2001, Anchorage, Alaska, USA, Apr. 2001, pp. 13881397.
[12] R. Ramanathan and R. Rosales-Hain, Topology control of multihop wireless networks using transmit power adjustment, in Proc., IEEE INFOCOM 2000, Tel Aviv, Israel, Mar. 2000, pp. 404413.
[13] A. Cerpa and D. Estrin, ASCENT: Adaptive self-conguring sensor networks topologies, in Proc., IEEE INFOCOM 2002, vol. 3, New York, NY, USA, Jun. 2002, pp. 12781287.
[14] V. Kawadia and P. Kumar, Power control and clustering in ad hoc networks, in Proc.
IEEE INFOCOM 2003, San Francisco, CA, USA, Apr. 2003, pp. 459469.
[15] O. Younis and S. Fahmy, Distributed clustering in ad-hoc sensor networks: A hybrid, energy ecient solution, in Proc., IEEE INFOCOM 2004, Hong Kong, Mar. 2004.
[16] S. R. Gandham, M. Dawande, R. Prakash, and S. Venkatesan, Energy ecient schemes for wireless sensor networks with multiple mobile base stations, in Proc., IEEE GLOBECOM 2003, vol. 22, no. 1, San Francisco, CA, USA, Dec. 2003, pp. 377381.
[17] H. Kim, Y. Seok, N. Choi, Y. Choi, and T. Kwon, Optimal multi-sink positioning and energy-ecient routing in wireless sensor networks, Lecture Notes in Computer Science (LNCS), vol. 3391, pp. 264274, Jan. 2005.
[18] Y. T. Hou, Y. Shi, H. D. Sherali, and S. F. Midki, Prolonging sensor network lifetime with energy provisioning and relay node placement, in Proc., Second Annual IEEE Communi- cations Society Conference on Sensor and Ad Hoc Communications and Networks, Santa Clara, Ca, USA, Sep. 2005.
[19] W. Hu, C. T. Chou, S. Jha, and N. Bulusu, Deploying long-lived and cost-eective hybrid sensor networks, in Proc., First Workshop on Broadband Advanced Sensor Networks, Santa Diego, USA, 2004.
[20] L. Qiu, R. Chandra, K. Jain, and M. Mahdian, Optimizing the placement of integration points in multi-hop wireless networks, in Proc., International Conference on Network Pro- tocols(ICNP), Berlin, Germany, 2004.
[21] P. Maulin, R. Chandrasekaran, and S. Venkatesan, Energy ecient sensor, relay and base station placements for coverage, connectivity and routing, in Proc., 24th IEEE Interna- tional Performance, Computing, and Communications Conference(IPCCC 2005.), Phoenix, USA, 2005.
[22] E. I. Oyman and C. Ersoy, Multiple sink network design problem in large scale wireless sensor networks, in Proc., ICC 2004, Paris, France, Jun. 2004.
[23] A. Bogdanov, E. Maneva, and S. Riesenfeld, Power-aware base station positioning for sensor networks, in Proc., IEEE INFOCOM 2004, Hong Kong, Mar. 2004.
[24] R. C. Shah, S. Roy, S. Jain, and W. Brunette, Data MULEs: Modeling a three-tier archi- tecture for sparse sensor networks, in Proc., IEEE Workshop on Sensor Network Protocols and Applications (SNPA), Anchorage, alaska, USA, May 2003, pp. 3041.
[25] L. Tong, Q. Zhao, and S. Adireddy, Sensor networks with mobile agents, in Proc., IEEE MILCOM 2003, vol. 22, no. 1, Boston, MA, USA, Oct. 2003, pp. 688693.
[26] S. Jain, R. C. Shah, G. Borriello, W. Brunette, and S. Roy, Exploiting mobility for energy ecient data collection in sensor networks, in Proc., 2nd IEEE/ACM Workshop on Model- ing and Optimization in Mobile, Ad Hoc and Wireless Networks (WiOpt), Cambridge, UK, Mar. 2004.
[27] H. S. Kim and T. F. Abdelzaher, Minimum-energy asynchronous dissemination to mobile sinks in wireless sensor networks, in Proc., ACM SENSYS, Los Angeles, CA, Nov. 2003.
[28] A. Chakrabarti, A. Sabharwal, and B. Aazhang, Using predictable observer mobility for power ecient design of sensor networks, in Proc., 2nd Int. Workshop on Information Processing in Sensor Networks (IPSN), Palo Alto, CA, USA, Apr. 2003, pp. 129145.
[29] J. Luo and J.-P. Hubaux, Joint mobility and routing for lifetime elongation in wireless sensor networks, in Proc., IEEE INFOCOM 2005, Miami, FL, USA, Mar. 2005.
[30] A. Kansal, M. Rahimi, W. J. Kaiser, M. B. Srivastava, G. J. Pottie, and D. Estrin, Con- trolled mobility for sustainable wireless networks, in Proc., IEEE Sensor and Ad Hoc Com- munications and Networks (SECON), Santa Clara, CA, Oct. 2004.
[31] A. Kansal, A. Somasundara, D. Jea, M. B. Srivastava, and D. Estrin, Intelligent uid infrastructure for embedded networks, in Proc., ACM MOBISYS 2004, Boston, MA, USA, Jun. 2004, pp. 111124.
[32] W. Zhao and M. Ammar, Message ferrying: Proactive routing in highly-partioned wireless ad hoc networks, in Proc., 9th IEEE Workshop on Future Trends in Distributed Computed Systems (FTDCS 2003), San Juan, Puerto Rico, May 2003, pp. 308314.
[33] F. Kuhn, R. Wattenhofer, and S. A. Zollinger, Worst-case optimal and average-case ecient geometric ad-hoc routing, in Proc., 4th ACM international symposium on Mobile ad hoc networking computing, 2003, Annapolis, Maryland, USA, 2003, pp. 267278.
[34] N. Megiddo, The weighted Euclidean 1-center problem, Mathematics of Operations Re- search, vol. 8, no. 4, pp. 498504, Nov. 1983.
Publication of new results
[B] Books
[B1] Zoltán Vincze, D. Vass, R. Vida and A. Vidács. Energy Eciency in Wireless Sensor Networks Using Mobile Base Station. EUNICE 2005: Networks and Applications Towards a Ubiquitously Connected World, IFIP International Federation for Information Processing Series, vol. 196, (C. Delgado Kloos, A. Marín, D. Larrabeiti (Eds.)), pp. 173-186, 2006, XIV, ISBN 0-387-30815-6
[J] Journals
[J1] Zoltán Vincze, D. Vass, R. Vida, A. Vidács and A. Telcs. Adaptive Sink Mobility in Event- driven Densly Deployed Wireless Sensor Networks. Ad-Hoc and Sensor Wireless Networks (Old City Publishing), vol. 3, no. 2-3,pp. 255-284., 2007
[J2] Vincze Zoltán, Vida Rolland. Mobil eszközök alkalmazása szenzorhálózatokban. (Hungar- ian) Híradástechnika, vol. 56, no. 12, pp. 12-17, December 2006
[C] International Conferences and Workshops
[C1] Zoltán Vincze, R. Vida and A. Vidács. On the Eciency of Local Information-based Sink Deployment in Heterogeneous Environments. In Proc., 2nd International Workshop on Performance Control in Wireless Sensor Networks (PWSN 2007), Austin, USA, 23 October, 2007.
[C2] Zoltán Vincze, R. Vida and A. Vidács. Deploying Multiple Sinks in Multi-hop Wireless Sensor Networks. In Proc., IEEE International Conference on Pervasive Services (ICPS '07), Istanbul, Turkey, 15-20 July, 2007.
[C3] A. Vidács, R. Vida, and Zoltán Vincze. Ecient Information Dissemination in Wireless Sensor Networks using Mobile Sinks. Information Systems Technology Panel Symposium (IST-062/RSY-016) on Dynamic Communications Management (NATO/PfP), Budapest, Hungary, 9-10 October, 2006
[C4] Zoltán Vincze, D. Vass, R. Vida and A. Vidács. Adaptive Sink Mobility in Event-driven Clustered Single-hop Wireless Sensor Networks In Proc., 6th International Network Con- ference (INC 2006), Plymouth, UK, July 11-14, 2006.
[C5] Zoltán Vincze, D. Vass, R. Vida, A. Vidács and A. Telcs. Sink Mobility in Event-driven Multi-hop Wireless Sensor Networks. In Proc., 1st Int. Conf. on Integrated Internet Ad hoc and Sensor Networks (InterSense), Nice, France, May 30-31, 2006. (Best Paper Award) [C6] Zoltán Vincze, K. Fodor, R. Vida and A. Vidács. Electrostatic Modelling of Multiple Mobile Sinks in Wireless Sensor Networks. In Proc., IFIP Networking Workshop on Perfor- mance Control in Wireless Sensor Networks, Coimbra, Portugal, 15-19 May, 2006.
[C7] Zoltán Vincze and R. Vida. Multi-hop Wireless Sensor Networks with Mobile Sink In Proc., CoNEXT'05: Proceedings of the 2005 ACM conference on Emerging Network Exper- iment and Technology, Toulouse, France, October 24-27, 2005.
[C8] Zoltán Vincze, D. Vass, R. Vida and A. Vidács. Energy Eciency in Wireless Sensor Net- works Using Mobile Base Station In Proc., 11th Open European Summer School (EUNICE 2005),Colmenarejo, Madrid, Spain, July 6-8, 2005.
[C9] F. Németh and Zoltán Vincze. Refresh Message Based Severe Congestion Handling in Resource Management for Diserv. In Proc., Polish-Hungarian-Czech Workshop on Circuit Theory, Signal Processing and Applications, Prague, Czech, September, 2003.
[C10] F. Németh and Zoltán Vincze. Refresh packet based severe congestion handling in RODA. In Proc.,9th EUNICE Open European Summer School and IFIP WG6.3 Workshop on Next Generation Networks, Balatonfüred, Hungary, September 8-10, 2003.
[NC] National Conferences and Workshops
[NC1] Zoltán Vincze and R. Vida. Deploying Multiple Sinks in Multi-hop Wireless Sensor Networks In Proc., High Speed Networks Laboratory HSN Workshop 2007, Balatonkenese, Hungary, 31 May - 1 June, 2007.
[NC2] Zoltán Vincze, D. Vass, R. Vida and A. Vidács. Adaptive Sink Mobility in Single-hop and Multi-hop Wireless Sensor Networks In Proc., High Speed Networks Laboratory HSN Workshop 2006, Balatonkenese, Hungary, 23-24 May, 2006.
[NC3] Zoltán Vincze, D. Vass, R. Vida and A. Vidács. Prolonging Wireless Sensor Network Lifetime Using Mobile Base Station In Proc., High Speed Networks Laboratory HSN Work- shop 2005, Mátraháza, Hungray, 23-24 May, 2005.
[NC4] A. Császár, A. Takáts, F. Németh, Zoltán Vincze. Resource Management in DiServ:
Link Failure Handling. In Proc., Workshop on High Speed Networks 2003, Budapest, Hun- gary, November 2003.
[NC5] B. Bartos, F. Németh, Zoltán Vincze. Hybrid Admission Control for Resource Manage- ment in Diserv. In Proc., Workshop on High Speed Networks 2002, Budapest, Hungary, 25-26 November 2002.
[O] Other
[O1] Zoltán Vincze and B. Bartos. A Resource Management in DiServ keretrendszer kiegészítése hibrid hívásengedélyezési eljárással. TDK (Students Scientic Conference, First prize) paper (Hungarian), Budapest, Hungary, 12 November, 2002.
[O2] Zoltán Vincze and B. Bartos. A Resource Management in DiServ keretrendszer kiegészítése hibrid hívásengedélyezési eljárással. OTDK (National Students Scientic Con- ference, Third prize) paper (Hungarian), Gy®r, Hungary, 17 April, 2003.
Citations
[CI1] R. Sarkar, X. Zhu and J. Gao. Double rulings for information brokerage in sensor net- works. In Proc., 12th annual international conference on Mobile computing and networking (MobiCom '06), Los Angeles, USA, September 24-29, 2006.
refers to
[27]. Z. Vincze and R. Vida. Multi-hop wireless sensor networks with mobile sink In Proc., CoNEXT'05: Proceedings of the 2005 ACM Conference on Emerging Network Experiment and Technologies, Toulouse, France, October 24-27, 2005.
as follows:
Information collection and delivery can explicitly use mobile nodes, such as data mules [12, 17, 27, 13, 23]. This is motivated by the observation that nodes around static sinks suer from unbalanced trac and energy consumption.
[CI2] J. Lee, W. Yu. Energy Ecient Target Detection in Sensor Networks using Line Proxies.
International Journal of Communication Systems., 2007.
refers to
[13]. Z. Vincze and R. Vida. Multi-hop wireless sensor networks with mobile sink In Proc., CoNEXT'05: Proceedings of the 2005 ACM Conference on Emerging Network Experiment and Technologies, Toulouse, France, October 24-27, 2005.
as follows:
Mobile sink environments have been studied for other applications in [13], [14]. They adopted the adaptive mobility solution where the sink moves inside the sensor network according to the current events, so as to minimize the energy consumption incurred by the multi-hop transmission of the event-related data.
[CI3] Y. Bi, L. Sun, J. Ma, N. Li, I. A. Khan and C. Chen. HUMS: An Autonomous Moving Strategy for Mobile Sinks in Data-gathering Sensor Networks. Accepted to EURASIP Journal on Wireless Communications and Networking., March 2007.
refers to
[36] Zoltan Vincze, Dorottya Vass, Rolland Vida, Attila Vidacs, Andras Telcs, "Adaptive sink mobility in event-driven multi-hop wireless sensor networks," in Proc. 1st Int. Conf.
on Integrated Internet Ad hoc and Sensor Networks (InterSense '06), Nice, France, 30-31 May, 2006.
as follows:
The authors of [36] proposed two strategies to move the sink adaptively to react to dynamic events that followed a correlated random walk mobility model, impracticable to the mobile devices that gather data periodically from all sensor nodes.
[CI4] M. Aly and A. Gopalan. TOLB: A Trac-Oblivious Load-Balancing Protocol for Next- Generation Sensornets. In Proc., 6th International Conference on AD-HOC Networks &
Wireless (AdHoc-Now'07), Morelia, Mexico, September 24-26, 2007.
refers to
[34] Zoltan Vincze, Dorottya Vass, Rolland Vida, Attila Vidacs, Andras Telcs, "Adaptive sink mobility in event-driven multi-hop wireless sensor networks," in Proc. 1st Int. Conf.
on Integrated Internet Ad hoc and Sensor Networks (InterSense '06), Nice, France, May 30-31, 2006.
as follows:
Next-Generation Sensor Nets (NGSN) to be composed of sensors deployed everywhere, together with gateways connecting sensors to Internet users/applications [26, 24]. Gateways may be stationary base stations [10], or mobile ones, such as robots, cell phones, and PDAs [35, 34].
[CI5] X. Wu and G. Chen. Dual-Sink: using mobile and static sinks for lifetime improvement in wireless sensor networks. In Proc., IEEE ICCCN 2007 - DSS, Honolulu, Hawaii, USA, Augustus 13-16, 2007.
refers to
[8] Z. Vincze, D. Vass, R. Vida, A. Vidacs, A. Telcs, "Adaptive sink mobility in event- driven multi-hop wireless sensor networks," in Proc. of the 1st International Conference on Integrated Internet Ad hoc and Sensor Networks, May 2006, Nice, France.
as follows:
Previous studies either assume that some global information of the network (e.g., location of the mobile sink) is already available (e.g., [2], [3], [4]) or let the sink spread the global information through repeated broadcasting across the network (e.g., [8], [9], [10]). ...
... Vincze et al. [8] propose adaptive sink mobility for energy eciency in WSNs for event- driven application.
[CI6] J. Ma, C. Chen and J. P. Salomaa mWSN for Large Scale Mo- bile Sensing The Journal of VLSI Signal Processing, online version:
http://www.springerlink.com/content/5u384um8kk458876/(January 2008) refers to
[27] Z. Vincze et al., "Adaptive Sink Mobility in Event-driven Multihop Wireless Sensor Networks," Proc. of InterSense 2006, Nice, France, 2006 (May 30-31).
as follows:
For controlled sink mobility [11-15, 22-27], the optimization problem can be generally classied into two categories: nding the optimal sink trajectory, i.e. the rendezvous based solution or traveling salesman problem that aims to minimize mobile sink visiting time for all the sensor nodes; nding the optimal sink location, i.e. to optimally place multiple sinks or relays in order to minimize the energy consumption and maximize network lifetime.
[CI7] J. -C. Chin, Y. Dong, W. -K. Hon and D. K. Y. Yau On Intelligent Mobile Target Detection in a Mobile Sensor Network in Proc., The 4th IEEE International Conference on Mobile Ad-hoc and Sensor Systems, Pisa, Italy, October 8-11 2007
refers to
[9] Z. Vincze, D. Vass, R. Vida, A. Vidacs, and A. Telcs, "Adaptive sink mobility in event-driven multi-hop wireless sensor networks," in Proc. InterSense, 2006.
as follows:
Rather than movement during the sensing process, a more limited form of sensor movement has been used in hybrid mobile/static sensor networks [9], [10]. In these networks, the
sensors may move to better deployment positions when the network is formed, but they do not move after the network is formed, or a mobile relay may move among static sensors to facilitate the data collection by acting as a sink of the data.
[CI8] S. Toumpis. Mother nature knows best: A survey of recent results on wireless networks based on analogies with physics Computer Networks: The International Journal of Com- puter and Telecommunications Networking, vol. 52, no. 2, pp. 360-383, 2007
refers to
[49] Z. Vincze, K. Fodor, R. Vida, A. Vidács, Electrostatic modelling of multiple mobile sinks in wireless sensor networks, in: Proceedings of the IFIP Networking Workshop on Performance Control in Wireless Sensor Networks, Coimbra, Portugal, May 2006, pp. 30- 37.
as follows:
In [49], the authors propose using an analogy with Electrostatics. In particular, sinks and nodes with low energy reserves are assigned positive charges, and nodes with high energy reserves are assigned negative charges. The sinks are left to move according to the combined electrostatic force they experience. As same-sign charges repel each other, and dierent- sign charges attract each other, the sinks will try to move far away from each other (and so cover the area uniformly), and also away from the sensors with low energy reserves. On the other hand, mobile sinks will be attracted by sensors with high energy reserves, who are willing to participate heavily in the collection of data. As energy levels of sensors will be changing, so will the charges associated with them, and so will the position of the mobile sinks. Studies by simulation show that this method outperforms all other approaches as far as balancing the energy consumption is concerned.
