Energy Optimization With Mobi Sink

Published Date: 02 Nov 2017

A wireless sensor network (WSN) usually consists of a set of sensors, deployed in hundreds or thousands over a region. Each sensor is capable of measuring acoustic, magnetic, spatial, or seismic data and the wireless network enables performing distributed computations over the gathered data by individual sensor nodes (SNs). This makes meaningful inferences at the sink, which sends the data to end user for appropriate action. WSNs promise novel applications in several domains such as forest fire detection, battlefield surveillance and monitoring of human physiological condition, which are only in the vanguard of enhancement provided by WSNs. Sensor nodes (SNs) can be spread out in a dangerous or remote environment by low flying airplanes or unmanned aerial vehicles that opens up new application fields. A mobile sink (MS) moving through the network deployment region can collect data from the static sensor nodes over a single hop radio link when approaching within the radio range of the sensor nodes or with limited hop transfers if the sensor nodes are located further. This avoids long-hop relaying and reduces the energy consumption at sensor nodes near the base station, prolonging the network lifetime. One of the main challenges raised by these networks is the fact that are usually power constrained, since sensing nodes typically exhibit limited capabilities in terms of processing and communication, and especially run on battery power. WSNs power limitation is impacted by the fact that, once deployed, often left unattended for the lifetime. Thus, in order to maximize the network’s operational lifetime, energy conservation is of prime consideration in algorithms used for WSNs. The proposed protocol called Dynamic Trajectory and

Tryst Nodes (DTTN) aims at minimizing the overall network overhead and energy expenditure of sensor nodes (SNs) by making dynamic trajectories between sensor islands to cover all the nodes ensuring balanced energy consumption among SNs and prolonged network lifetime. This is achieved through building cluster structures consisted of member nodes that route their measured data to their assigned cluster head (CH). The CHs perform data filtering upon the raw data and forward the filtered information to their assigned tryst nodes (TNs), typically located in proximity to the MS’s trajectory. The remainder of this paper is organized as follows: Section 2 reviews related work in the field. Section 3 details the design principles for DTTN protocol.

II. RELATED WORK

Energy efficiency is an important issue in the applications of wireless sensor networks (WSNs). Sensor nodes are constrained in energy supply and bandwidth. Therefore, it is necessary to optimize the energy usage and prolong the network lifetime by energy-efficient strategies. Recent years, number of approaches is exploited for data collection in WSN. The following two solutions are obtained in case of communication in WSNs, since in the first solution mobile sink may visit each sensor nodes and gather data by single hop communication. In such solutions, energy consumption is reduced. Expense of data delivery is very high since single hop communication is used. And in the second solution the mobile sink may visit some of the WSN location and sensor nodes send the data to mobile sink through multi hop communication. In such solutions, energy consumption is relatively high. Expense of delay time is low since multi hop communication is used. In addition sensor nodes should be constantly kept updated about mobile sink current location herby creating routing overhead. The proposed protocol is a tryst-based solution that involves monitoring of isolated urban areas with respect to environmental parameters, surveillance, fire detection, etc. In such environments mobile sinks may take dynamic trajectory and provide maximum aggregation. In proposed protocol the sensor node is dynamically assigned as a tryst node (TN) in a cluster which is proximity with the mobile sink trajectory. In this context, the work presented in [1] is mostly relevant to the mobile sink trajectory and the route of the mobile element based on the sensor’s buffer overflow deadline and traveling distances between the sensor nodes. Computing the sensor next overflow deadline depends on knowing in advance its buffer size and sensing rate. Deadline misses which will lead to data loss because the relaying of data acts in a recursive fashion until it reaches the base-line nodes.

In [2], a specific type of a sensor network application where the sensor nodes are required to be placed in fixed locations. The determination of a route is formulated and defined as problem belonging to the class of Traveling Salesperson Problems. The complexity being NP-Hard, the problem is solved by reducing it to an instance of the well-known travelling salesperson problem with neighborhoods. The scheduling of the mobile sink trajectory to collect data from each source directly which leads to high delay data delivery. The sensor nodes should constantly be kept updated about the mobile sink’s current location thereby creating considerable routing overhead. Due to this additional work, frequently the sensor node will get lose its energy. This direct-contact scheme is impractical in time intensive WSNs. A solution in between is to have SNs send first their data to a certain number of nodes (TNs) which buffer the received data and send them to MS when MS is within their transmission range [3], [4], [5], [6] or when they receive a query from MS asking for the buffered data . In the second approach, the MS does not necessarily pass near the rendezvous node and the data stored at each rendezvous nodes are forwarded to MS by reversing the route of the received query packet.

In [3], the MS traversal is performed on a per region basis, it visits regions one after another and stopping at each region for an appropriate interval to collect data. In this method the stop times depend on the local density in each network region, towards balanced traversal, stopping more in regions with higher traffic. In the stop-and-collect mode, the mobile sink stops at certain fixed locations to collect data. This Path Stop Point (PSP) method has low latency and high delay data delivery rates. In [4], sensors can only communicate with others within a very limited range, packets from some sensors may need multi-hop relays to reach SenCar. In this context, collecting network knowledge incurs a significant overhead on the sensor nodes. Due to this additional overhead, the lifetime of the sensor network will goes down .In [5], Sensor nodes have to buffer the information sensed during two consecutive visits of the sink. The sensor nodes memory has limited capacity, if the sinks stay long periods without visiting a node, data losses are likely to occur. In this approach, to cover all the network nodes in a reasonable amount of time is not possible and inefficiency with respect to data collection latency.

In [6], a rendezvous-based data collection approach is determined. A subset of nodes serves as the rendezvous points that buffer and aggregate data originated from sources and transfer to the base station when it arrives. These approaches are centralized approaches that try to minimize an energy related cost function without paying proper attention to the selection of nodes that will serve as rendezvous nodes. The described routing structures are built once and cannot be modified for the whole lifetime of the WSN. This fixed routing structure cannot be adapted for dynamic changes in sensor network. Another issue in all previous schemes is that there is no provision in case that rendezvous nodes run out of energy. Rebuilding of the routing structures may be required in order to bypass dead rendezvous nodes. Bypassing the dead rendezvous nodes, the sensor node will transfer the data directly to the mobile sink. Due to this direct approach, the mobile sink has to visit individual sensor nodes.

In [7], the issues of practical communication protocol design and motion control is addressed. The MS is used to collect data from groups of SNs. During a training period, all the WSN edge nodes located within the range of MS routes are appointed as rendezvous nodes and build paths connecting them with the remainder of sensor nodes. Those paths are used by remote nodes to forward their sensory data to rendezvous nodes; the latter buffer sensory data and deliver them to the MS when it reproaches in range. The movement of mobile robots is controllable which is impractical in realistic urban traffic conditions. Most importantly, no strategy is used to appoint suitable nodes as rendezvous nodes while selected rendezvous nodes are typically associated with uneven numbers of SNs.

In [9], the mobile sink only needs to visit subset of nodes as long as it travels inside the communication ranges of all nodes. An approximation algorithm for the problem is developed and it finds near-optimal paths. The mobility model with acceleration constraint is employed. The mobile sink just follows the pre-programmed trajectory and collects data, if any dynamic changes in the network cannot be easily and quickly accommodated into the sink trajectory. It requires geographical location information of the sensors in order to compute the mobile sink trajectory. The periodic gathering of location information can pose a heavy burden on the energy resources of the sensors .It would not apply to networks where localization services are not available. In [10], joint mobility and routing algorithms uses a combination of round routes and short paths .The problem of load balanced data collection in WSN is investigated and make use of existing multi-hop routing protocols and to achieve further improvements in terms of network lifetime by exploiting the base station mobility. In this context, as the mobile sink goes around the network, sensors will continuously track the position of the sink and send their packets to the sink via multi-hop communication. Base stations often change their paths dynamically; additional overhead is incurred in maintaining efficient routing topology. In all these approaches, the mobile sinks may visit each sensor node or sensor nodes send their data to mobile sink through multi-hop communication.

In all the previous rendezvous schemes, there is no provision in case that rendezvous nodes run out of energy. And there is no previous study that investigates the use of dynamic paths with tryst node between sensors and mobile sinks for improving WSN performances.

The proposed protocol consists of dynamic path selection which is a very flexible one. DPF-kNN is the algorithm which is based upon selecting k nearest neighbor nodes for optimizing energy related with the networks. The proposed protocol selects as tryst nodes only nodes with sufficient energy and in proximity with mobile sink trajectory for long time. Also, only tryst nodes with no overlapping contact intervals with mobile sink are selected, so the collisions due to the concurrent transmissions are eliminated. Moreover, the operation of tryst nodes is well synchronized and the right amount of data is distributed to each tryst node according to the contact time and data delivery rate of each tryst node. In case that a tryst node runs out of energy, it is quickly replaced by other available tryst nodes and thus transmission is not disrupted. Improved data throughput is confirmed by flexible tryst nodes and individual tryst node in every cluster, that allowing sufficient time to deliver their buffered data and preventing data losses. Thus in turn leads to much lower energy consumption in the wireless sensor network and also less data buffered at tryst nodes, reducing the buffer overflow at a tryst node.

III. DTTN PROTOCOL

The proposed protocol called Dynamic Trajectory and Tryst Node [DTTN] Protocol has intention to reduce the energy expenditure of sensor nodes by clustering of sensor nodes and assigning tryst node (TN) at each cluster according to the trajectory of the mobile sink to retrieve the information from cluster head (CH) and forward it to the mobile sink. Additionally the DTTN protocol has high data aggregation ratio by making dynamic trajectory of mobile sink to cover the urban area. In the following sections, the DTTN protocol methods are described.

A. Clustering

WSN is deployed in large-scale and the need for data collection necessitates well-organized network topology for the purpose of load balancing and extending the network lifetime. Clustering is an efficient approach for organizing the network in the above circumstance. Moreover achieving energy efficiency, clustering also reduces the channel contention and packet collisions which lead to improved network throughput under heavy load [11].

To build a cluster structure of unequal clusters Chen et al [12] algorithm is used. The clustering algorithm in [12] constructs a multi-sized cluster structure, where the size of each cluster decreases as the distance of its CH from the base station increases. In the proposed protocol, Chen et al algorithm is modified depending on the distance of the cluster heads from the mobile sink’s trajectory.

In the initial stage, the mobile sink moves along its trajectory broadcasting periodically a BEACON signal to all sensor nodes. All the nodes proximity to the mobile sink trajectory receives the BEACON message. Then these nodes send this BEACON message to the rest of the network. The following algorithm describes the cluster head election which is executed after the mobile sink completes its first trip. As soon as the clustering phase confirms, each cluster head proceeds to the selection of the appropriate cluster members to serve as tryst nodes.

Algorithm CLUSTER HEAD _ELECTION

1: if m received the BEACON message directly from the

mobile sink

2: m.Cmprange= R

3: else

4: m.Cmprange = R’

5: broadcast Competition_Msg(m.Node_ID,m.Eresidual,

m.Cmprange)

6: On receiving a Competition_Msg from a node u

7: if dist(m,u) < max(m.Cmprange , u.Cmprange)

8: u is added to Nm

9: while the "tentative CH completion time" has not expired

10: if ∀ u Є Nm, m.Eresidual > u.Eresidual

11: broadcast Final_CH_Msg(m. Node_ID)

12: exit

13: if a Final_CH_Msg(u.Node_ID) is received and u Є Nm

14: broadcast Quit_Competition_Msg(m.Node_ID)

15: exit

16: if a Quit_Competition_Msg(u.Node_ID) is received and

u Є Nm

17: u is removed from N

18: end while

All the nodes that receive the BEACON message will start the clustering procedure at the same time. Once the final cluster heads (CH) have been elected then each CH broadcasts a message to announce its election. Each non-CH node uses the received signal strengths to join to the closest CH.

B. Selection of Tryst Nodes

This process is taking place in each cluster having tryst nodes and it starts after the cluster head has received the messages from the candidate TNs (tryst node) of its cluster. TNs guarantee connectivity of sensor islands with mobile sinks; hence, their selection largely determines network lifetime. TNs lie within the range of traveling sinks and their location depends on the position of the cluster head and the sensor field with respect to the mobile sinks trajectory. Suitable tryst nodes (TNs) are those that remain within the mobile sink’s range for relatively long time with sufficient energy.

Nodes with relatively high competence values are likely to be elected as tryst nodes (TNs). The algorithm executed by each sensor node receiving BEACON packets, directly follows (Algorithm TN_CANDIDACY):

Algorithm TN_CANDIDACY

1: initialize nb= 0; m.Tfirst=0; m.Tlast=0

2: Wait until a BEACON is received

3: record BEACON receipt time t1 and signal strength is s1

4: m.Tfirst=t1, m.Tlast= t1, nb=1,nb,r=1

5: start ‘Connection Dropped Timer’

6: while ‘Connection Dropped Timer’ has not expired

7: wait until next BEACON is received or ‘Connection

Dropped Timer’ is expired

8: if a BEACON i is received

9: record BEACON receipt time ti and signal strength si

10: nb = nb + ti - ti-1

Tbeacon

11: nb,r= nb,r +1

12: m.Tlast = ti

13: reset ‘Connection Dropped Timer’

14: end if

15: end while

16: m.Compval = a1.Eresidual + a2.nb + a3.∑nbi=1 si

Emax nb

Send

17: TN_Cand_Msg(m.Node_ID, m.Compval , m.Tfirst ,m.Tlast )

In step 16, αi, i=1,2,3 are weight coefficients of the residual energy, the nb value and the average signal strength of received BEACON messages. Note that the first 16 steps of the algorithm are executed during the first mobile sink trip while the last one is executed right after the execution of the clustering algorithm and the election of the CHs. Each CHu receives TN_Cand_Msgs from the entire candidate TNs of its cluster, and then it proceeds to the selection of the appropriate tryst nodes (TNs) to build the set Ru of the final tryst nodes (TNs) associated with it.

The following TN_Selection algorithm executed by each cluster head for selecting the set of final TNs attached to it.

Algorithm TN_SELECTION

1: CRu= {mi | u has received a TN_Cand_Msg from mi and

mi.Compval ≥ T }, k = | CRu |

2: I = { I mi = [mi.Tfirst, mi.Τlast] | 1 ≤i ≤ k and mi Є CRu }

3: weight(Imi)= mi. Compval , 1 ≤ i ≤ k

4: sort the intervals’ endpoints m1.Tfirst , m1.Τlast ,m2.Tfirst ,

m2.Τlast, …, mk.Tfirst , mk.Τlast

run the algorithm of Hsiao et al on the sorted

5: endpoints list to compute a maximum weight independent

set U={ Imj1, Imj2,…., Imjl } C I, l ≤ k

6: Ru= {mji | Imji Є U and 1 ≤ i ≤ l }

It should be mentioned that in calculating the competence values of tryst nodes and the contact intervals [Tfirst, mi .Τlast ] (Algorithm TN_CANDIDACY), assumed that if a tryst node does not receive three consecutive BEACON messages from the mobile sink, this means that mobile sink has moved away from this node and thus the node can safely determine the final competence value and contact interval with mobile sink. Also implicitly assumed that if a node hears a BEACON message again in shorter than 3 BEACON broadcast periods, the loss of contact is due to transient communication problems (interference, collisions and current environmental conditions) and not due to departure of the mobile sink away from the node.

C. Data Aggregation and Forwarding to the Tryst Nodes

The data accumulated at individual source nodes are sent to local cluster heads (CHs). Cluster heads (CHs) perform data processing to remove spatial-temporal data redundancy. CHs then forward filtered data to tryst node (TN) attached to. In the case that multiple TNs exist in that cluster, data are not equally distributed among them. Instead, the CH favors the data delivery by the most suitable TNs, i.e., those with highest competence value (Compval). Data distribution among TNs should ensure that each TN will be able to accommodate its assigned data. Hence, CHu sorts the TNs in its Ru set in Compval decreasing order and deliver to each TN node mi Є Ru the maximum amount of data Di it can accommodate, minus an "outage prevention allowance" amount O. The Di value is calculated taking into account the TN’s data rate ri and the length li of the time interval [mi. Tfirst, mi. Tlast] that mi remains within the mobile sink’s range. The process is repeated for each mi Є Ru until all data available at u are distributed among its TNs. The algorithm executed by each CHu for distributing data to the TNs attached to it, follows (Algorithm DATA_ DISTRIBUTION):

Algorithm. DATA_DISTRIBUTION

1: D: amount of data available at u for distribution among

the TNs attached to u

sort the TNs in Ru in Compval decreasing order into a

2: sorted list m1, m2; ..., m | Ru | (mj . Compval ≤ mi.Compval,

∀i,j 1 ≤ i<j ≤ | Ru | )

3: i=0

4: while (D>0 and | Ru | >i+1)

calculate:

5: Ii=mi . Τlast - mi . Tfirst ; D’ data to TN mi

6: D’=min(Di,D)

7: transmit(D’,mi) // u sends D’ data to TN mi

8: D = D – D’

9: i= i +1;

10: end while

D. Communication between TNs and Mobile Sink

Data delivery occurs along an intermittently available link; hence, a key requirement is to determine when the connectivity between a TN and the mobile sink (MS) is available. Communication should start when the connection is available and stop when the connection no longer exists, so that the TN does not continue to transmit data when the MS is no longer receiving it. To address this issue, we use an acknowledgment-based protocol between TNs and MSs. The MS periodically broadcasts a POLL packet, announcing its presence and soliciting data as it proceeds along the path. The POLL is transmitted at fixed intervals Tpoll. This POLL packet is used by TNs to detect when the MS is within connectivity range. The TN receiving the POLL will start transmitting data packets to the MS. The MS acknowledges each received data packet to the TN so that the TN realizes that the connection is active and the data were reliably delivered. The acknowledged data packet can then be cleared from the TN’s cache. Thus, if the energy supply of a TN falls below a threshold, it may request the local CH to engage another node as TN so as to further extend the network’s lifetime without affecting the current clustered infrastructure. To enable TNs substitution, the CH polls the candidate TNs to be informed about their current residual energy status and then selects the new TN. In particular, a re-clustering scheme is proposed, which is part of the DTTN protocol and is applied locally within each cluster, whenever it is needed, in order to distribute energy consumption among sensor nodes by swapping the roles of the nodes.

IV. CONCLUSION

This paper introduced DTTN, a protocol that proposes the dynamic trajectory of mobile sink that retrieve information from isolated parts of WSNs. DTTN protocol mainly goals at maximizing connectivity, data throughput and enabling balanced energy expenditure among sensor nodes. The connectivity objective is addressed by employing MSs to collect data from isolated urban sensor islands using variable path and also through prolonging the lifetime of selected peripheral TNs which lie within the range of passing MSs and used to cache and deliver sensory data derived from remote source nodes. Improved data throughput is ensured by regulating the number of TNs for allowing sufficient time to deliver their buffered data and preventing data losses. Unlike other approaches, DTTN moves the processing and data transmission burden away from the vital periphery nodes (TNs) and enables balanced energy consumption across the WSN through building cluster structures that exploit the high redundancy of data collected from neighbor nodes and minimize inter-cluster data overhead.