Cluster Based Target Tracking And Recovery Computer Science Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

The rapid evolutions in wireless sensor network, it is possible to implement the wireless sensor network technology in different scenarios. Such WSNs is collection of thousands of tiny coordinative sensor nodes deployed in a physical environment for sensing respective events like speed, light, temperature etc. A wireless sensor nodes are battery operated. Such sensor node carries limited and generally irreplaceable battery power sources. So, the primary focus of WSNs must on power conservation. Due to such limitations Sensor nodes may leads to energy depletion and network failure. So it is very important for WSN to operate energy efficiently. In vast number of WSN applications, target tracking is one of the important application [1]. The sensors nodes in the vicinity (area) of an event must be able to continuously monitor it and report to the sink. A various traditional target tracking techniques for Wireless Sensor Networks use a centralized approach. In traditional target tracking methods, one node at a time usually performs sensing by resulting in heavy computation burden, less accuracy and more energy consumption where in WSNs each node has very limited power. Even if the more than one node performs sensing and node fails due to battery drainage then target can not detected. Hence, next predicted nodes who were supposed to be active for tracking target may not have the trace of target, causing loss of target. In this paper, we propose an energy efficient and fast recovery mechanism (RM) to recover the lost target during tracking using dynamic clustered network.

Rest of the paper is organized as follows:

Section II: Related works of existing cluster- based tracking techniques and recovery mechanisms.

Section III: Challenges in recovery mechanism

Section IV: Target tracking

Section V: Proposed recovery technique

Section VI: Conclusion


Wireless Sensor Network implementation has many design challenges like limited energy, communication failure, deployment, network lifetime, accuracy and secure communication etc. Dividing the network into groups of node as a cluster can be considered as a solution to handle these challenges [2]. The light source based tracking for terrain observation during night using clusters is discussed in [3]. DELTA is one of the distributive algorithm which tracks the object only at constant speed by dynamic clustering and selection of Cluster Head based on light measurement. The advantage of DELTA is that the communication range of the sensor is higher than their sensing range. But the main dispute of this method is that it can deal with constant speed only, and varying speed is not considered. Energy efficient approach, RARE [4] reduces the number of node involving in tracking. It consists of two sub algorithms, first is RARE-Area which ensures nodes with qualitative received data, and second is RARE-Node which reduces the unnecessary nodes participating in target tracking. For multiple moving targets tracking, a tree based approach STUN considers leaf nodes to track the targets as presented in [5]. Low-Energy Adaptive Clustering Hierarchy [6] reduces the energy consumption of each node. In LEACH, sensor network divided into small parts known as clusters and selects one of them as cluster-head. LEACH uses random selection and rotation of the cluster-heads to randomly distribute energy consumption in the network. LEACH operations can be divided into two phases first is Setup phase and another second is Steady phase. In the setup phase clusters network will formed and a cluster head is selected for each cluster. In the steady phase, data is sensed and sent to the central base station. As nodes are depleted of energy, all the approaches of WSN should have less communication among each other to conserve energy. Several tracking algorithms have already been developed [7]-[9].

Recovery mechanism for target recovery using static clustered WSN is discussed in [10], which is basically consists of four phases: (a) loss of target - when target's current location is not known, current CH declares target is lost and recovery mechanism is initiated, (b) search - current CH waits for some time to receive acknowledgement from downstream i.e. From next single hop clusters to reduce the chances of false initiation of recovery algorithm, (c) active recovery - this phase consists of three levels during recovery (d) sleep - once target is found, clusters which are participating currently in tracking remain awake and rest of them go to sleep state. However, RM presented in [10] requires large number of nodes to be in active state during recovery.

In this paper, we are proposing a dynamic clustered WSN which provides energy efficient recovery approach which results into less communication overhead and less number of active CHs for successful recovery. Instead of waking up all single hop clusters [10] or all double hop clusters during recovery, this approach will find single or double hop clusters nearest to target's next predicted location for saving time and energy of network.


Sometimes network fails to track the target trajectory accurately due to many reasons as follows:

1. Node Failures

2. Network Failure

3. Deployment

4. Localization Errors

5. Synchronization

6. Data Aggregation

7. Data Dissemination

8. Prediction Errors

9. Uncertain change in target's velocity/direction

1. Node Failures: Sensor nodes are basically battery operated and generally irreplaceable. So, even single node failure causes network failure.

2. Network Failure: The failure of network occurs due to communication breaks, overload of data, physical disasters etc.

3. Deployment: Setting up an executive sensor network in a real time or real world environment [11] is a Deployment. Sensor nodes can be deployed either by placing them sequentially in a sensor field or randomly by dropping it from height. Various deployment issues are [12, 13]:

3.1. Node dies due to either by normal battery discharge or due to short circuits.

3.2. Network congestion occurs due to many parallel data transmissions made by several sensor nodes simultaneously. So that even if the two nodes may very close to each other but still they can not communicate due to physical interference in the real world instead nodes that are far away from each other may communicate.

3.3. Low data yield is also one of the common problem in deployment. Low data yield occurs when network delivers insufficient amount of information.

3.4. Self Configured sensor networks is needed due to random or distributed deployment of sensor nodes in real world environment without human participation.

4. Localization Errors: Improper positioning of target from nodes in network.

5. Synchronization: For real time data collection, the clock synchronization is an important in sensor networks. Time Synchronization supports to maintain events synchronization between nodes in the network by providing common time scale for local clocks of nodes in the network. Global clock synchronization also require for some applications are navigation guidance, environment monitoring, vehicle tracking etc. In sensor network, need to properly coordinate and collaborate to perform a complex task like data fusion in which the data collected from number of sensor nodes and aggregated to generate a meaningful result. Lack of synchronization between sensor nodes then results into the inaccurate data estimation.

6. Data Aggregation: The main objective of the sensors nodes is to periodically sense the data from the surrounding environment process it and transmit it to the sink or base station. Thus a method for combining the sensed data into accurate and high quality information is required and this is accomplished through Data Aggregation. Data Aggregation is the process of collecting or aggregating data from multiple sensors and estimating the desired output by eliminating the redundant data and then providing fused information to the base station. So there are some issues fro data aggregation as to obtain complete and up to date information from neighboring nodes, improving clustering techniques for data gathering, eliminating redundant data transmission etc.

7. Data Dissemination: To obtain required data, the queries for the data are routed from node to node in the sensor network. Data dissemination is basically a two step process. In which initially node broadcasts its interests to its neighbors periodically and then through the whole sensor network. In the second step, the nodes having requested data will reply back to the source node with data. The key difference between data aggregation and data dissemination is that, in data dissemination all the nodes even the base station also can request for the data but in data aggregation periodic data transition of the aggregated data will occur is to the base station.

8. Prediction Errors: to save the energy all clusters should not be awake all time. In such condition advanced warning message has to be generated by working cluster heads for its upstream and downstream cluster heads. To do so current CH need to predict future location of target. In this prediction process minor errors can lead to target loss, since the target's trajectory may be lost due to a cluster not being woken up in advance.

9. Uncertain change in target's velocity/direction: Sudden change in target trajectory may not be handles by monitoring nodes and prediction errors will be generated.

Some of the applications like military require reliable and credible target tracking. To overcome such problems quick, time saving and energy efficient recovery mechanisms is needed.


Tracking algorithm [10] for target performs two steps:

i. Localization

ii. Prediction

1) Localization: Boundary nodes exchange their information with each other to identify the nodes who also have detected the target. Such three nodes if found, then trilateration mechanism initiated. Otherwise, wake message send by boundary nodes to its single hop neighbors to perform localization. The boundary node nearest to target sends this information to CH to perform prediction.

2) Prediction: The next predicted location of target depends on current and previous locations of target. Prediction will be performed by Current CH as follows:

• Predict target's next location.

• Find member nodes in the area of predicted location.

• Node nearest to target's predicted location will wake up only. If no node is present in the vicinity, then only single hop CH wake up, this will become current CH for further tracking.


In the clustered based tracking network, CH is itself responsible to predict next location of target and find out respective next CH to handle further tracking process. In static cluster based network, probability of CH failure is more than the failure of local (member) nodes as it because of communication overhead. The proposed recovery mechanism follows three steps (Figure 1 shows flowchart of tracking and recovery):

1. Announcement of lost of target

2. Recovery

3. Sleep

1) Announcement of lost of target: As continuous target moves away from current cluster, current Cluster Head sends warning massage to wake up to the next predicted or selected CH. If CH is failed due to any of the reason like Node Failures, Network Failure, Localization Error or Prediction Errors, then no acknowledgement will receive by current CH from selected CH. Current CH waits for acknowledgement for some period of time and finally target lost situation can be announced by current CH. This step is same as described in [10].

2) Recovery: As loss of target occurs, current CH initiates true recovery process immediately. Detail of recovery process is as given below:


NL: Next locations of target

PL: Predicted locations

SCH: Single hop CH


Current CH:

1. Wake up all member nodes.

2. Predict few NL based on previous PL.

3. As CH has information about its SCHs, it finds one of them which is not failed and nearest to NL.

4. Wake up this SCH.

Target entered into network

Cluster based n/w is formed and Cluster Head (CH) selected on the basis of nodes energy

Boundary nodes are found

Only Boundary nodes will remain active

Tracking/Localization of target

Prediction of Target's next location

Recovery Mechanism

Select next CH based on predicted location

Successful recovery of target



Figure 1: Target Tracking and Recovery Flowchart

5. If (any of NL is in vicinity of selected SCH) Selected SCH:

a. Wake up all member nodes.

b. Become current CH.

c. Locate target using trilateration and start tracking.

d. Send acknowledgement to current CH.


a. Make selected SCH as current CH temporarily.

b. Repeat step 1-5 till the target is recovered.

3) Sleep: Once target is recovered, only currently tracking nodes remain active and remaining will go in sleep mode, same as in [10].


This paper focused on the various problems of loss of target trajectory in a target tracking wireless sensor network. For WSNs to be used effectively and reliably, by considering energy and time saving issue for target tracking, fast and more accurate recovery mechanisms has been given in this paper.