We set our experiments in the same parameters as [6]. Deploying 100 sensor nodes in a square space (50m by 50m), whose coordinates are set randomly in this place. The sensing range of each node is 10 meters and each node knows neighbors’ position.
When a node turns off, all its neighbors can know this turning off decision. After all nodes have made decisions, the number of off-duty nodes and the current sensing coverage degree by on-duty nodes are compared with the original one where all nodes are on-duty. To calculate sensing coverage, we also use the same mechanism as [6], which divides the space into 1m by 1m unit cells. Assume an event occurs in each cell, with the event source located at the center of the cell. We investigate how many original nodes and how many on-duty nodes can detect every event, and also compute the average sensing degree before and after turning off nodes.
Table 4.1 shows the experimental results when we apply our protocol, [6] and probing-based [9] off-duty eligibility rules 100 rounds in 100 random topologies, respectively. As the results illustrated, by applying our proposed protocol, nodes can be turned off to eighty, which is better than [6]’s one half of original on-duty nodes. The sensing degree after turning off redundant nodes is reduced from 10 to 2, which also means more redundant nodes can be off-duty. The probing-based off-duty eligibility rule makes almost the same number of nodes be turned off as ours when the probing range is set as 7 meter, but blind points appear in 100 topologies in that case, i.e., blind points occur in every topology. It also shows the proposed coverage-based off-duty eligibility rule, such as [6] and ours can guarantee no blind points occurred.
104.06
Table 4.1: Comparison of three off-duty eligibility rules.
On-duty node number vs. Density
Another result of Table 4.1 shows that the larger probing range of [9] results in more nodes being turned off and more sensing coverage being reduced when the probing-based off-duty eligible rule is used. To investigate the relation between density and on-duty node number when the coverage-based off-duty eligible rule is used, we change node density by varying the node number and sensing range.
Although increasing node number and sensing range will increase the on-duty nodes, the proposed mechanism as illustrated in Fig. 4.1 also works well with coverage-based eligible rule. As shown in the figure, when the number of nodes increase to three times, the on-duty nodes proposed in [6] increase about 30% while ours K-CP protocol only increase about 10%. When the nodes are nearby the boundary, in this experiment, according to both [6] and our proposed K-CP eligible rule, nodes have no chance to turn off owing to all the other nodes cannot help to cover their sensing range, which is outside the boundary. However, our eligible rule still gets better results in the boundary situation.
Coverage degree vs. Density
One characteristic of our proposed protocol is that we can decide ”coverage degree”
with different application. In this experiment, we, firstly, investigate the change of sensing degree over node density. As shown in Fig. 4.2, the experiment show
Figure 4.1: on-duty number vs. deployed number. dtian means the protocol proposed in [6]
that increasing the number of nodes will also leads to level of coverage. Let each node calculates its coverage degree of perimeter, and the total sum of each node’s coverage degree divided by total nodes is the average coverage degree of this network.
Compare the average coverage degree, we define the actual coverage as : after each node calculates its coverage degree, the lowest coverage degree of node in this sensor network is actual coverage degree. We then define obtained coverage degree as : after turning off some redundant nodes, the actual coverage degree is obtained coverage degree. Interestingly, although the average coverage degree increases almost three times as the increase of deployed node number, the actual network coverage doesn’t have the same proportion. It means the inequality deployment of the sensors cause unbalanced level of coverage. Because our goal is to maintain the coverage degree of a given system above K, the gap between actual k and the obtained k is the space, where we can use some mechanism to prolong the surviving time.
Coverage degree vs. Surviving time
To investigate the survival time with the different coverage degree, we set the original power of each node ranges form 200 to 1000 unit, and deploy the sensor nodes in a square space (50m by 50m). To simplify, we assume each scheduling cycle consuming one unit power of on-duty node, and there is no other energy consumption. We define survival time as : in a given sensor network, how long can this network maintain the K before any point’s coverage degree is below than K. Fig. 4.3 shows the survival
time with different K. Compare with the Fig. 4.2, the survival time doesn’t increase as the same ratio as the gap between actual k and the obtained k. Because there are some area, even after the nodes performing the eligible rule , only a few nodes can be turned off in order to maintain K (i.e., the k in that area may greater than K ). These nodes are called critical nodes, which means although we can reduce original network’s level of coverage degree to K, once the critical nodes use up their power, they can’t find any neighbors to wake up to cover their sensing area. It also means there is no disjoint relationship between level of coverage, which will be another consideration for deploying the sensors. Fig. 4.4 shows the correlation between off-duty number and survival time. It also demonstrates that more off-duty nodes indeed increase more survival time. As we can tell from the increase of the slope, the more deployed node numbers can the system prolong more system survival time of K-CP than [6], because it increases the probability of disjoint situation of coverage degree.
The other characteristic of our proposed protocol is that even the network can’t maintain K, it still provides more stable coverage degree. If the sensor network calculates its average coverage degree after each cycle, as Fig. 4.5 illustrated, the slope whether perform the algorithm shows great variations in intensity, though both coverage degree can not remain above K. When we deploy 200 sensor nodes in a square space (50m by 50m), the vertical line means the time when the system can not maintain the K coverage degree, and we can see the slope of average coverage degree descends much more than the eligible rule algorithm is performed. It means that although some area of the sensor network occur blind points, most area can still be sufficiently covered.
Figure 4.2: deployed number vs. coverage degree
Figure 4.3: deployed number vs. survival time
Figure 4.4: [6] and K-CP deployed number vs. survival time
200 800 1400 2000 2600 3200 3800 4400
time unit
200 800 1400 2000 2600 3200 3800 4400
time unit
averagecoveragedegree
before scheduling after scheduling
Figure 4.5: average coverage degree vs. time unit
Chapter 5 Conclusions
In this paper, we have proposed a mechanism, namely eligible rule for node’s turning off calculation, safely turning off procedure to prevent blind points occurred, and scheduling rules for nodes to be on- or off-duty for power saving, which is based on distributed method to calculate coverage degree of each node. With the proposed techniques, we can guarantee the network is fully covered at least K level of cov-erage and also let maximum nodes to sleep. Our proposed solution can also work out with the ability to adapt the topological change. Applying the mechanism to asynchronous sensor network is currently our work.
Bibliography
[1] C.-F. Huang and Y.-C. Tseng. The coverage problem in a wireless sensor net-work. In ACM Int’l Workshop on Wireless Sensor Networks and Applications (WSNA), pages 115–121, 2003.
[2] D. Li, K. D. Wong, Y. H. Hu, and A. M. Sayeed. Detection, classification and tracking of targets in distributed sensor networks. IEEE Signal Processing Mag., 19(2):17–29, Mar. 2002.
[3] D. Nicules and B. Nath. Ad-hoc positioning system (APS) using AoA. In IEEE INFOCOM, 2003.
[4] A. Savvides, C.-C. Han, and M. B. Strivastava. Dynamic fine-grained localiza-tion in ad-hoc networks of sensors. In ACM Int’l Conf. on Mobile Computing and Networking (MobiCom), pages 166–179, 2001.
[5] S. Slijepcevic and M. Potkonjak. Power efficient organization of wireless sensor networks. In IEEE Int’l Conf. on Communications (ICC), pages 472–476, 2001.
[6] D. Tian and N. D. Georganas. A coverage-preserving node scheduling scheme for large wireless sensor networks. In ACM Int’l Workshop on Wireless Sensor Networks and Applications (WSNA), 2002.
[7] X. Wang, G. Xing, Y. Zhang, C. Lu, R. Pless, and C. Gill. Integrated coverage and connectivity configuration in wireless sensor networks. In ACM Int’l Conf.
on Embedded Networked Sensor Systems (SenSys), pages 28–39, 2003.
[8] F. Ye, G. Zhong, S. Lu, and L. Zhang. PEAS: A robust energy conserving protocol for long-lived sensor networks. In Int’l Conf. on Distributed Computing Systems (ICDCS), 2003.
[9] F. Ye, G. Zhong, S. Lu, and L. Zhang. Energy Efficient Robust Sensing Coverage in Large Sensor Networks. In UCLA Technical Report, 2002.
[10] R. Min, and A. Chandrakasan Energy-Efficient Communication for Ad-Hoc Wireless Sensor Networks. In Asilomar Conference on Signals, Systems, and Computers, vol. 1, pp. 139-143., Nov. 2001.