无

Motes propagate backwards the new routing status. When any of the above changes occur a new clustered tree network topology is formed.

Fig. 1 Network Formation as a result of (Cloud B) Node A mobility

A node before advertising its presence must consult the already participating nodes. For this reason it broadcasts over the WMN (Cloud B) a cluster availability request. All nodes will propagate the request to the whole network. When this request reaches the coordinator node it will reply by denoting the PHY channels that are used by the participating nodes and their IDs. The coordination node will additionally denote its role. When this information is available the node will choose one of the remaining channels and will send its Presence Entry. If it does not receive an association request from at least one mote within the interval set by the coordinator, then either there are no (interested) motes in its range or it is using an occupied channel. The node will try again broadcasting in the WMN at random intervals. Once it receives an Association Request, it will acknowledge it and broadcast the new cluster information to the WMN. This will also support locking the channel for other interested nodes. The cluster information is maintained by the Cloud A coordinator. All cluster internetworking, including formation and release of clusters is performed in Cloud B.

Although the IEEE 802.15.4 MAC Disassociation Command may be used for nodes leaving, the most usual

reason for disassociation of Cloud B nodes is mobility. Therefore, such a command is not very useful (power shut down could be one case where this Command could be useful; exploitation of an accelerometer functionality is another one). Our CL protocol uses the following mechanism for cases where the Disassociation Command can’t be issued.

A node must inform all connected motes every τ seconds on its status. If a child hop 0 mote does not receive a Presence Entry from the node at which it is connected for t>2τ then it transmits a Query Beacon to that node. If after 2τ the mote still has not received a Presence Entry it concludes that the node is no longer available and tries to connect to other available nodes stored in the look up table it maintains. If such nodes are not available or if the look up table is empty the mote establishes its default Cloud A link and informs its default parent mote. Then it also informs all its neighbor motes (at child status by definition of the protocol) that had previously assigned themselves to its cluster so that they link to another node or return to their default status following a similar procedure to the one described above for hop 0 motes. Cloud A broken links (due to e.g. power exhaustion, hardware failure or mote movement) are treated as in IEEE 802.15.4 cluster-tree topologies [4]

III. ANALYTICAL MODEL

We assume our WSN has a sleeping policy for each mote and aim to calculate the event detection reliability R (minimal prescribed packet rate) as in [7]. Solving the Markov model for one Cloud B node we have

R =

(1) where 1/λ is the packet interarrival time, n the number of

motes and

Pb = 1 -

(2)

with v the probability for vacation period after Markov point (return from sleep and end of transmission) and nA is the average distance between two consecutive Markov points in sec of Cloud A motes, which depends on the random sleep period that determines the power consumption per mote. Thus, the cluster lifetime depends on the required event reliability and the number of motes.

Eq. (1) and (2) is the result of the fact that the idle state of the Markov chain is reached when the bu er is empty after transmission. Since the packet arrival rate to the node in the source cluster is λ, then probability of zero Poisson arrivals during unit backo period can be approximated with Taylor series.

When a second Cloud B node arrives, then

R=R1+R2 (3)

无

R1 and R2 being the packet rates for the two clusters. One can derive from (1-3) that, because the number of motes of Cloud A remains constant, the arrival of additional nodes lowers the optimal packet rate increasing the mote lifetime. We now change the Markov model so that we assume a multihop star topology. That would change eqs. (1) and (2). In this case v will lower as it is linearly dependent to the steady state probability that are k packets in each mote buffer

intermediate hopes. Thus, as additional nodes enter consumption drops because (a) the motes are assigned different clusters and (b) the average number of decreases.

In fig. 2 we present the results of our analytical number of motes assigned in each one of them, i.e. n-1=n2=n/2, n1=n2= n3=n/3 and a one level decrease of average number of hops.

Fig.2 Average sleep probability for the case of one (P1), two (P2)

and three (P3) Cloud B nodes as a function of Cloud A motes

IV. MODEL VALIDATION AND DISCUSSION

We define presence as a new way of routing. information identifies a mote in terms of its participation in a route (tree) in a sensor network. Our solution focuses on exploiting the collaboration of the two networks, achieving lower network formation traffic. The network formation protocol described here resembles design and performance issues of cluster interconnection for beacon-enabled 802.15.4 clusters. Our approach, in which the cluster coordinator is used to bridge clusters is known to be superior in terms of traffic. Note also, that in a simplified case, where enough nodes exist to cover the network area of Cloud A fully, i.e. so as all motes become hop 0 motes, then our protocol operates as a static LEACH network, which is known to have superior energy savings compared to other existing WSN MAC approaches [6].

Fig. 3 shows Opnet simulation results for a Cloud A of 50 sensors (Radio range of a sensor node 40 m, Packet length 30 bytes, IFQ length 65 packets, Transmit Power 0.660W, Receive Power 0.395W) randomly distributed in an area of 100mX20m and Cloud B nodes moving through at 2m/s on random paths. It is seen that on the average only three Cloud

B nodes are enough to minimize required hops. Standard deviation shows not significant statistical error.

V. CONCLUSION

The main contribution of the work is the presence algorithm that defines a new way of routing. A realistic application scenario for use of this algorithm has been described. Implementation testing of the algorithm and hops (upper) for nodes moving through the Cloud A area (averages of 250 repetitions). Cases of one (diamonds), two (squares) and three (dots) moving Cloud B nodes.

REFERENCES

[1] http:// [2] /

[3] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, A

survey on sensor networks, IEEE Communications Magazine, vol. 40, pp. 102–114, Aug. 2002.

[4] Rico Radeke, et al., On Reconfiguration in Case of Node Mobility in

Clustered Wireless Sensor Networks, IEEE Wireless Communications, pp. 47-53, Dec. 2008

[5] Jelena Misic, and Vojislav B. Misic, Bridging Between IEEE

802.15.4 and IEEE 802.11b Networks for Multiparameter Healthcare Sensing, IEEE J. Selected Areas in Communications, vol. 27, pp. 435-7, May 2009

[6] Kazem Sohraby, Daniel Minoli and Taieb Znati, Wireless Sensor

Networks: Technology, Protocols, and Applications, (John Wiley & Sons, 2007) .

[7] Jelena Misic and Carol J Fung, The impact of master-slave bridge

access mode on the performance of multi-cluster 802.15.4 network, Computer Networks, vol. 51, pp. 2411-2449, 2007.