how to select the gateway node closest to the cluster head node using dijkastra algorithm

let us suppose there is 'n' cluster head nodes with there respective positions and id's and 10 gateway nodes are present randomly.....i want to search closest gateway node to each cluster head so that CH may sends its data to the nearest gateway........................plz reply me

Answer 1

[KSSV]

doc knnsearch....with the help of knnsearch you can get the nearest neighbours.

Answer 2

[Walter Roberson]

Dijkstra's algorithm is only for the case where something is to travel from node to node to node, passing through as many nodes as necessary to keep the cost as low as possible.

That is not the situation that you have. You have divided into clusters and each node in the cluster is permitted to send only to the cluster head, and the cluster head is to send directly to nearest gateway node even if sending to a bunch of intermediate hops would be less costly.

With that kind of set up all your cluster heads need to know is their own coordinates, and the coordinates of each of the gateways, and the formula for calculating the cost of transmission given coordinates (which might well come down to standard euclidian formula, but not necessarily. For example in fog, attenuation is faster than just the square of the distance.)

[kiranpreet kaur]

sir for my problem which algorithm i used to find nearest gateway node to a cluster head......plz reply. Instead of knn which another alg. is to be used

[Walter Roberson]

pdist2() for any of the standard distance measures.

Note: the standard distance measures are not useful if you have obstacles.

   1   tree    A

               B

Now it might be less expensive for "1" to transmit to "B" than to "A"

[kiranpreet kaur]

sir tell me can i use eucledian distance to find minimum distance b/w cluster head nodes and gateway nodes, so that cluster head nodes can transmit data to respective nearest gateway nodes............is it efficiently improves energy

[Walter Roberson]

It depends on how you model transmission costs.

In the ideal free-space (air, no obstacles) case with undirected antennas, power required to transmit a message increases with the square of the distance, so you can use the square of the euclidian distance to estimate the power. For relative comparisons to know which is going to be least costly you can use plain euclidian distance in such a model. When every node can transmit as effectively to every other node, sending directly is never more expensive of total energy compared to using shorter hops.

However, in real sensor networks you do have obstacles and so the cost of sending point to point can be much higher than the cost of using hops.

Also when I say that the total cost will not be higher for going direct (in the no-obstacles model) this does not take into account that if you keep sending by way of the cluster heads directly, that they are going to be depleted of stored energy faster than the other nodes. Either you need to keep redesignating cluster heads so none run out quickly, or else you need to switch transmission strategies.

The argument about the total energy not being higher for sending point to point from cluster heads to the gateway nodes also means that the total energy could potentially be even lower if every node transmitted directly to a gateway node instead of going first to a cluster head.

This argument does imply that each node can directly reach each gateway just by raising transmission power, which is not correct: transmission power is limited by law and by physical limitations. So you have to know which nodes are within effective range of the gateways, and any node that is not in range has to relay through one that is in range. But that applies to cluster heads as well: you cannot guarantee that a cluster head will be within effective range of the closest gateway, so you really should be planning a relay system anyhow. And that relay system might involve dijkstra's algorithm.