I am not that expert, but I think can be done by comparing their GPS locations.

usually, it is assumed that sensors are not equipped with GPS as this will add more cost. i want to ensure if there is any other way??  

I am too interested to know if there is any other way.

As sensor nodes are mostly stationary, it can use GPS one time a day to activate GPS, renew the location, then disactivate GPS then use the location in all calculation needed till next day for example.

Or when ordered to do so by the master node ( on-demand ).

By this it could be cost effective.

Thanks for your reply and your idea sounds good if we assume that sensors have GPS. 

i hope if anyone can answer us and tell if there is any other method 

Definitely, the GPS information is used for location information, you will read my article about VANET but same for WSN. It will help you to understand all about your question. 

i am not expert in WSN but , I think you may need to configure( manually) the gateway sensor node in which it can have the information of all nodes connected to it , such as location of a node and how to reach it (direction).

no need to use geographical based routing, however topology based routing is what you need to use.

Hope it works :-)

Each node knows it own location. So every node can broadcast its own location.

actually in many applications it is assumed that nodes don't know their location. adding something like GPS is considered a more cost that has to be avoided. so many papers search in the localization problem. i read some of them. but i actually ask for a determined point. the direction of the coming signal from the neighbor.how can it is recognized by the receiver. so for example, the node can know if one neighbor is located clockwise it. another one in counterclockwise.

Hi Walaa Ali. We can first assume that your sensor nodes do not have a GPS available for the sensor own location. So, in this case, you need to establish the initial position of each node and than treat the possible movements each node can do. In my paper, you can find an algorithm to identify the displacement of each node with rounds of analysis. After each round you can find the new displacement in the node trajectory. I hope it can help you for the first time approach in your research. 

Location finding system: Most of the sensor network routing techniques and sensing tasks require the knowledge of location with high accuracy.

mobilizer: May be needed to move sensor nodes when it is required to carry out the assigned tasks.

I think the sensor node can select its neighbor based on range ,  energy and distance.

Initially the neighbor list of a node is set to Null, In order to  select the neighbor we determine the energy of each node E(v), so the node list E(v) is sorted in decreasing order. After computing the energy we find the distance of each node ‘v’ from ‘n’ (i.e. the number of nodes) in order to determine the neighbor list of every node.

adding a GPS can add extra overhead for the sensors.

and you should remember that  sensors care about the data more than addresses because the redundant deployment ..just the opposite of the traditional network.. 

please check the edge detection in WSN it may help..