At some level of your simulation you have x and y coordinates for each node. When you initialize the whole simulation you can use your global knowledge of all of the coordinates to make assignments of nodes to macrocell. After that, each cluster head could know where its components are because your simulation could take advantage of its knowledge of all of the positions in order to just assign the position information in each cluster head.
This will not make for the most realistic simulation, but it works as a simulation.
To be more realistic, each node should need to figure out where it is, and each cluster head should need to figure out where its members are.
There are multiple ways for a node to figure out where it is. One of the ways is for the node to receive GPS updates and to decode them. Your simulation could go as far as to simulate multiple GPS transmissions and have each node go through the trouble of decoding them. Or you could keep an external array of positions and just simulate informing each node of where it is (possibly only to within 3 meters of your global knowledge of where it really is, to simulate the lack of precision in GPS calculations.)
GPS is only one method for nodes to figure out where they are; you should research other methods.
Once a node knows where it (thinks) it is, it could communicate that knowledge to other nodes. There are multiple possible protocols for that.
More common would be for nodes to figure out where other nodes are by measuring received signal strength together with information about how strong the transmission was. This gives distance rather than absolute coordinates or rather than distance and direction. (If you are simulating transmission equipment that actively transmits a focused beam rather than using a mostly omnidirectional transmission then you need a more complicated arrangement.) With at least three nodes that can hear each other you can start doing triangulation and so start figuring out relative positions. If absolute positions are known for at least two of the nodes then you can put the triangulation information together with the two known positions to assign absolute positions in two-space (if you only know the absolute position of one node, the position calculations can be wrong by both rotation and reflection. Knowledge of two absolute positions eliminates rotation, but might not eliminate reflection along the line joining the two.)
There are multiple protocols developed for nodes to exchange triangulation information and so build up maps. If you need to simulate that, then you need to research the possibilities and choose one to simulate.
But at this level, these are not MATLAB questions: these are questions about mobile technologies. You should be reading about those and making decisions about what exactly you are going to simulate. Only then should you worry about matters like whether you should use cell arrays or structures or whatever.
