According to my survey i found a software like Matlab

you can use MultiUAV2 simulation platform. Please see the appendix A of  the attached book: "UAV Cooperative Decision and Control", edited by Ral Shima and Steven Rasmussen. 

It really depends on the requirements of the scenarios you are studying, in particular in terms of how realistic you want your modelling of UAV core parameters (e.g. yaw/pitch/roll, forces applied, coordinates/speed/acceleration, etc.) to be. As an example, if you are investigating UAV mobility control / path planning strategies that are essentially network-driven and that require only to compute mobility pattern, speed and a rough estimation of forces applied and acceleration, you could be perfectly happy with a traditional discrete-event network simulator (e.g. ns-3 could do the job, among others). So, unless you have specific needs for UAV modelling in terms of aerodynamics, a generic network simulator may work for you. If your applicative scenarios & system requirements incur a more detailed model, then you need to go for specific tools here. Can't say more with your input assumptions. Hope it helps, though.

If you do need aerodynamic simulation, X-Plane works quite well. We used that along with matlab to simulate a few rotorcraft uavs. X-plane has a simple network based API that you can use to provide inputs to controls and get back pretty much any variable used in the aerodynamic simulation.

I agree with Jeffrey Craighead. Since you mention things like efficiency and collision avoidance you may need the increased fidelity that a full aerodynamic simulation gives you. I would recommend connecting X-Plane to MATLAB/Simulink through the UDP API.