This thesis is focused on dynamic modeling, control system design and vision based collision avoidance for Multi-Rotor UAVs. Multi-Rotor UAVs are defined as the rotary-winged unmanned aerial vehicles with multiple number of rotors. These Multi-Rotor UAVs can be utilized in military situations such as surveillance and reconnaissance. Also, Multi-Rotor UAVs can be used in obtaining vision information from steep terrains or disastrous sites. In this thesis, Quad-Rotor, Tilt Tri-Rotor, and Coaxial Tri-Rotor models are introduced as well as their attitude and position stabilization simulation. Control system is designed based on PID autopilot and vision-based collision avoidance control technique. Additionally, Optical Flow method, which is one of the vision-based collision avoidance techniques, is introduced along with the depth map calculation, which compensates the distance calculation from the optical flow estimation. Integrating the Optical Flow and depth map calculation methods, so-called Reactive Collision Avoidance System is developed having synergetic effect from both the optical flow and the depth map calculations. Finally, Quad-Rotor UAV’s collision avoidance simulations in 3D virtual spaces are conducted.