Current Unmanned Vehicle (UV) navigation systems are capable of autonomous navigation among disperse obstacles. However, these systems may fail to guide vehicles through highly confined environments because they do not explicitly consider the geometry of the vehicle in the navigation task. This paper presents a methodology that enables the navigation of Unmanned Vehicles (UVs) in such 3D environments. The proposed approach uses a hybrid navigation architecture which employs a global path planner and a local obstacle avoidance methodology in parallel and combines them utilizing an improved Model Predictive Control (MPC) approach that incorporates the geometry of the UV in the cost function. Using MPC enables the UV to generate complex maneuvering trajectories while avoiding obstacles, respecting the dynamic characteristics of the UV and preventing state and input saturation. Simulations in 2D and 3D demonstrate the effectiveness of the proposed method for the navigation of a highly maneuverable Rotary Unmanned Aerial Vehicle (RUAV) in a highly confined environment.