MPC is "powerful" controller if it is in hand of expert. Great power in hands of expert comes from huge number of DOF where its tiny details can be optimized or used for better control and we known that with big power comes big responsibility to justify use of "expensive" type of control. In this master thesis we process the problem of using electrical energy from battery "as little as possible" for control of moving masses and successfully accomplish desire mission of MORUS project. For this purpose we use MPC control to get the most out of it. For lack of MPC optimizer we modified algorithms for neural network so that we can use them for optimizing MPC cost function at every discrete time step. One of the hardest problem in MPC control is finding "optimal" control matrices. For this purpose we use GA and PSO algorithms to find "optimal" matrices for desire problem. In this thesis we implement linear and nonlinear MPC for roll, pitch, yaw as well as x, y and z axes control and compare pros and cons for using nonlinear versus linear MPC. At the very end we implement linear MPC for linear x/y velocity of non-tilting UAV. For executing any "complex" mission we need trajectory planner. For that purpose we implement local trajectory planner from article. To get one step closer to real system we implement simulation in GAZEBO environment. To create real conditions we implement drag GAZEBO plugin.