Decentralized Nonlinear Model Predictive Control for 3D Formation of Multirotor Micro Aerial Vehicles with Relative Sensing and Estimation

In recent years, extensive research is conducted on the coordination and cooperation strategies of multirotor Micro Aerial Vehicles (MAVs) to perform high-level missions such as scientific exploration, search and rescue, intelligence gathering etc. [1], [2]. The main motivator for this interest is the fact that the deployment of multiple vehicles reduces the risk of mission failures and provides higher performance and flexibility through parallelism [3]. Among the main subproblems of cooperative control, formation control is usually an essential component and Model Predictive Control (MPC) is a promising tool to carry out this task deliberately. Since MPC is architecturally flexible and handles the performance and constraints systematically in parallel, it is drawing more attention nowadays [4]. Among MPC methods, especially Nonlinear Model Predictive Control (NMPC) is particularly suitable to control the robots whose fast dynamics are needed to be predicted by nonlinear models and constraints as in multirotor MAVs. Additionally, for large scale systems, Decentralized NMPC (D-NMPC) strategies are advantageous since they address the computational complexity by dividing the overall optimization problem into decoupled subproblems and by reducing communication requirements [4]. Furthermore, in order to deploy highly autonomous multi-rotor MAVs in non-trivial environments, several researchers focus on elaborating local and relative sensing in formation control and try to solve its limitations [5].