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

The coordination and cooperation strategies of Micro Aerial Vehicles (MAVs) become increasingly popular to carry out high-level, complex tasks such as object transporta- tion, patrolling, inspection robustly and with high performance. One of the crucial component of these strategies is formation control. Most formation control approaches assume absolute localization and/or leverage communication between robots. In contrast, the main focus of this paper is three-dimensional decentralized formation control of multi-rotor MAVs by using exclusively relative sensing and estimation, eliminating any explicit communication between robots. The proposed method- ology is Decentralized Nonlinear Model Predictive Control (D- NMPC) approach, to aim optimality while satisfying constraints and scalability, in a leader-follower scheme in conjunction with Pose-Graph Moving Horizon Estimator (PG-MHE), to estimate neighbour state while filtering noise effectively. After intro- ducing a realistic six degrees of freedom (DOF) mathematical model of MAVs, the problem is formulated based on the local coordinate frames of robots. This type of formulation makes the formation independent of the full knowledge of global or common reference frames. Then, PG-MHE is presented to filter out and estimate the required states by D-NMPC. The next step is to propose a D-NMPC strategy, construct an Optimal Control Problem (OCP) and solve it by modified Real-time Iteration (RTI) scheme. A comprehensive simulative scenario consisting of three quadrotors is designed to observe and validate the convergence of the method. The initial results of this study show that satisfactory performance and convergence are achieved under noise in all local sensors and in cases where the dynamics of the formation quickly changes.