Decentralized Navigation of Multiple Quad-rotors using Model Predictive Control

Abstract: In this thesis, we develop a model predictive control (MPC) scheme for the navigationof multiple quadrotors in an environment with obstacles. The overall controlscheme is decentralized, since each quadrotor calculates its own signal based on localinformation. The MPC constraints take care of collision with the static obstacles,inter-agent collisions as well as input saturations. Firstly, we formulate and solvethe problem using a nonlinear MPC framework, where the agents and the obstaclesare modelled as 3D spheres. Secondly, to deal with complexity issues, we linearizethe model and constraints by employing polyhedral sets, and we solve the problemwith linear MPC. Thirdly, we use a mixed logical dynamical (MLD) framework tosolve our problem, which is then incorporated into a hybrid MPC problem. The performanceof the proposed solutions is demonstrated through computer simulationsand real-time experiments.