Multiple cooperative UAVs target tracking using Learning Based Model Predictive Control

In this paper, formation of a group of multiple cooperative unmanned aerial vehicles (UAVs) in a desired geometrical pattern while tracking an aerial target is implemented using decentralized Learning Based Model Predictive Control (LBMPC). The LBMPC is a new control technique that combines statistical learning along with control engineering providing guarantees on safety, robustness and convergence. The controller derived in this paper demonstrates the ability of the vehicles to cooperate, in a decentralized manner, to solve the formation problem in the presence of system uncertainties. The proposed controller respects the general formation constraints known as Reynolds rules of flocking during simulations. Our main contribution in this paper lays in the use of decentralized LBMPC in solving the problem of formation for a group of cooperative UAVs tracking an aerial target in the presence of unmodeled dynamics. A theoretical proof for stability will support our proposed controller.