Non-Linear Model Predictive Control for autonomous landing of a UAV on a moving platform

This work proposes a real-time Model Predictive Control (MPC) solution for the landing problem of a quadrotor on an moving platform whose dynamics is unknown. The aerial vehicle is capable of acquiring only bearing measurements and of retrieving its attitude and elevation; its objective is to autonomously track the target and safely land over it. To perform the design of the control strategy, a fast prototyping approach is proposed, in which Matlab is used in conjunction with Acado toolbox in order to attain both a low development time and a computationally efficient MPC solution suitable for the on-board deployment on resource constrained hardware. Performances are assessed by laboratory experiments with an indoor aerial platform in which the controller is implemented on an embedded device (Raspberry Pi 4) with limited computational power, carried on-board. The obtained results show that even in this scenario, the adopted approach and the Acado generated MPC solver are able to attain real-time performances and safely completing the required task.