A real-time model predictive position control with collision avoidance for commercial low-cost quadrotors

Abstract

Unmanned aerial vehicles (UAVs) are the future technology for autonomous fast transportation of individual goods. They have the advantage of being small, fast and not to be limited to the local infrastructure. This is not only interesting for delivery of private consumption goods up to the doorstep, but also particularly for smart factories. One drawback of autonomous drone technology is the high development costs, that limit research and development to a small audience. This work is introducing a position control with collision avoidance as a first step to make low-cost drones more accessible to the execution of autonomous tasks. The paper introduces a semilinear state-space model for a commercial quadrotor and its adaptation to the commercially available AR.Drone <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> system. The position control introduced in this paper is a model predictive control (MPC) based on a condensed multiple-shooting continuation generalized minimal residual method (CMSCGMRES). The collision avoidance is implemented in the MPC based on a sigmoid function. The real-time applicability of the proposed methods is demonstrated in two experiments with a real AR.Drone quadrotor, adressing position tracking and collision avoidance. The experiments show the computational efficiency of the proposed control design with a measured maximum computation time of less than 2ms.