Deep Neural Network Based Model Predictive Control for Standoff Tracking by a Quadrotor UAV

Abstract

The standoff tracking requires an unmanned aerial vehicle (UAV) to loiter in a circular orbit above a target of interest. To achieve it, we propose a deep neural network (DNN) based model predictive control (MPC) for a quadrotor UAV by taking into account the full UAV model and input constraints. Moreover, we propose a new Lyapunov guidance vector (LGV) with tunable convergence rates to plan a reference trajectory for the MPC. The computation latency on the field-programmable gate array (FPGA) at 200MHz is significantly reduced to a constant of 0.12ms. The hardware-in-the-loop (HIL) experiments verify the effectiveness and robustness of our method.