Adaptive object tracking of quadrotor in presence of disturbance

In this paper, a moving object tracking problem for quadrotor systems, containing model uncertainties, flying in dynamical and more practical environments such as GPS-denied and wind disturbance, is considered. For moving object detection, the maximum likelihood method is adopted to recover the trajectory of it based on features obtained from ORB-SLAM. For flight control, a discrete-time adaptive control method is developed in two steps. In the first step, the sliding mode and discrete-time adaptive control are adopted to overcome model uncertainties and plan desired attitudes. In the second step, the discrete-time adaptive control is adopted to overcome model uncertainties and external disturbance, and to track desired positions. The results are verified by two experiments. In the first experiment, a quadrotor with external 20 grams disturbance is used to track a 100 centimeter sinusoidal wave, and then the root mean square (RMSE) is calculated for comparison: by proposed method, the RMSE is 40 centimeter; by robust backstepping sliding mode control, 127 centimeters. In the second experiment, the mean error of human detection for a person standing 700 centimeters in front of a quadrotor is 40 centimeters.