2019/05/26 by Andriy Sarabakha, Sarabakha, Andriy, Erdal Kayacan +1
Computer Science · Engineering · #Adaptive Control of Nonlinear Systems #FOS: Computer and information sciences #Machine Learning (cs.LG) #Robotic Path Planning Algorithms #Robotics (cs.RO) #Target Tracking and Data Fusion in Sensor Networks
paper · pdf · doi:10.48550/arxiv.1905.10796
openalex publication_date 2019/05/26 · openalex created_date 2022/07/13 · openalex updated_date 2026/07/28
This work presents an online learning-based control method for improved\ntrajectory tracking of unmanned aerial vehicles using both deep learning and\nexpert knowledge. The proposed method does not require the exact model of the\nsystem to be controlled, and it is robust against variations in system dynamics\nas well as operational uncertainties. The learning is divided into two phases:\noffline (pre-)training and online (post-)training. In the former, a\nconventional controller performs a set of trajectories and, based on the\ninput-output dataset, the deep neural network (DNN)-based controller is\ntrained. In the latter, the trained DNN, which mimics the conventional\ncontroller, controls the system. Unlike the existing papers in the literature,\nthe network is still being trained for different sets of trajectories which are\nnot used in the training phase of DNN. Thanks to the rule-base, which contains\nthe expert knowledge, the proposed framework learns the system dynamics and\noperational uncertainties in real-time. The experimental results show that the\nproposed online learning-based approach gives better trajectory tracking\nperformance when compared to the only offline trained network.\n