Gaussian Process-based Learning Control of Aerial Robots for Precise Visualization of Geological Outcrops

To generate 3D virtual maps of outcrops in geoscience, manual flights of aerial robots are often employed which is challenging due to various reasons: 1) piloted flight over curved/uneven surfaces requires auto-focusing, 2) wind disturbances make it difficult to precisely maintain the desired overlap, and 3) hiring a skilled pilot is expensive as the process requires hours of data collection. In this work, we propose to fully automate the visualization process using a learning-based control framework, i.e., position tracking nonlinear model predictive controller in conjunction with Gaussian process (GP)-based disturbance regression which facilitates a precise tracking of the generated path. Thanks to the long-short term memory feature of the designed GP model, the disturbance forces are accurately estimated even for increasing magnitude levels and time-periods. The simulation and real-world tests manifest that the proposed method could provide a time- and cost-saving yet reliable visualization framework.

[1]  Erdal Kayacan,et al.  Online Learning-based Receding Horizon Control of Tilt-rotor Tricopter: A Cascade Implementation , 2018, 2018 Annual American Control Conference (ACC).

[2]  C. Rasmussen,et al.  Gaussian Process Priors with Uncertain Inputs - Application to Multiple-Step Ahead Time Series Forecasting , 2002, NIPS.

[3]  Efe Camci,et al.  Automated Tuning of Nonlinear Model Predictive Controller by Reinforcement Learning , 2018, 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).

[4]  Haikun Wei,et al.  A Gaussian process regression based hybrid approach for short-term wind speed prediction , 2016 .

[5]  Erdal Kayacan,et al.  Online Deep Learning for Improved Trajectory Tracking of Unmanned Aerial Vehicles Using Expert Knowledge , 2019, 2019 International Conference on Robotics and Automation (ICRA).

[6]  Erdal Kayacan,et al.  Receding horizon control of a 3 DOF helicopter using online estimation of aerodynamic parameters , 2018 .

[7]  Angela P. Schoellig,et al.  Learning‐based Nonlinear Model Predictive Control to Improve Vision‐based Mobile Robot Path Tracking , 2016, J. Field Robotics.

[8]  Claire J. Tomlin,et al.  Learning quadrotor dynamics using neural network for flight control , 2016, 2016 IEEE 55th Conference on Decision and Control (CDC).

[9]  H. Jin Kim,et al.  Model predictive flight control using adaptive support vector regression , 2010, Neurocomputing.

[10]  Erdal Kayacan,et al.  A Simple Learning Strategy for Feedback Linearization Control of Aerial Package Delivery Robot , 2018, 2018 IEEE Conference on Control Technology and Applications (CCTA).

[11]  Girish Chowdhary,et al.  Learning-Based Fast Nonlinear Model Predictive Control for Custom-Made 3D Printed Ground and Aerial Robots , 2018, Handbook of Model Predictive Control.

[12]  Marc Peter Deisenroth,et al.  Efficient reinforcement learning using Gaussian processes , 2010 .

[13]  M. Diehl,et al.  Real-time optimization and nonlinear model predictive control of processes governed by differential-algebraic equations , 2000 .

[14]  Zhile Yang,et al.  Time series wind power forecasting based on variant Gaussian Process and TLBO , 2016, Neurocomputing.

[15]  Dogan Kircali,et al.  An Intelligent Hybrid Artificial Neural Network-Based Approach for Control of Aerial Robots , 2019, Journal of Intelligent & Robotic Systems.

[16]  Claire J. Tomlin,et al.  Learning-based model predictive control on a quadrotor: Onboard implementation and experimental results , 2012, 2012 IEEE International Conference on Robotics and Automation.

[17]  Erdal Kayacan,et al.  A constrained instantaneous learning approach for aerial package delivery robots: onboard implementation and experimental results , 2019, Auton. Robots.