MASHINE REINFORCEMENT LEARNING FOR NAVIGATION OF MOBILE ROBOTS
Article Sidebar
Main Article Content
Abstract
New algorithm of machine learning for navigation of mobile robot navigation is introduced. It based on combination of Deep-Q-Learning and Double Q-Learning. Model is considered the movement of a mobile robot in some environment (environment is set Gazebo program package), known robot location and prevented obstackle collisions by navigation. Mobile Robotics Stimulation Toolbox and Gazebo visualization packages are used as Software. It is shown that the testing of new algorithm more than 10 times improved time characteristics in comparative with traditional algorithms of machine learning. The present algorithm may be to integrate in the apparatus means.
Article Details
This work is licensed under a Creative Commons Attribution 4.0 International License.
References
Назарова, А. В. Методы и алгоритмы мультиагентного управления робототехнической системой / А. В. Назарова,
Т. П. Ригова // Вестн. МГТУ им. Н. Э. Баумана. Сер. Приборостроение. – 2012. – С. 93–105.
Ростовцев, П. С. Обучение роботизированных систем с помощью нейронных сетей / П. С. Ростовцев, Д. Н. Васильев, М. И. Озерова // Россия молодая: передовые технологии в промышленности. – 2017. – № 2. – С. 123–125.
Neural Network-Based Learning from Demonstration of an Autonomous Ground Robot / Y. Fu, [et al.] // Machines. –2019. – V. 7, № 2. – DOI: https://doi.org/10.3390/machines7020024.
Thanh, T. Deep Reinforcement Learning for Multiagent Systems: A Review of challenges, Solution and Applications / T. Thanh, N. Nguyen, S. // IEEE Transactions on Cybernetics. – Vol. 50, no. 9. – P. 3826–3839, 2020. – DOI: 10.1109/TCYB.2020.2977374.
Описание пакета ROS Toolbox {Электронный ресурс]. – Режим доступа: https://www.mathworks.com/matlabcentral/fileexchange/66586-mobile-robotics-simulation-toolbox. – Дата доступа: 23.11.2020.