Autonomous Robot Navigation System Based on Double Deep Q-Network

Abstract

Various types of mobile robots with different engines, sensors, operating systems and continuous development in them. The first aim of this paper, use reinforcement learning algorithm for implement a working agent capable Explore the destination to deal with sparse reward signals controllers motion planning autonomous mobile robot navigation system. with one CPU multi core with store the result of agent to avoid run the experiment in same environment more than one time and store the result, to save time with fast process to ability move independently use different input of the data localization using RTK-GNSS, navigation system with grid maps or monocular camera to 2D.