AUTONOMOUS NAVIGATION OF MOBILE ROBOT IN DYNAMIC ENVIRONMENTS USING DEEP REINFORCEMENT LEARNING

Abstract

The aim of this work is to develop an autonomous navigation algorithm based on artificial intelligence called Deep Reinforcement Learning for human-robot mobile interaction. The development of the robot has been considered in dynamic environments, which have been scanned by means of a 2D Lidar sensor, taking into account the sudden change of direction of objects. The proposed research has been exploratory and experimental, carrying out a search of the state of the art in terms of autonomous navigation and human-robot mobile interaction. Artificial intelligence has been used as the main technique, specifically the union of deep neural renders with Reinforcement Learning. This allows the robot to learn based on the reward or penalty according to the results sent by the deep neural network.