Research on indoor robot navigation algorithm based on deep reinforcement learning

Abstract

Highly intelligent robot is the trend in the new round of development, and it is also the ultimate goal of robot navigation technology research.The ability to learn is an important embodiment of robot intelligence.Autonomous navigation is the basic function of indoor mobile robots <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[1]</sup> and it is the premise and key technology to realize the autonomy and intelligence of robots.With the continuous development of robot-related technologies and the continuous expansion of application scope, the tasks faced by robots are more challenging, and the work scenes are increasingly complex and diversified, which puts forward new requirements and challenges for navigation technology. Indoor robot navigation systems need not only to maintain accuracy and stability in fixed scenes, but also to be able to handle dynamic changes in the environment, and to take into account safety and the impact on human comfort in an environment that coexists with humans.This paper takes the indoor mobile robot platform as the research object <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[2]</sup> , introduces the machine learning algorithm based on neural network into the navigation task <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[3]</sup> , utilizes the existing research achievements in the field of deep learning to build a robot navigation system based on different learning modes, and endows the robot with intelligent perception ability and navigation strategy learning ability <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[4]</sup> .