Reinforcement Learning Approach for Navigation of Ground Robotic Platform in Statically and Dynamically Generated Environments

摘要

This paper considers robotic platform navigation in terms of logistics, movement and track routing within indoor environments. Smart navigation and platform routing using a neural network are investigated. The paper discusses environment modeling with Unity ML software suite in static (prefabricated) and dynamically generated environments. Along with reinforcement learning, a procedural generation approach and its possible industrial applications are considered. The proposed algorithm for environment generation is characterized by higher performance comparing to analogues and allows to avoid model overfitting.