Reinforced-SLAM for path planing and mapping in dynamic environments

Abstract

In this work, an artificial intelligence approach to the problem finding a path for exploring an unknown environment and at the same time creating a map with uncertainties in robot pose and measures, while locating itself with this map (SLAM problem) is used to create an intelligent, robust and efficient navigation system for robots. We propose the integration of two of the most widely used approaches for the implementation of autonomous systems, the reinforcement learning for navigation in unknown and dynamic environments, along with the SLAM (Simultaneous Localization and Mapping) type algorithms for localization and mapping the environment. Experiments in section IV also confirms the algorithm performance in presence of uncertainties on mapping and sensor readings for the path planing problem.