Robot Navigation Based on Situational Awareness

Abstract

Autonomous navigation for mobile robots is a prerequisite to complete the tasks. In this article, we propose a novel navigation method based on situational awareness. The scene predictor, scene interpreter and the topological map-based planner are the main components, where the scene interpreter is used to realize the mapping of perception information, and the scene predictor predicts the information in the neighborhood of current position. The basic idea of this research is the feeling of path. Based on the definitions of path and the amount of perception, the situational awareness fitting network is designed as the scene interpreter, and the position of the robot in a path can be implicitly described by a situational awareness value. With the global guidance of the planner, the robot achieves navigation. The proposed navigation method does not rely on Cartesian-based global position and its effectiveness is verified by simulations and experiments.