A Review of Visual Perception for Mobile Robot Navigation: Methods, Limitation, and Applications

Abstract

Successful robot navigation in unknown environments relies mostly on the surrounding information they perceive. This paper presents a review of various state-of-the-art vision-based methods that deal with the perception problems for mobile robot navigation and control in unknown environments. These methods are thoroughly discussed in a few groups, including visual Simultaneous Localization and Mapping (SLAM), bio-inspired vision, and machine learning, and in terms of their working principles, advantages and disadvantages for robot navigation, and current research trends. Since each group has different characteristics and challenges, the selection of an appropriate group should be made based on the purpose of the application. This review shows that the current research focus has been shifted toward lightweight perception solutions that could reduce computational time and improve accuracy, enabling real-time processing for robot navigation.