Image-Based Localization and Navigation Using Deep Learning for an Omnidirectional Human Assistive Robot

In recent years, nursing care has faced severe labor shortages, highlighting the need for robots capable of autonomous navigation and assistance. Accurate self-localization is crucial for autonomous movement; however, existing indoor navigation methods often rely on expensive equipment and suffer from delays caused by intensive processing demands. This paper proposes an autonomous navigation system utilizing a deep learning model that requires only visual input. The proposed method has been implemented on a newly developed omnidirectional mobile assistive robot and its performance has been validated through experimental testing. The results show that the robot could follow its designated path with minimal deviation while providing standing assistance. This approach offers a cost-effective and versatile alternative to conventional indoor navigation systems.