[3K1-GS-9] Human interface

[3K1-GS-9-05] Proposition of Adaptive Planning for Visually Impaired Walking Assistance System Using Proactive Sensing

〇Yuya Osaki¹, Yukie Nagano², Daiki Takamura¹, Yoshimune Hiratsuka³, Yuichi Nagaoka⁴, Satoshi Kurihara² (1. Graduate School of Science and Technology, Keio University, 2. Faculty of Science and Technology, Keio University, 3. Department of Ophthalmology, Juntendo University School of Medicine, 4. Tokyo Independent Living Support Center for the Visually Impaired)

Keywords:Support for the Visually Impaired, Proactive Sensing, Adaptive Planning

Image-based object detection technology plays an important role in assisting the visually impaired to walk. Recently, neural network-based detection models have improved the accuracy of this technology, but there is a problem that it is difficult to detect objects that appear small in the input image. As an approach to this problem, it is necessary to make the object appear larger in the input image by moving the angle of view of the camera. In this study, we propose a proactive sensing system that performs appropriate processing depending on the object. The system estimates the range where the object can be located by integrating the knowledge of the object obtained from a large-scale language model with the context of the current environment already obtained. This enables a range-focused detection process and an appropriate approach for small objects. In addition, to assist the visually impaired in walking in dynamic and complex real-world environments, object selection is based on Agent Network Architecture, a form of adaptive planning.