Authors: Kawade Sanskruti Avinash, Professor Bangar Abhishek Popat, Professor Bhosale Sachin Bajirao, Professor Karad V.A.
Abstract: Visual impairment restricts independent mobility and exposes people to collisions, falls and unsafe road situations. White canes and guide dogs remain the main aids, but they give little information about what an obstacle is, how far away it is, or when the user needs help. Artificial intelligence (AI) combined with wearable sensing and the Internet of Things (IoT) is increasingly proposed as a route to smart navigation and safety systems that perceive the scene, guide the user by audio or touch, and alert caregivers in an emergency. This paper presents a focused narrative review of five verified studies, published between 2018 and 2026 in IEEE, MDPI and River Publishers outlets, that cover sensor-based safety alerting, RGB-D and deep-learning navigation, machine-learning smart cane and glasses, edge-TPU indoor navi-gation, and an AIoT eyewear system with LiDAR ranging and an SOS function. The review examines sensing, AI per-ception, feedback and IoT integration. The literature shows that recent systems report object-detection accuracy of about 95 % and centimetre-level distance error in controlled tests, and that low-cost edge hardware such as the Raspber-ry Pi with a Google Coral TPU can run navigation models on the wearable itself. However, evidence is dominated by small prototypes, controlled or indoor settings, heterogeneous metrics and limited testing with visually impaired users, and few systems join navigation, perception and caregiver-facing safety in one design. The paper contributes a compar-ative analysis of the studies, a proposed conceptual “Third Eye” architecture that links wearable sensing, edge AI, IoT safety alerts and field validation, and a prioritised agenda of research gaps and directions.
