An Intelligent Robotics Device (IRD) is developed in order to support safe mobility for disabled or elderly people with various impairments. The proposed system offers five main functions: obstacle detection and avoidance through bone conduction, live tracking, geofencing, GPS navigation. First, through a combination of an ultrasonic sensor and bone conduction phone, it detects diverse obstacles and produces that describe environmental information, including the positions and sizes of obstacles, which is then given to the learning-based algorithm. By learning the common patterns among assigned to the same directions, the IRD can automatically find paths to prevent collisions with obstacles. Second, it distinguishes a situation whereby the user is standing on a sidewalk, traffic intersection, or roadway through analysing the texture and shape of the images, which aids in preventing any accidents that would result in fatal injuries to the user, such as collisions with vehicles. In this project, a novel IRD system is presented that provides more safety for people with impairments and for elderly people. In order to assure safe mobility, our product offers gps live tracking, GPS navigation and geofencing. With these three functions, it can perceive obstacles of various types and recognize dangerous situations, and then recommend viable paths to evade them. First, obstacles are identified using a combination of ultrasonic ones, and then the avoidable directions are determined using learning based algorithms. Second, in order to prevent collisions with vehicles at traffic intersections, the situation recognition component distinguishes the type of place where the user is currently located as a sidewalk, an intersection. This work aims to provide safe mobility to wheelchair users while they control the wheelchair toward a destination. In order to support safe mobility, the wheelchair must detect a range of obstacles and dangerous situations in real environments and generate avoidable paths to prevent collisions with them. In order to achieve this, a hybrid obstacle avoidance method and a situation recognition method are proposed. - View it on GitHub