In today's busy world, it is hard to get assistance and dependability not being an effective solution it is very difficult for people with mobility challenges to travel to their choice of destination. To address this challenge, we developed an autonomous mobility scooter than can navigate both indoors and outdoor using different sensors. However, one major challenge in autonomous navigation is obstacle avoidance. To provide safe navigation, we chose AI based obstacle avoidance system that can identify the obstacle class and behave accordingly while navigating autonomously. The project focuses on using low cost sensors to provide a cost effective solution that is affordable to people under moderate income group. Our autonomous mobility scooter uses a 2D-LIDAR, wheel encoder, IMU for indoor navigation and GPS, android application for outdoor navigation. The scooter has a vision based AI system for obstacle avoidance that uses ZED stereo camera to provide RGB image data for AI system and depth data under 10 meters for the obstacle avoidance system. The top level software includes Ubuntu operating system that runs ROS kinetic which handles the sensor data, decision making, providing commands to the scooter for navigation. Certain challenges faced using low cost sensors were address to the best of our capability to provide feasible solutions that have been implemented effectively. Compared to its counter part, the mobility scooter has certain limitations on regions that it can navigate in and speed of travel as a result of low cost of construction.
A detailed architecture diagram and explanation is provided in the document attached to this project