Navigation of Cart Follower Using Artificial Intelligence

Abstract

Wheelchair users encountered difficulties in carrying their items. A person who uses a wheelchair will get easy to move around with a wheelchair cart follower should be implemented to track and follow the wheelchair automatically while maintaining wireless connect 40 cm with the wheelchair to avoid any unnecessary collision. This project is developed using the camera and ultrasonic sensor inputs as the input to artificial intelligence (AI) to achieve such specifications. In this project, the target object, the wheelchair, is illustrated using a frisbee, while a robot car model illustrates AI. An obstacle avoidance algorithm and an object tracking algorithm using AI are developed. The obstacle avoidance algorithm detects the distance between the target object and the cart follower and controls the cart follower to move backwards if the distance is less than 40 cm. Object tracking algorithm uses deep neural network (DNN) transfer learning in object detection application with pre-trained Single Shot Detector (SSD.) The model was developed to track the target object’s position and react accordingly once the target object moves. Based on this proposed system, the obstacle avoidance algorithm achieved 99.66% accuracy in distance ranging. In comparison, the object tracking algorithm provided a complete navigation system with 90% efficiency for the cart follower to track and follow the movement of the target object.