In this project, we build a bot that can follow a person in real-world challenging situations. This bot has many applications such as autonomous carts in grocery stores, personal guides in hotels and hospitals, or as an autonomous suitcase. A person following robots in dynamic environments needs to address the tracking problem under different challenging situations (appearance changes, varying illumination, occlusions, pose changes such as crouching, exchanging jackets, etc.). 

An online convolutional neural network (CNN) is used to track the given target under different situations. The target being tracked might move around corners making it disappear from the field of view of the robot. We address this problem by computing the recent poses of the target and have the robot replicate the local path of the target when the target is not visible in the current frame. We made a robot using DC encoder motors with Arduino, Raspberry Pi, and Realsense Depth Camera.

We are following a research paper Integrating Stereo Vision with a CNN Tracker for a Person following Robot.

Project by: 

Gaurav Bansal(20165142),  Anuj Tiwari(20176032),  Ayush Agrawal(20175064)