Human Tracking and Following using Machine Vision on a Mobile Service Robot

Service robot technology is rapidly improving to give rise to robust and reliable machines operating alongside humans. This paper presents a human-following system that can identify a target human in a crowded environment and track the person’s motion, simultaneously avoiding obstacles while navigating through the environment. We implement the system on a mobile service robot platform with light detection and ranging (LIDAR) and RGBD sensors. The system uses a Discriminative Generative network (DG-net) for human detection. After detection, the localization module will locate the target person’s position in the environment. The navigation module generates a cost map of the surroundings for path planning. It allows the robot to navigate the changing environment avoiding obstacles while tracking the target person. Experimental results showed that the robot could identify and follow the target person reliably. At the same time, the robot navigates the crowded environment safely, avoiding other people and obstacles in the environment. Despite all that, the recovery module could not recover reliably after losing the target person. The demonstration video is available at https://github.com/LeoYong95/human_following.git