Autonomous Person-Specific Following Robot

Wesley P. Chan,Sina Radmard,Zhao Quan Hew,Jon Morris,Elizabeth Croft,H.F. Machiel Van der Loos

DOI: https://doi.org/10.48550/arXiv.2010.08017

2020-11-11

Abstract:Following a specific user is a desired or even required capability for service robots in many human-robot collaborative applications. However, most existing person-following robots follow people without knowledge of who it is following. In this paper, we proposed an identity-specific person tracker, capable of tracking and identifying nearby people, to enable person-specific following. Our proposed method uses a Sequential Nearest Neighbour with Thresholding Selection algorithm we devised to fuse together an anonymous person tracker and a face recogniser. Experiment results comparing our proposed method with alternative approaches showed that our method achieves better performance in tracking and identifying people, as well as improved robot performance in following a target individual.

Robotics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that in human - machine collaborative applications, service robots need to be able to specifically follow a certain user. However, most of the existing people - following robots do not know who they are tracking when tracking people. This paper proposes an identity - specific person tracker, which can identify and track nearby people, thereby achieving the following of specific users. Specifically, the method proposed in the paper uses the Sequential Nearest Neighbour with Thresholding Selection (SNNTS) algorithm, combined with an anonymous person tracker and face recognition technology, to achieve accurate tracking and identification of specific individuals.

The paper verifies the effectiveness of the proposed method through experiments. The experimental results show that, compared with other methods, this method performs better in tracking and identifying people, especially in the task of following a specific target individual, and the performance of the robot has been significantly improved.