Follow me: an architecture for user identification and social navigation with a mobile robot

TL;DR

The paper tackles socially responsible navigation for service robots by proposing a ROS2-based architecture that identifies a user via gesture signaling and facial recognition and then guides them while continuously monitoring the distance to ensure safety. The approach combines computer vision modules (gesture, skeleton, and face recognition) with a distance-based control loop, implemented across a three-node ROS2 architecture to maintain a user within a desired range, demonstrated on a MiR100 platform with a RealSense sensor. Experimental validation shows the robot can initiate guidance via gesture, maintain tracking through facial identification, and stop when the distance exceeds $2$ meters, with Exponential Moving Average smoothing used to mitigate depth-sensing noise. The work contributes a concrete, modular framework for intuitive, socially aware human-robot interaction and provides a foundation for future enhancements such as collision avoidance and environment mapping, expanding practical deployment in public venues.

Abstract

Over the past decade, a multitude of service robots have been developed to fulfill a wide range of practical purposes. Notably, roles such as reception and robotic guidance have garnered extensive popularity. In these positions, robots are progressively assuming the responsibilities traditionally held by human staff in assisting customers. Ensuring the safe and socially acceptable operation of robots in such environments poses a fundamental challenge within the context of Socially Responsible Navigation (SRN). This article presents an architecture for user identification and social navigation with a mobile robot that employs computer vision, machine learning, and artificial intelligence algorithms to identify and guide users in a social navigation context, thereby providing an intuitive and user-friendly experience with the robot.

Figures (3)

• Figure 1: Mobile robot MiR100 during experimental validation.
• Figure 2: Proposed architecture.
• Figure 3: Plot representing the measured distances from the RealSense with and without the use of EMA in parallel with the robot speed during experimental validation.