Follow Pedro! An Infrared-based Person-Follower using Nonlinear Optimization

TL;DR

This paper demonstrates a person-following system using infrared markers and AI algorithms like particle filters and nonlinear optimization on a ROS2 platform, showcasing integrated perception and navigation for robotics.

Contribution

It presents a complete infrared-based person-following system utilizing nonlinear optimization within ROS2, integrating perception and navigation on an embedded Linux platform.

Findings

Successful implementation of IR marker detection and tracking

Effective use of nonlinear optimization for SE(3) estimation

Integration of perception and navigation in ROS2 environment

Abstract

We used ROS2 as a platform to conduct AI research for developing a Follow-Me capability as a proof-of-concept on a wheeled robot, demonstrating that AI research is possible in the ROS2 framework. We developed a complete system that uses perception and navigation components based on a sensor suite of fisheye cameras, lidar, and IMU running on an ARM64 Embedded Linux platform that runs ROS2 natively. The perception package detects AR markers and/or IR beacons to track a person. The tracker uses AI algorithms such as particle filters and nonlinear optimization to extract the SE(3) information of the 2D feature.