Vision-Based Person-Following Algorithm for Assistive Elderly-Care Quadruped Robots

Area of Science:

• Robotics
• Computer Vision
• Human-Robot Interaction

Background:

• Aging population requires advanced mobility and care solutions.
• Commercial quadruped robots have limitations in safe and predictable user following.
• Existing following modes lack necessary safety margins and consistency.

Purpose of the Study:

• To develop a robust, vision-based algorithm for reliable person-following with quadruped robots.
• To enhance safety and predictability in human-robot interaction for mobility assistance.
• To address the limitations of native following modes in commercial quadruped robots.

Main Methods:

• Utilized a ZED 2 stereo camera and Robot Operating System (ROS) on a Unitree Go1 platform.
• Implemented a finite state machine for deterministic target tracking.
• Employed a velocity control strategy with distinct motion zones based on depth data.

Main Results:

• Achieved a mean processing latency of 66.5±4.3 ms in headless mode.
• Demonstrated 0.0% intrusion into the intimate safety zone, ensuring user safety.
• Maintained consistent operational stability and effective velocity synchronization (0.47-0.54 m/s).

Conclusions:

• The developed algorithm provides a stable technical foundation for robotic walking partners.
• The system shows promise for assisting with mobility in aging populations.
• Further clinical testing with elderly users is recommended for deployment.