Abstract:
As offices become increasingly automated, the demand for service robots that can seamlessly
navigate and interact with people is rapidly growing. Yet, developing such robots with reliable
tracking and smooth mobility remains challenging when constrained to low-cost hardware. To
fulfil this requirement, this paper presents the design and evaluation of a low-cost autonomous
human-following robot for office automation. The system combines AprilTag fiducial markers
for robust human tracking with a hybrid control architecture integrating a Raspberry Pi 4
for real-time computer vision and an Arduino Uno for motor control. Motion smoothing is
achieved through PID algorithms, while ultrasonic sensors enable dynamic obstacle avoidance
in cluttered indoor environments. Two central research questions guided the study: (1) How
can a human-following robot maintain reliable, real-time tracking in dynamic indoor environments
using low-cost hardware? and (2) Which vision-based approach offers the best balance
of accuracy, computational efficiency, and robustness? Comparative evaluation of vision-based
tracking, QR codes, and AprilTags showed that AprilTags achieved superior performance, with
over 98% tracking success, high pose estimation accuracy, and effective recovery after tracking
loss. Real-world testing demonstrated consistent following behaviour, maintaining an average
distance of 84±7cm with minimal latency (<4s). The results confirm that affordable hardware,
combined with fiducial marker tracking and PID-based control, can deliver reliable and natural
human-following performance suitable for office logistics. The research provides a scalable
foundation for service robotics in office, healthcare, and industrial environments.
