Experimental Evaluation of a Nav2 Regulated Pure Pursuit-PID Architecture for Trajectory Tracking of a Four-Wheel Mecanum Mobile Robot

Quang Vi Ngo; Van Vu Duong; Guanqiu Qi; Zhiqin Zhu

doi:10.37965/jait.2026.1632

Experimental Evaluation of a Nav2 Regulated Pure Pursuit-PID Architecture for Trajectory Tracking of a Four-Wheel Mecanum Mobile Robot

Authors

Quang Vi Ngo Department of Robotic and Intelligent Systems Engineering (RISE), Faculty of Electrical & Electronics Engineering, Thuyloi University, Hanoi, Vietnam https://orcid.org/0009-0002-2571-6470
Van Vu Duong Faculty of Electrical & Electronics Engineering, Thuyloi University, Hanoi, Vietnam
Guanqiu Qi Computer Information Systems Department, State University of New York at Buffalo State, Buffalo, NY, USA https://orcid.org/0000-0001-9562-3865
Zhiqin Zhu College of Automation, Chongqing University of Posts and Telecommunications, Chongqing, China

DOI:

https://doi.org/10.37965/jait.2026.1632

Keywords:

kinematic model, Mecanum mobile robot, path tracking, PID control, Pure Pursuit, Regulated Pure Pursuit, ROS 2 Navigation2

Abstract

This paper reports an experimental evaluation of a Nav2 Regulated Pure Pursuit (RPP)-PID architecture for trajectory tracking of a four-wheel Mecanum mobile robot. The high-level RPP controller runs on an NVIDIA Jetson Nano at 50 Hz, while four proportional-integral-derivative (PID) loops on an STM32 microcontroller regulate individual wheel speeds at 100 Hz. PP and RPP were evaluated using the same robot hardware, map, PID gains, start pose, goal pose, and reference path. Each quantitative tracking condition was repeated five times. RPP reduced the mean RMSE from 0.12 ± 0.03 m to 0.08 ± 0.01 m for curved-path tracking and from 0.10 ± 0.02 m to 0.05 ± 0.02 m for goal reaching. Obstacle-avoidance experiments were also conducted with both controllers under matched conditions and are discussed as system-level navigation tests because planning, localization, perception, and control remained active. The controller uses nominal Mecanum kinematics without explicit slip or delay compensation. The results demonstrate improved geometric tracking accuracy for RPP under the tested indoor conditions and confirm stable online operation of the distributed RPP-PID architecture.