Dynamic modeling and ROS 2-control architecture for a six degree of freedom robotic manipulator with performance comparison of PID, gravity compensation, and full inverse dynamics controllers

Abstract

This paper presents the dynamic modelling and control evaluation of a six-degree-of-freedom serial robotic manipulator simulated in Robot Operating System 2 Jazzy and Gazebo. The rigid-body dynamic model is derived via the Denavit–Hartenberg convention and includes the full inertia matrix M(q), Coriolis and centrifugal matrix C (q, q·), and gravity vector G(q). In addition to the dynamic modeling of the manipulator, three controllers are implemented and compared: (1) a baseline PID, (2) PID with feedforward gravity compensation, and (3) a full inverse-dynamics controller using the recursive Newton–Euler algorithm. All three controllers are implemented and evaluated entirely within ROS2 Jazzy and the gz_ros2_control stack, without reliance on ROS1 or offline simulation. Gravity compensation alone reduces steady-state error on the heaviest joint from 0.237 rad to 0.001 rad, at negligible computational cost. The recursive Newton–Euler algorithm controller achieves sub 0.02 rad error on four of six joints. The results show that the largest gains come from the simplest model augmentation, with diminishing returns at higher complexity.