A Virtual Mecanum Wheeled Robot ROS Simulator for Multi-view and Self-Following Motion Capture

Abstract

Motion capture (Mocap) on the go based on a mobile platform is valuable for clinical studies and rehabilitation. For multi-view gait analysis, Mecanum wheeled robots offer advantages over traditional differential drive robots. However, control issues in multi-view human tracking using Mecanum robots remain unexplored and lack a suitable virtual environment. This paper introduces a virtual Robot Operating System (ROS) environment with a Gazebo simulator as a research tool for multi-view human tracking on a Mecanum wheeled robot. The simulation incorporates a Proportional–Integral–Derivative (PID) controller and Kalman filter to maintain expected positional distance and relative viewing angles to the target. Our case study presents a quantitative evaluation of results obtained from the virtual environment for two specific tracking modes on a Mecanum wheeled robot: back-view following and side-view following with and without Kalman filtering. By optimizing the system, we decreased the distance error in backward following from 0.22 m to 0.12 m, and the angle error from 12.2° to 5.3°. Similarly, for side following, the distance error decreased from 0.32 m to 0.14 m, and the angle error reduced from 13.4° to 6.2°. These experimental results demonstrate that our approach enhances the accuracy of both tracking methods by over 50%. This work provides a necessary steppingstone for the development of human-tracking Mecanum wheeled robots for use in a clinical setting, providing a virtual environment for algorithmic development thereby eliminating wear on the hardware to be used in the clinical setting.