Handling Vision Noise Through Robot Motion Control in a Real-Time Teleoperation System

Abstract

Teleoperation allows robots to perform complex actions in tasks and environments where robot cognition and motion capabilities cannot support autonomy or where cooperation with humans is necessary. In this work, we present a teleoperation system for controlling the motion of a non-anthropomorphic robot arm through a natural user interface that tracks human movements through a single RGB-D visual sensor. A proportional-derivative (PD) controller is formulated to regulate robot end-effector commanded accelerations and to deal with the noisy and low-frequency human information. The system is integrated in the robot operating system (ROS) and a Universal Robots UR3 cobot is used for the evaluation. The results show that the PD controller achieves a superior performance in terms of both tracking accuracy and robot motion smoothness compared to using a proportional controller which regulates the commanded velocities. Furthermore, the same gains of the PD controller can be used to handle a wider range of human speeds. Finally, we provide real-time demonstrations of the system and the ROS pipeline.