Julian Salt Ducaju is defending his PhD thesis at the Department of Automatic Control.

Title: Control Strategies for Physical Human-Robot Collaboration
Speaker: Julian Salt Ducaju
Opponent: Professor Olav Egeland, NTNU
Committee: Professor Ulrike Thomas, TU Chemnitz
Professor Ángel Gaspar González Rodríguez, University of Jaen
Dr Mikael Norrlöf, ABB Robotics, Sverige
Advisor: Universitetslektor Björn Olofsson
Where: Lecture Hall M:B in M-huset, Ole Römers väg 1, Lund
When: Friday June 7th, 10:15
Zoom: https://lu-se.zoom.us/j/63759419777

Abstract: Recent industrial interest in producing smaller volumes of products in shorter time frames, in contrast to massproduction in previous decades, motivated the introduction of human-robot collaboration (HRC) in industrial settings, to increase flexibility in manufacturing applications. As a consequence, industrial environments would lose their fixed structure, thus increasingthe uncertainties present in this workspace shared between humans and robots. This thesis presents robot control methods to mitigate such uncertainties and to improve the involvement of human operators in industrial settings where robots are present, with aparticular focus on manual robot guidance, or kinesthetic teaching. 

First, the accuracy of manual robot guidance was increased by reducing the joint static friction without alteringthe robotic task execution, using additional degrees of freedom (DOFs) available in collaborative robots. Additionally, previous methods for a fast identification of the source of robot-environment physical contact in partially-unknown industrial environmentswere evaluated, extended, and modified to perform effective manual corrections of the robot motion. Then, an iterative learning method was proposed to achieve a more accurate use of manually-defined trajectories, while allowing a safe physical robot-environmentinteraction.

Moreover, safety is a major concern in uncertain scenarios where humans and robots collaborate. Regulating the robot-environmentinteraction forces, e.g., using impedance control, would improve safety, yet undesired parts of the collaborative workspace might need to be entirely avoided. To this purpose, a stable online variation of robot impedance during the manual guidance of the robotwas proposed. This proposal was later extended to further improve safety by considering a prediction of human guidance with coordinated robot control. Furthermore, the additional DOFs in collaborative robots were used to develop a stable online impedance variationmethod for robot obstacle avoidance without requiring modification of the main robot task.

All methods presented were tested experimentally on a real collaborative robot.