POSTED: 16 Feb, 2023

Australian Cobotics Centre researchers have three papers accepted for publication at the upcoming International Conference on Robotics and Automation (ICRA) 2023 in London. ICRA is the IEEE Robotics and Automation Society’s flagship conference and the premier international forum for robotics researchers to present and discuss their work.

Postdoctoral Research Fellow, Dr Fouad Sukkar gave is a brief summary of two of the papers appearing at the conference mid this year.

“Guided Learning from Demonstration for Robust Transferability”, by Fouad Sukkar, Victor Hernandez Moreno, Teresa Vidal-Calleja and Jochen Deuse, contributes to the idea of humans teaching robots how to carry out tasks in the real world. A common approach is to physically move the robot around and record its motions. However, this can be inconvenient if the robot is already busy doing work and even dangerous/ not possible in the case of large industrial arms. Ideally, we should be able to demonstrate the task more naturally using the original tool, such as a welder, without the physical robot present. This is where the proposed guided learning from demonstration (LFD) framework comes in handy. It visually displays regions in the workspace that the robot is capable of moving within to the demonstrator. That way the user can be confident that the robot will be able to execute the demonstrated task and avoid having to carry out repeated trial and error. To validate their method, a user study was carried out which involved users teaching a collaborative robot (cobot) how to carry out a welding task with obstacles in the environment. Results showed that with the visual guidance users were always able to achieve a successful execution of the task on the robot in comparison to only 44% of the time without it. While the framework is in a promising proof of concept state, the authors strongly believe this idea will be important for industry as they adopt more flexible methods for manufacturing. As for future work, they are currently working towards testing their framework in real world applications and exploring different guidance mediums such augmented reality and haptic feedback. A preprint for the paper can be found here: https://arxiv.org/abs/2302.03901.

“Optimal Workpiece Placement Based on Robot Reach, Manipulability and Joint Torques”, by Baris Balci, Jared Donovan and Peter Corke, looks at how to best place a workpiece for a robot to physically interact with. This is particularly important for cobots, such as the Universal Robot’s UR arms, which are designed to be safe around humans. Cobots are sensitive to external forces in order to be safe around people, however, this sensitivity complicates cobots’ physical interaction with objects for manufacturing purposes. The arm will deviate away from where you tell it to go due to the external forces that are required for the manufacturing processes. What the authors in the paper show is that depending on how you place a workpiece, it is possible to increase cobots’ physical interaction performance. They present a novel method for choosing this placement and demonstrate it works effectively in a simulated surface finishing application for a few different objects with interesting surface geometries.