Personalization of Robot Teleoperation Methods and Interface

Abstract

Teleoperation of a robot manipulator can be achieved in a myriad of ways, which typically involves an interface that a human operator interacts with to accomplish a teleoperation task. Since this can be achieved in various ways, an interface for teleoperation may not be suitable for some operators, which can then lead to errors and cause damage to equipment and materials. A teleoperation interface therefore requires personalization, which can be defined by tuning the interface parameters, controls, and other features to a human operator’s preferences. Personalization of these interfaces is especially important in human-robot collaborations where the operator is actively working closely with a robot in the same space. A background project relating to this thesis is an industrial use-case, and the goal of it is to implement a human-robot collaboration workplace to replace time consuming manual labor, which will require a personalized teleoperation interface. However, research done about the topic of personalization has been made sparingly and therefore needs to be studied further; how can it be achieved specifically, how does it affect a human’s and robot’s performances, and how useful some of the personalization features to be included are generally. This thesis explored research done into robot manipulators and then studied what types of teleoperation interfaces, control methods, and other useful features had been implemented to find out if there were correlations in these methods and features. The research was done in the form of a literature review for this thesis, and it yielded many valuable results and findings. While an interface can be personalized in many ways, certain features appeared in the research papers many times. These features include multimodality (meaning that the interface has multiple choices for control methods) as a mix of manual and automated controls, data feeds relevant to the operation task such as camera sensors with extra indicators for successful grasps, support for VR/AR goggles and other extra equipment to add immersion, and various customization options for interface layout, appearance, and control feature settings. As part of this thesis was a project relating to the use-case. The project was to design and implement a personalized GUI for an FR3 robot in both simulations and real life. The GUI was equipped with personalization features discovered in the literature review, such as a multimodal teleoperation system, data panels, customization options, and skill tiers with different features available in higher tiers, while locked away from lower tiers. After the GUI was fully operational with extensive testing done in the simulation, it was moved to the real robot for an experiment including five participants. These participants tested three of the control methods developed for the GUI: gamepad, keyboard, and an automated “go-to-goal” solution. The experiment also involved a metal sheet resembling a mould involved in the background use-case to test the robot’s compliance control. The experiment highlighted the usefulness and importance of a multimodal teleoperation system, as the participants had difficulties teleoperating with some of the control features, while still being able to teleoperate well with others. In conclusion, personalization of a teleoperation interface can be achieved most usefully by including multimodality, relevant data feeds of the teleoperation task, and customization options. These specific features also help humans and robots improve their performance. While many features can also be added in for more personalization, their usefulness to the teleoperation task should always be considered to avoid cluttering the interface and cause sensory overload issues.