Augmented Reality for Safe Human-Robot Collaboration

The primary concern in an environment where humans and robots coexist is ensuring human safety. However, the implementation of physical safety measures and barriers around the robot creates distance between humans and robots. This research aims to visualise virtual safety zones around a real robot arm using Augmented Reality (AR), thereby replacing the cages and bringing humans and robots closer together in a safe way. This thesis addresses this in four phases.

In phase one, the focus is on the development of an AR system and the calculation of AR safety zones for use in the system whilst considering how ISO standards used in industry can be adapted -- there is currently no international standard for visualising safety areas with AR for use in industry. This analysis also takes into account hardware and software delays, a factor previously overlooked in other studies, and explores the incorporation of these delays into the safety zone calculations.

The second phase addresses the display of the calculated safety zones around a moving robot arm. To demonstrate this, a collaborative pick and place application is used, involving a Universal Robots 10 (UR10) robot arm and a Microsoft HoloLens 2 for control and visualisation. This setup mimics a real task in an industrial robot cell.

In the third phase, the research examines the impact of visualised safety areas on users' feelings of safety and trust in the robot arm. The shape, size, and appearance of the AR safety zones are tested through three experiments. Experiments 1 and 2 are conducted with a real UR10 robot arm, while Experiment 3 is conducted in a virtual environment. The results indicate that visualisations using sectional cylinders, a safety zone shape that mimics the robot arm, together with virtual cage bars were a preferred safety zone combination for a real robot arm. However, the results were mixed in the case of a virtual robot arm experiment.

In the final phase, the investigation focuses on the transferability between real and virtual environments. A statistical analysis of equivalent experiments in a virtual and real environment revealed that users preferred different safety zone configurations. Subsequent discussion considers the question of whether safety visualisations can be initially tested in a virtual scenario and subsequently transferred to a real robot arm scenario.