Ayesha Jena

With Industry 5.0, it will become common to see robots actively collaborating with humans withinworkspaces. The challenge is to develop robots operating in shared spaces within unstructuredenvironments. A promising approach to optimizing the operational performance of Human-RobotInteraction (HRI) is to consider the Mixed-Initiative approach in which tasks and authority of humans andagents are dynamically defined based on abilities.Our aim is to have an interaction and reasoning framework for effective HRI collaboration. In order tomake these judgements, the system needs a reliable understanding of the overall situation (situationawareness), including underlying intentions of observable actions and provide a framework forclassifying various intentions, including initiation, movement, retrieval, idle observation, and compositeintentions. This classification is a significant step in addressing one of the major challenges in HRI, whichis bi-directional communication. This capability is pivotal in shaping the robot's behavior, ensuringeffective communication of the user's intentions. In turn, it enhances the robot's ability to comprehendhuman intentions during goal execution, leading to natural and safe interactions between humans androbots. Intention reading, as we define it, encompasses the ability to precisely comprehend the objectivesmotivating the observable actions executed by others.Currently I am using VR systems to have a framework for mixed initiative interaction where we cancollaborate and access humans’ intervention in these scenarios and provide useful insights into theseinteractions.

Publications

Esranur Ertürk

The goal of her Ph.D. project is to create a robotic system that can accurately hand over objects to humans without re-gripping. The project addresses three main challenges: effective scene understanding, recognizing user intentions, and executing collaborative tasks. Initially focusing on tools like scalpels and screwdrivers, the robot uses hand and tool tracking to ensure precise handovers. By integrating high-level knowledge and sensor information, the system predicts and executes the appropriate handover actions.

The project also involves evaluating different motion models to enhance the robot's performance, ensuring smooth and precise movements in various scenarios. This work aims to improve intuitive robot instruction and responsive collaboration in medical and industrial contexts, contributing to more efficient workflows. The project is funded by WASP: Wallenberg AI, Autonomous Systems and Software Program at Lund University.