Parag Khanna

PhD in Advanced Human-Robot Interaction at Division of Robotics, Perception and Learning, KTH.

I am a robot enthusiast who wants to bring the robotic development in labs to daily lives of people. My motivation has been those school days when many theoretical concepts of physics and mathematics urged me to ponder about their implementation and feasibility in real-world scenarios.

Currently I am involved in Human Robot Interaction (HRI) and in particular,Robot Human handovers. Robots already occupy a large part in our industries. However, with current advancements in technology, they are expected to be a part of our daily surroundings soon enough, collaborating with us and serving us. Sadly, though, unlike in movies, robots find it hard to perform simple yet essential tasks. Tasks that we perform seamlessly yet don’t realize. One such task is handover: the simple task of passing a thing to another person. My research focuses on getting seamless handovers between robots and humans.

Several collaborating robots largely rely on the person to adapt for handover. They will drop an object in your hand or release it after you pull at it hard enough. This makes handover unnatural, and the entire responsibility of handover falls on the person. Now, we don’t really want to play tug of war with a barista robot for our coffee. In a more serious scenario, consider service robots in an old age home or around children.

My approach is to study human-human handovers and the involved forces. As a part of initial study, I collected a dataset of over 2000 handovers and am using data driven Machine Learning methods to inspire new techniques for robot-human handovers. We humans have perfected our handovers by doing them unlimited times over several years, so we indeed are the best examples to learn from. 

One key aspect of my research focuses on adaptive grip release strategies—ensuring robots release objects naturally as humans begin to take them, rather than forcing an unnatural tug-of-war. Additionally, I investigate how object weight affects human motion in handovers, enabling robots to anticipate weight differences and refine their movements accordingly.

Beyond physical interaction, my research also explores social Human-Robot Interaction (HRI), specifically failure explanations in robots. As robots become more capable, their presence in human environments will expand, leading to increased physical and social interactions. In these shared spaces, failures in robot execution will inevitably occur. My research investigates how human-robot handovers can be leveraged to resolve robotic failures through adaptive explanation strategies that respond to human behavioral cues.

Additionally, I examine non-touch modalities, such as EEG brain signals and gaze tracking, to better understand human intentions during handovers—particularly distinguishing intentional handovers from other motions. By refining grip release strategies and exploring motion-based adaptations for varying object weights, I aim to make robot-human handovers more natural and intuitive.

This work is part of the Digital Futures - Advanced Adaptive Intelligent Systems project at KTH, which aims to develop innovative technologies that grant elderly individuals and people with disabilities greater autonomy, helping them live more active lives without needing constant assistance from others.

LinkedIn: parag-khanna

Handovers Dataset: Handovers@RPL

All Handovers Dataset here: Khanna-RPL-Handovers
Dataset of Human Reaction to Robotic Failure and Explanations: REFLEX Dataset