Leeds University is leading FREEpHRI, a project that aims to improve human-robot interaction by enabling robots to estimate human behavior in real time.
FREEpHRI: Flexible, Robust and Efficient Physical Human-Robot Interaction with iterative learning and self-triggered role adaptation, funded by an EPSRC Fellowship, aims to enable robots to intelligently detect the changes of human behavior and human-robot relationship automatically.
“Over the past few decades, robots have demonstrated their superiority in increasing productivity, efficiency and consistency in the production of products,” said lead researcher Dr. Zhenhong Li, a research associate in Leeds’ School of Electronic and Electrical Engineering†
“Now we want to bring people back into the development and production process. By leveraging the precision of robots and the creativity of humans, we can create a better system that can deliver individualized products and services.”
The fellowship will focus on developing a human-robot interaction control strategy that allows robots to flexibly respond to their human partners, Li said. The mechanic, for intended applications in sectors such as manufacturing and healthcare.
MORE ROBOTICS NEWS HERE
“Using the latest development of optimization theory and machine learning algorithms, we will investigate three fundamental problems in physical human-robot interaction,” said Li.
These, he explained, are how to model complex human interaction behaviors; how to efficiently update the robot’s control strategy to ensure the desired interactions; and how to deal with uncertainties in the human-robot system.
According to the team’s EPSRC grant summary, the control strategy will use optimization theory to model human-robot interaction behavior and learning techniques to compensate for the effects of unknown dynamics and external perturbations.
The human partner is assigned a cost function that implies motor ability, allowing the robot to adapt its role (employee or competitor) to the real-time estimate of the human cost function.
A self-activated role adaptation mechanism will use the human-robot system’s performance and estimated human behavior to detect the human’s role changes, causing the robot to change its role when needed.
Reliability and functionality of the proposed techniques will be evaluated, according to the team, through application in physical robot-assisted rehabilitation, using the techniques to achieve typical training strategies, initially in laboratory settings and then in hospital settings supported by the Leeds Teaching Hospital.
A strategic group will be established to inform the design and testing of the technologies, drawing on the expertise of academic and industrial partners.