Robot learning from demonstration (LfD) with scene understanding for Human-Robot Collaboration

BACKGROUND

Robot learning from demonstration (LfD) is a method that allows end users to teach a robot new skills without knowledge of programming. LfD has been actively researched in the field of human robot interaction (HRI) because it provides a means for a robot to acquire new capabilities through interactive communication. Also, there is a growing interest in employing LfD for Human Robot Collaboration (HRC) tasks, which would allow robots to work together with humans in a shared workspace.

HRI and HRC are similar, except that HRI focuses on the learning capabilities through communication while HRC also considers human factors, such as human intention understanding in collaborative tasks.

LfD covers a broad range of subjects, from theoretical research in robot learning (e.g probabilistic methods, such as HMM, GMM), perception (e.g. vision based motion understanding), and human factor issues (e.g interpreting human intention), to practical challenges, such as the correspondence problem of configurations between demonstrator and robot. In addition to motion learning, we are also interested in understanding human intentions at the task level. These are known as predictable and legible robot motions, where legible robot motions can be different from demonstrations because the main drive is how to choose the right actions to achieve the goal.

Limited research is found using Deep Learning based algorithms.

PROJECT AIMS AND METHODS

This project is expected to focus on some of the following research subjects:

•To develop algorithms for human operator motion and workspace understanding using vision or RGB-D sensors.
•To investigate if Deep Learning based approaches can be employed in LfD to support high level human intention understanding.
•To develop a means to integrate semantic domain knowledge for action representation and information reasoning.

The student will have the access to some cutting-edge facilities, and is expected to validate the research on these hardware platforms.

ELIGIBILITY

You should have obtained, or be about to obtain, a First or Upper Second Class UK Honours degree, or the equivalent qualifications gained outside the UK.

Applicants with a Lower Second Class degree will be considered if they also have a master’s degree. Applicants with a minimum Upper Second Class degree and significant relevant non-academic experience are encouraged to apply.