Physics-informed Learning for Robotic Manipulation of Granular Media

EPSRC DTP studentship in Engineering

Duration 3.5 years full time / 5 years part time

Project Summary

Granular media are ubiquitous, ranging from food like beans to construction materials like soils or sands. With the rapid development of robotics, robots need to interact with such materials in various settings, e.g. flour pouring in kitchen or seed/sands scattering for farming or construction. However, research in autonomous robotic manipulations of these materials is at its infancy.

 One main challenge is material deformation, due to the granularity. Traditional robotic motion planning is unscalable in unstructured environments due to explicitly designed models. To autonomously plan motions in real-time, granular characteristics need to be considered. However, physics-based numerical modelling is computationally demanding and prohibitive from efficient robot decision making.

 This project aims to unlock an efficient, physically plausible, and safe framework for autonomous robots so they can learn to manipulate this media. The research investigates in new learning-based approaches informed by physics laws, which hope to greatly accelerate the simulation process by allowing optimal and physically plausible actions to be computed efficiently. If successful, novel learning-based methods will support robots to learn to manipulate such materials, and generate motions autonomously in real-time. The work will focus on several related aspects, including efficient modelling of dynamic environments, reinforcement learning, and 3D vision processing.

 Research Environment

The research is tightly related to recent EPSRC and Royal-Society funded projects, providing specialised equipment to support this research. The research area is well aligned with the key research areas of the research group of Robotics and Autonomous Intelligent Machines (RAIM) of the school of engineering. The PhD studentship will be well integrated into the research activities of RAIM to maximise the research outcomes. It is expected that the work will produce a significant contribution to fundamental scientific research and strong impact for proposing solutions for real-world problems. The successful applicant will be collaborating with the visual computing group in the school of computer science and informatics, which has a strong track record of research publications in areas of 3D computer vision, deep learning, and so on.

 The student will have access to cutting-edge facilities of mobile and collaborative robots in the Robotics and Autonomous Systems Lab, including Kuka iiwa lbr, Kuka youbot, RobotNik Vogue+, turtlebots, state-of-the-art 3D cameras/scanners and high-end high-performance computers that can support all related research activities.

Learning and Development Opportunities

The student will closely work along with a full-time postdoctoral research associate and academic supervisors, who will directly provide specialist knowledge to the PhD student. New research will be carried out at the forefront of robotics and deep/reinforcement learning, under supervision from different areas. The student will gain experience in 1) robot programming through access to advanced sensors and robots in the lab, 2) state-of-the-art knowledge in cross-disciplinary areas of robot manipulation, deep learning, physics modelling, reinforcement learning and 3D vision, etc.

The project is a collaboration between the RAIM group in the school of engineering and the visual computing group in the school of computer science and informatics. The student will be expected to develop novel ideas to solve problems encountered in the project, and to present and report findings in seminars, conferences, and journals. Through the project, the student will develop a full spectrum of transferrable skills including critical thinking, problem solving, technical writing and presentation.

The expertise gained in foundational robot learning is highly demanded in the market and will pave a straight avenue for the student towards a future career in both academia and industry. 

Academic Criteria

Candidates should hold or expect to gain a first-class degree or a good 2.1 (or their equivalent) in Engineering, computer science, physics, or a related subject. 

Desirable skills:

·        Strong programming skills in Python (and/or C++)

·        Strong background in mathematics/numerical analysis

·        Knowledge of Machine Learning, ideally reinforcement learning, would be highly desirable, but not essential

·        Knowledge of robotics/computer vision would be desirable, but not essential

·        Background in programming robots using tools, such as ROS or similar, under Linux, would be desirable, but not essential

Applicants whose first language is not English will be required to demonstrate proficiency in the English language (IELTS 6.5 or equivalent)

Contact for further information

Please contact Dr Ze Ji ([Email Address Removed]) to informally discuss this opportunity

How to apply

Applicants should submit an application for postgraduate study via the Cardiff University webpages (http://www.cardiff.ac.uk/study/postgraduate/research/programmes/programme/engineering ) including;

·        an upload of your CV

·        a personal statement/covering letter

·        two references (applicants are recommended to have a third academic referee, if the two academic referees are within the same department/school)

·        Current academic transcripts

Applicants should select Doctor of Philosophy (Engineering), with a start date of October 2023.

In the research proposal section of your application, please specify the project title and supervisors of this project and copy the project description in the text box provided. In the funding section, please select "I will be applying for a scholarship / grant" and specify that you are applying for advertised funding, reference ZJ EPSRC 23

Deadline for applications

17^th Feb 2023. We may however close this opportunity earlier if a suitable candidate is identified.