Research Associate in Dexterous Robot Learning

Location: South Kensington Campus

About the role:

The Robot Learning Lab in the Department of Computing at Imperial College London is seeking a talented Research Associate (post-doc), to take a key role in an ambitious new project on dexterous robot learning. In this project, we will be studying co-design of robot hardware and control for robot manipulation. “Co-design” aims to simultaneously optimise the robot hand design (e.g. the number of fingers, the shapes of the fingers, and the positions of tactile sensors), and the control policy for that hand, when given a particular task or set of tasks. Through this, we aim to develop a framework that can automatically generate creative new robot hands with dexterous control policies, for human-level or even super-human performance on real-world tasks. You will be collaborating with a small team at Imperial College led by Dr Edward Johns, as well as a larger team across the UK as part of the ARIA-funded Robot Dexterity programme (see: https://www.aria.org.uk/opportunity-spaces/smarter-robot-bodies/robot-dexterity/). This post-doc position will be fully funded until March 2029.

For further information on Dr Edward Johns’ research and projects, see: 

www.robot-learning.uk.

What you would be doing:

You will be studying both the algorithms for robot co-design, and the real-world evaluation of the designs that emerge. You will investigate evolutionary algorithms to explore creative new hand designs, and reinforcement learning for policy optimisation, all within a huge GPU-based simulation with thousands of robots learning synchronously in parallel. The best solutions will then be 3D-printed, and policies will be fine-tuned with reinforcement learning in a robot "arm farm" we are building at Imperial College. You will also be assisting Dr Johns in leading the overall project, such as helping to supervise the research of PhD students on this project, collaborating with other partners in the ARIA programme, assisting with the delivery of project milestones, and managing the project’s codebase. We expect novel algorithms and results to come from this project, and so you will have opportunities to submit papers to conferences and journals, and travel to conferences to present these papers.

What we are looking for:

You must have a PhD and a strong background in both robotics and machine learning, with first-authored publications in the leading international conferences/journals in robotics. Candidates should have experience in some or all of the following six topics:

1. Reinforcement learning.
2. Evolutionary algorithms.
3. Training physical robots with learning algorithms.
4. Training multi-fingered robots with dexterous policies.
5. Transferring robot policies from simulation to reality.
6. Designing and building robot hands.

Please see job description for a full list of requirements.

*Candidates who have not yet been officially awarded their PhD will be appointed as Research Assistant, salary range £43,003 - £46,297 per annum.

Further Information

Full-time, Fixed term contract to start ASAP until March 2029.

Candidates should attach:

• A full CV, with a list of all publications.
• A 1-page document explaining:
  • (1) why you are interested in this position, and
  • (2) how your experience is relevant to some or all of the 6 topics listed above.

Informal enquiries related to the position should be directed to:

Dr Edward Johns: e.johns@imperial.ac.uk

For queries regarding the application process contact:

Jamie Perrins; j.perrins@imperial.ac.uk

Advert information