Making robots learn by trial and error

Project Level: Summer

Project Duration: 6 weeks

Hours of Engagement: 20-36 hours

Project Description:

Have you ever picked up a skill without any trial and error? Perhaps rarely, if not no. This project explores algorithms that robots can be used to learn skills like walking and swimming by trial and error. This has been extensively studied in an area of artificial intelligence called reinforcement learning (RL). The successful applicant of this project will apply some state-of-the-art RL algorithms to robot learning using fast and accurate simulators. There will be opportunity to develop new algorithms if time permits.

Expected Outcomes:

• Gain knowledge on some state-of-the art RL algorithms and a robotics simulator.

• Develop the ability to implement RL-based AI for robot learning.

• Develop skills in research design, implementation, experimentation, and communication.

• A report documenting the work done and the findings.

Suitable for:

Essential: knowledge of deep learning, strong programming skills

Desirable: knowledge of Markov decision processes and reinforcement learning

Contact for further information:

Dr Nan Ye: nan.ye@uq.edu.au