This is a one year project to investigate how Partially Observable Markov Decision Processes (POMDPs) could help improve generalisability in imitation learning. We will seek to integrate some of our new approximate POMDP solver with existing method in Bayes-adaptive Imitation learning.
The ideal candiate would:
- Have a good understanding of basic probability
- Fluent in programming
- Have done well in Intelligent Robotics course (COMP4620/8620 in 2023 / 2025 or COMP4680/8650 in 2024)
- Have a GPA >= 6.5/7.0
- Understanding of basic methods for Bayesian inference is a plus.
If you are interested, please send me a copy of your CV and transcript via e-mail.