Statistical Machine Learning For Robotics
The research in this area focusses on
- Investigating human and machine
perception (or representations) of the surrounding environment allowing
us to understand how intelligent autonomous agents can be
robustly designed for a variety of tasks.
- Online and offline learning of
features and models of the complex environment grounded with respect to
the robot (machine perception) and/or to the human designer (human
perception).
- Statistical and probabilistic techniques have proven to be
powerful methods for both unsupervised and supervised machine learning.
Probability theory ensures that algorithmic design incorporates the
inherent stochasticity of noisy
sensors or noisy data on an an autonomous agent.
A reading group is held every week which concentrates on various aspects of statistical machine learning. For papers that have been discussed click here |
