I am currently at the end of my first year as PhD student in the university of Essex. My work is mainly around Reinforcement Learning algorithms.
From a general point of view, the idea of reinforcement learning is very well grounded theoretically. However, practically it's converging to optimal solutions with small state space problems. But with large or continuous state spaces, the standard reinforcement learning is hard to implement due to the memory and time needed. Which make the use of the Function Approximation a crucial point in any Reinforcement Learning application.
Beside Function Approximation I'm interested in planning issues, learning in partial observable environments, multi-objective and multi-level reinforcement learning.
All these topics are ideal for the development of car racing games where there are a huge number of inputs to be leaned, the environment is either noisy or partially observable and in which the driver (in this case the controller) has to be clever, far-sighted, and well aware of all opponents behaviour.
I'm honoured to do my work under the supervision of Prof. Simon Lucas where I learned/ still learning how to question any possible/ impossible hypothesis, always aim for more, and how to share my work as a PhD student (show the statistics of this claim, how is this compared to others). With him, and the rest of the group we have a weekly meeting where we discuss the state of the art methods, innovations and our latest improvements.
Ahaa, about me I'm a Kenyan citizen although I've never been there except once or twice in my entire life. Spent most of my life in Saudi Arabia, went to Yemen for education and now I ended up in the UK. In life, I enjoy every new activity under one condition, It HAS to be FUN!
I believe intelligence is not the IQ level, is how to be "passionately curious", at least Einstein was!
And somehow this is the case for Robotics too!