My research interests are in machine learning, pattern recognition, robotics, artificial player for strategic games (hex in particular), discrete mathematics and optimization. The projects below illustrate my research activities (not an exhaustive list). If you are looking for a project (Honours, MSc, or PhD) in one of those areas, contact me to arrange a meeting for a chat on current opportunities.
Robot soccer is far from a frivolous activity. There is a long list of key research innovations (in vision, pattern recognition, locomotion, multi-agent systems, etc..) resulting from robot soccer related research activities.
We participate in the Kheperasot league.
Robot soccer has become a major stimulant for AI and intelligent robotics research by providing a standard problem where wide range of technologies can be integrated and examined. Significant research outcomes are derived from these activities which at first sight may seem frivolous to some. In our case, we are interested in real-time pattern recognition, reasoning, image processing, tracking, sensor-fusion and self-localization.
The navigation systems of commercially available autonomous lawnmowers rely on sensors measuring the magnetic field created by a perimeter wire. Some experimental systems rely on even more expensive sensing devices, like differential GPS or laser tracking systems that help locate the mowers exactly within a yard, but are considered too expensive for a domestic robot. Henry Huang and I have proposed a navigation system that requires only a standard web camera and the existence of some landmarks visually recognisable. The key idea is to induce the absolute positions of the landmarks from apparent angles derived from panoramic views taken from a few observation points. In his Honours dissertation, Henry described the localization system and demonstrated its accuracy is in simulation. Henry is continuing this work (PhD started in 2005).
It is desirable that vision impaired people have access to a user friendly interface to computer file systems. The system that we propose is an interface to the Explorer program of Windows XP operating system. The files and directories are distributed on a virtual screen similar to the standard Windows Explorer. A web camera monitors motions from the user to navigate a focus point in the file system with an abstract pointer. The motions monitored can be hand gestures or head movements. Voice commands are used to speed up navigation in the directory hierarchy. The project is brought to life by Frank Loewenich (Honours student).
With the increasing programmability of commodity graphics processing units (GPUs), these chips are now capable of performing more than the specific graphics computations for which they were designed. They are underused powerful coprocessors whose high-speed makes them potentially useful for a wide range of applications. The aim of this project is to develop very fast real-time vision systems using only hardware found in typical mass-produced personal computers. We will adapt and design new computer vision algorithms by efficiently mapping mathematical operations of computer vision onto modern computer graphics architecture. Our targeted applications are driver assistance and human-computer interface. The project is brought to life by Julius Ohmer (MSc student) and is co-supervised by Ross Brown.
Hex is a beautiful game with simple rules and a strategic complexity comparable to that of Go. I am developing artificial players for Hex with Rune Rasmussen (PhD student) and Simon Yuen (Honours student).
I have also the honour of supervising Cameron Browne on a PhD thesis on connection games. Cameron wrote the first book dedicated to Hex. His 2nd book ("Connection Games: Variations On A Theme") has just been published
Pattern recognition systems for temporal inputs with lexical constrained (like speech and hand-writing). Non-Deterministic Automata for Hidden Markov Models (with Alain Lifchitz)
Reinforcement Learning (with Vadim Bulitko)
Kernel machines (from theory to applications).