The ARIA project is a fundamental research project designing intelligent systems of the future.
The STaR Project, working to improve traffic flow in Sydney, is one of a number of projects within NICTA which require an intelligent control system. These systems must understand and respond to their environments as those environments change. The ARIA project was initiated in 2008 to develop technology underpinning complex intelligent systems, and to transfer insights between particular end-use projects.
ARIA researchers are currently working closely with STaR researchers to make sure that our basic research helps solve their real-world problems. We are also working with other researchers to make sure that ARIA technologies can be applied to domains beyond traffic control.
Fundamental Research
There are two aspects to building an intelligent system. The system must be able to perceive its environment, and must be able to effect some change in that environment based on what it perceives. Both the perception and control aspects of an agent have previously been studied using simple models of the world. We are extending this research using a higher-order logic which allows us to efficiently represent, and reason about, more complex environments.
Technical Details
The ARIA Project has developed a system for performing lifted probabilistic inference using a higher-order logic. This can be used to implement Bayesian tracking in complex domains, such as tracking a platoon of cars passing through an intersection.
We have also developed various techniques for learning hierarchical control policies. These allow us to learn to control much larger problems than previous techniques.
Exciting research into perception, control and the integration of the two is ongoing.
Researchers
Publications
A number of publications are currently undergoing peer review and will be listed here when published.