- Research
- Research Themes
- Business Areas
-
Projects
-
Completed Projects
- Adaptive Middleware Platforms
- Advanced Nonlinear Gradient Methods
- Shape Analysis of Brain Structures
- Broadband to the User
- Component Architectures
- Drivers Drowsiness Detection
- Dynamic Planning, Optimisation
- FM for Performance Evaluation
- Fundamental Limits of Wireless
- Last Metre Problem
- Layered Learning
- Learning and Representation
- LIXI Business Processes
- Macro-programming for WSN
- MAMI
- Mapping Genetic Components in Crops
- Office in a Box
- Ontology Toolkit
- Open SensorWeb Architecture
- SAIL
- Secure Embedded L4
- Smart Cars
- Validating Network Semantics
- Video Analysis for Surveillance
- Visualisation of Complex Networks
- Water Information Networks - WIN
- xQoS Systems
-
Completed Projects
- Collaboration
- Case Studies
Hierarchical Abstract Learner
The Hierarchical Abstract Learner project was building an automatic reasoning system based on higher order logic. The target domain for this system is intelligent systems; that is, agents, robots or other devices that are able to interact with their environment to autonomously achieve goals supplied by humans.
These intelligent systems must have sensors and actuators to affect their world, and unfortunately, nothing the real world is ever as exact as a logical system. To handle this uncertainty, we are building our system with strong probabilistic reasoning capabilities as well as strong logical reasoning capabilities.
When complete our system should be able to reason about its world at a variety of levels of abstraction, from relatively high level concepts (like knowing that when booking travel plans for someone between Australia and the US, it is usual to travel by aeroplane), to more detailed knowledge (a car turns to the left if it is moving and you turn its steering wheel to the left). We are also conducting research into how the system can acquire this some or all knowledge by itself. In particular, we would like a human to be able to outline an area of knowledge and the system discover the details using its own interaction with the world.
The DARPA Urban Challenge
One of the test projects for our ideas about agent systems is a collaborative entry into the DARPA Urban Challenge. The Sydney-Berkeley Driving team is a collaboration between the Australian Centre for Field robotics at the University of Sydney, our group at NICTA, the University of Technology, Sydney and the University of California, Berkeley.
Our research is built upon ideas developed in our previous collaboration with the University of New South Wales on the rUNSWift entry into the RoboCup international robot soccer competition. We were very successful in that collaboration, never failing to reach the playoffs and being in the final most years.
We are also pursuing other testbeds for this research, including collaborations with the Australian National University, and the Centre for Autonomous Systems.
Status
The HAL project has finished.