CSIRO GREMlab is a new cross-Divisional initiative within CSIRO which operates on a Laboratory model with participating projects contributing resources. GREMlab's aim is to produce directed design and control of multi-agent systems. THE GRAND PROBLEM is how to design agent properties to get desired emergent behaviour from multi-agent systems. This is a non-trivial, but very important, problem. It is common to many situations from large scale applications to the nano and molecular scales.

Learning in multi-agent systems

The creation of emergent system behaviour is being investigated through multi-agent genetic and other learning algorithms. Specific applications include mesoblock self assembly simulation and herding. Herding is a simple example: sheep do, cats don't ... What "emerges" depends on the agent properties.

Macroscale: Ageless Aerospace Vehicle (AAV) Smart Skin

Intelligent sensors on AAV skin are locally connected to detect and analyse damage. They must take appropriate action (assess, report, repair) and they need to work even when severely damaged.

This is to be achieved by agent design and excellent results have been obtained. Measurements of performance will come from a Concept Demonstrator which is currently being built.

Nanoscale: Smart Dust

Intelligent nanoparticles, Eg. "TV dust". Just add water and a TV self-assembles! This may sound far-fetched, but limited self-assembly of simple circuits is possible through "Directed self-assembly". Success in this area could lead to totally new ways of electronic circuit design and manufacture.