Sitting at Gordon Wyeth's feet, under his desk, is Corgi the robo dog - a University of Queensland international advance in robotics and the key to understanding intelligence.
Mr Wyeth has invented Corgi, believed to be the world's first robot that can learn to see and react rather than merely be programmed to perform or repeat identical tasks.
'What we teach Corgi is how to behave, in a similar way to how you teach a child,' Mr Wyeth said.
Factory robots taught or programmed to repeat an action are 'just dumb in intelligence', according to Mr Wyeth.
Mr Wyeth of the University's Electrical and Computer Engineering Department says his invention has enormous industry or community potential, and could be developed for tasks such as collecting litter, retrieving rocks on Mars or fruit picking.
'The other exciting part about this work is that we're actually starting to uncover the tips of the roots of intelligence. We're really starting to find out what intelligence is about,' Mr Wyeth said.
The robot was named Corgi because its configuration - a $20 microprocessor, frame/video grabber which turns the camera signal into a computer understandable picture, mouthpiece, batteries and other robotics technology - evolved into the shape of a squatty little dog, albeit run on three wheels.
'When I first proposed this idea, a lot of people said ?it can't be done. It's too hard'. We've shown it can be done, and what's more it can be done cheaply,' he said.
Mr Wyeth said it could cost as little as $1000 to create a reliable robotics system for a range of industry or community applications.
Once Corgi observes its environment via the camera, the images/objects are recorded on an external computer and, using a mouse, relevant images are highlighted either to target or to avoid.
This information is processed by an artificial neural network, an electronic computer program which acts like the neurons of the human brain, then downloaded to the microprocessor which rests on Corgi's back.
Mr Wyeth created Corgi as part of his PhD topic to build a robot that could be taught how to behave, if familiar with its environment. Corgi successfully demonstrates the project's aim by finding tennis balls.
'We experimented with finding tennis balls in an office/building environment but the methods could be applied to diverse tasks such as retrieving rocks on Mars, sorting macadamias or picking apples in an orchard. If you familiarise Corgi with a new environment, it can learn,' Mr Wyeth said.
'For instance, I teach it not to run into things as long as the environment is reasonably similar to what it's seen before.
'Retraining it to cope in a different environment simply involves driving it through the relevant environment.'
Mr Wyeth has been working on the Corgi concept since 1994 and is keen for industry support to continue to develop the technology.
While Corgi learns as a child learns, it cannot learn for itself, only what the programmer wants to teach it.
The next stage of the project is to train Corgi to retrieve objects, such as fetching tennis balls and returning them to a specific location.
'Currently it can only operate on what it sees and has no memory,' Mr Wyeth said.
'But there are lots of jobs, such as fruit picking that don't require memory.
'It could suit macadamia nut farmers who currently employ about 14 people to work 10-hour days sorting nuts. It's one of the biggest on-farm expenses in macadamia farming, whereas a system based on Corgi can do the job cheaply.'
Mr Wyeth has worked with fourth-year student Michael Jenvey to develop a prototype of this system.
He also is supervising PhD student Phillip Chan's research which aims to apply honey bee navigation to improve Corgi's navigational abilities.
'Bees can find their hive from any point in an area that they've flown over before. So, although there's not a lot happening in a bee's brain, it navigational mechanism must be reasonably simple.
'There's a lot of excitement in combining biology and robotics and in combining neurosciences and robotics.'
Mr Wyeth is also supervising Brett Browning PhD research investigating rat navigation, and how to apply the rodent's purposeful scurrying to robotics.
During his studies, Mr Wyeth took his own pet project, a robotic rodent, Cuqee III, to win the international American Power Electronics Conference Micromouse Championship held in the USA earlier this year.
Having conquered national and international micromouse competitions, Mr Wyeth is preparing fourth-year Electrical and Computer Engineering Department students to build a robot soccer team to compete in Robo Cup in Paris next year.
The robots play soccer on a field the size of a ping-pong table and a golf ball is used in lieu of a soccer ball.
For more information, contact Mr Wyeth (telephone 07 3365 3770).
