About Robots Australia

This site is a collection of notes, links and info relating to my research interests in intelligent robotics and cognitive computing. 

The current project is the Quadruped4 robot "PEDRO" which uses a custom controller with WiFi and many sensor interfaces, and small networked servo motors (Robotis AX-12 servos) to implement a four legged robot about 20cm high and 30cm long with stereo vision and other sensing capabilities. Software based on the Microsoft Robotics Studio package is used to both simulate the robot in a virtual environment, and to control the actual physical robot via WiFi.

Latest News

I maintain a blog of progress which will contain the most up to date news, but there is also dedicated pages for individual topics on this site which get updated when major changes occur.

The most recent blog entry is below (click on the link to be taken to the full blog);

See more RobotsAustralia blog entries 

Concepts behind the robots on Robots Australia

The aim of the robots I build is to help in gaining an understanding of the nature of intelligence and cognition, whether in animals or machines, and to see whether man-made creations can be given the rudiments of intelligence.

Over the years I have been interested in robotics, I have homed in on a few approaches and ideas that form the main basis of the robots I am working on at the moment.

General Concepts behind Hardware/Firmware

• Hardware is as simple and lightweight as possible, using low power design in all mechanical and electronic aspects.
  
• Limited computing power on the robot itself, merely enough to perform lowest level communications and control. I look on the on-board electronics and communications as a very rough equivalent of the biological spinal cord and brain stem, providing fast reflexive or patterned rythmic control under direction of higher levels.
  
• radio link from low power computing robot platform to high computing power desk platform. This approach enables lightweight, highly mobile robots that still can have essentially unlimited computing resources in the higher levels of control. Single desktop/laptop computers can be used for simple systems, or networked systems of computers acting in parallel can be used to scale up the control paradigm.
• Passive/Dynamic mechanical behaviour. Some very good work has been done recently by a variety of universities in the area of passive dynamic biped robots. I think there is possibilities even in the home experimenter realm to use some of these ideas. I specifically intend to investigate using powered joints with torque/loading sensory capability to reduce power consumption by entering a 'powered free run' mode where the joint is actively servo controlled, however is driven in the direction of least resistance for example during the swing phase of a biped gait.

Concepts behind Intelligent Control

• Embodiment is a concept I think is important, whereby I believe that a rich sensory input, and real body dynamics can help in developing real cognition and intelligence. I don't necessary think it is an essential component, however at the very least it would help to demonstrate the effectiveness of any particular approach.
• I was very impressed by the ideas contained in the book On Intelligence by Jeff Hawkins and Sandra Blakeslee. Whilst far from being the end of the story, it is a fairly complete theory of cortical operation that has some sound logic behind it, and has areas of pure speculation clearly labelled as such. I intend to use some of the heirarchical models and ideas contained in this book, along with some evolutionary algorithm and directed design modular 'building block' network elements to produce cortical like behaviour and control of the robots.
  

Robots.net Headlines

Robots.net is a great site, updated very often with the latest news and press releases from around the world regarding robotics.