Robotics Research

Applications: World Modelling

A world model is a representation of the external world contained inside the intelligence of a robot. This representation can either be built online by the robot itself using some kind of mapping technique based on its perception of the environment (SLAM for example), or it can take the form of a priori knowledge given to the robot offline. Both paradigms have advantages and disadvantages, and indeed some solutions combine elements of online mapping to allow for dynamic and unknown environments, while incorporating some a priori knowledge about important static conditions.

Another critical idea related to world modelling is localisation. Localisation is the process by which a robot establishes its own position (location) and orientation within its internal world model. In order for a robot to successfully make use of a world model or to map an environment accurate localisation is required. Many navigation techniques also rely heavily on successful localisation. Effective localisation in arbitrary environments over long periods of time is by no means a trivial task, and is still the subject of extensive research.

The goal of a large amount of current modelling research is to achieve Simultaneous Localisation and Mapping, or SLAM, where a mobile robot is able to locate itself in an environment while at the same time building a map of that environment.

World Modelling Research in Australia

A number of groups around Australia are involved specifically in SLAM research.

• David Austin at ANU is working on SLAM for indoor mobile robots, using vision as the primary sensor.
  
• Gamini Dissanayake at UTS is also developing SLAM capabilities, and is working on localisation of mining vehicles.
  
• The ACFR is developing systems for a number of SLAM applications, including indoor, land vehicle, subsea and airborne. They are also working on combining SLAM techniques with inertial navigation systems.
  
• Gordon Wyeth and the Mobile Robot Lab at the University of Queensland are developing an indoor mobile robot SLAM system based on a rat hippocampus.

• 

  The RatSLAM mobile robot at the University of Queensland. The robot attempts to achieve SLAM using a modelling approach inspired by the rat hippocampus.

• The Intelligent Robotics Research Centre (IRRC) at Monash University also has expertise in SLAM techniques.
  

Research is also being conducted on other areas of modelling.

• The ACFR is attempting to develop effective models of the dynamics of various systems including slip and traction in land vehicles, hydrodynamics and mass entrainment in undersea vehicles, and aerodynamics in aircraft.
• The Intelligent Systems and Automation Group at Deakin university has been involved in the modelling and simulation of power and manufacturing systems, as well as the application of ensembles of models to various systems.