Department of Engineering >>
Australian National University >>
Robotics and Control webpage
Maintained by Rob Mahony
Robotics
Robot manipulators are a widely used item of manufacturing equipment. For example, manipulators find application in rapid repetitive tasks such as welding and spray painting on production lines.
One project deals with the modelling and robotic control of flexible materials. Because of the flexibility, there are a number of difficult problems to be resolved, such as the use of force and vision sensing.
Humans are adept at learning and adapting when carrying out tasks. Work is underway which aims to study such learning and adaptive behaviour and to apply the findings to the intelligent control of robots.
One of the SCARA robot manipulators in the Department of Engineering.
A basic problem in robotics is to control the robot so that it follows a specified path as fast as possible. Robust fast path tracking controllers are being designed to achive accurate high-speed tracking of geometric paths in the presence of uncertainty and modelling errors.
Research into legged vehicles is motivated by the need for vehicles that can traverse terrain that is too rough for existing wheeled vehicles. A biped walking machine has been designed and fabricated and control algorithms for it are being developed.
One of the XORBOT robot manipulators in the Department of Engineering.
Computer Vision
Computer vision describes the automatic deduction of the structure and properties of a static or dynamic three dimensional world from either single or multiple two-dimensional images of the world. For a machine to interact "intelligently" with the environment, it must have sensing capabilities for gathering information about its surrounding. Equipped with a video camera and computer vision algorithms, robots and computers can interact within and as part of their environment; they can make decisions on what objects are where in the camera view. The automatic recognition of objects from images has attracted significant attention for many years and although apparently effortless for the human visual system, the task has proved a very challenging one for computers.
Views from a robot's "eye" view point with recognized objects being marked. (S. Abdallah)
Students at the ANU have the opportunity to work with an internationally renowned computer vision group based at the Department of Engineering (DE) and the Research School of Information Science and Engineering (RSISE). Projects conducted by our students include: visual based navigation of mobile robots, vision systems for underwater robots, and human gesture recognition.
A computer equipped with a camera can interpret human gestures. For example, the computer allows user to play virtual musical instruments based on the movement in the user hands. (S. Abdallah)
Control Systems
Control theory is the science of controller analysis and design. Research ranges from implementation issues which are important in engineering practice to the development of new theory.
High-speed digital control of complex dynamical systems can lead to numerical sensitivity issues which should be addressed in both the design and implementation of digital controllers. Some such issues are the level of numerical accuracy required and the appropriate structure for implementing the chosen control algorithm.
A theory of feedback system analysis, design and synthesis is being developed that is able to optimise the performance and robustness of control systems. This is the subject of H-infinity controller synthesis. Two issues are considered paramount: (i) The theory should offer a quantitative measure of performance and robustness that leads directly to an optimisation problem for which a synthesis procedure is available. (ii) The theory must be accessible to practising engineers.
An advanced theory for nonlinear robust control is being developed based on a generalisation to nonlinear systems of the H-infinity criterion. In the general case of imperfect measurement feedback, the theory is built on the concept of an information state, which is related to the statistical concept of sufficient statistic. The theory is essentially one of controlling an infinite dimensional nonlinear dynamical system.
Dynamics in nonlinear robust control [animation below] (M.R. James)
Animation