Teams from the Australian Centre for Field Robotics are working on a range of projects to improve the capabilities of robotics and computational models and allow for greater planning, control and learning.
Robots and autonomous vehicles must sense the world around them and move safely. In planning such movements there are usually far too many possibilities to check, so we need clever search methods. This is the problem of planning. The models used in planning don’t match reality, therefore we need to ensure the desired motion is achieved. This is the problem of control. Associate Professor Ian Manchester is working on projects to address these issues.
Generating predictive computational models from recorded data is a common requirement in engineering and natural sciences, especially when first-principle models are unavailable or too complex. Our team is addressing the major challenges of modelling dynamic systems, such as dealing with uncertainty, ensuring model stability, and identifying long-term dependencies between inputs and outputs.
We are developing dynamic walking robots, a class of humanoid bipedal robots that can walk efficiently by taking advantage of physical dynamics. They are often underused and some can be passive. There are many challenges for systematic design which come from the complexity of the dynamics.