The Leeds VWE group have developed software capable of displaying very complex configurations of physical objects and modelling movements of these objects subject to various constraints. However, the simple operations needed to perform a high-level task must always be given explicitly as input to the system. What would be extremely useful would be a module capable of computing an appropriate sequence of operations from a high-level goal description.
In general this is a very difficult problem requiring reasoning about many kinds of spatial and kinematic information. Hence, it will be sensible to focus on a more restricted problem. A limited sub-problem which may be suitable for a case study is that of reasoning about how to pick up and carry different kinds of object.