| Popis: |
Planning under uncertainty techniques are hard to apply to autonomous robotics problems: they require a tedious modeling effort when dealing with large state spaces. Factored representations, using state variables, are more compact, but they cannot cope very efficiently with variables that can take a large number of values, e.g. the robot localization variables. Search and rescue problems combine these with mission variables. We propose a generic hierarchical abstract model in order to simplify the modeling stage for robotics planning problems with action uncertainties. An algorithm automatically instanciates our abstract model, which is shown and assessed on several instances of search and rescue autonomous rotorcraft problems. |
