tutorial-notes-3-dearden

Planning and Learning in Hybrid Discrete ContinuousModelsRichard DeardenJune 20, 2005AbstractMany real world problems require richer representations than are typically studied in planning andlearning. For example, state estimation in complex systems such as vehicles or spacecraft often requires arepresentation that captures the rich continuous behaviour of these kinds of systems. Similarly, planningfor such systems may require a representation of continuous resource usage, particularly whenunder uncertainty. In this talk I will discuss some commonly used representations of these systems ashybrid systems, examine some approaches to planning and state estimation in them, and finally discusssome first steps toward learning a hybrid model, or at least parameters of such a model, directly fromdata.These notes are intended as a set of background information for an IJCAI tutorial talk. They are madeup of pieces of text taken from a variety of places, without necessarily having any continuity1 IntroductionIn these notes I discuss classical AI problems such as planning, diagnosis, and learning but applied ina much richer space of models than is used in traditional AI applications. The motivation for this isapplications such as Mars rovers, automated factories, chemical process plants, and spacecraft for whichdiscrete models such as those traditionally used in planning and learning are inadequate. In particular,we will look at systems represented using hybrid ...