In this and the following sections, I'm working towards real robots or animals living in the real world, and I'm going to show just how big a leap this is from classical planning.
The previous two sections have concerned planning in some kind of ideal world, where an agent always has enough time and knowledge to predict the effect of a plan. That need not be so. One may lack time to form a good plan, so have to improvise with a bad; one may lack knowledge needed to choose one plan over another (or the time necessary to apply that knowledge). To get an idea of the approaches that have been tried, read Introduction to Chapter 4 of Readings in Planning, pages 187--188; Introduction to Artificial Intelligence by Charniak and McDermott, pages 518--519; and if you want to go further, Decision Theory and AI II in Readings in Planning, especially pages 207--208 and 212--217.