search-perspective[1] - Search, Planning,& CSPs...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: 2/18/10 Search, Planning, & CSPs Making Sense Of It All Ron Parr CPS 170 How Do Search, Planning & CSPs fit together? •  Keep in mind these are all very general frameworks •  We typically think of search as the most general: –  –  –  –  Start Goal AcNons Costs •  We can formulate almost anything as search, even in a not enNrely unnatural way: –  –  –  –  Shortest path SorNng Planning CSPs •  Not everything that can be solved as search should be solved as search. The fact that you are holding a hammer doesn’t make everything a nail. 1 2/18/10 Algorithm vs. Concept •  There are Nmes when we will talk about search as a specific algorithm, i.e., something maintains a queue, pops things off the queue, expands them, etc. •  Other Nmes we will talk about search as a more abstract concept, e.g., finding a minimum of a funcNon by gradient descent can be thought of as a kind of search, even though we don’t maintain a queue CSPs •  Can formulate CSPs as search –  Goal = saNsfying assignment –  States = parNal assignments –  AcNons = assigning values to variables •  Using a generic search may not be a good idea: –  We don’t care about the path –  We don’t care about costs –  We have a largish branching factor –  We may miss opportuniNes to exploit structure in the problem, e.g., noNcing the structure of the constraint graph 2 2/18/10 Planning •  We can formulate planning as search –  Goal = plan goal –  States = SituaNons reachable from start state –  AcNons = plan acNons •  This seems like a be[er fit for search than CSPs (and it is), but –  The branching factor is huge –  The goal is usually a state set –  Difficult to come up with good heurisNcs •  We need to do something more clever than simply applying generic search techniques NoNonal View of Problem Classes CSPs Planning Search NB: To make this rigorous we would need to be a bit more precise and rigorous in our definiNons than what is expected/required for this class. 3 ...
View Full Document

{[ snackBarMessage ]}

Ask a homework question - tutors are online