DOC PREVIEW
UNCC ITCS 3153 - Uninformed Searches

This preview shows page 1-2-15-16-31-32 out of 32 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 32 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 32 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 32 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 32 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 32 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 32 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 32 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

ITCS 3153 Artificial Intelligence Lecture Lecture 33 Uninformed Uninformed Searches Searches mostly mostly copied copied from from Berkeley Berkeley Outline Problem Problem Solving Solving Agents Agents Restricted Restricted form form of of general general agent agent Problem Problem Types Types Fully Fully vs vs partially partially observable observable deterministic deterministic vs vs stochastic stochastic Problem Problem Formulation Formulation State State space space initial initial state state successor successor function function goal goal test test path path cost cost Example Example Problems Problems Basic Basic Search Search Algorithms Algorithms Problem Solving Agents Restricted Restricted form form of of general general agent agent function function SIMPLE PROBLEM SOLVING AGENT SIMPLE PROBLEM SOLVING AGENT percept percept returns returns an an action action static static seq seq an an action action sequence sequence initially initially empty empty state state some some description description of of the the current current world world state state goal goal aa goal goal initially initially null null problem problem aa problem problem definition definition state state UPDATE STATE UPDATE STATE state state percept percept if if seq seq is is empty empty then then goal goal FORMULATE GOAL FORMULATE GOAL state state problem problem FORMULATE PROBLEM FORMULATE PROBLEM state state goal goal seq seq SEARCH SEARCH problem problem action action RECOMMENDATION RECOMMENDATION seq seq state state seq seq REMAINDER REMAINDER seq seq state state return return action action Note Note This This is is offline offline problem problem solving solving solution solution with with eyes eyes closed closed Online Online problem problem solving solving involves involves acting acting without without complete complete knowledge knowledge Example Romania On On holiday holiday in in Romania Romania currently currently in in Arad Arad Flight Flight leaves leaves tomorrow tomorrow from from Bucharest Bucharest Formulate Formulate Goal Goal be be in in Bucharest Bucharest Formulate Formulate Problem Problem states states various various cities cities actions actions drive drive between between citites citites Find Find Solution Solution Sequence Sequence of of cities cities e g e g Arad Arad Sibiu Sibiu Fagaras Fagaras Bucharest Bucharest Example Romania Problem Types Deterministic Deterministic fully fully observable observable single state single state problem problem Agent Agent knows knows exactly exactly what what state state itit will will be be in in solution solution is is aa sequence sequence Non observable Non observable conformant conformant problem problem Agent Agent may may have have no no idea idea where where itit is is solution solution if if any any is is aa sequence sequence Non deterministic Non deterministic and or and or partially partially observable observable Percepts Percepts provide provide new new information information about about current current state state Solution Solution is is aa tree tree or or policy policy Often Often interleave interleave search search execution execution Unknown Unknown state state space space exploration exploration problem problem online online Example Vacuum World Single state Single state start start in in 5 5 Solution Solution Example Vacuum World Single state Single state start start in in 5 5 Solution Solution Right Right Suck Suck Conformant Conformant start start in in 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 E g E g right right goes goes to to 2 4 6 8 2 4 6 8 Solution Solution Example Vacuum World Single state Single state start start in in 5 5 Solution Solution Right Right Suck Suck Conformant Conformant start start in in 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 E g E g right right goes goes to to 2 4 6 8 2 4 6 8 Solution Solution Right Right Suck Suck Left Left Suck Suck Contingency Contingency start start in in 5 5 Murphy s Murphy s Law Law Suck Suck can can dirty dirty aa clean clean carpet carpet Local Local sensing sensing dirt dirt location location only only Solution Solution Example Vacuum World Single state Single state start start in in 5 5 Solution Solution Right Right Suck Suck Conformant Conformant start start in in 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 E g E g right right goes goes to to 2 4 6 8 2 4 6 8 Solution Solution Right Right Suck Suck Left Left Suck Suck Contingency Contingency start start in in 5 5 Murphy s Murphy s Law Law Suck Suck can can dirty dirty aa clean clean carpet carpet Local Local sensing sensing dirt dirt location location only only Solution Solution Right Right ifif dirt dirt then then Suck Suck Single state problem formation A A problem problem is is defined defined by by four four items items Initial Initial state state E g E g at atArad Arad Successor Successor function function S x S x set set of of action state action state pairs pairs E g E g S Arad S Arad Arad Arad Zerind Zerind Zerind Zerind Arad Arad Sibiu Sibiu Sibiu Sibiu Goal Goal test test can can be be Explicit Explicit e g e g at at Bucharest Bucharest Implicit Implicit e g e g NoDirt x NoDirt x Path Path cost cost additive additive E g E g aa sum sum of of distances distances number number of of actions actions executed executed etc etc C x a y C x a y is is the the step step cost cost assumed assumed to to be be non negative non negative A A solution solution is is aa sequence sequence of of actions actions leading leading from from the the initial initial state state to to aa goal goal state state State Space Real Real world world is is absurdly absurdly complex complex state state space space must must be be abstracted abstracted for for problem problem solving solving Abstract Abstract state state set set of of real real states states Abstract Abstract action action complex complex combination combination of of real real actions actions e g e g Arad Arad Zerind Zerind represents represents aa complex complex set set of of possible possible routes routes detours detours rest rest stops stops etc etc Abstract Abstract solution solution set set of of real real paths paths that that are are solutions solutions in in the the real real world world Each Each abstract abstract action action should should be be easier easier than than the the original original problem problem Example Vacuum World state space graph States States Actions Actions Goal Goal test test Path Path cost cost Example Vacuum World state space graph States States Integer Integer dirt


View Full Document

UNCC ITCS 3153 - Uninformed Searches

Download Uninformed Searches
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Uninformed Searches and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Uninformed Searches 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?