search

9
search exploring the consequences of possible actions

Upload: fox

Post on 04-Jan-2016

40 views

Category:

Documents


0 download

DESCRIPTION

search. exploring the consequences of possible actions. search context. good for simple problem solving leads to... game playing planning expert-systems ...etc. jargon. state (node) a static problem state state space all possible problem states operator/move - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: search

search

exploring the consequences of possible actions

Page 2: search

search context

• good for simple problem solving

• leads to...game playing

planning

expert-systems

...etc...

Page 3: search

jargon

state (node) a static problem state

state spaceall possible problem states

operator/moveFn to generate one state from another

legal move generatorFn to generate all successor states from a given state

search strategy approach to exploring state space

Page 4: search

a basic search algorithm

search( start, goal )put the start state onto readyuntil goal is found or ready is empty do

select a state S from readyremove S from readyif S is a goal then

finished !else

add all unvisited successor states of S to ready

add S to visited

Page 5: search

issues

what kind of result is result preferred?

how can paths between states be represented?

what is the strategy for state selection?

how is the legal move generator specified?can it be of a standard form?

what about different search strategies?efficiency / costs / etc

Page 6: search

strategies 1

breadth-first• layer by layer through search tree• ready is a queue

depth first• exhausting one limb of tree before going to next• ready is a stack

best first• explore from least cost (maps.google.co.uk)

Page 7: search

strategies 2

heuristic (?)• explore from closest to goal

heuristic (ok)• explore from least (cost + closeness to goal)

admissible searches• will find solutions if they exist

optimal searches• will find solutions with minimal effort

Page 8: search

trad. game playing

• based on search

• assumes 2 players, trying to win

basics• static evaluation fn (+/- numeric value)• minimax search routine• alpha-beta pruning

Page 9: search

add-ons (mostly to minimax)

• heuristic growth

• heuristic pruning

• use different eval fns at different stages

• library moves (open game / end game)