CS 188: Artificial Intelligence

Search

Instructors: Stuart Russell and Dawn Song

University of California, Berkeley
[slides adapted from Dan Klein, Pieter Abbeel]
 Today
 Agents that Plan Ahead

 Search Problems

 Uninformed Search Methods

 Depth-First Search
 Breadth-First Search
 Uniform-Cost Search
 Planning Agents

 Planning agents decide based on evaluating

future action sequences
 Search algorithms typically assume
 Known, deterministic transition model
 Discrete states and actions
 Fully observable
 Atomic representation
 Usually have a definite goal
 Optimal: Achieve goal at least cost
Move to nearest dot and eat it
Precompute optimal plan, execute it
Search Problems
 Search Problems
 A search problem consists of:

 A state space S
 An initial state s0
 Actions A(s) in each state
N -9
 Transition model Result(s,a)
 A goal test G(s) E -9
 s has no dots left
 Action cost c(s,a,s’)
 +1 per step; -10 food; -500 win; +500 die; -200 eat ghost

 A solution is an action sequence that reaches a goal state

 An optimal solution has least cost among all solutions
Search Problems Are Models
Example: Traveling in Romania
But then…
Example: Traveling in Romania
 State space:
 Cities
 Initial state:
 Arad
 Actions:
 Go to adjacent city
 Transition model:
 Reach adjacent city
 Goal test:
 s = Bucharest?
 Action cost:
 Road distance from s to s’
 Solution?
Models are almost always wrong
 State Space Sizes

 World state:
 Agent positions: 120
 Food count: 30
 Ghost positions: 12
 Agent facing: NSEW

 How many
 World states?
120x(230)x(122)x4
 States for pathing?
120
 States for eat-all-dots?
120x(230)
State Space Graphs and Search Trees
 State Space Graphs

 State space graph: A mathematical

representation of a search problem
 Nodes are (abstracted) world configurations
 Arcs represent transitions (labeled with actions)
 The goal test is a set of goal nodes (maybe only one)

 In a state space graph, each state occurs only

once!

 We can rarely build this full graph in memory

(it’s too big), but it’s a useful idea
 State Space Graphs

 State space graph: A mathematical

a G
representation of a search problem
 Nodes are (abstracted) world configurations b c
 Arcs represent successors (action results) e
 The goal test is a set of goal nodes (maybe only one) d f
S h
 In a state space graph, each state occurs only p r
q
once!
Tiny state space graph for a tiny
 We can rarely build this full graph in memory search problem
(it’s too big), but it’s a useful idea
 State Space Graphs vs. Search Trees

Each NODE in in
State Space Graph the search tree is Search Tree
an entire PATH in
the state space S

a G graph. e p
d
b c
b c e h r q
e
d f a a h r p q f
S h We construct the
tree on demand – p q f q c G
p q r
and we construct as q c G a
little as possible.
a
Quiz: State Space Graphs vs. Search Trees

Consider this 4-state graph: How big is its search tree (from S)?

S G

b
Quiz: State Space Graphs vs. Search Trees

Consider this 4-state graph: How big is its search tree (from S)?

a s
a b
S G
b G a G
b a G b G

… …

Important: Those who don’t know history are doomed to repeat it!
Quiz: State Space Graphs vs. Search Trees

Consider a rectangular grid: How many states within d steps of start?

How many states in search tree of depth d?

Tree Search
Search Example: Romania
Creating the search tree
Creating the search tree
Creating the search tree
 General Tree Search

 Main variations:
 Which leaf node to expand next
 Whether to check for repeated states
 Data structures for frontier, expanded nodes
 Systematic search
frontier

reached =
unexplored expanded U frontier
expanded

1. Frontier separates expanded from unexplored region of state-space graph

2. Expanding a frontier node:
a. Moves a node from frontier into expanded
b. Adds nodes from unexplored into frontier, maintaining property 1
Depth-First Search
 Depth-First Search
Strategy: expand a a G
deepest node first b c

Implementation: d
e
f
Frontier is a LIFO stack S h
p q r

d e p

b c e h r q

a a h r p q f

p q f q c G

q c G a

a
Search Algorithm Properties
 Search Algorithm Properties
 Complete: Guaranteed to find a solution if one exists?
 Optimal: Guaranteed to find the least cost path?
 Time complexity?
 Space complexity? b
1 node
… b nodes

 Cartoon of search tree: b2 nodes

 b is the branching factor m tiers
 m is the maximum depth
 solutions at various depths
bm nodes
 Number of nodes in entire tree?
 1 + b + b2 + …. bm = O(bm)
 Depth-First Search (DFS) Properties
 What nodes does DFS expand?
 Some left prefix of the tree down to depth m. 1 node
b
 Could process the whole tree! … b nodes
 If m is finite, takes time O(bm) b2 nodes
m tiers
 How much space does the frontier take?
 Only has siblings on path to root, so O(bm)

 Is it complete? bm nodes
 m could be infinite
 preventing cycles may help (more later)

 Is it optimal?
 No, it finds the “leftmost” solution, regardless
of depth or cost
Breadth-First Search
 Breadth-First Search
Strategy: expand a a G
shallowest node first b c
Implementation: e
d f
Frontier is a FIFO queue S h
p q r

d e p
Search
b c e h r q
Tiers
a a h r p q f

p q f q c G

q c G a

a
 Breadth-First Search (BFS) Properties
 What nodes does BFS expand?
 Processes all nodes above shallowest solution 1 node
b
 Let depth of shallowest solution be s … b nodes
s tiers
 Search takes time O(bs) b2 nodes

 How much space does the frontier take? bs nodes

 Has roughly the last tier, so O(bs)

 Is it complete? bm nodes
 s must be finite if a solution exists, so yes!

 Is it optimal?
 If costs are equal (e.g., 1)
Quiz: DFS vs BFS
 Quiz: DFS vs BFS

 When will BFS outperform DFS?

 When will DFS outperform BFS?

[Demo: dfs/bfs maze water (L2D6)]

Example: Maze Water DFS/BFS (part 1)
Example: Maze Water DFS/BFS (part 2)
 Iterative Deepening
 Idea: get DFS’s space advantage with BFS’s
time / shallow-solution advantages b
 Run a DFS with depth limit 1. If no solution… …

 Run a DFS with depth limit 2. If no solution…

 Run a DFS with depth limit 3. …..

 Isn’t that wastefully redundant?

 Generally most work happens in the lowest
level searched, so not so bad!
Uniform Cost Search
 Uniform Cost Search
2 a G
g(n) = cost from root to n b c
1 8 2
Strategy: expand lowest g(n) 2 e
3 d f
Frontier is a priority queue 9 2
S h 8
sorted by g(n) 1
1 p q r
15

S 0

d 3 e 9 p 1

b 4 c e 5 h 17 r 11 q 16
11
Cost a 6 a h 13 r 7 p q f
contours
p q f 8 q c G

q 11 c G 10 a

a
 Uniform Cost Search (UCS) Properties
 What nodes does UCS expand?
 Expands all nodes with cost less than cheapest solution!
b g1
 If that solution costs C* and arcs cost at least  , then the …
“effective depth” is roughly C*/ g2
C*/ “tiers”
 Takes time O(b ) (exponential in effective depth)
C*/ 
g3

 How much space does the frontier take?

 Has roughly the last tier, so O(bC*/)

 Is it complete?
 Assuming C* is finite and  > 0, yes!

 Is it optimal?
 Yes! (Proof next lecture via A*)
Video of Demo Maze with Deep/Shallow Water --- BFS or UCS? (part 1)
Video of Demo Maze with Deep/Shallow Water --- BFS or UCS? (part 2)
 Summary
 Assume known, discrete, observable, deterministic, atomic
 Search problems defined by S, s0, A(s), Result(s,a), G(s), c(s,a,s’)
 Search algorithms find action sequences that reach goal states
 Optimal => minimum-cost
 Search algorithm properties:
 Depth-first: incomplete, suboptimal, space-efficient
 Breadth-first: complete, (sub)optimal, space-prohibitive
 Iterative deepening: complete, (sub)optimal, space-efficient
 Uniform-cost: complete, optimal, space-prohibitive