Announcements

• Course TA: Danny Kumar ([email protected])
Course TA: Danny Kumar (ndkumar@cs unc edu)
• Guest lectures next week
• Assignment 1 out, due Tuesday, September 14
i d d S b
• Course discussion board 
– You should be able to find the discussion thread 
“AI in the news/on the Web” on Blackboard
– For participation points: at least two posts during the 
semester
Solving problems by searching
gp y g
Chapter 3
 Search
• We will consider the problem of designing goal‐
based agents in observable, deterministic, discrete, 
based agents in observable deterministic discrete
known environments
• Example:
Start state

Goal state
 Search
• We will consider the problem of designing goal‐
based agents in observable, deterministic, discrete, 
based agents in observable deterministic discrete
known environments 
– The solution is a fixed sequence of actions
q
– Search is the process of looking for the sequence of 
actions that reaches the goal
– Once the agent begins executing the search solution, it can 
ignore its percepts (open‐loop system)
 Search problem components
Search problem components
• Initial state Initial
state
• Actions
A i
• Transition model
– What
What is the result of 
is the result of
performing a given action 
in a given state?
• Goal state
Goal state
• Path cost
– Assume that it is a sum of  Goal
nonnegative step costs state

• The optimal solution is the sequence of actions that gives the 
p q g
lowest path cost for reaching the goal
 Example: Romania
• On vacation in Romania; currently in Arad
• Flight leaves tomorrow from Bucharest
Flight leaves tomorrow from Bucharest

• Initial state
– Arad
• Actions
– Go from one city to another
• Transition model
– If you go from city A to 
city B you end up in city B
city B, you end up in city B
• Goal state
– Bucharest
• Path cost
– Sum of edge costs
 State space
State space
• The
The initial state, actions, and transition model 
initial state actions and transition model
define the state space of the problem
– The set of all states reachable from initial state by any 
The set of all states reachable from initial state by any
sequence of actions
– Can be represented as a directed graph where the 
Can be represented as a directed graph where the
nodes are states and links between nodes are actions
• What is the state space for the Romania problem?
What is the state space for the Romania problem?
 Example: Vacuum world

• States
– Agent location and dirt location
– How many possible states?
H ibl ?
– What if there are n possible locations?
• Actions
– Left, right, suck
• Transition model
Vacuum world state space graph
 Example: The 8‐puzzle
Example: The 8 puzzle
• States
– Locations of tiles 
Locations of tiles
• 8‐puzzle: 181,440 states
• 15‐puzzle: 1.3 trillion states
• 24‐puzzle: 1025 states
• Actions
– Move blank left, right, up, down 
• Path cost 
– 1 per move

• Optimal solution of n‐Puzzle is NP‐hard
 Example: Robot motion planning
Example: Robot motion planning

• States
– Real‐valued coordinates of robot joint angles
• Actions
– Continuous motions of robot joints
C ti ti f b tj i t
• Goal state
– Desired final configuration (e.g., object is grasped)
• Path cost
P th t
– Time to execute, smoothness of path, etc.
 Other Real‐World
Other Real World Examples
Examples
• Routing
• Touring
• VLSI layout
S l
• Assembly sequencing
• Protein design
 Search
• Given:
– Initial state
– Actions
– Transition model
– Goal state
– Path cost

• How do we find the optimal solution?
– How about building the state space and then using 
g p g
Dijkstra’s shortest path algorithm?
• The state space is huge!
• Complexity of Dijkstra’s
p y j is O(E
( + V log V), where V
g ), is the size 
of the state space
 Tree Search
Tree Search
• Let’s
Let s begin at the start node and expand
begin at the start node and expand it by 
it by
making a list of all possible successor states
• Maintain a fringe
Maintain a fringe or a list of unexpanded states
or a list of unexpanded states
• At each step, pick a state from the fringe to expand 
• Keep going until you reach the goal state
• Try to expand as few states as possible
y p p
 Search tree
Search tree
Starting
• “What if” tree of possible actions and  state
outcomes
• The root node corresponds to the  Action
starting state Successor
• The children of a node correspond to  state
the successor states of that node’s state
• A path through the tree corresponds to 
A path through the tree corresponds to …
a sequence of actions
– A solution is a path ending in the goal 
state … … …
• Nodes vs. states
– A state is a representation of a physical  Goal state
configuration, while a node is a data 
g
structure that is part of the search tree
 Tree Search Algorithm Outline
Tree Search Algorithm Outline
• Initialize the fringe using the starting state
the fringe using the starting state
• While the fringe is not empty
– Choose a fringe node to expand according to search strategy
g p g gy
– If the node contains the goal state, return solution
– Else expand the node and add its children to the fringe
Tree search example
Tree search example
Tree search example
Tree search example
Tree search example
Tree search example
Fringe
g
 Search strategies
Search strategies
• A search strategy is defined by picking the order of node 
expansion
• Strategies are evaluated along the following dimensions:
– Completeness: does it always find a solution if one exists?
does it always find a solution if one exists?
– Optimality: does it always find a least‐cost solution?
– Time complexity: number of nodes generated
– Space complexity: maximum number of nodes in memory
Space complexity: maximum number of nodes in memory
• Time and space complexity are measured in terms of 
– b: maximum branching factor of the search tree
– d: depth of the least‐cost solution
d d h f h l l i
– m: maximum length of any path in the state space (may be infinite)
 Uninformed search strategies
Uninformed search strategies
• Uninformed search strategies use only the 
search strategies use only the
information available in the problem definition

• Breadth‐first search
• Uniform‐cost search
• Depth‐first search
p
• Iterative deepening search