2. Offline vs. online search
Offline search: With the transition
model, we can
1. Initial state
2. Possible actions: ACTIONS
3. Transition model: RESULT
4. Goal test: GOAL-TEST
5. Cost: STEP-COST, PATH-COST
Online search: Without the transition model,
we need to
3. Online search is useful in
There is a penalty for
Nondeterministic environments:
No environment model:
Examples of online search:
4. Competitive ratio
There are 2 types of path costs in online search:
Actual path cost:
Shortest path cost:
Competitive ratio = Actual path cost / Shortest path cost
5. Dead-end state
Some actions are irreversible, they lead to a dead-end state:
For example,
Safely explorable state space:
For example,
7. Why depth-first?
In online search: the agent actually walks in the environment.
In depth-first search: the agent only moves to its next position (with 1 or a few
actions)
However,
8. Online A* search: LRTA* (Learning Real-Time A*) algorithm
Ideas:
At each step, from state s, agent moves to the successor s’
which has the best estimated cost:
H(s):
13. Previous algorithms use atomic state representation:
Now, we move to factored representation:
Constraint satisfaction problems:
Constraint satisfaction problems (CSP)
20. Can CSPs be solved using searching on state space?
Yes, but
CSP solvers (e.g., constraint propagation methods) can be faster than state-space
searchers because
For example, once we have chosen {HaNoi = blue}
22. Local consistency: the key idea of constraint propagation
Enforcing local consistency in each part of the graph causes inconsistent values to be
eliminated throughout the graph.
Types of local consistency:
25. Arc consistency
Enforcing domains to satisfy binary constraints.
Xi is arc-consistent with respect to Xj if
For example, consider the constraint 𝐘 = 𝐗𝟐
To make the problem arc-consistent:
Example 2: Map coloring problem: