El golpe

Deusto University
ESIDE

.............................
Artificial Intelligence and Knowledge Engineering

11/06/2007

Question 1 - [2 points - 30 minutes]
Assume that in the Treasure Hunt Project, episode I, the knowledge available about the
environment has changed in such a way that neither the treasure location nor the rocks – or
corals – location are known initially and at each stepit is only possible to find out what there
is, by asking the user.
In accordance to this supposition and assuming that the goal is, again, to find the treasure,
answer the following questions:
1. Problem environment properties
 Partially observable because every element of the environment that is relevant to
finding a solution can’t be observed at the beginning.
 Deterministic becausethe problem environment state is completely determined
by the current state and the action executed by the system.
 Static because the problem environment will not change WHILE the system is
deliberating which action to take.
 Sequential because previous actions affect the current state.
 Discrete-state, discrete-percepts, and discrete-actions. Discrete-state because the
problem environmentis made up of a limited number of cells and each problem
state is determined by the cell occupied by the adventurer. Discrete-percepts
because the problem environment is limited in the number of elements that make
it up (rocks, adventurer, treasure, empty cells). Discrete-actions because there is a
limited number of actions.
2. Problem solving strategy or method, indicating whether there areany variations
with respect to what you decided for your Project #1. Explain your answer.
 In project 1, an offline search method was selected which FOUND a complete
sequence of actions for the problem so that afterwards it would be carried out
straight down without observing the environment. This will not be possible
anymore because the problem environment is partially observable.Partially
observable environments appear to be stochastic from the point of view of the
system; therefore an ONLINE search method is needed. This implies the following
general strategy:
 OBSERVE ENVIRONMENT (ask user about the immediate surroundings of the
current cell).
 COMPUTE/DETERMINE ACTION TO EXECUTE (invoke the GENERAL ONLINE DEPTH
FIRST SEARCH on the current state. The first step in thisalgorithm is to check if the
current cell contains the treasure chest. In this case the algorithm finishes).
 CARRY OUT ACTION ON PROBLEM ENVIRONMENT
1

3. If the problem solving method you have answered for question 2 required changes in
your implementation (APART FROM CHANGES IN THE PROBLEM ENVIRONMENT
DESCRIPTION) explain what kind of changes they would be.

(You must answer theabove questions according to your Project #1 scenario, land or
water).
 The objects describing the problem environment would change. In particular, the
list of rocks wouldn’t be needed nor the location of the treasure.
 For each operator used in project 1, we would need to add a corresponding
reverse operator.
 It isn’t possible to use a heuristic so we wouldn’t have the object Heuristic.
A new object would be need for the General ONLine Depth First Search. The most
important method in this object would implement the General Online Depth First
Search Algorithm.

Question 2 - [0,5 points - 20 minutes]
Given the following supposition:


A given car manufacturer is going to install a system which identifies speed limit signs
in order to prevent drivers exceeding the speedlimit. To this end, the car will have two
long range cameras, one on each of the two wing mirrors.

Answer the following questions:
1. Which AI approach or general technique would be appropriate?. Explain your
answer.
 The problem type is CHARACTER RECOGNITION, in particular: Digit Recognition.
Usually this type of problem can’t be solved by KBS because it isn’t possible to
verbalize the...