Quiz Th 41207 CompSci 171 Intro AI Max Welling

1
Quiz Th 4-12-07CompSci 171Intro AIMax
Welling
2
Crossword Puzzle Robot. 10pts Imagine a
company build a the following program into the a
cell-phone. Using the camera it would make a
picture of a crossword puzzle Software then
(perfectly) recognizes the white and black
squares and the numbers corresponding to the
questions. You are required to enter the
questions manually using the keyboard, starting
with its number. The device then connects to a
server which contains a huge amount of crossword
puzzle questions and their answers and tries to
solve it. 1) 2pts Describe the Task
Environment for this robot (explain all letters
in PEAS). 2) 2pts What are the natural
abstract states for this problem? 3) 2pts
Using these abstract states, choose between the
following properties of the task environment
(you may assume the recognition software works
perfectly). a) fully/partially observable, b)
deterministic/stochastic/strategic, c)
episodic/sequential, d) static/dynamic/semi
dynamic, e) discrete/continuous, f) single/multi
agent. 4) 2pts We will now search for a
solution. Describe a) the initial state, b) the
goal state, c) a successor function (the set of
all possible actions starting at some state x),
d) a path cost. 5) 2pts We will use
breadth-first-search to solve this problem.
Are we guaranteed to find a solution using BFS?
Explain why or why not.
3
1) 2pts Describe the Task Environment for this
robot (explain all letters in PEAS). Performance
Measure Is the puzzle solved Environment
Puzzle printed on paper Actuators screen that
shows the solution Sensor keyboard, camera 2)
2pts What are the natural abstract states for
this problem? Partial filled out puzzles. Order
questions and at level d there are possible
answers to the first d questions. 3) 2pts Using
these abstract states, choose between the
following properties of the task environment a)
fully/partially observable, b)
deterministic/stochastic/strategic, c)
episodic/sequential, d) static/dynamic/semi
dynamic, e) discrete/continuous, f) single/multi
agent. 4) 2pts We will now search for a
solution. Describe a) the initial state empty
puzzle b) the goal state solved puzzle c) a
successor function all possible answers to the
next question in some ordering consistent with
previous answers in that branch. d) a path cost.
constant cost per solved question 5) 2pts We
will use breadth-first-search to solve this
problem. Are we guaranteed to find a
solution using BFS? Explain why or why not. YES
we will always find a solution at level d where d
is the number of questions. All states higher in
the tree are partial solutions. BFS scans all
possible solutions to the puzzle.