Artificial Intelligence and Knowledge Engineering Laboratory

Task 2. Constraint Satisfaction Problem

Authors: H. Kwaśnicka, M. Piasecki, M. Zięba, M. Paradowski, M. Przewoźniczek, M. Laszczyk

Task Objectives

Getting familiar with algorithms for Constraint Satisfaction Problems in practice through individual
implementation.

Subtasks

• Get familiar with backtracking search algorithm and forward checking heuristic
• Analyse both problems shortly described below
• Formulate them as a CSP problems, i.e. define: variables, their domains and constraints in a formal or
at least semi-formal way
• Implement backtracking algorithms for the chosen problem and enhance them with forward
checking heuristics
• Test the implementations with problems of different sizes of the boards (​n​x ​n​)
• Present most interesting results: solutions and numerical results (e.g. execution times, also for
problems that cannot be solved)
• Discuss the results
• The report should contain all the above points (except the implementation)

Warming up Problem: Letter Mosaic

Write a CSP-based program to find a solution to the following problem. There is a rectangular board of
n​x ​n​fields. Fill the fields with two different letters (i.e. x and y) given as a part of the problem
definition in such a way that no three horizontally or vertically consecutive fields are filled with the
same letter (two same consecutive letters are allowed) and all columns and all rows are unique.
If there is no solution, then the program should inform about this.

Entertaining Problem: Hitori

Problem consists of a ​n x n​grid filled with numbers. The objective is to shade some of the cells to
satisfy the following requirements:
- no number appears in a row or a column more than once (considering only unshaded cells)
- shaded cells do not touch each other vertically or horizontally
- all unshaded cells create a single continuous area (can be fully traversed when moving only
horizontally or vertically)
Sample Hitori puzzles can be created manually or downloaded.

Task rating

2 points – formulate the problems as CSP

3 points – implement backtracking and forward checking CSP for them
3 points – compare efficiency of processing for various problem parameters
2 points – give a correct interpretation of the results

Exciting Problem for 2 Extra Points​.

This as an extension of the previous problem. Often there are multiple solutions to the Hitori problem
(especially for larger boards). The goal is to find the solution that minimizes the number of shaded
cells.
Bibliography
1. Lecture notes
2. Russell and Norvig: http://aima.cs.ber​
keley.edu/2nd-ed/newchap05.pdf