An Adaptive Genetic Algorithm for Solving N-

Queens Problem
Uddalok Sarkar1, *
Sayan Nag1
1
Department of Electrical Engineering
Jadavpur University
Kolkata, India
[email protected], *[email protected]

Abstract--In this paper a Meta-heuristic approach for solving the N-Queens Problem is introduced to find the
best possible solution in a reasonable amount of time. Genetic Algorithm is used with a novel fitness function
as the Meta-heuristic. The aim of N-Queens Problem is to place N queens on an N x N chessboard, in a way so
that no queen is in conflict with the others. Chromosome representation and genetic operations like Mutation
and Crossover are described in detail. Results show that this approach yields promising and satisfactory results
in less time compared to that obtained from the previous approaches for several large values of N.

Keywords--genetic algorithm, N-Queens problem, mutation, crossover.

1. Introduction:
Optimization problems in diverse domains including Engineering Design, Structural Optimization, Economics
and Scheduling Assignments need mathematical models and Objective Function. These problems can be of two
types: Unconstrained (without constraints) and Constrained (with Constraints). They involve both continuous as well
as discrete variables. The usual non-linear nature of these problems in association with a handful of restraints being
active at the global optima make the task of finding the optimal solutions all the more troublesome. Optimization
techniques and approaches have been successfully applied to various systems, namely, Stochastic Systems, Trajectory
Systems, and Deterministic Systems. Stochastic Systems comprises of the optimal control problem in the presence of
uncertainty. Trajectory Systems involves aerial paths but is mostly restricted to the computation of optimum
trajectories of the rockets and space flight trajectories. Deterministic Systems includes the computations of the
performance of several mathematical models of diverse engineering design problems for various physical systems
modelling with germane objective functions. Traditional methods which are used previously to solve these Multi-
objective Optimization problems include Gradient Descent, Dynamic Programming and Newton Methods. But they
are not devoid of shortcomings- they are computationally less efficient. The meta-heuristic algorithms came into
existence. These meta-heuristic approximate algorithms tend to provide better solutions in a reasonable amount of
time. There exists a number of meta-heuristics like Genetic Algorithm (GA) [1], Particle Swarm Optimization (PSO)
[2], Gravitational Search Algorithm (GSA) [3], Ant Colony Optimization (ACO) [4,5], Stimulated Annealing (SA)
[6,7], Plant Propagation Algorithm (PPA) [8,9] and so on [10,11].
NP (nondeterministic polynomial time) class of problems are those each of which is solvable in polynomial time
by a non-deterministic Turing machine. In other words, problems having no deterministic solutions which run in
polynomial time are called NP-class problems. They cannot be solved in real time using deterministic approaches
because of their high complexity of the order of 2n or (n!). Heuristic methods are used to solve these kinds of problems
in real time.

_____________________________
*
Corresponding Author, Email: [email protected]
 N Queens Problem is such an optimization problem attributed to the class of NP-Complete Problems. The goal
of N Queens Problem is to suitably place N number of Queens on an N x N chessboard in a way that there is no conflict
between them due to the arrangement, that is, there is no intersection between them vertically or horizontally or
diagonally. Eight Queens Problem is a just a special case of N Queens Problem, where N is equal to 8. Jordan B. and
Brett S. [12] considered extensions of the N Queens Problem including various board topologies and dimensions at
the same time surveying already known solutions for the N Queens Problem of placing N Queens on an N x N
chessboard. They simultaneously provided a simple solution for finding the intersections of diagonals. They explored
diverse areas of the problem, asserting various existing and novel hypotheses. Salabat K. et. al. [13] applied Ant
Colony Optimization for solving the 8 Queen Problem thereby sowing that ACO can yield better solution in decent
amount of time. Farhad S. et. al. [14] devised a novel technique for solving the N Queen Problem by using graph
theoretical approach- a hybrid of Depth First Traversal and Breadth First Traversal techniques.
Heuristic approaches are required to solve N Queens Problem in real time with optimal solutions. Genetic
algorithms (GA) is one such powerful heuristic method which is capable of efficiently solve the problem in real time
by virtue of its extensively developed exploration and exploitation properties. In this paper we propose a modified and
state-of-the-art Genetic Algorithm with a novel, efficient and robust Fitness Function for generating the best solution
for N Queens Problem with different random initial candidate solutions and calculating and comparing the fitness
values for each of the trial solutions. The paper is organized as follows: Section 2 contains a brief description of the
N-Queens Problem, Section 3 contains a detailed description of the Genetic Algorithm used along with the fitness
function, crossover and mutation operators used. Section 4 contains the Results and Table showing the optimal
solutions. Conclusion is provided in Section 5.

2. N-Queens Problem:
In 1848, A German Chess player Max Bezzel composed the 8-Queens Problem which aims to place 8 Queens in the
chess board in such a way that no two Queens can attack each other. In 1850 Franz Nauck gave the 1st solution to this
problem and generalized the problem to N-Queen problem for N non- attacking Queens on an N x N Chessboard. Time
complexity of an N-Queen problem is O(n!). Here, we are proposing a heuristic approach to obtain the best solutions
for this problem. We are depicting all the arrangements of an N x N board as an N-tuple (c1, c2, c3… cN), where ci
represents the position of the queen to be in ith column and cth row. Fig.1 shows an arrangement of 8 x 8 chessboard
and its 8-tuple representation.

Fig.1: Arrangement signifying tuple: (8, 6, 4, 1, 3, 5, 7, 2)

 3. Genetic Algorithm:
Genetic Algorithm is a widely accepted popular Evolutionary Optimization technique that imitates the process of
natural selection. Firstly, in Genetic Algorithm a population of individual chromosomes is selected randomly. These
chromosomes are better known as phenotypes in terms of genetics each of which has its individual set of characteristics
or traits widely known as genotypes. These genotypes can be altered by adopting various strategies including mutation
and crossover. These chromosomes are evolved towards better (eventually optimal) solutions for a combinatorial
optimization problem. This optimal solution is dependent on the termination criterion. Genetic Algorithm gives the
optimal solution depending on the nature of the fitness function and also on the structure of the algorithm used. Here,
Genetic Algorithm is used to solve the N Queens Problem. A general outline how Genetic Algorithm (GA) works is
given below:

1. A random population of candidate solutions is created and the fitness scores of the individuals are calculated
and the chromosomes are sorted in the population and ranked according to the fitness values.

2. Certain number of chromosomes will pass onto the next generation depending on a selection operator,
favoring the better individuals based on their ranking in the population.

3. Selected chromosomes acting as parents will take part in crossover operation to create children whose fitness
values are to be calculated simultaneously. Crossover probability is generally kept high, because it is seen to
give better children in terms of fitness values.

4. Then based on a mutation probability, a mutation operator is applied on new individuals which randomly
changes few chromosomes. Mutation probability is generally kept low.

5. Evaluated off-springs, together with their parents form the population for the next generation.

Steps 2-5 are repeated until a given number of iterations, that is, generations have been evaluated or solution
improvement rate falls below some threshold, that is, the difference between two best solutions from consecutive
generations is below a certain tolerance limit. Below mentioned are some detailed explanation of the population
initialization, the fitness function and the crossover and mutation operators used.

 Population Initialization: Firstly we initialize a random population of chromosome of length 1000. Every
chromosome here is actually a vector of length N, which actually is a random permutation of (1, 2, 3… N).

 Fitness Function: We design the Fitness of each chromosome in such a way that k number of queens on the same
diagonal situation will accumulate k-1 points to the fitness value. All these points are summed which obtains the
fitness value.

Let’s take the N x N chess board arrangement (c1, c2, c3… cN). We’ll first check whether more than one Queen
is on the same (↗) directed diagonals. This happens if,

For 𝑖 ∈ {1,2 … 𝑁} and 𝑗 ∈ {1, 2 … 𝑁} and 𝑖 ≠ 𝑗;

(c𝑖 − 𝑖) == (c𝑗 − 𝑗)

Similarly, for our next checking whether more than one Queen is on the same (↖) directed diagonals, this happens
if,

For 𝑖 ∈ {1,2 … 𝑁} and 𝑗 ∈ {1, 2 … 𝑁} and 𝑖 ≠ 𝑗;

 (𝑐𝑖 + 𝑖) == (𝑐𝑗 + 𝑗)

Thus we amass the fitness value if any of the above situation occurs. Which clearly shows that the fittest
chromosome is signified as 0 fitness value.

The fig.2 and fig.3 shows an ordinary and an optimal arrangement respectively with corresponding fitness values.

Q Q

Q Q

Q Q

Q Q

Q Q

Q Q

Q Q

Q Q

Fig.2: Fitness Value: 7 Fig.3: Fitness Value: 0

Pseudo code:

Function fitness (chromosome) {

t1 = 0; //number of repetitive queens in one diagonal while seen from left corner
t2 = 0; //number of repetitive queens in one diagonal while seen from right corner
size = length (chromosome);
for i= 1 to size:
f1 (i) = (chromosome (i)-i);
f2 (i) = ((1+size)-chromosome (i)-i);
end
f1=sort (f1);
f2=sort (f2);
for i=2 to size:
if(f1 (i) == f1 (i-1)) //checks whether two Queens are in same diagonals seeing from left corner or not
t1 = t1+1;
end
if(f2 (i) == f2(i-1)) //checks whether two Queens are in same diagonals seeing from right corner or not
t2 = t2+1;
end
end
fitness_value = t1 + t2;
return fitness_value;
}

 Crossover Function: For obtaining global optimum for a non-convex function the cross over function performs
a very important role for obtaining better offspring. As we are considering each of the chromosomes actually a
 random permutation of (1, 2, 3… N), so it was needed to design a permutation crossover. Here we are using the
order 1 crossover.

Order 1 Crossover is one sort of very simple permutation crossover in which 2 random points are selected from
parent-1 and the alleles between these points are carried over to the child and the other alleles from parent-2 which
are absent in child are carried and placed in the child in the order which they appear in parent-2.

Parent1: 52316487
Parent2: 18647532

Child: 87316452

Step1: In the above scenario, the allele points 3 and 4 are selected randomly from the parent-1 and the alleles
between these two points are carried over to the Child chromosome.

Step2: The alleles equivalent to (or equal to) the alleles between the crossover points of parent-1 are eliminated
from parent-2.

Step3: The rest of the alleles of parent-2 are placed in the child chromosome according to their order.

Actually we so designed our Crossover Function that it generates two child from the parent chromosome at the
same time. Let the parent chromosomes are A and B. Then the Child-1 is obtained by having A as Parent-1 and B
as Parent-2 and Child-2 is obtained by having B as Parent-1 and A as Parent-2.

 Bad Population repository: We Have made a repository of weak chromosomes. The repository is a size of ⌊√𝑁⌋,
where ⌊𝑥⌋ defines maximum integer less or equal to 𝑥.

1. In every iteration any 2 chromosomes are selected randomly and gone through crossover and mutation and
the offspring competes with the chromosomes of main population. If these offspring are fitter than the worst
chromosomes of the main population then the offspring substitutes the worst chromosomes of main
population.

2. In every iteration one chromosome is selected randomly and crossed over with any randomly selected
chromosome from main population. This offspring also competes with the chromosomes of main population
for its existence.

This idea helps us to improve the avoidance of local optimum. In every run we are keeping this repository
constant. No updates of chromosome is occurring here.

 Mutation: Mutation is very important in genetic algorithm for not to stuck the process in local optimum. Here
we have taken mutation probability 0.8. We have also applied double mutation with a probability of 0.4.

Before mutation: 26834175

After mutation: 26734185

In a single mutation function we are randomly selecting 2 alleles and swap them. By running this single mutation
two times in a succession we achieve the double mutation function.
 4. Experimental Results:
Table 1 shows the obtained optimal solutions to N-Queens problem for N = 10, 11… 25. We have obtained 3 optimal
solutions for each N using our metaheuristic algorithm. For each run the number of iterations taken by our algorithm
to reach the global optimum is also been shown accordingly. The obtained solutions i.e. the elements of N-tuples are
shown as table format under the Optimal Solution column. Results clearly shows that the optimal results are obtained
with much lesser value of iterations.

Table 1: Optimal Solutions and iterations taken

N Iterations Optimal Solution

25 1431 22 6 11 14 24 15 19 8 4 17 1 3 12 16 7 21 18 2 25 23 10 20 5 9 13
382 12 4 16 23 11 24 22 20 2 5 15 10 7 19 3 8 17 25 6 1 18 13 9 14 21
281 17 5 22 8 12 21 1 13 20 16 19 24 11 23 4 7 3 15 9 2 14 10 18 25 6
24 4043 6 8 15 17 14 10 23 18 24 2 4 11 7 3 16 13 19 9 22 5 21 12 1 20
392 13 3 22 6 18 2 23 11 4 14 17 19 24 8 20 5 1 15 10 7 21 12 9 16
420 6 14 21 17 24 16 1 7 4 10 3 18 15 23 11 19 2 5 8 13 22 20 9 12
23 1363 5 14 22 3 8 13 21 4 6 23 19 2 18 7 17 11 9 16 10 20 15 1 12
113 18 16 21 17 3 6 13 1 4 7 5 22 14 19 11 15 8 20 9 23 2 10 12
347 4 16 14 23 1 10 22 15 11 5 2 18 21 8 13 17 7 3 19 6 20 9 12
22 282 8 13 1 9 14 19 4 22 15 12 6 11 21 18 3 17 20 10 2 16 5 7
2044 13 18 12 5 8 1 11 16 2 6 21 9 15 19 22 3 17 4 7 10 20 14
1082 18 3 12 2 16 5 11 1 15 20 6 4 9 19 17 10 14 7 22 8 21 13
21 404 6 11 16 10 15 21 3 1 9 14 5 19 2 12 18 7 4 17 20 8 13
807 7 19 13 16 12 5 3 1 6 2 11 20 14 17 4 18 21 8 10 15 9
592 16 5 13 21 7 18 11 3 1 4 10 17 19 12 15 6 14 9 20 8 2
20 453 12 7 11 3 20 18 14 4 6 8 5 15 17 9 2 16 19 10 1 13
279 9 17 1 10 2 16 18 8 12 3 15 6 20 13 5 7 19 14 11 4
170 4 16 9 6 1 17 20 5 13 11 18 2 7 12 15 3 8 10 14 19
19 3166 1 5 9 13 17 8 12 15 6 3 19 4 14 18 10 2 11 16 7
1462 2 15 9 11 13 8 1 3 19 10 14 17 6 18 12 5 16 4 7
382 8 5 2 13 9 16 12 19 17 6 4 1 7 14 10 18 15 3 11
18 5730 12 3 11 13 5 9 1 4 7 16 10 17 15 18 8 2 14 6
271 13 4 16 1 10 7 3 12 17 2 6 18 11 14 8 15 5 9
1828 3 11 13 5 1 18 14 2 8 15 9 7 17 4 12 16 6 10
17 1244 11 5 3 6 13 10 14 17 15 4 2 7 9 12 8 1 16
338 15 2 12 17 8 5 11 9 3 16 13 10 1 6 4 7 14
114 16 10 7 5 8 12 14 17 2 4 13 11 9 6 1 3 15
16 98 5 15 2 11 6 16 3 10 7 4 14 1 13 9 12 8
330 14 10 7 4 12 1 9 5 2 15 3 8 11 13 16 6
255 3 13 11 8 4 1 12 15 2 16 9 6 14 10 7 5
15 378 12 8 5 13 6 10 2 4 14 9 11 15 7 1 3
 25 11 15 6 12 5 7 1 13 2 14 8 3 9 4 10
2749 3 10 6 2 5 15 13 9 4 14 7 11 1 8 12
14 49 11 1 10 5 9 2 12 14 7 13 4 6 8 3
1077 1 14 7 2 12 9 6 3 10 4 13 8 5 11
31 2 6 11 9 7 1 4 14 12 8 5 3 13 10
13 2272 12 6 3 10 4 11 8 2 7 13 1 9 5
224 9 12 8 3 1 13 5 10 6 11 2 4 7
8 8 13 3 6 11 5 12 9 4 2 7 10 1
12 2865 6 2 5 10 12 4 8 11 3 1 7 9
82 2 8 5 9 1 6 11 3 12 7 4 10
52 4 9 1 3 11 8 12 2 6 10 7 5
11 320 10 8 6 3 9 11 1 5 7 2 4
339 8 10 2 4 9 7 3 11 6 1 5
97 3 9 6 4 11 1 8 2 5 7 10
10 2383 6 8 1 4 9 5 2 10 3 7
125 4 10 7 2 6 3 1 8 5 9
251 5 7 4 1 8 2 9 6 3 10

5. Conclusion and future works:

This paper reveals that the N-Queens problem can be solved within a reasonable time by using metaheuristics
like Genetic Algorithm and its modified forms. N-Queens Problems have a modicum of practical usage yet they
represent a major class of NP problems that cannot be solved in a reasonable amount of time using deterministic
methods- the reason of their importance. Genetic Algorithms were framed as heuristics for solving problems with
good and bad solutions depending on the fitness values of the solutions, yet here they proved their ability to solve
combinatorial optimization problems with simple binary (yes and no) answers. In the previous literatures, attempts
were made to solve N-Queens Problems with Genetic Algorithms. In this paper with modifications in the fitness
functions, hereby giving rise to a novel modified GA, it is seen that this approach yields promising and satisfactory
results in less time compared to that obtained from the previous approaches.

The future works include the use of various other metaheuristics algorithms for solving the N-Queens
Problem with more efficiency. Also application of the proposed algorithm and other variants of it in solving other
scheduling problems will lead to interesting results.
Appendix-A:

Fig.4: An optimal Solution for 50x50 Chessboard

Solution in Tuple: (18, 34, 12, 25, 28, 35, 44, 26, 49, 13, 21, 14, 33, 41, 48, 50, 5, 3, 23, 15, 47, 37, 17, 7, 36, 39, 20,
29, 45, 42, 31, 16, 19, 9, 32, 43, 2, 11, 6, 46, 30, 8, 1, 40, 24, 10, 38, 22, 4, 27)
Appendix-B:

Fig.5: Another optimal Solution for 50x50 Chessboard

Solution in Tuple: (19, 16, 27, 46, 50, 3, 22, 28, 43, 2, 15, 10, 25, 35, 20, 29, 4, 41, 44, 21, 49, 12, 31, 9, 32, 42, 38,
34, 18, 13, 30, 14, 6, 37, 7, 24, 47, 40, 48, 8, 36, 33, 1, 23, 11, 39, 17, 5, 45, 26)
Appendix-C:
Solution of 100-Quuens Problem:

(34, 27, 61, 99, 2, 80, 66, 11, 54, 47, 51, 85,
83, 60, 45, 7, 37, 41, 18, 70, 17, 87, 10, 48,
79, 75, 88, 90, 98, 1, 63, 24, 42, 44, 91, 22,
100, 49, 28, 5, 8, 57, 65, 73, 76, 36, 21, 31,
84, 52, 59, 29, 94, 23, 81, 95, 50, 30, 6, 96,
39, 43, 69, 77, 92, 25, 72, 3, 97, 19, 14, 86,
71, 56, 38, 16, 78, 82, 4, 26, 12, 89, 15, 35,
40, 9, 53, 64, 74, 58, 32, 20, 68, 55, 33, 46,
67, 13, 93, 62)

References:
1. S.-J. Wu and P.-T. Chow, “Genetic algorithms for nonlinear mixed discrete-integer optimization problems
via meta-genetic parameter optimization,” Engineering Optimization, vol. 24, no. 2, pp. 137–159, 1995.

2. R. Eberhart and J. Kennedy, “A new optimizer using particle swarm theory,” in Proceedings of the 6th
International Symposium on Micro Machine and Human Science (MHS ’95), pp. 39– 43, IEEE, Nagoya,
Japan, October 1995.

3. Rashedi, E., Nezamabadi-pour, H., Saryazdi, S.: GSA: A Gravitational Search Algorithm 179(13), 2232–
2248 (2009).

4. M. Dorigo and G. D. Caro, “Ant algorithms for discrete optimization”, Artificial Life, vol. 5, no. 3, (1999),
pp. 137-172.

5. M. Dorigo and L. M. Gambardella, “Ant colony system: a cooperative learning approach to the traveling
salesman problem”, IEEE Transactions on Evolutionary Computation, vol. 1, no. 1, (1997), pp. 53-66.

6. C. Zhang and H.-P. Wang, “Mixed-discrete nonlinear optimization with simulated annealing,” Engineering
Optimization, vol. 21, no. 4, pp. 277–291, 1993.

7. E. H. L. Aarts, J. H. M. Korst, and P. J. M. van Laarhoven, “Simulated annealing,” in Local Search in

Combinatorial Optimization, pp. 91–120, 1997.

8. Nag, Sayan. "Adaptive Plant Propagation Algorithm for Solving Economic Load Dispatch Problem." arXiv
preprint arXiv:1708.07040 (2017).

9. Nag, Sayan. "A Type II Fuzzy Entropy Based Multi-Level Image Thresholding Using Adaptive Plant
Propagation Algorithm." arXiv preprint arXiv:1708.09461 (2017).

10. JC Bansal, H Sharma and SS Jadon “Artificial bee colony algorithm: a survey.” Int. J. of Advanced
Intelligence Paradigms 5.1 (2013): 123-159.

11. X.S. Yang, “Firefly algorithms for multimodal optimization”, LNCS, vol. 5792, pp. 169–178. Springer,
Heidelberg (2009).

12. Jordan B. and Brett S, "A survey of known results and research areas for n-queens", Discrete Mathematics
309, 2009.
13. Salabat K., Mohsin B., M. Sharif, Malik S., R. Baig, "Solution of n-Queen Problem Using ACO", IEEE,
2009.

14. Farhad S., Bahareh S. and G. Feyzipour, "A New Solution for N-Queens Problem using Blind Approaches:
DFS and BFS Algorithms", International Journal of Computer Applications (0975 – 8887) Volume 53– No.1,
September 2012.

15. Darrel Whitley and Nam-Wook Yoo, “Modeling Simple Genetic Algorithms for Permutation Problems”.

16. Naveen Kumar, Karambir, Rajiv Kumar, “A Comparative Analysis of PMX, CX and OX Crossover operators
for solving Travelling Salesman Problem”, International Journal of Latest Research in Science and
Technology, Vol.1, Issue 2 :Page No.98-101, July -August (2012)