Yugoslav Journal of Operations Research

21 (2011), Number 1, 93-102

DOI: 10.2298/YJOR1101093S

SOLVING THE TWO-DIMENSIONAL PACKING PROBLEM

WITH m-M CALCULUS

Aleksandar SAVIĆ,
Faculty of Mathematics, University of Belgrade
aleks3rd@gmail.com
Tijana ŠUKILOVIĆ,
Faculty of Mathematics, University of Belgrade
tijana@matf.bg.ac.rs
Vladimir FILIPOVIĆ
Faculty of Mathematics, University of Belgrade
vladofilipovic@hotmail.com

Received: July 2010 / Accepted: May 2011

Abstract: This paper considers the two dimensional rectangular packing problem. The
mathematical formulation is based on the optimization of a non-linear function with
piecewise linear constraints with a small number of real variables. The presented method
of m-M calculus finds all optimal solutions on small instances. Computational
performance is good on smaller instances.
Key words: Non-linear optimization, m-M calculus, two dimensional packing.
MSC: 90C30, 90C56

1. INTRODUCTION

Packing problem and the closely related problem of cutting, encompass a wide
range of industry originated problems. A successful optimal solution, or even finding an
approximately good solution, significantly facilitate in both saving money and raw
materials.
Most of the contributions in literature are devoted to the case where the items to
be packed have a fixed orientation with respect to the stock unit(s), i.e., it is not allowed
to rotate them. This case, which is the object of the present article, reflects a number of
94 A. Savić, T. Šukilović, V. Filipović / Solving 2-dim. Packing Problem

practical contexts, such as the cutting of corrugated or decorated material (wood, glass,
clothing stripes), or the newspapers paging.
For variants allowing rotations (usually by 90º) and/or constraints on the items
placement (such as the ‘‘guillotine cuts’’) the reader is referred to [1, 8], where a three-
field classification of the area is also introduced. General surveys on cutting and packing
problems can be found in [10, 11, 12]. Results of the probabilistic analysis of packing
algorithms can be found in [6] and [7].
Let us introduce the problems in a more formal way. We are given a set of n
rectangular items j ∈ J = {1,K , n} each defined by a width, wj, and a height, h j:
(i) in the Two-Dimensional Bin Packing Problem (2BP), we are further given an
unlimited number of identical rectangular bins of width W and height H, and the
objective is to allocate all the items to the minimum number of bins;
(ii) in the Two-Dimensional Strip Packing Problem (2SP), we are further given
a bin of width W and infinite height (hereafter called strip), and the objective is to
allocate all the items to the strip by minimizing the height to which the strip is used.
In both cases, the items have to be packed with their w-edges parallel to the W-
edge of the bins (or strip). We will assume, without loss of generality, that all input data
are positive integers, and that w j ≤ W and hj ≤ H (j = 1, . . . , n).
Both problems are strongly NP-hard, as is easily seen by transformation from
the strongly NP-hard (one-dimensional) Bin Packing Problem (1BP), where n items, each
having an associated size hj, have to be partitioned into the minimum number of subsets,
so that the sum of the sizes in each subset does not exceed a given capacity H [20].
A special case of the aforementioned problems is packing a set of rectangles
into a bounding rectangle.
The problem of determining whether a set of rectangles can be packed into a
bounding rectangle can be related to some practical applications in floor planning,
placement, or job scheduling problems [9]. The rectangle packing problem belongs to a
subset of classical cutting and packing problems and has been shown to be NP-complete
[17, 19].
Cutting and packing problems [1, 2, 4, 8, 15, 25] have a wide range of
applicability that has been studied for more than 40 years. The two-dimensional cutting
stock problems with rectangle shapes are closely related to our problem. In [5], the basic
formulation issues and solution procedures of linear programming, sequential heuristic
and hybrid solution procedures are addressed. Based on the usage of guillotine or non-
guillotine cuts, the problems can be roughly divided into two styles of cutting.
Constrained cutting refers to the employment of orthogonal and guillotine cuts under
certain copy constraints of the rectangles [13]. Optimal algorithms for orthogonal two-
dimension cutting were proposed in [2, 15]. However, they might not be practical for
large problems. In the orthogonal packing problem, a set of rectangles are to be packed
into a rectangle board with rectangle edges parallel to the x- and y-axes of the board,
respectively, and the height is to be minimized. In order to reduce the number of possible
orthogonal packing patterns, a so-called BL-condition was introduced, and a combination
algorithm using genetic and deterministic methods was proposed [3, 14, 18]. In [22], a
new and improved algorithm with level heuristics was suggested. In [16, 21, 25], a
various range of metaheuristics was suggested for solving the two dimensional rectangle
packing problem. A very effective quasi-human heuristic was presented in [26]
 A. Savić, T. Šukilović, V. Filipović / Solving 2-dim. Packing Problem 95

This work presents an application of m-M calculus [23] on the mathematical

model given in [24]. The proposed method of solving will allow finding all optimal
solutions for the problem, which can contribute to the quality of the solution, because
some different optimal solutions can be more or less preferable by some other criterion.

2. MODEL

The problem expounded is the two dimensional packing of a big rectangle with
a given number of small rectangles. It is important to consider that the sides of small
rectangles must be parallel with the sides of a big rectangle. Two cases are possible: in
the first case, a 90° degree rotation for smaller rectangles is not allowed, while the second
case allows rotation. Smaller rectangles must be placed completely inside the big
rectangle and they must not overlap. Let the dimensions of the big rectangle be
respectively A and B. There are n small rectangles. Let their dimensions be respectively ai
and bi i = 1,. . . ., n. It is allowed for the small rectangles to have the same dimensions.

b1
a1
b2
B a2
M

bn
A an
Figure 1 Big rectangle and array of smaller rectangles

Let us consider a strip of width A and infinite length. Rectangles will be packed
on this strip, and let bounding rectangle be at the beginning of the strip, as shown on
Figure 2.

A
Figure 2 Infinite strip with big rectangle at its beginning packed with smaller
rectangles
96 A. Savić, T. Šukilović, V. Filipović / Solving 2-dim. Packing Problem

Let xi and yi, i = 1, . . . , n be coordinates of the centers of rectangles in the

coordinate system, where strip starts from x-axis and is spreading in the direction of y-
axis. Bounding rectangle then has vertices in points with coordinates (0,0), (0,B), (A,B)
and (A,0). Model is presented and proofed in [24]. Now the mathematical model for
packing rectangles into bounding rectangle is as follows.
n
max(0, B − yi )
max ∑ ai bi (1)
i =1 B − yi

x j − xi y j − yi x j − xi y j − yi
− + + ≥ 1, i = 1, 2, K , n − 1, j = i + 1,K , n (2)
ai + a j bi + b j ai + a j bi + b j

bi
yi − B ≥ , i = 1,2,K , n (3)
2
ai a
≤ xi ≤ A − i , i = 1, 2, K, n (4)
2 2
bi
yi ≥ , i = 1,K, n (5)
2
The model is based on the fact that the rotation of rectangles is not allowed.
Conditions (2) do not allow rectangles to overlap. Conditions (3) stipulate that a rectangle
should be put either in bounding rectangle, or outside without overlapping the
boundaries. Conditions (4) and (5) stipulate that rectangles stay inside the strip of width
A and infinite length without overlapping the boundaries. Objective function (1) is such
that only rectangles put in bounding rectangle are considered.
It is easy to see that this is a non-linear mathematical model. All conditions are
at the most piecewise linear functions of variables xi and yi. The number of variables is
relatively small and is equal to 2n. The number of conditions equals

n ( n − 1) 1
+ 3n = ( n 2 + 5n ) (6)
2 2
The number of conditions is relatively large; however, this fact is not significant
in proposing a method for solving the problem.

3. m-M CALCULUS

The method of m-M calculus is presented in [23]. This paper will briefly present
its general ideas. Let the variables x1, x2, …, xn be from some segment
D = [a1 , b1 ] × [a2 , b2 ] × K × [an , bn ] ⊆ R n . Then let there a system of inequalities be given

f1 ( x1 , x2 ,K, xn ) ≥ 0, f 2 ( x1, x2 ,K , xn ) ≥ 0,K , f k ( x1 , x2 ,K , xn ) ≥ 0 (7)

where functions fi : D → R, i = 1, 2,K , k .

 A. Savić, T. Šukilović, V. Filipović / Solving 2-dim. Packing Problem 97

Now, for any Δ ⊆ D , m( f i )( Δ), M ( fi )( Δ) can be determined. By definitions

given in [23], segment ∆ is feasible if and only if

( ∀i )( M ( f i )(Δ) ≥ 0) (8)
Also, segment ∆ is solutional if and only if

( ∀i )( m( f i )(Δ ) ≥ 0) (9)
and finally, segment ∆ is indeterminate if and only if

( ∀i )( M ( f i )(Δ) ≥ 0), (∃i )( m( f i )(Δ ) < 0) (10)

Now, idea of solving with m-M calculus is that in a tree cell strategy starting set
D is decomposed in segments. In every iteration, all infeasible segments are rejected,
until either there are no more segments left, which means there is no solution; or there are
no more indeterminate segments, which means all that is left are solely solutional
segments. A union of solutional segments gives the solution set. As decomposition
progresses, iteration includes a better approximation of solution set. Finally, the results of
this procedure are all the solutions for the system of inequalities (7).
This last fact determined using this method for solving problems (1)-(5).
Based on the previously presented mathematical model for packing, we had to

solve a problem of constrained optimization on an n-dimensional segment.

For that purpose, first we had to write the appropriate m-M functions for
inequality formulas. m-function for objective function and inequalities (2) and (3) are:
n max ( 0, B − byi )
∑ B − ayi
aibi − AB (11)
i =1

⎛ bx j − axi ay j − byi ax j − bxi by j − ayi ⎞

1 − max ⎜ − , − ⎟−
⎜ ai + a j bi + b j ai + a j bi + b j ⎟⎠
⎝
(12)
⎛ bx j − axi ay j − byi bx j − axi ay j − byi ⎞
− max ⎜⎜ + ,− − ⎟
⎝ ai + a j bi + b j ai + a j bi + b j ⎟⎠

bi
− max ( byi − B , B − ayi ) . (13)
2
and respectively M-functions are
n max ( 0, B − ayi )
∑ B − byi
aibi − AB (14)
i =1
98 A. Savić, T. Šukilović, V. Filipović / Solving 2-dim. Packing Problem

⎛ bx j − axi ay j − byi ax j − bxi by j − ayi ⎞

1 − min ⎜⎜ − , − ⎟−
⎝ ai + a j bi + b j ai + a j bi + b j ⎟⎠
(15)
⎛ bx j − axi ay j − byi bx j − axi ay j − byi ⎞
− min ⎜ + ,− − ⎟
⎜ ai + a j bi + b j ai + a j bi + b j ⎟⎠
⎝
bi
− min ( byi − B , B − ayi ) (16)
2

where ayi and byi are lower-left and upper-right y-coordinate of ith rectangular,

respectively. Constraints (4) and (5) are simple inequalities, and their m and M functions
are the same as constraints.
Also, since we have been using dyadic tree cell-decomposition strategy, we had

to establish some feasibility criteria for the n-dimensional segment. The Cartesian

product of [X1,Y1] × [X2,Y2] × … × [Xn,Yn] n-cells is feasible if it satisfies two conditions:

(1) ( ∀i ∈{1,K , k}) m( f i )([ X1 , Y1 ] × [ X 2 , Y2 ] × K × [ X n , Yn ] ≥ 0

(2) (∀ i ∈ {1,K , k }) m( fi )([ X1 , Y1 ] × [ X 2 , Y2 ] ×K × [ X n , Yn ] < 0 ⇒

M ( fi )([ X1 , Y1 ] × [ X 2 , Y2 ] × K × [ X n , Yn ] ≥ 0

Therefore, our algorithm is simple: we start from a reshaped bounding box and
perform cell-decomposition until there are no more feasible cells, or some predefined
level of accuracy is accomplished.

4. EXPERIMENTAL RESULTS
All experiments were performed on a PC with P4 2G CPU and 960Mb RAM.
Experiments were performed on a small number of rectangles, to be more specific, 4 and
5. The bounding rectangle in all tests had dimensions 5 by 7.
In the first test we packed 4 rectangles with dimensions 3×4, 2×6, 3×2 and 5×1
into the 5×7 rectangle. There are 8 possible solutions shown in Figure 3, and our
proposed method was able to detect all of them. The results are presented in Table 1.
 A. Savić, T. Šukilović, V. Filipović / Solving 2-dim. Packing Problem 99

Figure 3 Packing 4 rectangles

Table 1 Centers’ coordinates of 4 packed rectangles for all possible solutions

rectangle
number 1 2 3 4
solution (yellow, 3×4) (blue, 2×6) (red, 3×2) (green, 5×1)
number
1 4.5, 2.0. 1.0, 3.0. 4.5, 5.0. 3.5, 6.5.
2 1.5, 2.0. 4.0, 3.0. 1.5, 5.0. 3.5, 6.5.
3 4.5, 4.0. 1.0, 3.0. 4.5, 1.0. 3.5, 6.5.
4 1.5, 4.0. 4.0, 3.0. 1.5, 1.0. 3.5, 6.5.
5 4.5, 3.0. 1.0, 4.0. 4.5, 6.0. 3.5, 0.5.
6 1.5, 3.0. 4.0, 4.0. 1.5, 6.0. 3.5, 0.5.
7 4.5, 5.0. 1.0, 4.0. 4.5, 2.0. 3.5, 0.5.
8 1.5, 5.0. 4.0, 4.0. 1.5, 2.0. 3.5, 0.5.
Total processing time was 61 second, while the memory cost was 398Mb.
As shown in Figure 3, most solutions were obtained from one solution using
symmetric properties. Because of this, it was necessary to make the method modification
so that the stopping criterion would be finding the first optimal solution.
Results of the method modified with this criterion performed on two sets of
rectangles, first with 4 and second with 5 rectangles can be seen on Figure 4. The
depicted solutions are the first found solutions in both cases.

Figure 4 Left: first optimal solution for packing 4 rectangles; right: first optimal solution
for packing 5 rectangles (the gap between orange and green rectangle is unoccupied
space)

Results obtained by finding the first optimal solution are presented in Table 2.
100 A. Savić, T. Šukilović, V. Filipović / Solving 2-dim. Packing Problem

Table 2 Results of packing with modified stopping criterion

Test 1 Test 2
Number of rectangles 4 5
Dimensions of rectangles 2×6 1×7
3×2 2×2
3×4 2×2
5×1 2×5
4×2
Time for detecting first optimal solution 2 69
(s)
Memory cost (MB) 61 681
Detected centers 1.0, 3.0. 4.5, 3.5.
4.5, 5.0. 3.0, 4.5.
4.5, 2.0. 3.0, 6.0.
3.5, 6.5. 1.0, 4.5.
2.0, 1.0.

As shown in Table 2, memory requirements are greatly increasing with the

number of small rectangles. Because of this, each cell was represented by only three
pieces of information. Each cell consists of two pointers to the lower left and upper right
end of the cell and id of the cell – indicator to efficiently determine whether the solution
has been detected, and from which cell the decomposition solving process should
continue. Endpoints of a cell were structured as two n-dimensional arrays of floats and id
is the integer type. Although the process of cutting-off non-feasible cells is very fast,
number of solutional segments grows exponentially, and so does the memory cost. Build-
up of the number of solutional segments is in one moment drastically cut-off, and after
that point in process the number of feasible segments plummets. If procedure reaches this
point, all optimal solutions are quickly found afterwards. Otherwise, memory
requirements stopped the execution of the algorithm.

5. CONCLUSION

This paper presented a method for finding all optimal solutions for the two
dimensional rectangle packing problem. The method applied m-M calculus and dyadic
decomposition of starting bounding rectangle. Performance was relatively good for small
instances, but for greater instances, the method needs much memory. The modification
for finding the first optimal solution was presented as well.
Based on the presented results, we can conclude that the method is applicable
for finding all solutions, but has drawbacks. The most promising direction would be in
the better optimization of the decomposition process, which can lead to improvement in
method performance, especially for greater instances.
 A. Savić, T. Šukilović, V. Filipović / Solving 2-dim. Packing Problem 101

REFERENCES

[1] Baldacci, R., and Boschetti, M.A., “A cutting-plane approach for the two-dimensional
orthogonal non-guillotine cutting problem”, European Journal of Operational Research, 183
(3) (2007) 1230-1248.
[2] Beasley, J.E., “An exact two-dimensional non-guillotine cutting tree search procedure”,
Operations Research, 33 (1985) 49–64.
[3] Binkley, K.J., and Hagiwara, M., “Applying self-adaptive evolutionary algorithms to two-
dimensional packing problems using a four corners' heuristic”, European Journal of
Operational Research, 183 (3) (2007) 1230-1248.
[4] Christofides, N., and Hadjiconstantinou, E., “An exact algorithm for orthogonal 2-D cutting
problems using guillotine cuts”, European Journal of Operational Research, 83 (1995) 21–38.
[5] Cintra, G.F., Miyazawa, F.K., Wakabayashi, Y., and Xavier, E.C., “Algorithms for two-
dimensional cutting stock and strip packing problems using dynamic programming and
column generation”, European Journal of Operational Research, 191 (1) (2008) 59-83.
[6] Coffman, E.G., Jr., and Lueker, G.S., Probabilistic Analysis of Packing and Partitioning
Algorithms, Wiley, Chichester, 1992.
[7] Coffman, E.G., Jr., and Shor, P.W., “Average-case analysis of cutting and packing in two
dimensions”, European Journal of Operational Research, 44 (1990) 134–144.
[8] Correa, J.R., “Near-optimal solutions to two-dimensional bin packing with 90 degree
rotations”, Electronic Notes in Discrete Mathematics, 18 (2004) 89-95.
[9] Diegert, C., “Practical automatic placement for standard-cell integrated circuit”, American
Journal of Mathematical and Management Sciences, 8 (34) (1988) 309–328.
[10] Dowsland, K.A., and Dowsland, W.B., “Packing problems”, European Journal of Operational
Research, 56 (1992) 2–14.
[11] Dyckhoff, H., Finke, U., Cutting and Packing in Production and Distribution, Physica Verlag,
Heidelberg, 1992.
[12] Dyckhoff, H., Scheithauer, G., and Terno J., “Cutting and packing (C&P)”, in: M.,
Dell’Amico, F., Maffioli, S., Martello, (eds.), Annotated Bibliographies in Combinatorial
Optimization, Wiley, Chichester, 1997, 393–413.
[13] Epstein, L., Levin, A., and Van Stee, R., “Two-dimensional packing with conflicts”, Acta
Informatica, 45 (3) (2008) 155-175.
[14] Gonzalves, J.F., and Resende, M.G.C., “A parallel multi-population genetic algorithm for a
constrained two-dimensional orthogonal packing problem”, Journal of Combinatorial
Optimization, (2010) 1-22 Article in Press.
[15] Hadjiconstantinou, E., and Christofides N., “An optimal algorithm for general orthogonal 2-D
cutting problems”, Technical report MS-91/2, Imperial College, London, UK.
[16] Harwig, J.M., Barnes, J.W., and Moore, J.T., “An adaptive tabu search approach for 2-
dimensional orthogonal packing problems”, Military Operations Research, 11 (2) (2006) 5-
26.
[17] Hochbaum, D.S., and Maass, W., “Approximation schemes for covering and packing
problems in image processing and VLSI”, Journal of the Association for Computing
Machinery, 32 (1) (1985) 130–136.
[18] Kim, J., Moon, B.-R., “A hybrid genetic algorithm for a variant of two-dimensional packing
problem”, Proceedings of the 11th Annual Genetic and Evolutionary Computation
Conference, GECCO-2009, 287-292.
[19] Leung J., Tam, T., Wong, C.S., Young, G., and Chin, F., “Packing squares into square”,
Journal of Parallel and Distributed Computing, 10 (1990) 271–275.
[20] Lodi, A., Martello, S., and Monaci, M., “Two-dimensional packing problems: A survey”,
European Journal of Operational Research, 141 (2002) 241–252.
102 A. Savić, T. Šukilović, V. Filipović / Solving 2-dim. Packing Problem

[21] Lu, S., Lu, Y., Zha, J., and Lu, S., “Cross-entropy method for two-dimensional packing
problem with rectangle pieces”, Beijing Jiaotong Daxue Xuebao/Journal of Beijing Jiaotong
University, 33 (2) (2009) 39-43.
[22] Ortmann, F.G., Ntene, N., van and Vuuren, J.H., “New and improved level heuristics for the
rectangular strip packing and variable-sized bin packing problems”, European Journal of
Operational Research, 203 (2) (2010) 306-315.
[23] Prešić S.B., “A survey of the m-M calculus”, Yugoslav Journal of Operations Research, 8 (1)
(1998) 137-168.
[24] Savić A., “Packing and cutting problems“, Master thesis, Faculty of Mathematics, University
of Belgrade, 2000.
[25] Toro, E.M., Garcs, A., and Ruiz, H., “Two dimensional packing problem using a hybrid
constructive algorithm of variable neighborhood search and simulated annealing”, Revista
Facultad de Ingenieria, 46 (2008) 119-131.
[26] Wu, Y.L., Huang, W., Lau, S., Wong, C.K., and Young, G.H., “An effective quasi-human
based heuristic for solving the rectangle packing problem”, European Journal of Operational
Research, 141 (2002) 341–358.