VIETNAM NATIONAL UNIVERSITY, HO CHI MINH CITY

UNIVERSITY OF TECHNOLOGY
FACULTY OF COMPUTER SCIENCE AND ENGINEERING

MATHEMATICAL MODELING (CO2011)

Assignment

“Cutting stock problem”

Instructor(s): Mai Xuân Toàn

Students: Lê Anh Quân - 2352992 (Group CC04 - Team 91)

Nguyễn Trung Kiên - 2252396 (Group CC04 - Team 91)
Nguyễn Bùi Huỳnh Khương - 2352637 (Group CC04 - Team 91, Leader)

HO CHI MINH CITY, DECEMBER 2024

 University of Technology, Ho Chi Minh City
Faculty of Computer Science and Engineering

Member list & Workload

No. Fullname Student ID Problems % done

- Algorithm + Case study
1 Lê Anh Quân 2352992 - Question 2: 100%
- Question 3:
- Code + Analyze Results
2 Nguyễn Trung Kiên 2252396 100%
- Question 4 + 5:
- Intro + Acknowledge
3 Nguyễn Bùi Huỳnh Khương 2352637 100%
- Question 1:

Table 1: Member list & workload

Assignment for Mathematical Modeling - Academic year 2022 - 2023 Page 1/14
 University of Technology, Ho Chi Minh City
Faculty of Computer Science and Engineering

Contents
Member list & Workload 1

1 Introduction to Cutting-Stock Problem 3

1.1 Problem Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Computational Complexity and Classification . . . . . . . . . . . . . . . . . . . . 3
1.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Acknowledge 4
2.1 Cutting-Stock Problem (CSP) Overview . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 CSP Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Mathematical Formulation of CSP . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.4 Integer Linear Programming (ILP) for Cutting-Stock Problem . . . . . . . . . . . 5

3 Algorithm 5
3.1 Genetic Algorithm (GA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1.2 Advantages and Disadvantages for 2D Cutting Problems . . . . . . . . . . 6
3.2 Column Generation Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2.2 Advantages and Disadvantages for 2D Cutting Problems . . . . . . . . . . 7
3.3 Greedy Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

4 Modeling 7
4.1 Problem Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.2 Integer Linear Programming (ILP) Model . . . . . . . . . . . . . . . . . . . . . . 8
4.2.1 Decision Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.2.2 Objective Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.2.3 Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

5 Case Study 9
5.1 Real-World Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.2 Model and Solution Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.2.1 Problem Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.2.2 Algorithm Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

6 Analyze Results 10
6.1 Metrics for Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
6.2 Results from Best-Fit Algorithm and Genetic Algorithm . . . . . . . . . . . . . . 10
6.2.1 Best-Fit Algorithm Results . . . . . . . . . . . . . . . . . . . . . . . . . . 10
6.2.2 Genetic Algorithm Results . . . . . . . . . . . . . . . . . . . . . . . . . . 11
6.3 Comparison of Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
6.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

7 Conclusion 12
7.1 Achievements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
7.2 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
7.3 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Assignment for Mathematical Modeling - Academic year 2022 - 2023 Page 2/14
 University of Technology, Ho Chi Minh City
Faculty of Computer Science and Engineering

1 Introduction to Cutting-Stock Problem

The Cutting-Stock Problem (CSP) is a classic optimization challenge in operations research, fo-
cused on cutting standard-sized pieces of stock material (such as paper rolls, sheet metal, pipes,
or steel bars) into smaller, required sizes while minimizing material waste. This problem is of
significant importance in industries such as paper production, metalworking, textile manufactur-
ing, and packaging, where raw materials are produced in fixed sizes, and these materials must
be efficiently cut to meet specific production demands or customer requirements.
The goal of this exercise is to explore methods for solving the Cutting-Stock Problem, with a
focus on **(state the focus of your exercise, e.g., minimizing waste, using specific algorithms)**.
Specifically, we aim to apply optimization techniques such as **Linear Programming** and
**Heuristic Methods** to evaluate cutting strategies that minimize waste and improve material
utilization in industrial settings.

1.1 Problem Description

The basic idea of the Cutting-Stock Problem is simple: Given a set of raw materials or stock
pieces of fixed dimensions (such as large rolls of paper or sheets of metal), the objective is to
determine how to cut these stock materials into smaller pieces, meeting specified demands, while
minimizing the leftover or wasted material. This problem is typically formulated as an integer
linear programming (ILP) problem, where the objective is to minimize the waste, and the decision
variables represent the number of cuts made from each stock piece. The constraints ensure that
all demand pieces are satisfied.

1.2 Computational Complexity and Classification

From a computational perspective, the Cutting-Stock Problem is classified as **NP-hard**.
This means that finding an optimal solution becomes computationally difficult as the problem
size increases. The CSP can also be reduced to the **knapsack problem**, another NP-hard
problem. Despite its complexity, various solution techniques, such as dynamic programming,
column generation, and heuristic algorithms, have been developed to handle practical instances
of the problem.
The problem can be classified based on the **dimensionality of the cuts**:
• One-dimensional (1D): Stock materials are linear, such as pipes or cables, which need
to be cut into smaller lengths. This is common in industries like steel manufacturing and
paper production.
• Two-dimensional (2D): Stock materials are flat and rectangular, such as in the furniture,
textile, or glass production industries, where large stock sheets need to be cut into smaller
pieces of varying sizes.

• Three-dimensional (3D): Cutting problems involve three-dimensional materials (e.g.,

cutting large blocks into smaller pieces), closely related to the **3D packing problem**,
which focuses on packing objects into containers efficiently.

1.3 Conclusion
The Cutting-Stock Problem is a critical optimization challenge in industries that rely on cutting
raw materials into smaller pieces. Solving this problem efficiently allows companies to reduce

Assignment for Mathematical Modeling - Academic year 2022 - 2023 Page 3/14
 University of Technology, Ho Chi Minh City
Faculty of Computer Science and Engineering

waste, lower costs, and improve operational efficiency. Given the NP-hard nature of CSP, devel-
oping efficient algorithms and heuristics is essential for solving large-scale practical problems in
operations research and industrial engineering.

2 Acknowledge
2.1 Cutting-Stock Problem (CSP) Overview
The **Cutting-Stock Problem (CSP)** is a classical optimization problem in operations research.
It involves cutting standard-sized pieces of raw material (stock material), such as rolls of paper,
sheet metal, or steel bars, into smaller pieces of specified sizes while minimizing the waste mate-
rial. This problem arises in many industrial applications, including the paper, textile, and metal
industries, where raw materials are produced in standard sizes, and cutting them into smaller
pieces is necessary to meet specific production or customer demands.

2.2 CSP Types

The Cutting-Stock Problem can be classified based on the dimensionality of the cuts:
• One-Dimensional (1D): Involves cutting long materials (like pipes or metal rods) into
smaller lengths.
• Two-Dimensional (2D): Cutting flat materials like metal sheets or paper into smaller
rectangular pieces.
• Three-Dimensional (3D): Although less common, it involves cutting materials into
three-dimensional shapes (e.g., cutting large blocks of metal into smaller parts).

2.3 Mathematical Formulation of CSP

The Cutting-Stock Problem can be formulated as a Linear Programming or Integer Linear
Programming (ILP) problem. Below is the general formulation:
Objective: Minimize the waste material (leftover stock) after cuts have been made.
n
X m
X
Minimize W = x i li − dj wj
i=1 j=1

Where:
• W is the total waste.
• xi represents the number of cuts made from stock piece i.
• li is the length or size of stock piece i.
• dj is the number of pieces required for demand j.
• wj is the size of the demand piece j.
Constraints:
n
X
aij xi ≥ dj
i=1
Where:

Assignment for Mathematical Modeling - Academic year 2022 - 2023 Page 4/14
 University of Technology, Ho Chi Minh City
Faculty of Computer Science and Engineering

• aij is the number of pieces of demand j that can be obtained from stock piece i.
• dj is the demand for size j, ensuring that the required pieces are met.

2.4 Integer Linear Programming (ILP) for Cutting-Stock Problem

Integer Linear Programming (ILP) is a method for solving optimization problems where the
decision variables are restricted to integer values (e.g., the number of cuts). ILP is particularly
useful in the Cutting-Stock Problem because the decision variables represent the number of cuts
or the number of times a stock piece is used, which cannot be fractional.
ILP Formulation: The Cutting-Stock Problem can be solved using ILP, where:

• Decision Variables: xi represents the number of times stock piece i is cut, and xi must
be an integer.
• Objective Function: Minimize the waste of material.
• Constraints: Ensure that all demands are met, and the cuts do not exceed the stock
dimensions.

Mathematically, the ILP model for CSP can be formulated as:

n
X
Minimize xi (minimizing waste)
i=1

Subject to:
n
X
aij xi ≥ dj (satisfy demand for each piece size)
i=1

Where:
• aij represents the number of pieces of size j that can be obtained from stock piece i.
• xi represents the number of cuts made from stock piece i.

3 Algorithm
3.1 Genetic Algorithm (GA)
3.1.1 Overview
The Genetic Algorithm (GA) is an optimization technique inspired by the process of biological
evolution. Each feasible solution to a problem is treated as an individual in a population. The
algorithm evolves the population through three main mechanisms: natural selection, crossover,
and mutation, as detailed below:

• Encoding: Solutions (e.g., cutting patterns) are encoded as binary strings or other suitable
formats. For example, in a 2D cutting problem, an individual might be represented as a
matrix depicting the arrangement of pieces on a large sheet of material.
• Population Initialization: A population of individuals is generated either randomly or
based on some initial feasible solutions.

Assignment for Mathematical Modeling - Academic year 2022 - 2023 Page 5/14
 University of Technology, Ho Chi Minh City
Faculty of Computer Science and Engineering

• Fitness Evaluation: Each individual is assessed using a fitness function. For the material
cutting problem, the fitness function can be defined as:
Total area of pieces cut
Fitness =
Area of the large sheet
The goal is to maximize material usage efficiency.
• Selection: The best-performing individuals (with higher fitness values) are selected as
parents to produce the next generation. Common selection methods include roulette wheel
and tournament selection.
• Crossover: Combines information from two parents to produce new offspring. For example,
this might involve merging cutting patterns from two different individuals.
• Mutation: Randomly alters some genes of an individual to increase diversity. For example,
changing the position of a piece within the cutting pattern.
• Iteration: Selection, crossover, and mutation are repeated over multiple generations until
a good solution is found or a maximum number of iterations is reached.

3.1.2 Advantages and Disadvantages for 2D Cutting Problems

Advantages:
• Flexibility and Versatility: GA can handle various problem variations (e.g., different
piece sizes and shapes).
• Capability to Explore Large Solution Spaces: GA can identify near-optimal solutions
in complex, large search spaces.
• Avoidance of Local Optima: Mutation helps escape local minima.
Disadvantages:
• Computationally Expensive: Due to the large population size and high iteration count.
• Parameter Sensitivity: Performance depends on parameters such as population size,
crossover rate, and mutation rate.
• No Global Optimality Guarantee: May not converge to the best solution if the popu-
lation lacks diversity.

3.2 Column Generation Algorithm

3.2.1 Overview
Column generation is an efficient optimization technique, especially suitable for problems with
very large solution spaces. The 2D cutting stock problem is a prime example, with millions of
potential cutting patterns.
Core Idea: Instead of considering all possible cutting patterns from the beginning, the
algorithm works with a smaller subset of patterns and iteratively adds more as needed.
Steps:
• Master Problem: Optimizes based on the current subset of patterns.

Assignment for Mathematical Modeling - Academic year 2022 - 2023 Page 6/14
 University of Technology, Ho Chi Minh City
Faculty of Computer Science and Engineering

• Pricing Problem: Identifies additional cutting patterns (columns) that can improve the
solution.

Procedure:
• Initialization: Start with a basic set of patterns, e.g., cutting one piece per sheet.
• Solve Master Problem: Find the optimal solution based on the current pattern set.
• Solve Pricing Problem: Identify a new pattern that improves the objective function
(e.g., reducing waste).
• Update: Add the new pattern to the master problem and repeat until no improving pattern
can be found.

3.2.2 Advantages and Disadvantages for 2D Cutting Problems

Advantages:

• Computational Efficiency: Significantly reduces the number of patterns to consider.

• Scalability: Handles large-scale problems with millions of potential patterns.
• Linear Optimization: Focuses on finding optimal solutions using linear programming.
Disadvantages:

• Complex Pricing Problem: Requires designing an efficient algorithm to generate new

patterns.
• Dependency on Initialization: Poor initial patterns can lead to suboptimal solutions.
• Integer Solutions Not Guaranteed: The master problem often yields fractional solu-
tions, requiring rounding to ensure integer feasibility.

3.3 Greedy Algorithm

The greedy algorithm was used to minimize the waste during the cutting process. It selects the
best immediate solution at each step.

4 Modeling
4.1 Problem Description
The 2D cutting stock problem involves optimizing the arrangement of smaller pieces on a large
material sheet to maximize resource utilization, minimize waste, and fulfill demand for each
piece.
Inputs:

• W, H: Dimensions of the large sheet (width and height).

• wi , hi , di :
– wi : Width of piece i.

Assignment for Mathematical Modeling - Academic year 2022 - 2023 Page 7/14
 University of Technology, Ho Chi Minh City
Faculty of Computer Science and Engineering

– hi : Height of piece i.
– di : Number of pieces i required.
• A set of feasible cutting patterns j (manually created or generated automatically).
Outputs:
• A list of cutting patterns j, each detailing the positions of pieces cut from the large sheet.
• The number of times each pattern j is used.
Objective: Optimize material usage by minimizing the number of large sheets used and
reducing waste area.
Assumptions:
• Pieces do not overlap.
• Pieces cannot be rotated (fixed width and height).

4.2 Integer Linear Programming (ILP) Model

4.2.1 Decision Variables
xij : Number of pieces i cut from pattern j
yj : Number of times pattern j is used

4.2.2 Objective Function

Minimize the number of large sheets used, equivalent to minimizing the total number of pattern
uses:
X
Minimize Z = yj
j

4.2.3 Constraints
• Demand Satisfaction: Ensure the total number of pieces i cut across all patterns j meets
or exceeds the demand di : X
xij ≥ di ∀i
j

• Pattern Size Limitation: Ensure the total area of pieces in each pattern j does not
exceed the large sheet’s area:
X
wi · hi · xij ≤ W · H ∀j
i

• Linking Variables: If pattern j is used (yj > 0), the total number of pieces cut from this
pattern must not exceed the maximum capacity of the pattern:
xij ≤ Cij · yj ∀i, j
where Cij is the maximum number of pieces i that can fit in pattern j.
• Integer Variables:
xij ∈ Z+ , yj ∈ Z+ ∀i, j

Assignment for Mathematical Modeling - Academic year 2022 - 2023 Page 8/14
 University of Technology, Ho Chi Minh City
Faculty of Computer Science and Engineering

5 Case Study
5.1 Real-World Scenario
In furniture manufacturing, companies often need to cut large wooden sheets (e.g., 240 cm x 120
cm) into smaller pieces for assembling products like tables, chairs, and cabinets. For instance:
• 60 cm x 40 cm: Demand d1 = 50

• 80 cm x 60 cm: Demand d2 = 30
• 120 cm x 60 cm: Demand d3 = 20
In this scenario, the goal is to efficiently arrange the smaller pieces on the large sheets while
minimizing waste and fulfilling the demand for each piece size.

5.2 Model and Solution Approach

5.2.1 Problem Modeling
The problem is modeled using the ILP formulation described above, with the dimensions W, H
as inputs.

5.2.2 Algorithm Application

Method 1: Genetic Algorithm (GA)

• Encoding: Each individual represents a way to cut pieces from the sheet.

• Fitness Function: Evaluate solutions based on material usage and waste reduction.
• Execution: Evolve solutions over generations to find the most efficient cutting pattern.

Method 2: Column Generation

• Initial Patterns: Start with a few simple cutting patterns.

• Iterative Process: Solve the master problem to find the best combination of patterns.
Generate new patterns by solving a subproblem until no improvement is possible.

5.3 Results and Discussion

• Material Efficiency: Minimizes waste and reduces the number of large sheets used.
• Cost Reduction: Reduces material procurement costs and improves production efficiency.
• Automation Benefits: Automates the planning process, ensuring accurate cuts and effi-
cient layouts.

Conclusion: Applying algorithms like Genetic Algorithm and Column Generation in the two-
dimensional cutting-stock problem improves efficiency, reduces waste, and optimizes production
processes in real-world applications.

Assignment for Mathematical Modeling - Academic year 2022 - 2023 Page 9/14
 University of Technology, Ho Chi Minh City
Faculty of Computer Science and Engineering

6 Analyze Results
6.1 Metrics for Evaluation
To evaluate the performance of the Best-Fit Algorithm and Genetic Algorithm, we used
the following key metrics:

• Material Utilization: This metric measures how efficiently the stock material is used by
the algorithm. The formula is:
Total Area of Pieces Cut
Utilization Rate = × 100
Total Area of Stock Used
A higher utilization rate signifies better material efficiency.
• Waste Percentage: The waste percentage is the complement of the material utilization
rate and shows the unused portion of the stock material.

Waste Percentage = 100 − Utilization Rate

Lower waste percentage indicates better optimization of material usage.

• Execution Time: This metric tracks the time taken by each algorithm to generate a
solution. Execution time is a critical factor in real-time or large-scale applications.

• Memory Usage: Measures the memory consumed by the algorithm during its execution.
It’s important to consider, especially when dealing with large datasets or in environments
with memory constraints.

6.2 Results from Best-Fit Algorithm and Genetic Algorithm

6.2.1 Best-Fit Algorithm Results
• Material Utilization: The Best-Fit Algorithm is effective for small to medium-sized
problems. It provides reasonable material utilization with relatively low waste for small
datasets. However, as the problem size increases, the algorithm struggles to maintain high
utilization, and the efficiency drops.
• Waste Percentage: Best-Fit generally leads to higher waste percentages when dealing
with large problems. The algorithm fails to find the optimal arrangement for all pieces
as the complexity grows, resulting in leftover material that could have been used more
effectively.
• Execution Time and Memory Usage: Best-Fit is very fast for small problems, with
minimal memory usage. It requires less computational power and memory, making it suit-
able for real-time applications with moderate data sizes.
• Scalability: As the dataset size increases (i.e., more pieces or larger stock sheets), Best-
Fit’s performance declines. The algorithm struggles to efficiently place all the pieces, and
the execution time increases. The solution quality also decreases, particularly in terms of
material utilization.

Assignment for Mathematical Modeling - Academic year 2022 - 2023 Page 10/14
 University of Technology, Ho Chi Minh City
Faculty of Computer Science and Engineering

6.2.2 Genetic Algorithm Results

• Material Utilization: The Genetic Algorithm (GA) significantly improves material uti-
lization, especially for larger and more complex problems. By evolving multiple generations
of solutions through crossover and mutation, GA explores a wider solution space and iden-
tifies optimal cutting patterns that reduce waste.
• Waste Percentage: The waste percentage is considerably lower in GA than in Best-Fit.
The algorithm’s ability to explore multiple combinations and solutions helps reduce waste,
making it more efficient in utilizing stock material.
• Execution Time and Memory Usage: GA requires more computation time and memory
than Best-Fit. The need to maintain and evaluate a population of solutions, as well as the
iterative nature of the algorithm, means that GA is more resource-intensive. However, the
trade-off is higher solution quality.
• Scalability: GA is highly scalable and performs well with large datasets. It can handle
more pieces and larger stock sheets, but the computational resources required increase as
the problem size grows. The execution time also scales exponentially with the number of
generations and population size.

6.3 Comparison of Results

Metric Best-Fit Algorithm Genetic Algorithm

Material Utilization Moderate (good for small problems) High (especially for large problems)
Waste Percentage Higher waste in larger problems Lower waste, more optimized
Execution Time Fast for small problems Slower, especially for large problems
Memory Usage Low High
Scalability Struggles with large problems Better scalability, slower runtime

Table 2: Comparison of Best-Fit Algorithm and Genetic Algorithm

6.4 Conclusion
Based on the analysis of the Best-Fit Algorithm and Genetic Algorithm:

• Best-Fit Algorithm is suitable for small to medium-sized problems where execution

speed is critical, and the solution quality is acceptable. It works well with fewer pieces or
less complex configurations. However, it is not the best choice for large problems, as the
waste percentage increases and material utilization decreases with problem size.
• Genetic Algorithm is ideal for larger, more complex problems where the goal is to mini-
mize material waste and maximize stock utilization. While it requires more computational
resources and takes longer to execute, GA provides better results in terms of material opti-
mization and waste reduction. It is more scalable and can handle large datasets effectively,
but at the cost of increased execution time and memory usage.

Thus, for real-world applications involving larger and more complex cutting problems, Ge-
netic Algorithm should be preferred despite its higher computational cost, as it offers better
overall performance in terms of material usage and waste minimization. On the other hand,

Assignment for Mathematical Modeling - Academic year 2022 - 2023 Page 11/14
 University of Technology, Ho Chi Minh City
Faculty of Computer Science and Engineering

Best-Fit Algorithm is more appropriate for simpler, smaller problems where speed is more
important than perfect optimization.

7 Conclusion
The Cutting-Stock Problem (CSP) is a critical optimization challenge faced by various
industries, requiring effective methods to minimize material waste and optimize material utiliza-
tion. In this assignment, we explored two key algorithms—Best-Fit Algorithm and Genetic
Algorithm (GA)—to solve the problem.

7.1 Achievements
Through this study, we successfully:

• Compared the Best-Fit Algorithm and Genetic Algorithm in terms of their ability to
optimize material usage, minimize waste, and scale with larger datasets.
• Demonstrated that the Best-Fit Algorithm is highly efficient for small to medium-sized
problems where speed and simplicity are important.

• Showed that the Genetic Algorithm provides superior results for larger and more complex
problems, significantly reducing waste and optimizing material utilization.
• Applied these algorithms to a real-world case study in the furniture manufacturing industry,
validating their practical use in improving operational efficiency and reducing costs.
• Integrated optimization techniques like Integer Linear Programming (ILP) and heuris-
tic methods to automate the decision-making process in industrial applications, highlight-
ing their potential for real-world use.

7.2 Limitations
While the study provided valuable insights, there are several limitations:
• Best-Fit Algorithm struggles with larger, more complex problems, as its performance
deteriorates, leading to higher waste and inefficient material usage.
• Genetic Algorithm (GA), while effective, requires significant computational resources
and time, especially with larger populations and multiple generations.
• The models focused primarily on two algorithms and did not explore other potential meth-
ods such as Column Generation or Dynamic Programming, which could offer further
optimization opportunities.

• Real-world complexities, such as piece rotation or machine constraints, were not fully con-
sidered in the models, which could affect the feasibility and accuracy of the results in
certain industrial settings.

Assignment for Mathematical Modeling - Academic year 2022 - 2023 Page 12/14
 University of Technology, Ho Chi Minh City
Faculty of Computer Science and Engineering

7.3 Future Work

To build upon this research, future work could focus on:
• Hybrid Algorithms: Exploring hybrid approaches that combine the strengths of heuristic
methods like Best-Fit with exact methods such as ILP or Column Generation to improve
both efficiency and solution quality.

• Parameter Tuning for GA: Improving the performance of the Genetic Algorithm
by refining parameters such as population size, mutation rate, and crossover methods to
balance execution time and solution quality.
• Incorporating Additional Constraints: Introducing more real-world constraints, such
as piece rotation, irregular shapes, or dynamic demand, to make the model more robust
and applicable across different industries.
• Case Studies Across Industries: Evaluating the algorithms in diverse real-world sce-
narios, such as textile manufacturing or metalworking, to validate their performance and
adaptability in various contexts.

• Real-Time Optimization: Developing real-time optimization tools for the CSP, inte-
grating both heuristic and exact optimization methods, to facilitate automated, efficient
decision-making in operational settings.
Overall, this assignment demonstrates the importance of combining computational techniques
and mathematical modeling to address complex, real-world challenges like the Cutting-Stock
Problem. The findings lay the foundation for the continued development of advanced solutions
that can significantly improve material utilization, reduce waste, and optimize production pro-
cesses in diverse industrial applications.

Assignment for Mathematical Modeling - Academic year 2022 - 2023 Page 13/14
 University of Technology, Ho Chi Minh City
Faculty of Computer Science and Engineering

References
[1] Holland, J. H. (1975). Adaptation in Natural and Artificial Systems. University of Michigan
Press.
[2] Goldberg, D. E. (1989). Genetic Algorithms in Search, Optimization, and Machine Learning.
Addison-Wesley.
[3] Mitchell, M. (1998). An Introduction to Genetic Algorithms. MIT Press.
[4] Eiben, A. E., & Smith, J. E. (2003). Introduction to Evolutionary Computing. Springer.
[5] Whitley, D. (2010). A genetic algorithm tutorial. Statistics and Computing, 4(2), 65-85.
[6] Banzhaf, W., Nordin, P., Keller, R. E., & Francone, F. D. (1998). Genetic Programming:
An Introduction. Morgan Kaufmann.
[7] Michalewicz, Z. (1996). Genetic Algorithms + Data Structures = Evolution Programs.
Springer.
[8] Vose, M. D. (1999). The Simple Genetic Algorithm: Foundations and Theory. MIT Press.
[9] Rechenberg, I. (1973). Evolutionsstrategie: Optimierung technischer Systeme nach Prinzip-
ien der biologischen Evolution. Frommann-Holzboog.
[10] Schwefel, H.-P. (1995). Evolution and Optimum Seeking. Wiley-Interscience.
[11] Barnhart, C., Johnson, E. L., Nemhauser, G. L., Savelsbergh, M. W. P., & Vance, P. H.
(1998). Branch-and-price: Column generation for solving huge integer programs. Operations
Research, 46(3), 316-329.
[12] Desaulniers, G., Desrosiers, J., & Solomon, M. M. (Eds.). (2005). Column Generation.
Springer.
[13] Lübbecke, M. E., & Desrosiers, J. (2005). Selected topics in column generation. Operations
Research, 53(6), 1007-1023.
[14] Gilmore, P. C., & Gomory, R. E. (1961). A linear programming approach to the cutting-stock
problem. Operations Research, 9(6), 849-859.
[15] Gilmore, P. C., & Gomory, R. E. (1963). A linear programming approach to the cutting-stock
problem—Part II. Operations Research, 11(6), 863-888.
[16] Vanderbeck, F. (1999). Computational study of a column generation algorithm for bin pack-
ing and cutting stock problems. Mathematical Programming, 86(3), 565-594.
[17] Coffman, E. G., Garey, M. R., Johnson, D. S., & Tarjan, R. E. (1984). Performance bounds
for level-oriented two-dimensional packing algorithms. SIAM Journal on Computing, 9(4),
808-826.
[18] Johnson, D. S. (1973). Near-optimal bin packing algorithms. PhD Thesis, Massachusetts
Institute of Technology.
[19] Garey, M. R., & Johnson, D. S. (1979). Computers and Intractability: A Guide to the Theory
of NP-Completeness. W. H. Freeman.
[20] Burke, E. K., Kendall, G., & Whitwell, G. (2004). A new placement heuristic for the or-
thogonal stock-cutting problem. Operations Research, 52(4), 655-671.

Assignment for Mathematical Modeling - Academic year 2022 - 2023 Page 14/14