Balancing Pareto Front exploration of Non-dominated Tournament Genetic Algorithm (B-NTGA) in solving multi-objective NP-hard problems with constraints

Michał Antkiewicz,Paweł B. Myszkowski

DOI: https://doi.org/10.1016/j.ins.2024.120400

2024-10-14

Abstract:The paper presents a new balanced selection operator applied to the proposed Balanced Non-dominated Tournament Genetic Algorithm (B-NTGA) that actively uses archive to solve multi- and many-objective NP-hard combinatorial optimization problems with constraints. The primary motivation is to make B-NTGA more efficient in exploring Pareto Front Approximation (PFa), focusing on 'gaps' and reducing some PFa regions' sampling too frequently. Such a balancing mechanism allows B-NTGA to be more adaptive and focus on less explored PFa regions. The proposed B-NTGA is investigated on two benchmark multi- and many-objective optimization real-world problems, like Thief Traveling Problem and Multi-Skill Resource-Constrained Project Scheduling Problem. The results of experiments show that B-NTGA has a higher efficiency and better performance than state-of-the-art methods.

Neural and Evolutionary Computing

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the inefficiency and imbalance in exploring the Pareto Front (PF) in multi - objective NP - hard combinatorial optimization problems. Specifically, the author proposes a new balanced selection operator applied to the improved Balanced Non - dominated Tournament Genetic Algorithm (B - NTGA) to more effectively explore the Pareto Front approximation (PFa) of multi - objective and multi - objective NP - hard combinatorial optimization problems with constraints. The following is the main problem description in this paper:

### 1. **Problem Background**
Multi - objective and multi - objective NP - hard combinatorial optimization problems (such as the Travelling Thief Problem (TTP) and the Multi - Skill Resource - Constrained Project Scheduling Problem (MS - RCPSP)) have the following characteristics:
- **NP - hardness**: The solution spaces of these problems are very large and cannot be fully searched by traditional algorithms.
- **Multi - objectivity**: Multiple conflicting objectives need to be optimized simultaneously.
- **Constraints**: Solutions must satisfy a series of constraints, otherwise they are infeasible solutions.

### 2. **Deficiencies of Existing Methods**
Existing multi - objective optimization algorithms have the following problems when exploring the Pareto Front:
- **Uneven exploration**: Some areas are over - sampled while others are ignored, resulting in uneven coverage of the Pareto Front.
- **Lack of adaptability**: They cannot dynamically adjust the balance between exploration and exploitation according to the current search state.
- **Inefficiency**: Some algorithms are not efficient in handling complex constraints, resulting in a waste of computing resources.

### 3. **Solutions Proposed in the Paper**
The author proposes B - NTGA, aiming to solve the above problems. The main improvements of B - NTGA include:
- **Balanced selection mechanism**: A selection mechanism based on Upper Confidence Bound (UCB) is introduced, combined with density measures (such as crowding distance), to dynamically balance exploration and exploitation.
- **Archive utilization**: Actively use the archive to record the visited areas and adjust the subsequent search strategy according to the visit frequency.
- **Reduce redundant exploration**: For areas that have been fully explored, further search is reduced, thereby improving the overall efficiency of the algorithm.

### 4. **Experimental Verification**
The paper verifies B - NTGA through two benchmark multi - objective optimization problems (TTP and MS - RCPSP). The experimental results show that B - NTGA has higher efficiency and better performance compared with existing methods, especially in exploring the "gaps" of the Pareto Front.

### Summary
The core problem of this paper is to improve the efficiency and balance of Pareto Front exploration in multi - objective NP - hard combinatorial optimization problems. By introducing the balanced selection mechanism and archive utilization, B - NTGA can explore the Pareto Front more effectively, reduce redundant search, and perform well in practical applications.