Constraint Solving and Optimizationn Using Nature-Inspired Techniques

Constraint solving and optimization is tackled by scientists in almost every area, including scheduling and planning, configuration, resource allocation, finance, computational biology and machine learning. Since classical systematic and mathematical methods cannot effectively provide suitable solutions for these types of problems, metaheuristic algorithms were introduced. Due to the fact that metaheuristic algorithms have different characteristics, they can address specific optimization problems more effectively than others. This is the main motivation for developing new robust metaheuristic algorithms with the objectives of addressing problems faster specially when they scale up. However, one challenge with metaheuristics is their immature convergence. This issue can be addressed by keeping the exploitation and exploration of the algorithms in balance. Metaheuristics can play a significant role in the area of Constraint Satisfaction Problems (CSPs) which is known as an appropriate framework to formulate and solve many real-world problems under multiple constraints. In order to address the above goals and challenges, this thesis proposes a new nature-inspired technique namely Mushroom Reproduction Optimization (MRO) algorithm. MRO is inspired and motivated by reproduction, growth and migration mechanism of the mushrooms in the nature. To assess the performance of MRO, we conduct several experiments on low and high dimensional benchmark functions and constrained engineering problems. This thesis also introduces a new fuzzy controlling system to tune dynamically the Firefly (FA) algorithm’s parameters in order to keep the exploration and exploitation in balance in each of the searching steps. This will address the immature convergence, a challenging issue in metaheuristic algorithms. To evaluate the performance of the proposed fuzzy FA, we conduct several experiments on high and low dimensional benchmark functions. By taking advantage of the existing powerful metaheuristic algorithms including Particle Swarm Optimization (PSO), Harmony Search (HS), Firefly (FA), Genetic Algorithm (GA), Artificial Bee Colony (ABC) and also MRO, this thesis proposes new efficient methods for solving CSPs and Constraint Optimization Problems (COPs). This thesis also presents new methods for solving Constraint Optimization Problems (COP) including Weighted CSPs (WCSPs) and Dynamic CSPs. A WCSP is a CSP in which preferences between solutions are considered and the optimal solution is the one with minimum weight. And also a DCSP is a CSP which suffers from evolving environment (over the time) and the main consequence of this situation is that the solution for the previous CSP may not be valid for the new CSP. For assessing the performances of the proposed methods in solving CSPs and COPs, we conduct several experiments on the problem instances generated using model RB. The results are appealing and demonstrate the superiority of some methods over others.

Name thesis:degree_name

Doctor of Philosophy (PhD)

Level thesis:degree_level

Doctoral -- first

Discipline thesis:degree_discipline

Computer Science

Grantor dc:publisher

Faculty of Graduate Studies and Research, University of Regina

Year dc:date.issued

2020

Author dc:creator

• Bidar, Mahdi

Advisors dc:contributor.advisor

• Mouhoub, Malek
• Sadaoui-Mouhoub, Samira

Committee members dc:contributor.committeemember

• Louafi, Habib
• Manashty, Alireza
• Volodin, Andrei

Language dc:language.iso

en

DOI

https://doi.org/10.82465/4756

OAI identifier oai:identifier

oai:uregina.scholaris.ca:10294/9164