Multi Threaded Evolutionary Algorithmic Frameworks for Solving Dynamic Optimization Problems

Abstract

Evolutionary computing (EC), a field within Artificial Intelligence (AI), has demonstrated its effectiveness in addressing a wide range of optimization challenges. Evolutionary Algorithms (EAs), the pool of algorithms in the domain of EC, emulate the biological evolution process of diverse species to devise optimization techniques. Over the past two decades, Differential Evolution (DE) and Genetic Algorithm (GA) have emerged as the two most popular and widely used algorithms in the research community. In general, EAs are population-based algorithms that aim to discover the global minimum within a specified search space through an iterative fashion. Integrating EAs with parallel and distributed approaches has demonstrated significant advantages in solving optimization problems across various domains. As a result, the global research directions are increasingly focusing on integrating EAs with parallel and distributed approaches to address optimization problems more effectively. The rapid growth of information technology has introduced multitude of optimization challenges across various domains. The real-world optimization problems often exhibit dynamic characteristics, they are termed as dynamic optimization problem (DOPs). With the constantly changing nature of these problems, rendering traditional approaches to solve them is found ineffective. newlineConsidering the above observation, this research work primarily focuses on addressing the challenges in solving the DOPs. The initial stages of the research work focus on developing a highly parallel and distributed EA (DEA) framework and testing its performance on various computing paradigms, with benchmarking and real-world optimization problems. A multi-threaded distributed evolutionary algorithmic framework (named as MTDEA) is proposed, with its two variants MTDDE (ie., MTDEA with DE in all the nodes) and MTDGA (ie., MTDEA with GA in all the nodes). Further, three fault tolerant algorithms are implemented and plugged in to the MTDEA framework. Then to ensure the robus