L-SAGA: A Learning Hyper-Heuristic Architecture for the Permutation Flow-Shop Problem

Overview

L-SAGA (Learning Simulated Annealing Genetic Algorithm) is a generative hyper-heuristic designed to address the Permutation Flow-Shop Problem (PFSP), an NP-hard problem significant in manufacturing and production environments.

Components

L-SAGA combines two main components:

• Low-Level Genetic Algorithm: Handles job sequencing with operations like initialization, selection, crossover, mutation, and replacement.
• High-Level Simulated Annealing: Optimizes hyperparameters of the genetic algorithm, incorporating a learning component to enhance the search process. The integration of these components allows L-SAGA to efficiently find optimal or sub-optimal solutions for PFSP instances, adapting to various problem sizes and maintaining high solution quality.

Key Features

• Generative Hyper-Heuristic: Automatically generates and selects heuristics for diverse PFSP instances.
• Hybrid Approach: Combines simulated annealing and genetic algorithms to balance exploration and exploitation.
• Adaptability: Can handle various problem instances with different numbers of jobs and machines.
• Performance: Demonstrates potential for high-quality solutions as shown by benchmark tests.