Adaptive dynamic self-learning grey wolf optimization algorithm for solving global optimization problems and engineering problems | JoVE Visualize

Area of Science:

Computational Intelligence
Metaheuristic Optimization Algorithms
Swarm Intelligence

Background:

The Grey Wolf Optimization (GWO) algorithm is a popular metaheuristic known for its simple structure and efficiency.
However, the standard GWO suffers from low population diversity, susceptibility to local optima, slow convergence, and imbalanced exploration-exploitation.
These limitations hinder its performance on complex optimization problems.

Purpose of the Study:

To propose an improved metaheuristic algorithm, the Adaptive Dynamic Self-Learning Grey Wolf Optimization (ASGWO), to overcome the shortcomings of the original GWO.
To enhance the convergence rate, population diversity, and ability to escape local optima.
To validate the effectiveness of ASGWO on classical test functions and engineering problems.

Main Methods:

Nonlinearizing and segmenting the convergence factor to balance global and local search.
Introducing a dynamic logarithmic spiral path for wolf movement to expand search range and enhance local development.
Designing a dynamic self-learning step size based on evolution success rate and iteration count to prevent oscillations and local optima entrapment.
Proposing a novel position update strategy using global optimum and random positions to increase population diversity and avoid premature convergence.

Main Results:

ASGWO demonstrated superior performance compared to GWO, PSO, and WOA on 23 classical test functions.
The proposed algorithm showed significant improvements in convergence accuracy and speed.
ASGWO exhibited a strong ability to escape local optima and avoid premature convergence.
Effective performance was also observed in engineering applications like the gear train, pressure vessel, and car crashworthiness problems, as well as in feature selection tasks.

Conclusions:

The ASGWO algorithm effectively addresses the limitations of the standard GWO, offering improved optimization capabilities.
The adaptive and self-learning mechanisms significantly enhance convergence speed, accuracy, and robustness against local optima.
ASGWO presents a promising metaheuristic approach for a wide range of optimization problems, including complex engineering applications and feature selection.