A Robust Coevolutionary Neural-Based Optimization Algorithm for Constrained Nonconvex Optimization.

Area of Science:

• Computational intelligence
• Optimization algorithms
• Neural networks

Background:

• Nonconvex optimization problems often ignore perturbations, increasing system burden and time.
• Existing methods struggle to efficiently manage system perturbations during optimization tasks.

Purpose of the Study:

• To propose a robust coevolutionary neural-based optimization algorithm for handling perturbations in nonconvex problems.
• To enhance the stability and efficiency of optimization by integrating particle swarm optimization with robust neural dynamics.

Main Methods:

• Hybridization of particle swarm optimization (PSO) with robust neural dynamics (RND).
• Neural agents guided by RND replace particles, enabling mutual search and stabilization against perturbations.
• Theoretical analysis to ensure global convergence with probability one.

Main Results:

• The proposed algorithm demonstrates inherent robustness against system perturbations.
• Effectiveness and robustness validated through illustrative examples and comparisons with existing methods.
• Successful application to source localization and redundant manipulator optimization with satisfactory performance.

Conclusions:

• The robust coevolutionary neural-based optimization algorithm offers a superior approach to handling perturbations in nonconvex optimization.
• The integration of RND with PSO provides a stable and efficient framework for complex real-world applications.
• The algorithm effectively addresses internal and exogenous perturbations in source localization and robotic manipulation tasks.