Gravitational Search Algorithms for Optimization Problems

Across the landscape of metaheuristic methods, gravitational search algorithms (GSA) stand out through their foundation in classical physics. Candidates for the optimal solution are conceptualised as particles with masses, whose mutual attraction, governed by an analogue of Newton’s law of universal gravitation, directs the search process. Each particle’s mass reflects its relative fitness, so that heavier particles exert stronger pull, guiding lighter ones towards promising regions of the search space. The iterative interplay of gravitational constant, distance, acceleration and velocity enables a dynamic balance between exploration of new regions and exploitation of known good solutions. This approach has been adapted and extended through chaotic perturbations, manifold learning, fuzzy controllers and hybrid schemes combining particle swarm, genetic and differential evolution strategies. Its flexibility has found application in engineering design, resource allocation, feature selection, neural network training and process parameter tuning. With advances tailored to high-dimensional landscapes, distributed population structures and real-world constraints, GSA remains a vibrant tool for global optimisation challenges, offering a robust framework for both theoretical insight and practical problem-solving.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Recent developments have focused on overcoming the limitations of standard GSA in high-dimensional and complex settings. A manifold-guided variant employs nonlinear dimension-reduction to extract task-relevant subspaces, adapting gravitational coefficients to steer particles along effective directions and suppress noise arising from redundant features. This method has demonstrated accelerated convergence and enhanced accuracy on benchmark and real-world high-dimensional problems. Another strand of research hybridises gravitational dynamics with particle swarm principles, resulting in a PSO-GSA framework that leverages the social learning of particles and gravitational interactions for improved process optimisation. Applications to biodiesel synthesis and industrial parameter identification have shown higher reliability and reduced computational cost. Additionally, a distributed multi-layered GSA introduces hierarchical information exchange among subpopulations, combining a historical knowledge layer with an elite tier to maintain diversity and intensify search around high-quality solutions. Benchmark tests and economic dispatch applications confirm its superior balance between convergence speed and solution quality compared with conventional and contemporary metaheuristics.

Gravitational Search Algorithms for Optimization Problems publication trend

The graph below shows the total number of articles in gravitational search algorithms for optimization problems across all publications each year (not limited to Nature Index journals).

Technical terms

Gravitational Search Algorithm (GSA): A physics-inspired metaheuristic that models candidate solutions as masses whose interactions, based on gravitational attraction, drive the search for optima.

Metaheuristic: A high-level problem-solving framework that guides exploration and exploitation strategies to find near-optimal solutions in complex search spaces.

Exploration: The phase of a search algorithm focused on investigating diverse regions of the search space to avoid local optima.

Exploitation: The phase of a search algorithm that intensifies the search around known good solutions to refine quality and convergence.

Particle: An individual solution in population-based algorithms, characterised by position and velocity in the optimisation landscape.

References

1. A Manifold‐Guided Gravitational Search Algorithm for High‐Dimensional Global Optimization Problems. International Journal of Intelligent Systems (2024).
2. Hybridized Particle Swarm—Gravitational Search Algorithm for Process Optimization. Processes (2022).
3. A Novel Distributed Gravitational Search Algorithm With Multi-Layered Information Interaction. IEEE Access (2021).