Switching from exploration to exploitation in gravitational search algorithm based on diversity with Chaos

The gravitational search algorithm (GSA) is one of the most promising algorithm in the physics-based metaheuristics category. However, GSA suffers from premature convergence due to rapid reduction in diversity, whereas a chaotic gravitational search algorithm (CGSA) can degrade the convergence speed and exploitation power. To address these issues, the current study proposes an algorithm that enhances the exploration capability of GSA using a disruption strategy with chaotic dynamics. If no significant change is observed in diversity values during the initial stages of the search process, disruption is performed using a sigmoid function. Then, the search process executes gradual exploitation using a sigmoid function without chaotic dynamics. The proposed algorithm is tested with GSA, CGSA, and PSO (particle swarm optimization) on 28 benchmark functions. It is observed that the algorithm outperforms GSA and PSO in 19 cases and CGSA in 20 cases. Diversity analysis shows that the algorithm generates superior exploration versus exploi-tation percentage with improved mean diversity values. To determine its robustness, the algo-rithm is applied to four unconstrained engineering problems. The results suggest that the algorithm can solve practical engineering problems in a reasonable number of iterations.