Event-triggered dynamic seed invasive weed optimization (ET-DSIWO): a nature-inspired approach for non-stationary optimization

This paper addresses the challenge of optimizing complex, dynamic environments where traditional optimization algorithms struggle to adapt efficiently to changing conditions. In a dynamic environment, conventional metaheuristic algorithms often exhibit premature convergence and lack the flexibility needed to maintain good performance, resulting in suboptimal solutions and inefficiencies. We propose a nature-inspired optimization strategy, the Event-Triggered Dynamic Seed Invasive Weed Optimization (ET-DSIWO) algorithm to overcome these issues. Drawing on principles of seed longevity, dormancy, and the natural balance between transient and persistent seed banks, ET-DSIWO introduces a dynamic seed management system designed to enhance adaptability and resilience in nonstationary environments. Unlike the Static Generation Production (SGP) model used in traditional Invasive Weed Optimization (IWO), our dynamic seed management system ensures a diversified and actively managed pool of solutions across multiple generations. By leveraging an innovative event-triggered re-evaluation mechanism, ET-DSIWO continuously monitors environmental shifts and adjusts search strategies accordingly, avoiding the common pitfall of being trapped in previous minima. This mechanism allows ET-DSIWO to dynamically balance exploration and exploitation, supporting a more comprehensive search and addressing the convergence limitations found in other methods. Extensive comparative evaluations validate that ET-DSIWO consistently achieves the highest Friedman rank across 12 benchmark functions, outperforming several recent meta-heuristic algorithms. These results affirm ET-DSIWO’s superior robustness, resilience, and adaptability, demonstrating its efficacy for real-world applications that require flexible optimization under fluctuating conditions.