RDEx-CASK: Cauchy Mutation, Archive, and Stagnation Kick for RDEx-CSOP

Researchers present RDEx-CASK, an enhanced optimization algorithm that extends RDEx-CSOP with three modifications targeting stagnation issues in constrained single-objective optimization. The method introduces Cauchy-sampled scale factors, a small feasible-only archive, and per-individual stagnation counters that trigger adaptive parameter adjustments, achieving competitive performance on CEC benchmark problems.

RDEx-CASK represents an incremental but methodologically sound advancement in constrained optimization algorithms, focusing on a persistent challenge in evolutionary computation: population stagnation during late-stage convergence. The enhancement demonstrates the discipline of targeted problem-solving within algorithm design, where three specific mechanisms address identified performance bottlenecks rather than attempting wholesale redesign.

The technical context reveals this work builds upon RDEx-CSOP, an existing framework within the broader landscape of differential evolution variants. Constrained single-objective optimization (CSOP) problems remain computationally relevant across engineering, manufacturing, and resource allocation domains. The integration of a JADE-style archive with adaptive sampling probabilities and Cauchy-distributed mutation parameters reflects current best practices in balancing exploration and exploitation during late convergence phases.

The practical impact centers on computational efficiency rather than algorithmic novelty. By reducing convergence time on benchmark problems while maintaining solution quality, the approach offers value to practitioners implementing optimization routines in production environments. The stagnation kick mechanism—triggered after 180 non-improving generations—provides an interpretable recovery strategy that could generalize to other optimization contexts.

The competitive positioning against UDE-III and CL-SRDE on the CEC suite (standardized benchmarks widely used in algorithm evaluation) suggests the work meets current standards but does not substantially exceed them. Future developments might explore scaling to higher dimensions or testing on real-world constrained problems beyond synthetic benchmarks. The modest modifications indicate incremental progress rather than breakthrough performance gains.

• →RDEx-CASK adds stagnation detection and recovery mechanisms to improve late-stage optimization performance on constrained problems.
• →The algorithm maintains competitive performance against existing methods (RDEx, UDE-III, CL-SRDE) on CEC benchmark suites.
• →Three targeted modifications address population stagnation without redesigning core algorithm components.
• →Practical improvements focus on time-to-target efficiency rather than solution quality breakthroughs.
• →The approach demonstrates incremental refinement in evolutionary computation rather than fundamental algorithmic innovation.