Computational Intelligence for Crop Improvement: A Critical Taxonomic Survey of Metaheuristic Optimization

Abstract

Crop improvement combines biological prediction, experimental design and management decisions under constrained resources and environmental variation. This narrative taxonomic survey organizes metaheuristic optimization by problem class, search mechanism, application role and evaluation design. It considers yield prediction, genomic selection, irrigation, pest detection, phenotyping and sensor deployment, while distinguishing optimization of a computational criterion from demonstrated agronomic benefit. A documented genetic-algorithm study of training-population selection illustrates a concrete link between genomic information, a prediction-error criterion and phenotyping decisions. A published crop-rotation framework illustrates scenario-based multi-objective search without establishing a universal algorithm advantage. The survey also audits the interpretation of a retained numerical compilation: missing instance-level records, invalid complete mean ranks, inconsistent sensitivity summaries and unreconciled literature percentages prevent it from supporting inferential conclusions. No new benchmark experiment or validated systematic corpus is asserted. Research priorities are transparent objectives and constraints, independent validation across environments and genotypes, matched computational budgets, reproducible data and uncertainty-aware decision support.