Niching Global Optimisation: Systematic Literature Review

Niching in global optimisation refers to a set of techniques designed to identify multiple optimal solutions within a nonlinear, multimodal landscape. These algorithms enhance the exploratory capabilities of conventional metaheuristics by maintaining diversity and supporting coexisting subpopulations across a search space, thereby allowing a more deterministic approach to the true global optimum. Niching algorithms can be categorised into three primary subfamilies: sequential or temporal niching, parallel or spatial niching, and hybrid models which integrate various niching subparadigms. This research paper aims to explore the effectiveness and limitations of different niching algorithms by providing a systematic literature review of the theoretical frameworks within these subfamilies. Eleven major niching native subparadigms have been identified: fitness sharing, crowding, clearing, speciation, restricted tournament selection, clustering, multiobjectivisation, embedded hybrid methods, ensemble hybrid methods, and other hybrid approaches. This study offers a detailed examination of each paradigm’s theoretical foundation, including template algorithmic layouts, and delineates the unique elements of each approach. Research contributions from the inception of niching to 2024 have been aggregated from the SCOPUS database and systematically classified. Data aggregation included journal articles, conference papers, review papers, and research reports published in English only following the PRISMA framework. Application papers with novel theoretical ideas were also taken into account. In all, 203 research works were retained under the inclusion and exclusion criteria. This study concludes with overarching high-level recommendations for future research in modern niching optimisation, emphasising the development of space and time-scalable methods to enhance the adaptability and efficiency of optimisation algorithms in diverse, increasingly multivariable multimodal problems. © 2024 by the authors.