A comprehensive survey on interactive evolutionary computation in the first two decades of the 21st century

Abstract

Interactive evolutionary computation (IEC) has demonstrated significant success in addressing numerous real-world problems that are challenging to quantify mathematically or are inadequately evaluated using conventional computational models. This success arises from IEC’s ability to effectively amalgamate evolutionary computation (EC) algorithms with expert knowledge and user preferences. These problems encompass the creative and personalized generation of products, art, and sound; the design optimization of communication systems, environments, and pharmaceuticals; and expert support in areas such as portfolio selection and hearing aid fitting, among others. Despite significant advancements in IEC over the past two decades, no major comprehensive survey encompassing all aspects of IEC research has been conducted since 2001. This article aims to address this gap by providing a comprehensive survey and an enriched definition and scope of IEC, along with innovative ideas for future research in this field. The proposed IEC definition more clearly reflects the mechanism and current research status of the IEC. Additionally, the survey categorizes IEC research into five distinct directions from a problem-oriented perspective: interactive evolutionary computation algorithms, IEC algorithm improvements, evolutionary multi-objective optimization with IEC, human perception studies with IEC, and IEC applications. Each direction is meticulously explored, elucidating its contents and key features, while providing a concise summary of pertinent IEC studies. Finally, the survey investigates several promising future trends in IEC, analyzing them through the lens of these five directions and considering the current perspective of computational intelligence, artificial intelligence, and human-machine interaction.