Benchmark multi-objective optimisation test problems with mixed encodings

摘要

The fields of multi-objective combinatorial and continuous optimisation have both experienced a significant and rapid growth in publications and research since the turn of the century. Despite this it is more often than not that methods in both fields are applied only to problems of one type of encoding. However, many real-world problems require parameters vectors that use multiple encodings. This paper presents a suite of novel multi-objective optimisation test problems (Exeter1 to 6) with mixed encodings (real and binary) and offer variable correlation between the objectives at different stages of the search. The problems are demonstrated using NSGA-II, SPEA2 and a (μ+λ) Evolution Strategy which were modified to operate on both encodings simultaneously.