Solving Dynamic Optimisation Problem with Variable Dimensions

摘要

Over the last two decades, dynamic optimisation problems (DOPs) have become a challenging research topic. In DOPs, at least one part of the problem changes as time passes. These changes may affect the objective function(s) and/or constraint(s). In this paper, we propose and define a novel type of DOP in which dimensions change as time passes. It is called DOP with variable dimensions (DOPVD). We also propose a mask detection procedure to help algorithms in solving single objective unconstrained DOPVDs. This procedure is used to try to detect ineffective and effective dimensions while solving DOPVDs. In this paper, this procedure is added to Genetic Algorithms (GAs) to be tested. The results in this paper demonstrate that GAs which use the mask detection procedure outperform GA without it especially Periodic GA 5 (PerGA5).