'Hang on a minute': Investigations on the effects of delayed objective functions in multiobjective optimization

We consider a multiobjective optimization scenario in which one or more objective functions may be subject to delays (or longer evaluation durations) relative to the other functions. We motivate this scenario from the viewpoint of experimental optimization problems, and derive several simple strategies for dealing with population and/or archive updates under these conditions. These are embedded in a ranking-based EMO algorithm and tested on the WFG test problems augmented with delayed objective(s). Results indicate that good performance can be achieved when the most recently generated solutions are submitted for evaluation on the delayed objective functions, and missing objective values are approximated using a fitness inheritance-based approach. Also, in general one should wait for all evaluations to complete before resuming search if the delay is short, while a non-waiting strategy should be preferred for longer delays. © 2013 Springer-Verlag.