Time-varying constraints and other practical problems in real-world scheduling applications

When an evolutionary algorithm is used as an optimizer in a scheduling software application that is destined for use in a real-world commercial setting, a number of time-variability issues are encountered. This paper explores several such issues and other practical problems that arose during the solution of a scheduling application in the area of wine bottling. Each hurdle was addressed by appropriately adjusting the candidate individual representation, the procedure used to decode an individual, or the objective function itself. Addressing these issues is critical when designing and constructing the evolutionary algorithm, in order to ensure that the resulting system is robust enough to meet the demands of day-to-day use. The approach described in this paper has been proven by implementation and vigorous sustained use in a complex business environment.