Automated storage and retrieval systems (AS/RSs) usually use a single storage and retrieval (S/R) rule to decide which task should be done first. This approach can not obtain better performance due to lack of adjustment with environment or system changes. In addition, the single rule approach can only obtain the better result for a single performance criterion. This study proposes a different approach which gives an opportunity to dynamically choose a rule from multiple SIR rules when a task decision comes. This research uses the genetic algorithm (GA) approach to develop the dynamic rule which has better performances in both total execution time and total tardiness time. Moreover, this study develops an integrated system which combines a simulation system software with a GA software. The parameters of the simulation model of an AS/RS are studied, namely, number of open storage locations, number of dual commands, number of due-times, and arriving rate of new task. The rules can be chosen are first-come-first-serve (FCFS), dynamic-nearest-neighbor (DNN), and SDDT(Shortest Delivery Due Date).