Hello, I am working on a dynamic scheduling problem and want to compare the performance of different metaheuristics for solving it. However, since rescheduling occurs at multiple points, a new schedule (solution) is selected from the Pareto front at each step.

Because the algorithms are stochastic, I cannot guarantee that they will generate the same solutions or select the same schedule at each rescheduling point. As a result, the next Pareto front—and the entire problem—can change significantly depending on the solution chosen.

Given that the sequence of problems evolves differently for each algorithm, how can I fairly compare their performance if the rescheduling process leads to completely different problem instances?