Exploring robustness of plans for simulation-based course of action planning: A framework and an example

Planning requires evaluating candidate plans multicriterially, which in turn requires some kind of a causal model of the operational environment, whether the model is to be used as part of evaluation by humans or simulation by computers. However, there is always a gap - consisting of missing or erroneous information - between any model and the reality. One of the important sources of gaps in models is built-in assumptions about the world, e.g., enemy capabilities or intent in military planning. Some of the gaps can be handled by standard approaches to uncertainty, such as optimizing expected values of the criteria of interest based on assumed probability distributions. However, there are many problems, such as military planning, where it is not appropriate to choose the best plan based on such expected values, or where meaningful probability distributions are not available. Such uncertainties, often called "deep uncertainties," require an approach to planning where the task is not choosing the optimal plan as much as a robust plan, one that would do well enough even in the presence of such uncertainties. Decision support systems should help the planner explore the robustness of candidate plans. In this paper, we illustrate this functionality, robustness exploration, in the domain of network disruption planning, an example of effect-based operations.