A Computational Method for Mapping the Decision Space of the Lunar Exploration Program
- Paper ID
5030
- author
- company
Massachussets Institute of Technology (MIT);
- country
United States
- year
2006
- abstract
This study uses a novel computational technique to identify, rank, and categorize key decisions and uncertainties in NASA's Lunar Exploration Program. Using a computational tool, Object-Process Network (OPN), our method is capable of dealing with hundreds of decision variables in the lunar program. These interacting decision variables can be both socio-political in nature, or involving technological expertise. For example, one of the most impacting (high-priority) Lunar Exploration Program decisions identified by this method is ``specifying the role of international participants for lunar hardware development”. This socio-political decision should be resolved early in the program because it is both nearly impossible to change and strongly impacts technical variables such as the design of the vehicle's sub-system interfaces. This paper discusses both the high- and low-priority decisions and the insights that are gained from using this computational architecting technique. There are three major elements of this approach. First, the solution space is described in a meta-language, which models the solution space as a finite number of solution space partitions using decisions variables as the branch points. These decisions variables are compiled into an executable meta-model, stored in a format called decision graph. Second, the decision graph is enumerated into a finite number of feasible choices and outcomes. Third, the graph is iteratively sorted and reduced by a combination of human interaction with the graph, simulation techniques and graph-reduction algorithms. The sorting and reduction processes simultaneously prune away obviously poor branches of the decision space and ranks the decisions and uncertain variables in the order of most-impacting to least-impacting. To present the results to a wide range of decision-makers, we have a flexible information visualization interface to show the decision graph in multiple formats. The current reporting interfaces include DSM, Decision-Tree, State-Transition Diagram, Equation Form, and dynamically customizable forms. These human-machine interfaces provide a bi-directional mechanism that allows different kinds of decision-makers to share and present their views and inputs via a common, synchronizable information system. This study extends a previous method using OPN that enumerated over a thousand mission mode options for Moon and Mars Exploration Architectures. The updated method implements new algorithms such as Reduced Ordered Binary Decision Diagrams (ROBDDs) and AND/OR Search Trees to expedite the computational tasks in reasoning about the interactive effects of decision variables.