Support for Space Architect

What would be the impact of technology resource allocation over a very wide breadth of NASA applications? How could it be done?

This pilot study was undertaken to assist the NASA Space Architect's Capability Requirements Analyses and Integration team (CRAI) in its mission to provide the Space Architect with a quantitative and defendable process to help make investment decisions.

We began with the premise that some kind of systematized optimization would be beneficial to the process of deciding which technologies to fund for development. We conceived of two possible approaches: (1) evaluating technologies grouped by Capability Area, Mission Type, Technology Category, or some other grouping, or (2) pooling as many technologies as possible into one large group for purposes of evaluation, regardless of how they may otherwise fit into the above-mentioned areas, types, or categories, to construct one super-portfolio. Our study compared the two approaches and found that the latter approach produces more technological improvement per investment dollar.

We also developed a spreadsheet tool that the Space Architect can use to assist the decision-making process.

Initial Choices

One possible policy for guiding investment decisions would be to select technologies that would adequately enable a particular mission. An alternate policy would be to treat all technologies or capabilities as equivalent, and invest in those that offer the maximum performance gain per dollar. Or a policy-maker could decide to combine the two in some fashion.

This study took the "democratic" approach. Each technology was considered of equal importance without regard to its utility for a particular mission. Our criterion for evaluating each Capability Area and technology was the amount of improvement it was likely to achieve per dollar invested in its development.

Our model was time-independent -- that is, all funding amounts were grand totals, not expenditures per year. We intend to follow up with a time-dependent model.

Data Collection

We grouped the multitude of technologies being developed at NASA into 21 Capability Areas.

Each Capability Area is divided into Mission Types (Mars Surface, Inner and Outer Solar System, and In-Space Assembly), which are each subdivided into Technology Categories (such as Mobility, Robotic Management, Sample Acquisition, Onboard Science), each of which contain individual technologies (such as aerial mobility, surface mobility, localization, etc.).

We initially focused on one Capability Area: Automation & Robotics. For each Technology Category within that Capability Area, we collected two key sets of data: state-of-the-art (SOA) values, and target values required by current and proposed NASA missions.

This table shows four measures of effectiveness of "Mars Mobility," which is one of the technology areas within the "Automation & Robotics" category. For "Mars Precision EDL, landing ellipse major axis" for example, it will take approximately $2 million to improve it from 200 km at TRL 8 (the current SOA) to 55 km at TRL 4 (the mean target value for the Mars Science Laboratory mission). Note that this particular attribute has a negative polarity, which means that smaller values are better. "TRL" refers to the Technology Readiness Level.

Derivation of a Unitless Metric

To enable us to compare technology areas whose values are measured in disparate units, we developed unitless metrics for each. We divided the value for each mission requirement by the SOA, yielding a unitless measure of how much the current SOA must improve to meet mission requirements.

Then we took the log base 2 of this number (or in the case of items with negative polarity, the negative log base 2), to produce the number of times the SOA must double to meet mission requirements, which is considered a handier way to consider the needed improvement. (Improvement is measured in "bits," where one bit equals one doubling of value.)

If a technology is to be utilized in more than one mission, unitless metrics were calculated according to each mission's requirements, and then added together to produce a weighted value to be used in ranking technologies and constructing development portfolios.

For example, Mars Precision EDL has a target value (for the Mars Science Lab mission) of 55 km and an SOA of 200 km. Dividing 55 km by 200 km yields a unitless value of 0.275. This particular technology area has negative polarity, so we take the negative log2 of this result. The negative log2 of 0.275 is 1.86, which means that the current state-of-the-art must double just under 2 times to match the requirements of the Mars Science Lab mission.

This technology is also applicable to the Mars Sample Return mission, and the unitless metric we calculated to meet the requirements of that mission value is 5.18. Adding the two unitless numbers yields 7.04, the technology's weighted value to be used in evaluating it as a candidate for a development portfolio.

Portfolio Analysis

We used Excel's Solver to optimize portfolios of technology areas, based on expected improvement, at various budget levels. Technologies were assumed to be either funded completely or not at all.

These 14 Technology Categories are the entire set that comprise the Capability Area, "Automation & Robotics" (which is one of the 14 Capability Areas into which NASA's technologies are grouped). Excel Solver calculated that for a budget of $250 million, from among the above list of 14 technology areas, the greatest amount of improvement (measured in "bits," where one bit equals one doubling of value) would be achieved by funding the seven areas that have "yes" in the "funded" column. For that level of funding (actually, $248 million), we could expect technological capability to double 17.48 times. Note that the data presented here is preliminary and subject to change.

Multiple Portfolio Analysis

We took two approaches to optimizing the technologies in the Automation & Robotics and the Formation Flying areas. First, we assigned equal budgets to each of the two Capability Areas and optimized the technologies within each of them, producing two optimized technology portfolios. Then we considered both sets of technologies together, ignoring which area they belonged to, and optimized the resulting portfolio for a budget equal to the combined amounts we had previously assigned to the two Capability Areas.

Our results show that optimizing all the technologies without regard to their areas yields the most total technology improvement (the highest bit return).

The lower two curves represent the Automation & Robotics area funded and optimized independently from the Formation Flying area. The next curve up (with squares) represents the sum of the lower two curves -- that is, the total amount of technology improvement obtained by optimizing the two Capability Areas separately. The top curve (with diamonds) represents the results obtained when all the technologies in those two areas are optimized together in one big pool. We see that the latter method yields the most technological improvement (the highest bit return). If additional Capability Areas were considered similarly, the gap between the two methods would be even more pronounced. (Note that some of the technologies considered here may already be funded.)

Note that the difference between the addition line and the overall optimization line is greatest on the left half of the graph, where funding is a constraint. As the budget increases, the technology portfolio approaches complete funding, and the difference in bit return diminishes.

Future Refinements

This study included two Capability Areas, Automation & Robotics and Formation Flying. Ultimately, we plan to include all 21 Capability Areas in our study.

Including time-based modeling in follow-up studies would allow the recommendation of technology budgets as a function of time. We also plan to introduce uncertainty and risk data, to permit maximizing of technology bit return while minimizing risk and staying within a specified budget.

Various other constraints will be addressed and made user-switchable, so the user can select the constraints to be included in the evaluation. Finally, more measures of portfolio effectiveness will be added, such as the average percent increase in technology level per year that is necessary to achieve mission success.