The START team has been recruited by various NASA entities to conduct two basic types of analysis. One is to assemble an optimal portfolio of technologies (or technology sets or missions, etc.) to be funded for development, given a specific data set and objective. The other is to allocate tasks between human astronauts and their robotic assistants, and/or to schedule those tasks optimally, given a specific data set and objective. The analyses are similar in that it is necessary first to capture in a database the considerations on which the decisions should be based. If the results of the analysis do not seem right to the decision-makers, the database should be revisited and modified if necessary. But each type of analysis has its own methodology.
Please note that some studies may require only part of this sequence, and that the direction of the sequence may change according to the needs of the particular study. For example, the developer of a particular nanotechnology might want to know how it could be put to use in NASA's programs. In such a case, we would employ a "bottom-up" approach, beginning at Step 4 and working upward to Step 2. Frequently, we are called upon to split the difference -- working top-down until we've derived the capability requirements for a particular mission, then switching to bottom-up to identify the capabilities of a particular set of technologies that were funded as basic research. (The case study, "Rover Autonomy Study #2," is a good example of this approach.) The action lies in matching capabilities with requirements.
- Develop a clear, complete statement of the problem to be studied.
The problem needs to be stated unambiguously, specifying what is to be maximized (such as science return) or minimized (such as cost or time), with all pertinent policy, schedule, and budget constraints. We probe to uncover any unstated assumptions that need to be taken into account, since unarticulated assumptions can undermine a study.
- Identify the decision-maker's goals and priorities and the associated metrics.
This includes relative priorities or a range of relative priorities among multiple goals, and is used to define an objective function that maximizes the overall value of the investment subject to specified constraints.
- Design or select one or more architectures (precise scenarios) to accomplish the goals.
A study may address mission architectures, system architectures, or both. For example, a mission architecture might include launching a spacecraft, landing it safely in a certain location on Mars, having a rover disembark and travel to where scientists suspect a pool of underground water, drilling to a depth of 1 km, retrieving a sample, analyzing the sample for signs of life, and reporting the results to Earth. A system architecture may be limited to the design and functions of the rover.
- If working with a mission architecture, allocate its constituent activities to the available agents (e.g., astronauts and robots) and resources, and calculate the optimal scheduling.
- Identify agents (astronauts and robots), activities (move, carry, deploy, etc.), and resources (tools, vehicles, power, time, etc.). The various information components, which are defined by the user, are described in a hierarchical structure.
- Identify parameters and constraints.
- Define the figure of merit (FOM) to be optimized (i.e., the objective function discussed in step 2).
- Define the starting configuration state, S (e.g., astronauts in the habitat module, pressurized vehicle docked to the habitat, none of the tasks accomplished, etc.).
- Define the goal configuration state, G (e.g., all tasks accomplished and astronauts back in the habitat module, etc.)
- Search for the optimal allocation of tasks to agents and the optimal sequence of events, given the figure of merit and all appropriate constraints. Our planner (called HURON) is based on the A* least-cost search algorithm. Various search techniques, including a hash map, min heap queue, and cycle loop detection, have been employed to shorten the search rate from as much as one node per second to about 100-200 nodes per second, depending on the computer.
- Starting from S, generate all the new possible configurations (subject to pruning techniques that expedite progress toward the optimal solution).
- Evaluate each new configuration using FOM. Select the best alternative that does not violate any constraint.
- Repeat until G is reached. This process generates a tree. The optimal task allocation and associated information are given by the path between S and G.
- The result is presented as a timeline that shows what tasks are executed, when, and by which agents.
- Adjust inputs as desired and repeat the analysis process until a satisfactory result is achieved.
- Identify and assess the capabilities and/or technologies required by the architecture.
- Determine the capabilities required to conduct the activities. Identify interdependencies between capabilities. For example, a Mars rover's sample-acquisition capability depends on coordination of its sensing and manipulation capabilities.
- Assign a value to each activity, based on its contribution to mission goals. Map activities to capabilities and derive a value for each capability based on the activities it supports. This provides unitless values that enable comparison of capabilities (and/or technologies) with dissimilar units (mass in kg, volume in cm3, cost in dollars, etc.).
- Divide capabilities into enabling (required) and enhancing (not required but helpful). Enabling capabilities must be fully funded or the mission cannot operate. Enhancing capabilities may be partially funded and still provide value.
- If desired, identify and assess technology candidates which purport to fulfill or partially fulfill the required capabilities.
- Characterize the capabilities and/or technologies.
A variety of metrics are used, including the state of the art, desired performance levels, development cost and risk, etc. We capture uncertainties in the capabilities, using performance attributes and their probability distributions.
The START team can also help sponsors identify the time horizon they wish to target for development of their capabilities and/or technologies.
- Evaluate and rank the capability or technology candidates to identify which to fund for development.
Using the START optimization tool, compute optimal portfolios of capabilities and/or technologies for the range of investment budgets and timelines that are of interest to the decision-maker, including (if desired) the optimal funding schedule to make the most efficient use of any expected variations in the available budget during the development process.
- Assess the relative return on investment for competing architectures (if comparing architectures).
Knowing which capabilities/technologies can be supported given a particular set of constraints (result of step 6), compute the relative values of the activities expected to be actualized. Estimate the life-cycle cost for each of the mission concepts which can be technologically enabled. Assess the relative productivity (expected value of activities divided by life-cycle cost) for the concepts being assessed to determine the relative return on investment.
- Validate the results.
Establish the robustness of the results through a validation process that includes both consistency checks of the data and automated sensitivity analyses of the results. For more on validation, click here.
- Recommend an optimal portfolio and present its trade-offs to the decision maker.
We present results in both tabular and graphical forms.
- Adjust inputs as desired, and repeat the analysis process until a satisfactory result is achieved.
Maintain an optimal portfolio as technologies mature and customer requirements change.
