A Generative Modeling Approach to Resource-Efficient Early Mission Concept Formulation

Projects in NASA’s Pre-Phase A aim to define a mission concept that is aligned with program goals, technically feasible, and affordable enough to merit continued development. However, this work is conducted under tight constraints—limited time, budget, and workforce—despite the need to generate, assess, and compare a wide range of mission concepts. Without structured tools to guide this process, this early concept formulation can become inefficient, increasing uncertainty and risk in later phases of mission development. Traditional concept formulation relies on discriminative tools—models that use features of a mission concept (e.g., mass, power) to predict target outcomes such as financial viability (i.e., cost). While effective for evaluating specific designs, these tools do not efficiently support the generation, assessment, nor comparison of a broad range of alternatives. Generative models are a promising alternative for overcoming these limitations. Rather than predicting outcomes from features, generative models identify the accessible design space of features from the desired target outcomes such as technical feasibility and financial viability. By guiding concept formulation to exploration within an informed population rather than by brute force or random trial-and-error search methods, generative models significantly reduce the time, labor, and finances required to identify selectable mission concepts. This paper presents the development and application of generative models at NASA’s Jet Propulsion Laboratory to support early-stage mission concept formulation. These results demonstrate the potential of generative models to transform early mission formulation into a more strategic, resource-efficient, and selection-ready process.