Resumen
The miniaturization of electronics, sensors, and actuators has enabled the growing use of nanosatellites for earth observation, astrophysics, and even interplanetary missions. This rise of nanosatellites has led to the development of an inventory of modular, interchangeable commercially-off-the-shelf (COTS) components by a multitude of commercial vendors. As a result, the capability of combining subsystems in a compact platform has considerably advanced in the last decade. However, to ascertain these spacecraft?s maximum capabilities in terms of mass, volume, and power, there is an important need to optimize their design. Current spacecraft design methods need engineering experience and judgements made by of a team of experts, which can be labor intensive and might lead to a sub-optimal design. In this work we present a compelling alternative approach using machine learning to identify near-optimal solutions to extend the capabilities of a design team. The approach enables automated design of a spacecraft that requires developing a virtual warehouse of components and specifying quantitative goals to produce a candidate design. The near-optimal solutions found through this approach would be a credible starting point for the design team that will need further study to determine their implementation feasibility.