Automated design of spacecraft power subsystems

This paper discusses the application of evolutionary computing to a dynamic space vehicle power subsystem resource and performance simulation in a parallel processing environment. Our objective is to demonstrate the feasibility, application and advantage of using evolutionary computation techniques for the early design search and optimization of space systems. With this approach, engineers specify several sets of conditional subsystem performance criteria to trade off subsystem goals of mass, cost, performance and risk. Once specified, the integrated evolutionary/simulation software will then automatically generate a design option for each criteria, selecting and sizing power elements based on the space system's anticipated performance in the simulated environment. Initial activity plans from two actual JPL missions, Mars exploration rovers (MER) and Deep Impact (DI) are used to test the software. Our results have shown human-competitive advantages by generating credible design concepts much faster than humans are able to and without the need for expert initial designs.