Abstract

The design of satellites is a labor-intensive, costly, and time consuming process. It is highly multi-disciplinary, involving several diverse and interdependent subsystems all critical to mission success. In the typical approach, subsystem experts gather and flesh out design concepts over the timescale of weeks, melding analyses from multiple heterogeneous modeling tools and iterating until a sound design is settled upon. However, due to the need for collocation, the incongruity of design information from disparate tools and the inefficiencies inherent in the iterative process, this process remains expensive, errorprone and slow. Technology that can automate and optimize the spacecraft design process to lower costs while maintaining reliability has been identified as one potential solution. However, due to the prohibitively expansive searchable space implied in spacecraft design, most tools of this kind can consider only a very limited subset of design variables and can be unreliable, inefficient and prone to generating unrealistic or suboptimal designs. The Systems Platform for Integrated Design in Realtime (SPIDR) i, a project underway at USC’s Information Sciences Institute and Space Engineering Research Center (SERC), presents a unique approach to this problem. SPIDR’s search and optimization engine utilizes innovative applications of artificial intelligence (AI) techniques to quickly generate provably optimal results. The application of AI planning techniques enables consideration of the entire design space by starting search from a high-level problem specification, while constraint propagation offers an analog to the concept of iteration, where …

Date

2010

Authors

Lucy Hoag, Tatiana Kichkaylo, David Barnhart

Journal

International Conference on Engineering and Meta-Engineering (ICEME), Orlando, FA