抄録
Concept of “Autonomous Controller Redesign” is proposed as a new approach to onboard spacecraft control system reconfiguration. In this concept, when some failures occur in control system, “Autonomous Controller Design System” redesigns the control system fully considering given control requirements and constraints. In order for various type requirements and constraints to be taken into account, search-based design procedure is indispensable, but this may cause difficult problem of combinatorial explosion of search space. To solve this problem, knowledge oriented search is utilized, and machine learning is essentially used to acquire such knowledge. Actual computer program has been developed, and its performance is verified by computer simulations. The capability of this system as a reconfiguration tool is also verified by simulations assuming attitude control system with 4 reaction wheels. It is indicated that the new methodology can more flexibly design control systems under the new physical constraints induced by failures.
