Using OPN for Comparing Fault-Tolerant GN&C System Architectures
- Paper number
IAC-09.D1.4.2
- Author
Mr. Gregor Hanuschak, Massachussets Institute of Technology (MIT), United States
- Coauthor
Prof. Edward Crawley, Massachussets Institute of Technology (MIT), United States
- Coauthor
Mr. Nicholas Harrison, The Charles Stark Draper Laboratory, Inc., United States
- Coauthor
Mr. Cornelius Dennehy, NASA Engineering and Safety Center, United States
- Coauthor
Dr. Alejandro Dominguez-Garcia, United States
- Coauthor
Dr. Steven Hall, Massachussets Institute of Technology (MIT), United States
- Coauthor
Mr. John West, The Charles Stark Draper Laboratory, Inc., United States
- Year
2009
- Abstract
In the paper, "A Comparison of GN\&C Architectural Approaches for Robotic and Human-Rated Spacecraft," different architectural approaches for fault tolerance in guidance, navigation, and control (GN\&C) systems were analyzed at the topmost level. The study broke down the GN\&C systems into simple subunits, i.e., sensors, computers, and actuators, and analyzed how the components were interconnected. This paper expands upon the previous paper. It uses the previous paper’s subunit / interconnection construct as a top-level abstraction for building a preliminary model of GN\&C system architectures. This model was implemented using the Object Process Network (OPN) modeling language in order to more easily enumerate possible architectures and ultimately identify which architectures have optimal properties. Partial 2x2 systems (i.e., systems with up to dual redundancy per component class for two component classes) and 3x2 systems (systems with up to triple redundancy per component class for two component classes) were modeled in OPN. Within the constraints of these models, all possible architectures were rigorously enumerated and the weight/reliability trade-offs of cross-strapping components and using more than one type of component were assessed. The described models assume perfect coverage – 100\% accuracy in detecting and isolating a failure. The models also assume that more reliable components tend to be heavier, more costly, and/or more complicated to deal with. Given these assumptions, it was found that more reliable components are only beneficial in single string systems or systems with single point failures. All optimal architectures employing component redundancy could be produced from generic connections and the least reliable type of component from each component class. Implementation of the findings in this paper could be extremely beneficial to the development of robotic and human-rated systems for Project Constellation; exploring commonality in GN\&C components can reduce both non-recurring and recurring cost and risk.
- Abstract document
- Manuscript document
(absent)