Software-based k-out-of-n systems: theory and example

Describes a design concept for constructing a static, component-level k-out-of-n control system (i.e. one in which at least k out of the n modules function correctly), and illustrates it with a spacecraft control problem. The static design is contrasted with a dynamic design in which a centralized controller must be used to perform proper system reconfiguration as actuator failures occur. In the static approach, the control function of the software is distributed to n (nonidentical) modules and no system reconfiguration is needed to tolerate module failures, while in the dynamic approach, central software is used to reconfigure the system as module failures occur. The system reliability of the static approach is compared with that of the dynamic approach, and the design conditions under which one approach may yield better system reliability than the other are identified. Advantages of the static design over the dynamic design are noted