Fault-tolerant design optimization: application to an autonomous underwater vehicle navigation system

A systematic, quantitative approach to optimizing a fault-tolerant system design is presented. Sensitivity analyses are used to provide: (1) systematic guidance in selecting the design modifications that are the optimization process, (2) optimization process stopping criteria, and (3) robustness measures for the resulting design. The optimization process permits trades among a variety of free parameters such as component quality, redundancy, operational/maintenance policies, and FDIR (fault detection, isolation, and reconfiguration) processes and their associated coverage probabilities. These are evaluated quantitatively to provide system metrics such as reliability, mission success probability, maintenance actions, and various cost measures. The optimization problem is one where the free parameters are adjusted such that a cost function is minimized (or maximized) without violating the given constraints. A systematic application of the evaluation is shown that enables a design to be optimized with respect to an arbitrary cost function, subject to constraints.<
>