A condition-based mission abort policy with random mission duration

The failure of safety-critical systems is catastrophic for their indispensable applications in important fields such as aerospace, automotive, medical devices, and nuclear power. Various mission abort policies have been proposed to enhance system survivability. A commonly used assumption is that the missions executed by the system are of fixed durations, which is unrealistic in many practical situations. This paper proposes a mission abort policy considering random mission durations. The mission duration follows a general distribution and is bounded by a lower and an upper bound. The degradation of the system is characterized as a multi-state continuous-time Markov chain. Operational states are observed at equidistant time epochs, while the failure state is self-announcing. At each decision epoch, the decision-maker must decide whether to abort or continue the mission, with the objective of minimizing the expected total cost over the mission duration. The decision-making model involves various types of costs, including system failure cost, mission failure cost, inspection cost, operation cost, and rewards. Stochastic dynamic programing is used to formulate the optimization problem and find the optimal solution. A numerical experiment is conducted along with a sensitivity analysis. A policy comparison is performed to demonstrate the effectiveness of the proposed policy.