Mission risk control via joint optimization of sampling and abort decisions

Area of Science:

• Decision theory
• Risk management
• Control systems engineering

Background:

• Safety-critical systems require robust risk management during operations.
• Partial observability of system health necessitates adaptive control strategies.
• Existing mission abort policies often lack dynamic adjustment based on real-time information.

Purpose of the Study:

• To develop optimal, dynamic sampling and mission abort policies for partially observable safety-critical systems.
• To minimize the total expected cost, including sampling, mission failure, and system malfunction.
• To integrate partial health information into joint decision-making for sampling and aborting missions.

Main Methods:

• Formulating the problem as a partially observable Markov decision process (POMDP).
• Devising dynamic sampling and mission abort policies based on the system's belief state.
• Analyzing the value function and control limit selection for optimality.

Main Results:

• The proposed policy jointly optimizes sampling and dynamic mission abort decisions.
• Structural insights into the value function and optimality conditions were derived.
• Numerical experiments demonstrated superior performance over heuristic abort policies in controlling mission loss.

Conclusions:

• The developed information-driven approach effectively manages failure risks in safety-critical systems.
• Dynamic, belief-state-based policies offer significant advantages over static or heuristic methods.
• This research provides a framework for enhancing the reliability and safety of complex systems through optimized decision-making.