Robust Satisficing Decision Making for Unmanned Aerial Vehicle Complex Missions under Severe Uncertainty

Area of Science:

• Robotics
• Artificial Intelligence
• Decision Theory

Background:

• Unmanned Aerial Vehicles (UAVs) face challenges in complex missions due to environmental uncertainty.
• Existing decision-making methods struggle to balance robustness and performance under high uncertainty.

Purpose of the Study:

• To develop a robust satisficing decision-making method for UAVs operating in uncertain environments.
• To maximize mission robustness while satisfying predefined requirements.

Main Methods:

• Formulation as a robust satisficing optimization problem using info-gap decision theory.
• Construction of an info-gap based Markov Decision Process (IMDP) incorporating Linear Temporal Logic (LTL) for mission specifications.
• A two-stage strategy involving a product IMDP and robust dynamic programming for policy generation.

Main Results:

• A robust satisficing policy was generated, maximizing robustness to uncertain IMDPs while ensuring LTL specification satisfaction.
• Monte Carlo simulations demonstrated effectiveness in UAV search missions under severe uncertainty.
• The proposed policy demonstrated superior uncertainty tolerance compared to existing methods.

Conclusions:

• The novel robust satisficing method effectively enhances UAV mission performance under uncertainty.
• The approach guarantees desired performance levels even with significant environmental uncertainties.
• This method offers a more effective solution for real-world UAV applications.