Risk-sensitive mean-field control of large-scale UAV Swarms for stochastic disaster tracking and robust first...

Area of Science:

• Robotics and Control Systems
• Artificial Intelligence
• Operations Research

Background:

• Existing methods for multi-UAV disaster response often rely on risk-neutral, deterministic models, limiting their effectiveness in uncertain environments.
• Heuristic, clustering, and swarm-based approaches lack robustness against stochastic uncertainties and communication constraints inherent in disaster scenarios.

Purpose of the Study:

• To develop a novel risk-sensitive mean-field optimal control framework for large-scale multi-UAV disaster tracking and response.
• To address stochastic uncertainty, communication limitations, and enhance robustness in UAV operations.

Main Methods:

• A risk-sensitive performance function with exponential form and modified Riccati recursion is employed.
• Mean-field approximation is used to decouple computational complexity from the number of UAVs, enabling distributed real-time control.
• Entropy-dual formulation ensures distribution robustness against worst-case disturbance realizations.

Main Results:

• The proposed framework demonstrates superior performance in coverage efficiency, mission time, energy consumption, connectivity, path length, and disturbance robustness compared to existing methods.
• Theoretical results on closed-loop Schur stability, risk-admissibility, and convergence to a mean-field equilibrium were rigorously proved and verified.
• Simulations in urban environments validated the algorithm's effectiveness.

Conclusions:

• The developed framework is the first integrated approach to risk-sensitive mean-field control, distributionally robust optimization, and communication-constrained consensus for UAV disaster relief.
• This approach significantly improves the reliability and efficiency of multi-UAV systems in complex disaster scenarios.