Resilient UAV Mission Management in Threat-Prone Environments using Markov Decision Processes

This paper proposes a unified Markov Decision Process (MDP)-based framework for resilient mission management of a single unmanned aerial vehicle (UAV) operating in threat-prone and resource-constrained environments. UAV missions are inherently challenged by limited battery capacity, stochastic threats, dynamic goal reprioritization, and potential actuator, sensor, or payload faults. Existing approaches typically address these factors in isolation, resulting in suboptimal decision-making under complex operational conditions.The proposed framework formulates mission execution as a finite-state MDP that integrates post-fault system capability assessment, state-of-charge-based range feasibility, dynamic goal updates, and adaptive navigation mode switching for threat evasion. A structured cost function captures trade-offs among mission commitment, fault mitigation, energy preservation, and risk-aware maneuvering. The optimal policy is computed offline using value iteration and implemented online via a precomputed lookup structure to enable low-latency execution.Simulation results demonstrate adaptive goal commitment, safe recovery under major faults, and recharge-aware planning under energy constraints, thereby enhancing mission reliability and operational resilience.