A Survey of Risk-Calibrated Certifiably Safe and Resource-Aware (RCSR) Path Planning for Unmanned Aerial Vehicles

Effective mission planning, path search, and path following are critical for unmanned aerial vehicles (UAVs) operating in complex, dynamic, and resource-constrained environments. Classical path planning approaches, including graph-based search, sampling-based methods, and trajectory optimization, provide structured solutions with performance guarantees but often exhibit limited adaptability to uncertainty, environmental disturbances, and evolving mission constraints. Reinforcement learning (RL) offers a complementary capability by enabling adaptive decision-making and online response to dynamic obstacles and partial observability. This paper examines UAV path planning and navigation within a Risk-Calibrated, Certifiably Safe, and Resource-Aware (RCSR) framework, with emphasis on its implications for mission planning, path search, and path following. Classical planning techniques are reviewed alongside recent advances in RL-based navigation for single-UAV and multi-UAV systems. Particular attention is given to safe reinforcement learning, constrained optimization, and runtime assurance mechanisms that address safety, regulatory compliance, and resource limitations in real-world deployments. Through a comparative analysis of classical, learning-based, and hybrid planning architectures, this work highlights key trade-offs among adaptability, safety, computational cost, and energy efficiency. The paper concludes by identifying hybrid learning–planning approaches as a practical direction for scalable, reliable, and deployable UAV mission planning systems.