An Explainable and Fair Hierarchical RL-Based Alternative Trajectory Proposal Framework for Autonomous U-Plan Approval in U-Space

The growing use of UASs for commercial and recreational purposes presents significant challenges for airspace management, particularly in the approval of U-plans within UTM systems. Manual approval processes are inefficient while existing automated methods frequently lack adaptability, fairness, and transparency. This paper proposes a framework that integrates automated U-Plan authorization with alternative trajectory suggestion, leveraging a Hierarchical Reinforcement Learning (HRL) architecture to suggest fair alternatives instead of outright rejections. Decision-making is structured into three levels: high-level approval, mid-level conflict resolution strategy selection (re-routing or re-scheduling), and low-level trajectory adjustments. The system aims to balance operator needs, airspace rules, and public impact by incorporating factors such as airspace demand, population density, and operator preferences. SHapley Additive exPlanations (SHAP) are used to enhance transparency, thereby fostering trust and fairness. Additionally, a decomposed reward function improves explainability within the framework structure. Simulations demonstrate the framework's scalability, fairness, and efficiency, establishing it as a robust solution for future UTM systems.