Explainable AI for UAV Mobility Management: A Deep Q-Network Approach for Handover Minimization

Black-box AI systems in critical infrastructure need transparency: The paper demonstrates that AI systems managing UAV cellular connections were previously unexplainable "black boxes," making it difficult for network operators to understand or trust automated handover decisions—a significant concern for any AI system controlling essential services.

Explainability tools can make AI decisions understandable without sacrificing performance: By applying SHAP (Shapley Additive Explanations) to a deep learning system, the researchers showed they could identify which specific factors—like signal strength (RSRP) and signal quality (RSRQ)—most influenced handover decisions, making the AI's reasoning transparent to human operators.

Real-world testing validates responsible AI approaches: The framework was validated using actual data from UAV flight trials rather than simulations alone, suggesting that explainability methods can work in practical, deployed systems where reliability and accountability matter.

Interpretability enables better human-AI collaboration: The paper explicitly states their goal was "bridging the gap between AI-driven models and human decision-makers," indicating that explainable AI isn't just about transparency—it's about enabling human operators to work effectively with automated systems and intervene when necessary.

Multiple factors influence critical decisions, and understanding their relative importance matters: By quantifying how different network parameters impact handover choices, the approach helps stakeholders understand whether the AI is making decisions based on reasonable, auditable factors rather than spurious correlations.