This paper investigated the integration of Explainable Artificial Intelligence (XAI) models into Unmanned Aerial Vehicle (UAV) systems to enhance decision making in complex and dynamic terrains. The motivation stemmed from the growing reliance on autonomous UAVs in mission critical operations where transparency, trust, and accountability are essential. The study presented a structured overview of various XAI techniques ranging from inherently interpretable models such as decision trees and rule based systems to post hoc methods like LIME, SHAP, and Grad CAM, as well as hybrid approaches including attention based networks and neuro symbolic AI.
Each model was evaluated in terms of its application to UAV tasks such as path planning, obstacle avoidance, mission adaptation, and threat detection. The analysis revealed a persistent trade off between interpretability and performance, with simpler models offering transparency but limited adaptability, and complex models achieving high accuracy at the cost of explainability. Key deployment challenges such as real time latency, computational overhead, lack of standardized datasets, and the stability of explanations under changing environmental conditions were discussed in depth.
The paper concluded by emphasizing the need for lightweight, real time XAI models optimized for edge deployment on UAVs, and highlighted the future potential of end to end explainable pipelines and interactive explanation interfaces to support human AI collaboration. This research contributes to the growing field of trustworthy autonomy in aerial robotics.