Optimizing UAV Payload Delivery Mechanisms: A Systematic Analysis of Mechanical Parameters Affecting Descent Rates

This paper explores unpowered, controlled payload delivery mechanism for UAVs. 108 experimental trials were conducted from a 30-meter drop height. A ML model was developed to analyze and predict how motor RPM, pulley diameter, rope diameter, and payload weight influence descent time of the mechanism. To strengthen the model’s generalizability across a wider range of real-world conditions, physics-based data augmentation was applied to extend the experimental dataset. The analysis identified motor RPM as the most impactful variable, with higher speeds reducing drop time by up to 86.2%. Among the machine learning models evaluated, Gradient Boosting delivered the highest predictive accuracy (R² = 0.9912).