Informetric method for determining the effective drop point of humanitarian cargo from UAVs under limited computational resources.

The subject of the research is the calculation of high-precision drop points for humanitarian cargo from unmanned aerial vehicles (UAVs) in challenging atmospheric conditions and under strict limitations on onboard computational resources. The object of the study is the airdrop process, which includes ballistic, aerodynamic, and informational factors that determine the final trajectory of the container. The author examines in detail aspects such as the integration of differential-geometric modeling of the atmosphere based on the Ricci flow, quantum-inspired global optimization of the drop point, and lightweight neural network-based real-time trajectory correction using the ESP32 microcontroller. Special attention is given to the distribution of computational load between the Raspberry Pi 5 single-board computer, which performs resource-intensive calculations, and the energy-efficient controller responsible for online corrections. Thus, the research aims to establish a unified infometric approach that minimizes the uncertainty of the landing coordinate and ensures a metric level of delivery accuracy for cargo. The research methodology is based on the combination of the Ricci flow for adaptive atmospheric modeling, quantum-inspired particle swarm optimization for CARP search, and TinyML corrections of the cargo trajectory on the ESP32 during descent. The main findings of the study include the confirmed feasibility of metrically accurate airdrop without heavy navigation systems and the demonstration of the effectiveness of the proposed infometric concept QRNA. The author's special contribution to the research is the development of a hybrid algorithm that combines methods of differential geometry, quantum-inspired optimization, and lightweight neural network training, as well as its practical implementation on accessible single-board devices. The novelty of the study lies in the integration of the Ricci flow for dynamic distortion of the metric model of the atmosphere directly in the drop point calculation task and the application of quantum swarm searching in the CARP coordinate space. Additional novelty is manifested in the use of a TinyML network for online trajectory correction of the cargo, which has not been previously applied in the context of humanitarian UAVs. The modeling results demonstrate a reduction in the root mean square error of landing to 0.15 m, which is an order of magnitude better than advanced ML approaches and two orders of magnitude more accurate than classical ballistic methods, confirming the high practical value of the developed algorithm.

Unmanned Aerial Vehicles, Cargo drop point, High-precision drop, Aerodrop, Infometric algorithm, Ricci's flow, Quantum Roe optimization, Neural network control, Limited computing resources, Humanitarian logistics