Characterizing Buffer Overflow Constraints in UAV Data Collection with Energy-Harvesting Sensors

This paper studies UAV-assisted data collection from battery-less, energy-harvesting (EH) sensors in precision agriculture, with a focus on how sensor buffer overflow constraints interact with UAV visitation frequency and mission cost. While minimizing the computing resources of a sensor reduces cost and provides better sustainability in precision agriculture, it also poses several problems, most notably the overflow of limited data storage. Instead of relying on cost-heavy multi-hop networking, unmanned aerial vehicles (UAVs) can be utilized to collect data from the sensors through a come-and-collect strategy, minimizing the load of sensor devices. However, this in turn increases the frequency and cost of UAV mission flights. This paper characterizes buffer overflow of battery-less sensors through conservative approximation of the steady-state overflow probability under different EH levels, and analyzes its impact on feasible visitation intervals of UAVs. Through our evaluation, we validate that EH prediction accuracy is critical for guaranteeing buffer stability, as the prediction error increases, the overflow probability rises, even when visitation intervals are based on the analytical overflow bound. In contrast, the overall UAV mission cost remains relatively insensitive to prediction inaccuracy under the considered cost structure, as the flight-cost component dominates the visit-frequency-dependent term.