Dynamic Safety Separation Calibration for Urban UAV Operations Under Stochastic Wind Disturbances and Joint Evolution of Multi-Source Errors

Abstract

To address the problem of safety separation calibration in urban low-altitude airspace, where wind disturbances are significant and UAV traffic density continues to increase, this paper proposes a dynamic safety separation calibration model that accounts for the joint evolution of multisource errors. The model integrates navigation system error, flight technical error, velocity estimation error, and stochastic windinduced disturbances, and establishes a unified representation of the time-varying covariance of position uncertainty. A twodimensional Gaussian probability density integration method is then employed to evaluate collision risk in typical encounter scenarios under a prescribed target level of safety. Simulation results show that wind disturbance can substantially increase the required dynamic safety separation. At a flight altitude of 30 m and under strong wind conditions of $8 ~\mathrm{m} / \mathrm{s}$, the required separation increases from 9.96 m in calm conditions to 30.45 m. Ignoring wind-induced control error leads to an underestimation of the safety boundary by about 55 percent. Sensitivity analysis further indicates that the required separation grows in an approximately log-linear manner as the target safety level becomes more stringent, and increases nearly linearly with the prediction horizon. Omnidirectional scanning results also show that the safety boundary expands significantly with increasing wind intensity. Within the scenario and parameter ranges considered in this study, the selected target safety level and prediction horizon provide a reasonable balance between flight safety and airspace capacity. The proposed method can provide quantitative support for dynamic safety separation calibration in urban low-altitude operations.