The rise of unmanned aircraft systems (UAS) for urban drone delivery introduces significant risks, particularly the potential for crash-induced fatalities on the ground. A crucial strategy to address this challenge is through risk assessment and mitigation of flight routes that consider the stochastic nature of urban populations. Traditional strategies treat drone flight route approval and execution independently, which fall short in such a dynamic risk environment where plans deemed safe at the strategic approval stage may later prove hazardous, and vice versa. To address these intricacies, this paper introduces a novel two-stage stochastic optimization model that integrates strategic route feasibility assessment with tactical route selection and timing adjustments. A unique aspect of our model is the implementation of a risk penalty that effectively bridges decisions between the two stages, thereby reducing the likelihood of decision errors caused by stochastic variations. Through extensive simulations within Singapore’s urban context, our model demonstrates a risk reduction by an average of 36.13%, which significantly outperforms traditional methods. This performance consistency across 100 simulated urban scenarios proved the robustness and broad applicability of our model. Furthermore, our model shows an 18% improvement in resolving potential decision errors, with the stochastic solution further affirming a notable risk decrease of 27.18%. Our research enhances the domain of UAS risk-based stochastic decision making and provides opportunities for automated flight approval, drone fleet management, and urban airspace management.

中文翻译：

---

动态地面风险不确定性下的随机路线优化，确保无人机安全投送作业


用于城市无人机送货的无人驾驶飞机系统 （UAS） 的兴起带来了重大风险，尤其是地面坠机事故引发死亡的可能性。应对这一挑战的关键策略是通过风险评估和缓解飞行路线，考虑到城市人口的随机性。传统策略独立处理无人机飞行路线的批准和执行，这在这样一个动态的风险环境中是不够的，在这种环境中，在战略批准阶段被认为安全的计划可能在后来被证明是危险的，反之亦然。为了解决这些复杂性，本文引入了一种新颖的两阶段随机优化模型，该模型将战略路线可行性评估与战术路线选择和时间调整相结合。我们模型的一个独特之处在于实施了风险惩罚，有效地在两个阶段之间架起了决策的桥梁，从而降低了随机变化导致决策错误的可能性。通过在新加坡城市环境中进行广泛的模拟，我们的模型表明风险平均降低了 36.13%，这明显优于传统方法。这种在 100 个模拟城市场景中的性能一致性证明了我们模型的稳健性和广泛的适用性。此外，我们的模型显示，在解决潜在决策错误方面提高了 18%，随机解决方案进一步确认了 27.18% 的显着风险降低。我们的研究增强了 UAS 基于风险的随机决策领域，并为自动飞行审批、无人机机队管理和城市空域管理提供了机会。